journal of pipeline engineering - pipemag.com - dec09 for... · leigh fletcher, miab technology pty...

December, 2009 Vol.8, No.4

ScientificSurveys Ltd, UK

ClarionTechnical Publishers, USA

Journal ofPipeline Engineering

incorporatingThe Journal of Pipeline Integrity

Sample

copy

Journal of Pipeline Engineering

Editorial Board - 2009

Obiechina Akpachiogu, Cost Engineering Coordinator, Addax PetroleumDevelopment Nigeria, Lagos, Nigeria

Mohd Nazmi Ali Napiah, Pipeline Engineer, Petronas Gas, Segamat, MalaysiaDr Michael Beller, NDT Systems & Services AG, Stutensee, Germany

Jorge Bonnetto, Operations Vice President, TGS, Buenos Aires, ArgentinaMauricio Chequer, Tuboscope Pipeline Services, Mexico City, Mexico

Dr Andrew Cosham, Atkins Boreas, Newcastle upon Tyne, UKProf. Rudi Denys, Universiteit Gent – Laboratory Soete, Gent, Belgium

Leigh Fletcher, MIAB Technology Pty Ltd, Bright, AustraliaRoger Gomez Boland, Sub-Gerente Control, Transierra SA,

Santa Cruz de la Sierra, BoliviaDaniel Hamburger, Pipeline Maintenance Manager, El Paso Eastern Pipelines,

Birmingham, AL, USAProf. Phil Hopkins, Executive Director, Penspen Ltd, Newcastle upon Tyne, UK

Michael Istre, Engineering Supervisor, Project Consulting Services,Houston, TX, USA

Dr Shawn Kenny, Memorial University of Newfoundland – Faculty of Engineeringand Applied Science, St John’s, Canada

Dr Gerhard Knauf, Mannesmann Forschungsinstitut GmbH, Duisburg, GermanyLino Moreira, General Manager – Development and Technology Innovation,

Petrobras Transporte SA, Rio de Janeiro, BrazilProf. Andrew Palmer, Dept of Civil Engineering – National University of Singapore,

SingaporeProf. Dimitri Pavlou, Professor of Mechanical Engineering,

Technological Institute of Halkida , Halkida, GreeceDr Julia Race, School of Marine Sciences – University of Newcastle,

Newcastle upon Tyne, UKDr John Smart, John Smart & Associates, Houston, TX, USA

Jan Spiekhout, NV Nederlandse Gasunie, Groningen, NetherlandsDr Nobuhisa Suzuki, JFE R&D Corporation, Kawasaki, Japan

Prof. Sviatoslav Timashev, Russian Academy of Sciences – Science& Engineering Centre, Ekaterinburg, Russia

Patrick Vieth, Senior Vice President – Integrity & Materials,CC Technologies, Dublin, OH, USA

Dr Joe Zhou, Technology Leader, TransCanada PipeLines Ltd, Calgary, CanadaDr Xian-Kui Zhu, Senior Research Scientist, Battelle Pipeline Technology Center,

Columbus, OH, USA

❖ ❖ ❖

Sample

copy

4th Quarter, 2009 225

The Journal ofPipeline EngineeringincorporatingThe Journal of Pipeline Integrity

Volume 8, No 4 • Fourth Quarter, 2009

Contents

Roland Palmer-Jones, Susannah Turner, and Dr Phil Hopkins ......................................................................... 229A new approach to risk-based pipeline-integrity management

Leanne M Tindall, Dr Julia M Race, and Jane Dawson ....................................................................................... 241Investigating the relative severity of dents in pipelines based on magnetic-flux-leakage inspection data

Robert B Francini and William Nik Baltz ............................................................................................................ 253Blasting and construction vibrations near existing pipelines: what are the appropriate levels?

Enrique Acuña C .................................................................................................................................................... 263Natural gas distribution integrity management in Chile: a new way of doing things

Professor Andrew Palmer and Andrew Ngiam .................................................................................................... 267Developing extraordinary talent

Gjertrud Elisabeth Hausken, Jørn Yngve Stokke, and Steinar Berland ............................................................. 271Designing offshore pipeline safety systems utilising flow and pressure in multi-design-pressure pipeline systems

F Podbevsek, H J Brink, and J Spiekhout ............................................................................................................ 283Horizontal directional drilling: the influence of uplift and downlift during the pull-back operation

❖ ❖ ❖

OUR COVER PICTURE shows the pipeline-repair system developed by Subsea 7 for a 1350-m deep offshorepipeline repair for Total Angola in its Girassol field. The repair to the12-in water-injection pipeline was recently

satisfactorily completed and the pipeline is now back in service after comprehensive testing.

Sample

copy

The Journal of Pipeline Engineering226

1. Disclaimer: While every effort is made to check theaccuracy of the contributions published in The Journal ofPipeline Engineering, Scientific Surveys Ltd and ClarionTechnical Publishers do not accept responsibility for theviews expressed which, although made in good faith, arethose of the authors alone.

2. Copyright and photocopying: © 2009 Scientific SurveysLtd and Clarion Technical Publishers. All rights reserved.No part of this publication may be reproduced, stored ortransmitted in any form or by any means without the priorpermission in writing from the copyright holder.Authorization to photocopy items for internal and personaluse is granted by the copyright holder for libraries andother users registered with their local reproduction rightsorganization. This consent does not extend to other kindsof copying such as copying for general distribution, foradvertising and promotional purposes, for creating newcollective works, or for resale. Special requests should beaddressed to Scientific Surveys Ltd, PO Box 21, BeaconsfieldHP9 1NS, UK, email: [email protected].

3. Information for subscribers: The Journal of PipelineEngineering (incorporating the Journal of Pipeline Integrity)is published four times each year. The subscription pricefor 2009 is US$350 per year (inc. airmail postage). Membersof the Professional Institute of Pipeline Engineers cansubscribe for the special rate of US$175/year (inc. airmailpostage). Subscribers receive free on-line access to all issuesof the Journal during the period of their subscription.

4. Back issues: Single issues from current and past volumes(and recent issues of the Journal of Pipeline Integrity) areavailable for US$87.50 per copy.

5. Publisher: The Journal of Pipeline Engineering ispublished by Scientific Surveys Ltd (UK) and ClarionTechnical Publishers (USA):

Scientific Surveys Ltd, PO Box 21, BeaconsfieldHP9 1NS, UKtel: +44 (0)1494 675139fax: +44 (0)1494 670155email: [email protected]: www.j-pipe-eng.com

www.pipemag.com

Editor and publisher: John Tiratsooemail: [email protected]

Clarion Technical Publishers, 3401 Louisiana,Suite 255, Houston TX 77002, USAtel: +1 713 521 5929fax: +1 713 521 9255web: www.clarion.org

Associate publisher: BJ Loweemail: [email protected]

6. ISSN 1753 2116

THE Journal of Pipeline Engineering (incorporating the Journal of Pipeline Integrity) is an independent, international,quarterly journal, devoted to the subject of promoting the science of pipeline engineering – and maintaining and

improving pipeline integrity – for oil, gas, and products pipelines. The editorial content is original papers on all aspectsof the subject. Papers sent to the Journal should not be submitted elsewhere while under editorial consideration.

Authors wishing to submit papers should send them to the Editor, The Journal of Pipeline Engineering, PO Box 21,Beaconsfield, HP9 1NS, UK or to Clarion Technical Publishers, 3401 Louisiana, Suite 255, Houston, TX 77002, USA.

Instructions for authors are available on request: please contact the Editor at the address given below. All contributionswill be reviewed for technical content and general presentation.

The Journal of Pipeline Engineering aims to publish papers of quality within six months of manuscript acceptance.

Notes

v v v

www.j-pipe-eng.comis available for subscribers

Sample

copy


Editorial

IT SEEMS A PARADOX: 10-15 years ago, as intelligentinspection was developing and the science was coming

into its own, a major point of discussion among all involvedwas to do with the technology itself. The big question was“high resolution or low resolution?”, relating to the capacityof the equipment to detect pipewall features. Putting thisanother way, did the operator want a ‘quick-and-dirty’ (andtherefore cheap) inspection, or was the Full Monty required?The issue revolved around the capacity of the intelligentinspection tools to inspect, and the cleanliness of thepipeline that the tool was to inspect was sometimesconsidered of less significance.

Now, however, intelligent-inspection tools are of anundreamt-of greater sophistication, and the general questionbeing asked is no longer to do with their capacity accuratelyand precisely to detect features, but it’s to do with howclean the pipeline is – a far more basic issue, and one thatdoes not take advanced equipment to resolve. A pipeline’sinternal cleanliness has, quite properly, become a questionof great significance: nevertheless, there are no publishedstandards of ‘cleanliness’ and although there are many waysin which deposits can be removed from a pipe wall, ensuringa pipeline is clean enough for an inspection to be carriedout remains very much a subjective process. It has oftenbeen said that the best cleaning tool is a magnetic-fluxleakage intelligent pig, and this remains true, although it isalso the most expensive. While it is clear that each pipelineis different, and its internal cleanliness is very muchdependant on the physical conditions of the fluid or gasthat it is transmitting, it seems surprising that the industryas a whole has not found it possible to establish some basicguidelines for achieving cleanliness. Under normaloperating conditions, minimization of pipewall depositswill obviously improve the flow conditions as well as thepipeline’s overall efficiency and cost-effectiveness, to saynothing of the effect on reducing the potential for corrosion.When it’s time for an inspection, deposits and other debris

Pipeline cleanliness and black powder: an increasing issuefor gas pipelines

must be removed, both to ensure that the tool’s sensors canhave unimpeded access to the pipe wall, and to remove thepossibility of debris clogging-up the tool, and even causingit to become stuck.

The question of ‘how clean is clean?’ is not unfamiliar and,in fairness, is being asked increasingly more frequentlynowadays. At three recent pipeline-industry events that theJournal of Pipeline Engineering has attended (in Ostend,Pittsburgh, and Aberdeen), the question was once againraised, although once again there were no particular answersother the general advice of establishing and maintaining aregular cleaning-pig programme, which would be enhancedprior to an inspection run.

One of the most pernicious cleaning problems for gaspipelines is the formation and accumulation of so-called‘black powder’. This material, which is as fine as flouralthough far more dangerous because it is pyrophoric, isone of the least understood but most prominentcontamination problems in gas pipelines. Black powder isthe name given to the mixture of iron oxides, carbonates,and sulphides found in gas lines; it can also incorporatesalt, sand, clay, mineral scales such as calcium carbonatesand gypsum, strontium and barium sulphates, metal powder,welding flux, and stubs of welding rods. The sources ofblack powder include millscale, corrosion products, saltsand scales from gas wells and wet gas gathering systems, andatmospheric corrosion, and the variability of its compositionis illustrated by reports of the powder ranging from beingcompletely iron sulphide to completely iron oxide.

Black powder can cause product quality problems andexcessive wear and erosion on internal pipe walls and manyother pipeline components including compressors, turbines,and valves. The accumulated solids can plug small orificesand consequently affect measurement equipment and, asthe particles settle out of the gas stream, they can fill-in

Sample

copy


surface pits and other internal pipe-wall anomalies,preventing accurate inspection; in sag bends, these build-ups can harbour corrosive bacteria.

Although difficult to deal with, the problems cased byaccumulations of this material can be overcome, as canmost pipeline problems, by careful planning and attentionto detail. Dr John Smart points out that the problem ofblack powder is as much one of powder movement as oneof the presence of the material in the first place. Oil and gaspipeline cleaning chemistries have been developed thathave superior cleaning characteristics that can dissolve theoil, glycol, or paraffin matrices that can hold black powderparticles to pipe walls. After being dispersed, the particlescan be pigged out of the pipeline, and frequent pigging andchemical treatments with biocides and corrosion inhibitorscan control internal corrosion caused by solids’ deposition.

Removal of the black powder from the pipeline is not theend of the affair: as the material is hazardous, necessaryarrangements for its disposal must be made, and obviouslythese should be in place before any pigging operationsbegin.

A special session on this problem is being organized at thePipeline Pigging and Integrity Management (PPIM)conference being held in Houston on 17-18 February, andorganized by our sister publication Pipelines Internationaland Clarion Technical Conferences (see www.clarion.org),and other papers at the event will also address the issue. Themore the subject of ‘how clean is clean?’ can be discussed,the more likely it is that shared experiences can lead to ashared solution; at the very least, ‘clean’ needs to be kept inthe spotlight of pipeline integrity management andoperations.

Funding announcedfor pipeline research

THE PIPELINE Research Council International (PRCI)2010 research programme has been approved by its

Board of Directors at a meeting in Banff, Alberta, Canada.Just over $8 million in funding has been allocated to the2010 programme, which includes research into pipelinecorrosion, integrity, operations, design, materials,construction, facilities, and underground storage. PRCIChairman Paul MacGregor said “This commitment of the

PRCI membership in the face of challenging financialconstraints in all sectors of the global economy, demonstrateshow industry leadership and collaboration can meet thechallenges facing the energy pipeline industry.

“The projects selected by our members reflect theircommitment to the safe, efficient, and reliable operation ofenergy pipelines worldwide,” Mr MacGregor continued. Inaddition to core programmes, the PRCI will also undertakea range of initiatives in 2010, many in conjunction with itsEuropean and Australian research partners. Some of theinitiatives include:

• The development of a co-ordinated research plan toaddress the specific technical issues associated withcarbon dioxide transport

• The development of guidelines to assist operators inevaluating damage to subsea pipelines

• The development of a comprehensive approach forthe management of unpiggable pipelines will beexpanded to include a ‘Base Resource’ document toprovide guidance for inspection and assessment ofunpiggable pipelines – a keynote paper on thisproject will be given at the Pipeline Pigging andIntegrity Management conference (see above).

• Improving the understanding of data produced bypipeline inspection technologies, with an emphasison determining the effects of uncertainty in thedata

• The re-examination of standard emissions factorsfor compressor station fugitive emissions, and thedevelopment of improved fugitives measurementmethods and reporting procedures

• Improving compressor and pump station facilityintegrity to evaluate the effects of vibration onstation piping components, and improve themethods and practices used for bolted joints.

As many readers will know, the PRCI is a non-profitcorporation comprised of 34 energy pipeline operatingcompanies located in the United States, Europe, Canada,South America, and the Middle East. The Council hasrecently moved its US-based administration headquartersfrom Arlington to Falls Church – see www.pric.org.

Sample

copy


PIPELINE-INTEGRITY MANAGEMENT requires theconsideration of many factors that may cause the

degradation of the pipeline: for example, externalinterference damage, external corrosion, environmentalcracking, fatigue loading. Action must be taken to preventor limit degradation, and optimize inspection andmaintenance. In doing this it is important to consider theconsequences of a failure.

Integrity management involves consideration of pipelinedesign, operation, leak detection, emergency response,training, etc. The approach to integrity management thatconsiders both the probability or likelihood of failure and

the consequences of a failure is known as ‘risk-based’.When the focus of an integrity-management system isinspection, it is known as ‘risk-based inspection’ (RBI).

Recent changes in legislation in the USA [1] have led to anincreased use of risk-based integrity management forpipelines. In the UK a risk-based approach to pipeline-integrity management has been included in legislation andwidely applied since 1996 [2]. Risk-based integritymanagement includes the following basic elements:

• Data collection and integration – to facilitate a riskassessment.

• Hazard identification – hazards that may result in arupture, leak or loss of serviceability are identified.Hazards typically include corrosion, third-partyinterference, ground movement, manufacturingdefects, mal-operation, etc.

A new approach to risk-basedpipeline-integrity management

by Roland Palmer-Jones*, Susannah Turner, and Dr Phil HopkinsPenspen Integrity, Newcastle upon Tyne, UK

THE USE OF risk-based integrity-management systems for pipelines is increasing in popularity, and nowchanges in legislation in the USA require operators to use risk assessment in high-consequence areas.

The methodologies used range from point-scoring qualitative schemes to detailed quantified systemsrequiring structural reliability analysis, release modelling, and post-incident behavioural modelling.

In the UK a risk-based approach to pipeline-integrity management has been included in legislation since1996, and is widely used. Experience with implementing systems and applying them to on- and offshorepipeline systems has led to the following conclusions: point-scoring systems cannot replace expertknowledge; point-scoring systems always need to be modified to suit a particular system and need updatingas parameters change; detailed automated systems generate a huge number of sections and range of risks– this can be confusing and cannot easily be accounted for in inspection planning; a clear link between risksand inspection or monitoring is needed; and simplicity and flexibility are critical.

This paper describes a radical new approach to using risk assessment for pipeline-integrity management.This new approach focuses on identifying whether hazards are time dependant (such as corrosion) orrandom (such as third-party damage), and then either estimating a time to failure or a probability ofoccurrence. These estimates can be based on experience, history, or specific detailed studies. The effectof inspection and monitoring is also considered. This methodology allows the user to manage the risksassociated with their pipeline in a way that is flexible, rational, consistent, and can be readily understoodby othersT:. It also allows the reasons for decisions regarding inspections to be recorded, and new usersto quickly learn the key safety issues for the pipeline.

*Author’s contact details:tel: +44 (0)191 238 2200email: [email protected]

Sample

copy


• Consequence evaluation – the consequences of arupture, leak, or loss of serviceability are evaluated.Consequences may include loss of life or injury,environmental damage, loss of revenue, damage toproperty, damage to reputation, etc.

• Section selection – the pipeline system is dividedinto sections where hazards, or consequences, aredifferent from the hazards or consequencesassociated with other sections. For example it iscommon to evaluate on- and offshore pipelinesseparately.

• Risk analysis – the probability of failure due to ahazard and the consequences of that failure areevaluated and multiplied together give a measure ofthe risk for each hazard. The risks for each hazardmay then be combined to give an overall estimate ofthe risk level for each section.

• Risk assessment – the calculated risk is comparedagainst an acceptable or target risk level or benchmarkvalue to determine the high risk sections/pipelines/hazards.

• Mitigation – a plan is made to control the risksidentified. This is a critical stage and must be linkedclearly to the relevant hazards.

• Review and update – the process is continuous andthe results of inspection and maintenance activitiesmust be for a repeat analysis.

The risk-management process is shown graphically in Fig.1.

There are a variety of different systems in use for conductingrisk assessments on pipelines. The systems that are used canbe placed into three generic methodologies:

• ranking• point scoring• quantified

In this paper these three methodologies are evaluated, anda logical methodology that can utilise the best of all threeapproaches is developed.

Ranking systemsRisk-ranking systems are simple and flexible. Crediblehazards for a pipeline are identified by an expert, or teamof experts, and the relative probability of failure for eachhazard is ranked, typically as high medium or low. Forexample, the probability of failure due to internal corrosionfor a flowline carrying oil with a high water cut, at hightemperature, and with no corrosion inhibitor, would beranked as high compared to the probability for a gastransmission pipeline, carrying sales’-quality gas, whichwould be ranked as low. The consequences of a failure fromeach hazard for the pipeline are also qualitatively ranked.For example the consequences of failure of a water-injectionpipeline would be ranked as low compared to theconsequences of failure of a gas pipeline in a densely-populated area.

The advantages of ranking systems are:

• They are relatively easy and quick to implement andunderstand.

Gathering, Reviewing & Integrating Data

IdentifyHazards

Risk Assessment

Mitigation

EvaluateConsequences

Fig.1. The risk-management process.

Sample

copy


• They are flexible and can take account of, the resultsof detailed studies, unusual hazards, and changes inindustry practice.

• They ensure input from experts.• They can be applied even where there is limited

data.

Disadvantages of ranking systems include:

• It is difficult to get consistent risk levels for differenthazards, consequences, years, assessors, and pipelinesections.

• Significant issues can be missed if expert assistanceis not sought.

• Links to mitigation (such as inspection frequencies)are subjective.

Point-scoring risk assessmentPoint-scoring systems have been developed by a number ofindustry experts. These involve assigning points relating todifferent aspects of pipeline design, operation, history andenvironment. So, for example, points would be assignedfor good quality coating, benign ground conditions, a well-maintained cathodic protection system, etc. Theaccumulated score would indicate a low probability offailure due to external corrosion. Points are also assigneddepending on the consequences of failure. For examplepoints would be assigned for low population density,duplication of supply, low stress operation, well-drilledemergency repair, etc. The accumulated score would indicatelow consequences of failure. The probability andconsequence scores are multiplied and the resulting productof the points gives a measure of the risk.

The advantages of point scoring systems include:

• They provide good consistency from one pipelinesection to the next, and from year to year.

• They provide good guidance on common pipelineissues.

• They can be automated, so that all that is needed isthe input of pipeline data.

• They are generally accepted in the pipeline industry.

The disadvantages of point-scoring systems include:

• They can need substantial modification for eachnew application.

• They can be inflexible and make it difficult toincorporate the findings of specific detailed studies.

• They require large amounts of information, whichmay not be available.

• They attempt to replace experience and expertisewhich can lead to significant issues being missed.

• They can require the consideration of some issuesthat may not be significant for the particular system,thus wasting time and effort.

• Automated systems that show significant changesalong a pipeline can confuse inspection andmaintenance planning.

Quantified risk assessmentQuantified risk assessment is a process for calculatingabsolute risk levels based on predicting failure frequency(failures per km per year), and the consequences of failure(the number of casualties, financial costs of a failure, etc.).Failure frequency may be predicted based on historicaldata, or structural reliability analysis. The consequencesare predicted using fire models, oil-dispersion models, lossmodels etc. The advantages of these systems include:

• Consistent comparison of risk levels for differentfailure modes.

• The benefits of reducing failure frequency can bequantified.

The disadvantages of this type of system include:

• ‘Acceptable’ risk levels and hence ‘target’ failureprobabilities must be agreed.

• Historical data may be limited and may not apply toparticular pipelines.

• Good quality data are required.• Specialist software may be needed.• Generally not practical for whole pipeline systems.• The effects of inspection and maintenance on failure

frequency can be difficult to quantify.

Combination risk assessmentThe three generic approaches outlined above each haveadvantages and disadvantages. An alternative is to combinethe different methods in a rational manner, and ensurethat mitigation activities such as inspection are appropriate.

The combined system is a qualitative risk-ranking systemthat provides flexibility and ensures expert input, andwhich is calibrated against quantified risk levels to providecredibility and consistency. The aim of the combinedsystem is to set inspection intervals that give an acceptableprobability of failure.

Any risk-management system must consider pipeline design,operation, and inspection and remedial actions, sincethese are the factors that control risk. The combined systemdirectly links risk-mitigation activities to the relevant hazardsor consequences. This combined risk assessment is nowillustrated by applying it to two different types of pipelinehazard:

• time-dependant hazards (such as corrosion);• random hazards (for example, external interference).

Sample

copy


Application of the new approachto time-dependent failure

Traditional approach to inspection

Time-dependent failures are where the condition of thepipeline degrades over time, and inspection is used tomonitor this degradation. If the condition of the pipelineis known precisely, and if the rate of degradation can beaccurately predicted, then it is possible to select an inspectionfrequency that will always allow timely remedial action tobe taken so that all failures can be prevented.

This approach is both simplistic and subjective as there issignificant uncertainty about the pipeline condition anddegradation rate. In cases where high degradation rates areexpected, extremely frequent inspection may be requiredto prevent failures using this method, and such inspectionsmay not be practical or cost effective. Inspection for time-dependent failure modes will always be aimed at monitoringdegradation, but the uncertainties mean that someprobability of failure must be accepted, and whether or notthis is tolerable should depend on the risk presented by thehazard.

Traditional approach todefining inspection frequency

Inspections aimed at monitoring degradation can be usedto prevent failures due to time-dependent failure modes.Figure 2 illustrates the degradation of pipeline conditionover time for a time-dependent failure example.

The required inspection interval (Tinsp

) that will allowfailures to be avoided is dependent on the time taken todegrade to failure (the remaining life,), and on the timerequired to take action to prevent failure (the action time,

Ta), after the damage has been detected. The remaining life

(Trem

) will depend on the degradation rate. For example anonshore pipeline that is 10mm thick subject to externalpitting corrosion at a typical rate of 0.1mm/yr will have aremaining life of approximately 100 years.

The action time (Ta) will depend on the hazard, and the

pipeline design. For example for internal corrosion of anoffshore pipeline the action required to prevent failuremight be pipeline replacement, and this could take 2 to 3years; for external corrosion pitting of an onshore pipelineit may be possible to carry out a repair within a few days ofdetecting the damage.

T T Tinsp rem a= − (1)

Note that Trem

is the remaining life from the the point atwhich some degradation of the pipeline can be measured tofailure.

This method does not account for the probability that thepipeline starts to degrade. It works on the basis that ifdegradation should start, then an inspection will always becarried out sometime between the start of degradation andthe point at which there is only time T

a left before failure.

However, in reality this method cannot guarantee that allfailures are prevented because it does not take any accountof the uncertainty associated with the prediction of T

rem. A

more realistic illustration of the degradation of a time-dependent failure is shown in Fig.3 which illustrates howT

rem might be expected to follow a statistical distribution.

Basing the calculation of Tinsp

on the minimum expectedvalue of T

rem should ensure that there are no failures.

However, if there is significant uncertainty over thedegradation rate this will lead to the minimum value of T

rem

being extremely small. The calculated required Tinsp

wouldthen also be small, possibly requiring inspection at

Remaining Life

Action time

Failed

Perfect

Degradation can be identified by inspection

Condition

Inspection interval

Time

Corroded Area

Fig.2. Prevention of time-dependentfailures by inspection.

Sample

copy


impractical intervals, with no indication given of the levelof risk represented by inspection at a longer interval.

An alternative is to calculate the required value of Tinsp

togive an acceptable probability of failure, given what isknown about the likely values of T

rem.

A new approach: time-dependant failure

The aim of the method is to maintin an acceptable andconsistent probability of failure. The probability of failuredepends on the inspection interval. So the method aims tofind the inspection interval that gives an acceptable failureprobability, and the acceptable failure probability dependson the acceptable risk level, which in turn depends on theconsequences of failure.

Calculation of failure probability

The probability of failure of a section of pipeline, Pfail

, dueto a particular time-dependent failure mode is calculatedfor any given value of T

insp as

P pT P dTfail rem IF rem

allTrem

= ∫ (2)

where:

p(Trem

) is the probability density function (PDF) of Trem

(theprobability that T

rem has any given value), and P

IF is the

probability that inspection is too late to allow failure to beprevented, for any given value of T

rem and the chosen value

of Tinsp

.

Failed

Perfect

Condition

Time Minimum Trem

Maximum Trem

Mean Trem

Fig.3. Illustrationof distribution of Trem.

Probability Density

Trem Best Estimate

0.9 x Trem BestEstimate

Shaded area represents percentage confidence that actual value of Trem is at least 0.9 x Trem.

Time

Fig.4. Illustration ofconfidence rank definition.

Sample

copy


Note that Pfail

is the probability of failure before the nextinspection, given that degradation starts before the nextinspection. It is therefore a conservative calculation offailure probability given that there is a finite chance that thedegradation will not have initiated.

Calculation of remnant life probability

The distribution of Trem

must be defined to effectively usethis method. In most cases statistical parameters describingthe PDF will not be available. A semi-quantitative methodis therefore proposed using a best-estimate value for T

rem,

and a level of confidence in this best estimate by use of high,medium, or low confidence rank.

The best-estimate value is then used to represent theexpected value of T

rem in a lognormal distribution, and the

confidence rank is used to infer the expected value ofln(T

rem) and the value of the variance of ln(T

rem) as required

to define the lognormal distribution. The lognormaldistribution has the benefit that the PDF is zero at zero T

rem,

and the distribution does not include negative values ofT

rem. This represents the practical case where the time to

failure cannot be negative.

In this proposed method, the confidence rank is used todefine the confidence that the actual value of T

rem is not less

than 0.9 times the best estimate of Trem

, as illustrated byFig.4.

Table 1 gives the probabilities proposed for the definitionof high, medium, and low confidence, and Fig.5 illustratesthe PDFs derived using these confidence ranks. Note that

by defining the confidence rank probabilities relative to aproportion of the mean (rather than an absolute value ora difference), the shape and position of the PDFs relative tothe origin remain unchanged with changing values of thebest estimate of T

rem.

Probability that inspection is too late to prevent failure

For each combination of Trem

and Tinsp

there is a probabilitythat the inspection will be too late to prevent failure. Oneinspection is performed in each time interval of T

insp. The

condition of the pipeline at the time of inspection dependson when the degradation started relative to the time of theinspection. If the remaining life at the time of the inspectionis less than the action time (T

a) then the pipeline may fail

before it can be repaired; if the remaining life is greater thanthe action time, then failure can be prevented. The densitydistribution of the time of inspection relative to the start ofdegradation takes a rectangular form as shown in Fig.5.Inspection is as likely to be at one time relative to the startof degradation as at another (given that degradation starts).The probability that inspection is too late to preventfailure, P

IF, is then given by the shaded area of the PDF in

Fig.6. The calculation of PIF is defined by Equn 3.

P if T T

T T T

Tif T T T

if T

IF rem a

insp rem a

inspinsp rem a

i

= ≤

=− −

> −

=

1

0

( )( )

nnsp rem aT T≤ −( )

(3)

The target failure probabilities listed in Table 2 would

knarecnedifnoCfoeulavlautcaehttahtytilibaborP

T(efilgniniamermer

nahtsseltonsi)Tfoetamitsetsebehtsemit9.0

mer

hgiH %59

muideM %57

woL %06 Table 1. Proposed definitions for

confidence ranks.

0

high

medium

low

Fig.5. Probability density functionsfor high, medium, and low

confidence.

Sample

copy


CknarecneuqesnoCeruliaftegratdesoporP

rymkrepytilibaborp

lavivrustegratdesoporP

ytilibaborp

1 %00.01 %00.09

2 %00.1 %00.99

3 %1.0 %09.99

4 %10.0 %99.99

5 %100.0 %999.99

PknarytilibaborP PdetaluclacfoegnardesoporPliaf

1 Pliaf

%1.0=<

2 P<%1.0liaf

%1=<

3 P<%1liaf

%01=<

4 P<%01liaf

%05=<

5 Pliaf

%05>

CknarecneuqesnoCeruliaftegratdesoporPytilibaborpsaytilibaborp

Pknar

knarksirtegratdeilpmI)PxC(

1 3 3

2 2 4

3 1 3

4 1 4

5 1 5

Table 2. Proposed target failure andsurvival probabilities.

Table 3. Proposed conversion ofcalculated Pfail to probability.

Table 4. Target probabilitiesconverted to probability rank

ProbabilityDensity

Time of inspection relative to start of degradation

(Trem - Ta) Tinsp

1/Tinsp

Fig.6. Illustrationof calculation of PIF.

Sample

copy


convert to probability ranks as shown in Table 4. This tablealso shows the calculated qualitative risk rank.

Calculation of Tinsp

for target failure probabilities

Equation 2 can be used to find the value of Tinsp

requiredto give a specific failure probability, P

fail. This requires an

iterative, numerical solution. For a risk-based system, therequired values of Pfail would be expected to vary with thedifferent consequence ranks. Table 2 lists the proposedtarget failure probabilities (and associated survivalprobabilities) for five consequence ranks ranging from 1 –low (e.g. no safety implications, low cost) to 5 – high (e.g.significant safety implications). Note that the calculatedand target failure probabilities are interpreted as failureprobabilities per km.

Calculated failure probabilityas a qualitative probability rank

The principle of the combined method is to provide aqualitative tool allowing simple comparison between therisks presented by each failure mode, in each pipelinesection. Therefore the calculated values of P

fail are converted

into qualitative probability ranks as defined in Table 3.

New approach: application to random failure

Inspections aimed at finding damage can be used to controlthe risk due to failure modes initiated by random eventsthat fail with time, but they can do nothing to control therisk due to random events that would result in immediatefailure of the pipeline. In these cases, risk can only be

reduced by changing the design (for example installingadditional pipeline protection), or by operational measures(for example, more patrols).

In the case where the damage does not fail immediately,inspection can be used to limit the probability that damagewill fail. The probability that damage has occurred increasesover time until the pipeline is inspected. Once the pipelineis inspected, the condition of the pipeline is revealed andthe probability that hidden damage has occurred is returnedto zero, either because it is shown that no damaging eventhas occurred, or because remedial action is taken to repairthe damage; this is illustrated in Fig.7.

A shorter inspection interval will reduce the probabilitythat undetected damage has occurred, as shown in Fig.8.

Failure criteria

Once a damaging event has occurred, the pipeline may faildue to a degradation of the pipeline condition at thedamaged site, by mechanisms such as corrosion or fatigue.Alternatively, subsequent events at the location which hasalready been damaged (and weakened), may cause failure.In this methodology it is proposed that events are consideredcoincident if they occur in the same 5-m length of pipeline.For each pipeline section, the overall probability of failureper km (P

fail), due to a particular randomly-initiated hazard,

within the inspection interval Tinsp

, is given by:

P Pfail = − −( )1 1 5

1000 5 (4)

where P5 is the probability of failure within the inspection

interval Tinsp

for each 5-m section, given by:

Probability of Undetected Incident having occurred

Time Tinsp

(Inspection Interval) Tinsp Tinsp

Max Probability

Fig.7. Illustration of inspection to control the probability of undetected damage having occurred.

Sample

copy


P P P D P5 1 1 1= ⋅ + >( ) ( ¦ ) ( ) (5)

where:

P(1) is the probability of one damaging event occurringin 5m within time T

insp

P(D|1) is the probability that the damage degrades tofailure within time T

insp, given that one damaging

event has occurred1, andP(>1) is the probability that more than one event occurs

in 5m (i.e. that multiple coincident events occurand the pipeline fails).

Random event probability and incident rates

The damaging pipeline incidents are modelled as eventswhich occur randomly at a constant average rate. Eachevent can then be described as a homogeneous Poissonprocess, where P(n), the probability of n events occurring intime T

insp, in 5m of the pipeline (as used in Equn 3) is given

by [1,4]:

P nT

nTinsp

n

insp( )( )

!exp( )= −

λλ (6)

where l is the incident rate, which is constant and isexpressed as incidents per 5m per year. lT

insp is then the

expected number of incidents in time Tinsp

in a 5-m section.

It follows that P(1) and P(>1) in Equn 5 are given as:

P T Tinsp insp( ) exp( )1 = −λ λ (7)

P P P

T Tinsp insp

( ) ( ( ) ( ))

( ) exp( )

> = − += − + −

1 1 1 0

1 1 λ λ (8)

Calculation of the probability of events occurring in a giventime period therefore requires knowledge of the incidentrates for each failure mode modelled, in each section ofpipeline assessed. To avoid the need for detailed numericalcalculation, and to allow a simplified lookup process, aqualitative description of the incident rate from a lowincident rate of 1, to a high rate of 5 is used. Table 5 givesthe proposed rates and as an example an indication oftypical pipeline incidents in the UK North Sea to whichthese relate. Note that it is important to consider that theseare the incident rates for events causing damage.

Probability of degradation of damage

The probability of damage degrading to failure within timeT

insp, given that an incident has occurred, P(D|1), is

calculated using the methodology proposed for time-dependent failures as:

P D pT P dTrem

allT

IF rem

rem

( ¦ )1 = ∫ (9)

where:

Trem

is the remaining life after the pipeline has beendamaged

p(Trem

) is the probability density function (PDF) of Trem

(the probability that Trem

has any given value), andP

IF is the probability that inspection is too late to allow

Probability of Undetected Incident having occurred

Time

Tinsp

ReducedMax

Probability

Fig.8. Illustration of the effect of more frequent inspection on the probability of undetected damage having occurred.

1 Note that P(1) x P(D|1) is then the probability that one event occursand that this event degrades to failure.

Sample

copy


failure to be prevented, for any given value of Trem

and the chosen value of Tinsp

.

The probability that at least one out of n incidents degradesto failure in time T

insp, P(D|n), is then given by:

P D n P D N n( ¦ ) ( ( ¦ ))= − −1 1 (10)

For time-dependent failures, a best estimate for remaininglife is required, together with a qualitative assessment of theconfidence in that estimate. That level of detail is excessivefor randomly-initiated failures; instead, incidents areclassified as likely to give damage of high, medium, or lowseverity. These incident-severity ranks are used to definethe best estimate of T

rem which is taken as the mean of a

lognormal distribution. The spread of the distribution isdefined according to the definition of medium confidencein the methodology for time-dependent failures, wherethere is a 75% confidence that the actual remaining lifeafter an incident will not be less than 0.9 times the bestestimate of remaining life.

The proposed definitions for the best estimate of remaininglife for high, medium, and low severity incidents aredefined in Table 6 and the PDFs are illustrated in Fig.9.The severity of an incident will depend on the nature of theincident as well as on the design of the pipeline and thelevel of protection provided.

ExampleAs an example, consider the risk associated with failure due

to anchor impact on two 18-in diameter parallel subseacrude oil pipelines, both with a consequence rank of 3(significant environmental and lost production cost, limitedsafety issues). The incident rate rank, which is the same forboth as they run parallel, is 3 (the pipelines are in an areawith significant oilfield development activity). The severityrank for one pipeline is medium, since it is pressure cycledand any damage caused would degrade due to fatigue; forthe second section, which is not cycled, it is low.

Calculations of the required inspection interval Tinsp

havebeen performed for each combination of incident-rate rankand incident-severity rank, for consequence rank 3, assumingthat coincident events will cause failure. These results areshown in Fig.10. Note that it is assumed that inspectionsare scheduled on a yearly basis, and the calculated values ofT

insp have been rounded down to the nearest year.

From Fig.10, the first section would require inspectionevery eight years for anchor impact, and the second wouldrequire inspection every 40 years.

Implementation and futuredevelopments

The methodology described here is still in development,and is being implemented for the operator of a network ofonshore and offshore pipelines in the UK. It is expectedthat fine tuning of the links between qualitative rankings,target probabilities, and failure rates will be required to suitdifferent applications. It is also anticipated that the impact

tnedicnIknaretar

etartnedicnIrymkrep

aeSNlacipytfoselpmaxE]5[noitacoldnatnedicni

1 6.0-E0.1 riapergniriuqerstnedicnitcejbodepporD

2 4.0-E0.1

riapergniriuqerstnedicnitcapmipihS

riapergniriuqerstnedicninoitcaretnilwarT

riapergniriuqerstnedicnirohcnA

3 3.0-E0.1

ni01nahtsselfosresirnoriapergniriuqerstnedicnI

retemaid

senozytefas-llewaesbusniriapergniriuqerstnedicnI

4 2.0-E0.1 sresirelbixelfnoriapergniriuqerstnedicnI

5 1.0-E0.1 atadelbacilppaoN

knarytirevestnedicnI efilgniniameretanmitsetseB

hgiH 3

muideM 01

woL 05

Table 5. Incidentrate rank definitions.

Table 6. Proposed incidentseverity rank definition.

Sample

copy


0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

0 10 20 30 40 50 60 70

high severity

medium severity

low severity

Probability Density

TremFig.9. Illustration of incident-severity rank probabilitydistributions.

of maintenance and monitoring (for example inhibitorinjection reliability) could be incorporated, to give a more-rounded assessment.

The key issue in risk assessments in any risk-based integritymanagement system is that the selected or calculated failure-probability scores, and failure-consequence scores, must befully justified, and the justification should be recorded toensure consistency in future updating.

ConclusionsThe methodology described here is the result of many years’experience in the development and application of risk-based inspection and integrity-management systems forpipelines. It attempts to retain the best aspects of flexibilityand simplicity that a risk-ranking scheme provides, whileproviding a much higher level of consistency, and clearjustifiable links to inspection frequency and mitigationactions.

1

10

100

1000

10000

0 1 2 3 4 5

Incident Rate Rank

)sry( psniT deriuqeR

High severityMedium severityLow Severity

Fig.10.Variation ofrequired Tinsp with incident raterank and severity rank forconsequence rank 3.

AcknowledgementsThe authors would like to thank their colleagues at Penspenand Andrew Palmer and Associates for their help andsupport in the preparation of this paper. In addition wewould like to thank our clients who have supported thedevelopment of these methods.

References1. Pipeline safety, pipeline integrity management in high

consequence areas (gas transmission pipelines), 2003. FinalRule, 49 CFR Part 192, Department of Transportation,USA, December.

2. The Pipeline Safety Regulations, 1996. SI 1996 No. 825. TheHealth and Safety Executive, UK.

3. Robert E. Melchers. Structural reliability analysis andprediction, 2nd Edn, Wiley.

4. Patrick D.T. O’Connor. Practical reliability engineering, 3rd

Edn, Wiley.5. PARLOC 2001: The update of loss of containment data for

Sample

copy


offshore pipelines. Prepared by Mott MacDonald Ltd for theHealth and Safety Executive, The UK Offshore Operator’sAssociation, and the Institute of Petroleum, 5th Edn, July2003.

6. Conference proceedings: Risk based and limit state designand operation of pipelines. IBC, Olso, October 1998.

7. David J. Smith. Reliability maintainability andrisk.Butterworth Heinemann.

8. John Moubray. Reliability-centred maintenance. ButterworthHeinemann.

TITLES OF

INTEREST

FROM

ASME PRESS

Technical Writing A–Z: A CommonsenseGuide to Engineering Reports and Theses,British English Editionby Trevor M. Young

Topics include: format and content of reports and theses;copyright and plagiarism; print and Internet reference cita-tion; abbreviations; units and conversion factors; significantfigures; mathematical notation and equations; writing stylesand conventions; frequently confused words; grammaticalerrors and punctuation; commonsense advice on issuessuch as getting started and holding the reader’s attention.

2005 256 pp. Softcover ISBN: 0-7918-0237-XOrder No. 80237X $29 (list)/$23 (ASME member)Order sets of 10 copies at a special price. Order No. 80236S $199

American Edition:2005 256 pp. Softcover ISBN: 0-7918-0236-1Order No. 802361 $29 (list)/$23 (ASME member)Order sets of 10 copies at a special price. Order No. 80236S $199

Pipeline Operation and Maintenance: A Practical Approachby M. Mohitpour, J. Szabo, and T. Van Hardeveld

Covering pipeline metering, pumping, and compression, thebook covers day-to-day concerns of the operators andmaintainers of the vast network of pipelines and associatedequipment and facilities that deliver hydrocarbons andother products. It is a useful reference for veterans and atraining tool for novices.

2004 600 pp. Hardcover ISBN: 0-7918-0232-9Order No. 802329 $125 (list)/$99 (ASME member)

Mister Mech Mentor, Volume I:Hydraulics, Pipe Flow, Industrial HVAC &Utility Systemsby James A. Wingate

Gain practical knowledge from frank, colorful cases andlearn to solve mechanical problems related to hydraulics,pipe flow, and industrial HVAC and utility systems withthese organized solutions to the problems involving: waterand steam hammer phenomena; gravity flow of liquids inpipes; siphon seals and water legs; regulating steam pres-sure drop; industrial risk insurers’ fuel gas burner pipingvalve train; controlling differential air pressure of a roomwith respect to its surroundings; water chiller decoupledprimary-secondary loops; pressure drop calculations ofincompressible fluid flow in piping and ducts; water chillersin turndown; hydraulic loops; radiation heat transfer; andthermal insulation.

2005 160 pp. Softcover ISBN: 0-7918-0235-3Order No. 802353 $45 (list)/$36 (ASME member)

Pipeline Design and Construction:A Practical Approach, Second Editionby M. Mohitpour, H. Golshan and A. Murray

This second edition includes updated codes and standardsinformation, solutions to technical problems, additional ref-erences, and clarifications to the text. It offers straightfor-ward, practical techniques for pipeline design and con-struction, making it an ideal professional reference, trainingtool, or comprehensive text.

2003 700 pp. Hardcover ISBN: 0-7918-0202-7Order No. 802027 $110 (list)/$88 (ASME member)

Order Now! North America: www.asme.org • Europe: www.ihsatp.com

Sample

copy


THIS PAPER DESCRIBES a method for determiningthe relative severity of dents on the basis of the signal

received from a MFL inspection. In the paper, the definitionof a dent is taken as “a permanent plastic deformation ofthe cross section of the pipe caused by external forces” [1,

2]. A dent that varies smoothly in cross section and containsno stress raisers, such as a gouges, crack, or weld, is definedas a plain dent.

Review of pipeline dent failuresIn order to understand the threat posed by plain dents topipeline integrity, a review has been conducted of thepublished failure data for oil and gas transmission pipelinesfor the US and Europe [3-7]. One of the problems inconducting an analysis of this type is that “failure at plaindents” is not a discrete category or failure cause in many of

This paper was presented at the 7th International Pipeline Conference,held on 29 September – 3 October, 2008, in Calgary, Alberta, Canada,and organized by the ASME Pipeline Systems Division.

*Author’s contact detailstel: +44 (0)191 230 6507email: [email protected]

Investigating the relative severityof dents in pipelines based onmagnetic-flux-leakageinspection data

by Leanne M Tindall*1, Dr Julia M Race2, and Jane Dawson3

1 Atkins Boreas, Newcastle-upon-Tyne, UK2 School of Marine Science & Technology, Newcastle University, Newcastle-upon-Tyne,

UK3 PII Pipeline Solutions Business of GE Oil & Gas, Cramlington, UK

DENT DAMAGE IN pipelines may result from either impact damage caused by third parties orconstruction damage. Third-party damage generally occurs on the upper half of the pipe (between the

8 o’clock and 4 o’clock positions) and has historically contributed to the highest number of pipeline failures.Dents caused during construction generally occur on the bottom half of the pipe, and tend to be constrainedby the indenter causing the dent, i.e. a stone or rock in the pipeline bed/backfill. However, all dents havethe potential to cause an increase in stress in the pipeline, and consequently increase the pipeline sensitivityto both static and fatigue loading.

Although there are extensive recommendations for the ranking and repair of dents, failures of dents thatare acceptable to pipeline codes have recently been reported. Guidance is therefore needed in order thatoperators can identify dents for which excavation and inspection is uneconomic and could potentially bedamaging to pipeline safety, and dents for which further action is required.

This paper provides a review of the published recommendations for the treatment of pipeline dents andgoes on to present a method that is being developed to determine the relative severity of dents in a pipelineusing magnetic-flux-leakage (MFL) signal data. The proposed method involves measuring MFL signalparameters related to the geometry of the dent and relating these to high-resolution caliper inspection data.This analysis enables a relationship to be established between the MFL signal data and dent depth and shapemeasurements. Once the model is verified, this analysis can then be used to provide a severity ranking fordents on pipelines where only MFL data is available.

Sample

copy


the failure databases. As a result, data relevant to pipelinedents has to be inferred from categories relating to externalinterference and construction damage, i.e. damagemechanisms by which denting may occur on a pipeline.However, it is recognized that external interference, forexample, often results in gouges or cracking which may ormay not be associated with residual denting, and thereforethe analysis will over-represent the actual situation.Construction damage may include dents caused by rocks inthe backfill or by excavation equipment impacting thepipeline. It is considered unlikely that rock dents couldcontain associated features such as gouges or cracks, andindeed Rosenfeld [2] states that if the coating remainsintact then it is “impossible for the dent to be anything butplain”. However, the number of rock-dent failures is alsonot recorded in the failure database information availablein the public domain.

In recognition of the relative coarseness of the publisheddata, a recent study by the Department of Transportation(DOT) [8] has reviewed the Office of Pipeline Safety’s(OPS) statistics for gas and hazardous liquid pipelines up to2003 in an attempt to determine the number of failuresthat have occurred as a result of damage from plain dents.The analysis indicated that <0.2% of the incidents onliquid lines and <<0.1% of the incidents on gas pipelineswere related to dents. The conclusion of the study wastherefore that failures from dents do not form a significantproportion of the total number of pipeline failures.

Although this review of failure data has concluded thatdents alone are not a major cause of pipeline failure, there

have, in recent years, been a number of in-service failuresat dents that were within the code limits, were reportedthrough standard ILI technologies, but were not identifiedas significant [4, 9-11]. This has raised concerns in theindustry regarding the perceived conservatism of the dent-assessment methods and the lack of coherent industryguidance on best practice with regard to dent-managementstrategies.

Dent-acceptance standardsMuch of the research into the failure of plain dents underpressure loading has indicated that plain smooth dents donot significantly reduce the burst strength of pipelines, andtherefore do not require repair [2,12,13] unless they arevery deep [1]. There is currently no analytical methodavailable for calculating the failure pressure of a plain dentand therefore the traditional and codified acceptabilitylevels for plain dents have been empirically derived on thebasis of dent depth from full-scale test results. However, ithas been suggested that the dent depth alone is not sufficientto define the severity of a plain dent, and recent coderevisions and guidance documents have recognized thatthe strain in the dent and the effects of pressure-cycleinduced fatigue should also be considered in an assessmentof dent severity.

A summary of dent-assessment requirements is presentedin Table 1. It is highlighted that, with the exception ofAPI1156, there is little industry guidance on the acceptabilityof dents in pressure-cycled pipelines even though past

stnednialP

deniartsnoC deniartsnocnU

]02,41[8.13BEMSA %6otpulevelniartsroDO%6otpU

]42[4.13BEMSA4SPN>sretemaidepipniDO%6otpU4SPN<sretemaidepipnimm6otpU

6511IPA ]62,52[ tnemssessaeugitafaseriuqerDO%2>,DO%6otpU

GRPE ]72[ SYMS%27fossertspoohataDO%7

]1[MADP DO%01otpU DO%7otpU

]82[266Z mm6.101>rofDO%6<roDOmm6.101rofmm6otpU

sdlewtastneDroskcarchtiwstneD

seguognoisorrochtiwstneD

8.13BEMSA ]02,41[niarts%4roDO%2otpU

toNsdlewelitcudnosdlewelttirbnodewolla

dewollatoN yllaudividnissessA

4.13BEMSA ]42[ dewollatoN dewollatoN dewollatoN

6511IPA ]62,52[elitcudnoDO%2otpUnodewollatoNsdlew

sdlewelttirbdewollatoN deredisnoctoN

GRPE ]72[ dewollatoN dewollatoN dewollatoN

266Z ]82[ dewollatoN dewollatoN dewollatoN

Table 1. Summary of staticdent-assessment methods.

Sample

copy


research – and, indeed, operational experience – clearlyindicates that pressure cycling increases the risk of dentfailure.

Dent-strain assessmentThe 2003 version of ASME B31.8 [14] introduced a strain-acceptance criterion as well as a method for estimating thestrain in dents in Appendix R. In this method, it is assumedthat there are two components of strain acting in the pipewall in the circumferential and longitudinal directions.Within each of these directions, the strain can be separatedinto membrane and bending components. The membranestrain components are constant through the pipe wall, butthe bending strain components vary linearly about theneutral axis at t/2 (Fig.1). In this analysis it is assumed thatthe through-thickness strain is zero, although this is opento debate [15].

The longitudinal and circumferential bending strains canbe calculated from measurements of the curvature of thedent, and are given by the following equations:

εlb

t

R=−

2 2(1)

εcb

t

R R= −

2

1 1

0 1

( ) (2)

It is well documented [15-19] that the equations in AppendixR of ASME B31.8:2003 for longitudinal and circumferentialbending strains are incorrect in that the factor of 1/2(shown in Eqns 1 and 2) is missing from the ASME B31.8code. This has been corrected in the newest release ofB31.8:2007 [20].

The calculation of the membrane strains is more difficult.In the ASME B31.8 strain methodology, the shear-membrane strain is neglected and the circumferential-membrane strain is assumed to be negligible as the pipe isassumed to accommodate the deformation due to the dentwithout significant deformation in this direction. In thelongitudinal direction, the deformation must involve someextension of the pipe wall. The ASME B31.8 code presents

the following expression for an approximation to thelongitudinal membrane strain, which was benchmarkedagainst a limited number of finite-element analyses [8].

εlm

d

L=

1

22( ) (3)

The codified methodology then presents equations forcombining the calculated membrane and bending straincomponents (using Equns 1 to 3) to obtain the net strainat the inside and outside diameters of the pipe surface(Equn 4):

ε ε ε ε ε ε ε

ε ε ε ε ε ε

02 2

12

= + − +( )+ − +( )

= − +( )+

cb cb lb lm lb lm

cb cb lb lm

and

llb lm+( )ε 2(4)

The above approach to the calculation of the membranestrains has recently been described by Czyz and Lukasiewicz[15, 21] as being simplistic and inaccurate, and they arguethat the assumption that the circumferential and shearstrains are negligible is not supported by FEA (finite-element analysis). Instead, they present a methodology forcalculating the membrane strains using a specialized andsimplified finite-element model. In addition, they presentthe following equation for combining the longitudinal andmembrane strains, remembering that in this analysis thecombination of strains, e

c includes the circumferential

membrane strain ecm

which is not assumed to be negligibleas in the equations above:

ε ε ε ε εγ

eff c c l l= + + +2

3 22 2

2

(5)

where ec = e

cm + e

cb and e

l = e

lm + e

lb with the positive or

negative sign indicating whether strain is being calculatedon the outside or inside diameters of the pipe, respectively.

Indeed, Czyz and Lukasiewicz [21] have demonstrated, and

WT

Membrane

Membrane

Bending Bending

Pipe Circumference Pipe Axis

Longitudinal Strain

Circumferential Strain

Fig.1. Strain components in thepipe wall [15].

Sample

copy


verified with FEA, that their approach predicts effectivestrains that may be significantly larger than those calculatedusing the method in ASME B31.8.

Although, the errors in the determination of the longitudinaland circumferential bending strains have now beencorrected in the latest version of ASME B31.8 (2007) therestill appear to be issues with the longitudinal membranestrain expression and, potentially, with other assumptionsas well. These other discrepancies have less of an effect onthe overall estimation of the total strain; however, they stillneed to be resolved. The implication of there beingdiscrepancies in the strain calculations presented in theB31.8 code is that once the strains are estimated, they mustbe compared to a strain criterion.

Rosenfeld [22] suggested that a suitable strain criterionmaybe between 3 and 12%. This is based on the ASMEB31.8 code that permits field bends that produce a coldstrain of 3% in the pipe wall and the observation that thelikelihood of puncture in dents increases where the materialstrain exceeds 12%. Therefore a recommendation wasmade that a value of 6% strain be conservatively adopted.This has now been codified in ASME B31.8 [14, 20] and isalso specified in the US Federal Regulations [23]. For dentson ductile seam or girth welds, the strain limit is reducedto 4%. Therefore, if the effective strain calculationspresented in ASME B31.8 are underpredicting the totalstrain, dents which would actually be unacceptable to thestrain criterion could remain unrepaired in pipelines. Asthe strain limit itself is a conservative value there is someallowance for such errors if they are relatively small.

Nevertheless, an independent review should be conductedto ensure the dent strain guidance given in ASME B31.8 iscorrect so that it is appropriately applied to dented pipelines.

Basis of investigationThe review of dent-acceptance standards has shown thatthe two main parameters used to establish the severity of aplain dent are dent depth and dent strain. This paperfocuses on how MFL data can be used to predict the severityof dents based on each of these two parameters.

Measurements are taken from MFL inspection data and arethen correlated with high-resolution caliper data. A modelis developed in which the severity of a dent is determinedfrom the MFL data alone. The model is then tested andvalidated against predictions determined from caliper data.

Basic introduction to MFL dataAn MFL inspection vehicle induces a localized magneticfield into the pipeline in two ways:

• Permanent brushes induce a magnetic field fromthe north to the south pole through the pipe wall.Any deviation from plain, undamaged, pipe changesthe magnetic field and is recorded by a ring ofprimary sensors (Fig.2).

• A small magnetic field is partially induced into the

Primary Sensor

Magnetic FieldExternal Metal Loss Defect

S NPrimary Sensor

Magnetic FieldExternal Metal Loss Defect

S NSecondary Sensor

Magnetic Field

External Metal Loss Defect

Secondary SensorMagnetic Field

External Metal Loss Defect

4 sensors on one finger

Dent indication

4 sensors on one finger

Dent indication

NS

NS

NS

NS

NS N

S

NS

NS N

SNS

NS

NSNS

NSNS

NSNS

NSNS

NSNS N

SN

S

NSNS

NSNS N

SNSN

SNS

NS

NS

Fig.2. Magnetic-flux leakage at an external metal-loss defect(primary sensors).

Fig.3. Magnetic-flux leakage at an external metal-lossdefect (secondary sensors).

Fig.4. Dent indication on primary sensors. Fig.5. Dent indication on secondary sensors.

Sample

copy


pipe wall via a magnet located on the back of eachsensor in a secondary ring of sensors (Fig.3).

When it was first developed, the primary objective of theMFL inspection tool was to identify and measure metal-lossdefects. The primary sensors are predominantly used todetect and size metal loss, and the secondary sensors areused to discriminate between defects located on the internalor external surface of the pipe wall.

A by-product of the technology was the ability to detectdent defects. The internal/external discrimination(secondary) sensors are the sensors that give an indicationof a dent. However, the main corrosion (primary) sensorsalso give an indication depending on the size and shape ofthe dent.

The signals produced from the primary sensors are mainlydue to the physical movement of the sensor arm as it passesover the deformation (Fig.4). However, if metal loss is alsopresent, this may also be detected. In this example, thechannels are grouped together in a set of four: the fourchannels represent one ‘finger’. It is the physical movementof the whole finger which allows identification of the dent.Therefore, if a dent is smooth and shallow, it is harder todetect with the primary sensors.

The signals produced from the secondary sensors give amagnetic response which appears to be similar to thatobserved with an internal metal-loss defect (Fig.5). However,the orientation of the bi-polar signal is axi-symmetrical, i.e.an internal metal-loss defect gives a signal that is firstnegative then positive; a dent gives a signal that is firstpositive then negative. The extent of a signal at a dent is

typically much larger in both the longitudinal andcircumferential directions than a signal caused by an internalmetal-loss defect. One theory suggests that the dent signalis caused by the change in the mechanical properties of thesteel that have occurred as a result of work-hardeningduring the denting process. This change in materialproperties affects the permeability of the steel and themagnetic response is altered.

Using MFL datato characterize dents

The parameters that can be predicted from the MFL datainclude the length, the width, and relative amplitude of thedent. These values will change depending on the amplitudescaling of the data. Consequently, the measurements takenare relative, and are used to infer the shape of the dent. Allmeasurements are taken from the secondary sensors, asthese show a magnetic response rather than the mechanicalresponse shown by the primary sensors. The followingsections explain how each of these parameters is measured.

Length

The length is measured as the distance along the pipe axiswhere the deflection is present (Fig.6). The ends of the dentare defined where the signal recorded on the secondarychannel is smooth, i.e. the typical response you wouldexpect from a pipe without any defects present.

The secondary channels will very rarely be completelysmooth due various factors, including the surface roughnessof the pipe. Consequently, an amount of engineeringjudgment and knowledge of the MFL data is required inorder to estimate when the dent signal has ended.

Width

The width is taken as the number of secondary channelsaffected by the dent, see Fig.7. This is then transformedinto an approximate measurement of the width based onthe size of the tool, outside diameter of the pipe, and thetotal number of sensors fitted.

Amplitude

The amplitude of the signal is measured from the channelwhich has recorded the largest change from positive tonegative throughout the dent (Fig.8).

Prediction of dentmeasurements from caliper data

A caliper inspection tool has sensor ‘arms’ which detectphysical movement in the curvature of the pipe. As the

LengthLength

No. of sensors

1

2

No. of sensors

1

2

Fig.6. Length measurement.

Fig.7. Width measurement.

Sample

copy


inspection tool reaches a deformation in the pipe the arrayof arms around the circumference is compressed. Theshape of the dent in terms of a depth, length, and width ispredicted from the deflection as each of the sensors passover the deformation. The profile of the dent predicted bythe caliper tool can be used to determine the relative strainaccording to ASME B31.8 as discussed earlier, and asillustrated by Dawson et al. [17].

Correlation of measurementsA data sample consisting of over 200 dents was investigated.The data used is from onshore pipelines covering a rangeof diameters and operating with gas and liquid products.Both top-of the-line and bottom-of-the-line dents wereincluded in the data set: 30% of the dents are top-of-the-line, and 70% are bottom-of-the-line.

Dent dimensions

The first stage of the analysis involves investigation of thebasic parameters associated with each of the dents, andFig.9 shows the relationship between the lengths predictedfrom each of the inspection technologies. For the majorityof dents, the prediction of the length from the MFL data ishigher than that from the caliper data. There are a numberof possibilities for this trend:

• The change in material properties at the location ofthe dent, detected in the MFL data, occurs beforethe change in curvature of the pipe wall.

• The caliper data is less sensitive to small changes inthe curvature of the pipe.

• Movement of the secondary sensors in the MFLdata causes the length to be over-predicted.

The correlation between the prediction of dent widths isnot as strong, and there is a large amount of scatter in thedata. Indeed, it is recognized that the resolution of dataaround the pipe circumference in most caliper tools is lesssuperior than the resolution along the axis of the pipe. Thiscorrelation could be improved with advances in inspectiontool technology.

Figure 10 shows the dent depth as reported by the caliper

inspection and the relative amplitude predicted from theMFL data. The relative amplitude of the dents is taken tobe the ratio of the amplitude and the outside diameter ofthe pipe. There are no dents with a depth less than 2% ofthe outside diameter (OD) reported by the caliper inspection.This is due to the dent-reporting threshold which was set at2% OD in each of the data sets investigated. The figureshows that there is no correlation; this indicates that thedepth of the dents is not directly related to the amplitudeof the signal in the MFL data.

Further investigation into the correlation of caliper dentdepth and measurements taken from the MFL data wasperformed using a ratio of the length and amplitudemeasured from the MFL data. This parameter is investigatedto give an indication of the shape of the dent. Shorterlengths combined with high amplitudes suggest that thedent will be sharp. A negative correlation is observed whenthis ratio is compared with the dent depth estimated fromthe caliper data (Fig.11). The figure shows that there is nota strong direct relationship between the two parametersand it would be difficult to reliably predict the dent depthfrom the length-to-amplitude ratio from the MFL data. Asimilar conclusion was drawn from a comparison of theratio of width and amplitude versus caliper dent depth.

Dent strain

Earlier in the paper it was discussed that dent depth alonewas not sufficient to define the severity of dents, and thatthe strain in a dent is considered as a better indicator ofseverity. So a more-critical relationship to investigate is thatbetween the MFL-measurable parameters and the strain inthe dent. The next stage of the analysis investigatedpredicting the longitudinal bending strain in the dentusing the measurements from the MFL data and comparingit with the estimated longitudinal bending strain in thedent determined from the caliper data.

AmplitudeAmplitude

Fig.8. Amplitude measurement.

0.0

0.5

1.0

1.5

2.0

0.0 0.5 1.0 1.5 2.0Caliper Length (m)

MFL

Len

gth

(m)

Fig.9. Comparison of dent length.

Sample

copy


The strain in a dent can be considered as the sharpness ofthe dent, either at the shoulders (for a short dent) or at thebase (for a long dent). The parameter investigated from theMFL data is the ratio of the length to the amplitude. Thefocus is directed to the length parameter rather than thewidth parameter, as this showed a stronger correlation withthe measurements taken from the caliper data.

Figure 12 shows the ratio of length to amplitude measuredfrom the MFL data in comparison to the longitudinalbending strain predicted from measurements taken fromthe caliper data. On the chart ‘Diameter 1’ is the smallestdiameter investigated and ‘Diameter 3’ is the largest. Itwould be true to say that Diameter 1 is a small-diameterpipeline, Diameter 2 is a medium-sized pipeline, and

Diameter 3 is a large-diameter pipeline. The figure showsthat there is the same general trend for each of the diametersinvestigated: higher longitudinal strains are observed indents where the length-to-amplitude ratio is lowest.Essentially these are dents which have the smallest lengthsand the highest amplitudes; this indicates that they aresharp and hence have higher bending strains. The figurealso sho�ws that there may be a slightly-different relationshipwith each diameter.

Proposed modelsOn the basis of the above findings, a model is proposed foruse in predicting the longitudinal bending strain as a

0.0

0.5

1.0

1.5

2.0

2.5

0 2 4 6 8 10

Caliper Depth (%OD)

Rel

ativ

e A

mpl

itude

0123456789

10

0 2 4 6 8 10

Caliper Depth (%OD)

Rat

io o

f Len

gth:

Am

plitu

de

Fig.10. Comparison of dent depth. Fig.11. Comparison of dent depth.

Fig.12. Comparison of longitudinal bending strain. Fig.13. Proposed model: diameter 1.

0

1

2

3

4

5

6

0 2 4 6 8 10Ratio of Length:Amplitude

Long

itudi

nal B

endi

ng S

train

(%) Diameter 1

Diameter 2Diameter 3

0

1

2

3

4

5

6


Long

itudi

nal B

endi

ng S

train

(%) Diameter 1

Sample

copy


measure of the relative severity of dents. The ratio of thelength to amplitude from the MFL data and the longitudinalbending strain showed a stronger relationship than thedent depth; therefore this parameter is used in determiningproposed models to predict dent severity.

Figures 13 to 15 show each of the diameters investigatedand the corresponding curves used to predict thelongitudinal bending strain in dents. For each diameter,the prediction curve is derived using a fit to power-lawcurves. Each curve takes the same form but with differentcoefficients, and shows a strong correlation to the data.

An investigation into the ‘actual’ versus predictedlongitudinal bending strains is shown in Fig.16. The ‘actual’strain is taken to be the longitudinal bending strain estimatedusing the caliper data. Although the curves in Figs 13 to 15appear to show a good fit, the comparison of actual andpredicted strains shows a certain amount of scatter.

In general, the higher longitudinal bending strains in eachof the data sets tend to be underpredicted: a plot of theresidual values further demonstrates the variation in thepredictions, as shown in Fig.17.

The variation in the actual and predicted strains is quantifiedfor each of the models and overall in Table 2. In each casethe strain is predicted to be within +1% of the value

predicted using high-resolution caliper data with at least an80% confidence level.

Additional investigation of the data in Figs 13 to 15 showsthat for a ratio of length to amplitude of approximately lessthan unity, large variations in the predicted strain couldoccur. In other words, the slope of the prediction curve ishigh, and any small variation in ratio will give a largevariation in strain. This implies that for small ratios, theprediction of strain may be less accurate. However, therelationship shows that the longitudinal bending strain islikely to be higher for smaller length-to-amplitude ratios,and this alone would give a conservative indication of dentseverity and a means of ranking dents when no caliper dataare available.

Verification of modelThe proposed model for the large-diameter model(‘Diameter 3’) is used to predict the longitudinal bendingstrain using data from an independent MFL data set. Theratio of the length to amplitude is calculated according tothe methodology described above, and the longitudinalbending strain is predicted using the curve fit presented inFig.15. Figure 18 shows the ‘actual’ longitudinal bendingstrain, which is estimated from the corresponding caliperdata, in comparison to the predicted longitudinal bending

0

1

2

3

4

5

6


Long

itudi

nal B

endi

ng S

train

(%) Diameter 2

0

1

2

3

4

5

6


Long

itudi

nal B

endi

ng S

train

(%)

Diameter 3

Fig.14. Proposed model: diameter 2. Fig.15. Proposed model: diameter 3

ledoM niarts%1-/+nihtiwatadfoegatnecreP

1retemaiD 09

2retemaiD 69

3retemaiD 18

llarevO 78 Table 2. Prediction tolerances.

Sample

copy


strain. The figure shows the same trend in prediction ofstrains as demonstrated in Fig.16: the higher longitudinalstrains tend to be underpredicted by the model.

Summary of resultsThe analysis has shown that although there was not a strongrelationship with dent depth, there was good correlationbetween the longitudinal bending strain predicted fromcaliper data and the ratio of length to amplitude measuredfrom the MFL data. As dent strain is considered to providea better indication of the dent severity, this lack ofrelationship with the dent depth is not important. Theproposed model, which is based on the dent longitudinalbending strain, provides a means of predicting the dentseverity when only MFL data is available. Although theprediction of the actual strain is inexact, the proposedmodels allow a priority ranking of the severity of the dentsfor investigation purposes.

Susceptibility topressure-cycle-induced fatigue

As discussed earlier in te paper, there is little industryguidance on the acceptability of dents in pressure-cycledpipelines, where there is an additional risk from fatigue.Previous work [17] has shown a strong correlation betweenthe stress-concentration factor resulting from the presenceof the dent (based on FE analysis) and a dimensionlessvolumetric parameter. The dent fatigue life is inverselyproportional to the stress-concentration factor, andtherefore also to the dent volumetric parameter:

Fatugue lifeLd W

D tf

∝1

2( ) (6)

where,

L = dent lengthd

f= final re-rounded dent depth

W =dent widthD = pipe diametert = pipe wall thickness

Since the volumetric parameter is dominated by the largestdent dimensions – namely the dent length and width –these dimensions can be predicted from the MFLmeasurements and used to provide an approximate rankingof the dents in terms of the relative fatigue risk. Using thisapproach, the dents with the largest predicted length xwidth dimension would be ranked highest in terms of therelative fatigue risk.

ConclusionsThe purpose of the developed methodology is to determinethe relative severity of dents reported on a pipeline usingMFL signal data alone. The analysis showed that althoughthere was not a strong relationship with dent depth, therewas good correlation between the longitudinal bendingstrain in the dent and the ratio of length to amplitudemeasured from the MFL data. Therefore, since dent strainis considered to provide a better indication of the dentseverity, the lack of relationship with the dent depth is notconsidered to be important. The proposed strain model

0

2

4

6

0 2 4 6Predicted Long. Bending Strain (%)

Actu

al L

ong.

Ben

ding

Stra

in (%

)

Diameter 1Diameter 2Diameter 3

-5.0 0.0 5.0Difference in Long. Bending Strain (%)

Diameter 1Diameter 2Diameter 3

Fig.16. Actual versus predicted strain. Fig.17. Actual minus predicted strain.

Sample

copy


provides a means of predicting the dent severity when onlyMFL data are available.

It is predicted that dents with the smallest length-to-amplitude ratios are the most likely to contain the highestlongitudinal bending strain. Indeed, as this ratio approachesunity the likelihood of the strain approaching the ASMEB31.8 strain limit increases. Consequently, the plain dentsthat fall nearest to this level should be considered to havea higher priority for remediation than other dents with ahigher length-to-amplitude ratio.

In terms of susceptibility to fatigue, the dents with thehighest length x width values should be considered to havethe highest risk (assuming the cyclic pressure loading isfairly constant along the pipeline).

There are two main applications for the developed models:

(i) on pipelines where multi-channel caliper data is notavailable; and

(ii) for dents below the calliper-reporting threshold.

Point (i) relates to lines where only a MFL tool has been runin a pipeline, i.e. a multi-channel caliper tool has not beenrun, and so information regarding the dimensions of thedent (the dent profile) is not available. The methodologycan be used in this case to ascertain the relative severity ofthe reported dents in terms of both static and fatigue (if theline is subjected to pressure cycling) behaviour from theMFL data alone. The prioritization obtained from applyingthis methodology can then be used to define a suitableremediation programme.

Point (ii) relates to the population of low-level dents notcurrently detected/reported by caliper tools. Although, as

calliper-tool technology advances, the number of dents thatfall into this category will reduce, the current industry-accepted reporting threshold is dents above 2% OD indepth. Note that previous work by Dawson et al. [17]showed that caliper tools typically detect approximatelyone-quarter of the dents detected by MFL tools. Theinference from this is that only one-quarter of the dents aregreater than the 2% OD caliper threshold, and whilst theselow-level dents are unlikely to be an integrity threat, theproposed method does enable an assessment of their severityrelative to other dents in the pipeline to be made.

However, it is highlighted that since the introduction ofmore-stringent industry regulations (such as in the US),and due to the increasing awareness around the potentialseverity of pipeline dents (and associated gouging and/orcracking), many pipeline operators insist on running acaliper tool at theÈ same time as a metal-loss or crack-detection ILI tool. The inspection industry is moving in thedirection of offering pipeline operators the opportunity torun integrated ILI tools that enable the detection and sizingof metal loss and dents in one inspection pass with a singleintegrated tool. Nevertheless the developed models fill acurrent gap in the industry for determining the relativeseverity of plain dents in terms of strain using MFL dataalone.

However, it should be stated that when good-quality caliperdata are available, these data should be the source of thepreferred method of predicting dent severity in terms ofboth the strains in the dent and the fatigue risk.

AcknowledgmentsPart of this work was conducted by Leanne Tindall duringher studies for the Masters degree of Pipeline Engineeringat Newcastle University. The authors would like to thankPII Pipeline Solutions Business of GE Oil & Gas forpermission to publish this paper.

Nomenclatured = dent depthL = deformed dent lengthL

o= initial dent length

OD = outside diameterR

0= radius of curvature of undeformed pipe

R1

= external radius of curvature in circumferentialdirection

R2 = external radius of curvature in longitudinaldirection

t = nominal pipe wall thicknessg = shear membrane straine

eff= effective strain on the pipe

ec

= net circumferential strain on the pipee

cb= circumferential bending strain

ecm

= circumferential membrane strain

0

1

2

3

4

5

6

0 1 2 3 4 5 6Predicted Long. Bending Strain (%)

Actu

al L

ong.

Ben

ding

Stra

in (%

)

Fig.18. Actual versus predicted strain.

Sample

copy


el

= net longitudinal strain on the pipee

lb= longitudinal bending strain

elm

= longitudinal membrane straine

i= biaxial strain on the pipe ID

eo

= biaxial strain on the pipe OD

References1. A.Cosham and P. Hopkins, 2004. The effect of dents in

pipelines - guidance in the Pipeline Defect Assessment Manual.Int.J.Pressure Vessels and Piping, 81, 2, p127.

2. M.J.Rosenfeld, J.W.Pepper, and K.Leewis, 2002. Basis ofthe new criteria in ASME B31.8 for prioritization and repairof mechanical damage in pipelines. Proc.InternationalPipeline Conference, IPC 2002, Calgary, Alberta.

3. PHMSA, 2007. Pipeline safety program: Pipeline safetyincident reports. Pipeline and Hazard Materials SafetyAdministration. Available from http://www.phmsa.dot.gov.

4. P.M.Davis, J.Dubois, A.Olcese, F.Uhlig, J-F.Larivé, andD.E.Martin, 2006. Performance of European cross-countryoil pipelines: Statistical summary of reported spillages, 2004.CONCAWE.

5. EGIG, 2005. Gas pipeline incidents: 6th Report of theEuropean Gas Incident Data Group.

6. D.Browne and R. Hicks, 2005. UKOPA pipeline faultdatabase - pipeline product loss incidents (1962-2004).UKOPA Fault Database Management Group.

7. D.Lyons, 2002. Western European cross country oil pipelines– 30-year performance statistics. CONCAWE, Report 01/02.

8. M.Baker, 2004. Dent study final report, TTO Number 10,Integrity Management Program, Delivery Order DTRS56-02-D-70036, DOT Research and Special ProgramsAdministration, Office of Pipeline Safety, November.

9. D.C.Johnston and T.G. Hrncir, 2002. Using in-line inspectionto address deformations containing near-neutral pH stresscorrosion cracking. ASME, Calgary, Alberta, Canada.

10. S.D.Ironside and L.B.Carroll, 2002. Pipeline dentmanagement program. Proc. International PipelineConference, IPC 2002, Calgary, Alberta.

11. J.McCoy and S.Ironside, 2004. Dent management program.Proc.International Pipeline Conference, IPC 2004, Calgary,Alberta.

12. C.R. Alexander, 1999. Review of experimental and analyticalinvestigations of dented pipelines. ASME Pressure Vesselsand Piping Division, Publication PVP.

13. J.R.Fowler, A.T.Katsounas, and R.Boubendier, 1992. Criteriafor dent acceptability of offshore pipelines. PRCI Report PR-201-927, Pipeline Research Council International CatalogNo. L51671, July.

14. ASME, 2004. ASME B31.8, Gas transmission anddistribution piping systems, 2004 Edition.

15. S.A.Lukasiewicz, J.A.Czyz, C.Sun, and S.Adeeb, 2006.Calculation of strains in dents based on high resolution in-line caliper survey. IPC2006 Paper 10101, Presented at 6thInternational Pipeline Conference, Calgary.

16. M.J.Rosenfeld, P.C.Porter, and J.A.Cox, 1998. Strainestimation using Vetco deformation tool data.Proc.International Pipeline Conference, Calgary, Alberta,Canada.

17. S.J.Dawson, A.Russell, and A.Patterson, 2006. Emergingtechniques for enhanced assessment and analysis of dents.Proc. 6th International Pipeline Conference, Calgary, Alberta,Canada.

18. D.B.Noronha, R.R.Martins, B.P.Jacob, and E.Souza, 2005.The use of B-splines in the assessment of strain levels associatedwith plain dents. Paper IBP1245-05, Rio Pipeline Conferenceand Exposition, Rio de Janeiro, October.

19. D.J.Warman, D.Johnston, J.D.Mackenzie, S.Rapp, andB.Travers, 2006. Management of pipeline dents andmechanical damage in gas pipelines. Proc. InternationalPipeline Conference.

20. ASME, 2007. ASME B31.8, Gas transmission anddistribution piping systems.

21. J.A.Czyz, S.A.Lukasiewicz, C.Sun, and S.Adeeb, 2008.Calculating dent strain. Pipeline and Gas Technology, January/February, pp38-45.

22. M.J.Rosenfeld, 2001. Proposed new guidelines for ASMEB31.8 on assessment of dents and mechanical damage.Topical Report GRI-01/0084, Gas Technology Institute.

23. 49CFR192, 2003. Pipeline safety: pipeline integritymanagement in high consequence areas (gas transmissionpipeline). Final Rule, Department of Transportation, 15December.

24. ASME, 1992. ASME B31.4, Liquid transportation systemfor hydrocarbons, liquid petroleum gas, anhydrous ammoniaand alcohols.

25. C.R.Alexander and J.F.Kiefner, 1997. Effects of smooth androck dents on liquid petroleum pipelines. API 1156,November.

26. J.F.Kiefner and C.R.Alexander, 1999. Effects of smooth androck dents on liquid petroleum pipelines (Phase II).Addendum to API 1156, October.

27. P.Roovers, R.Bood, M.Galli, U.Maerewski, M.Steiner, andM.Zarea, 2000. EPRG methods for assessing the toleranceand resistance of pipelines to external damage. PipelineTechnology, Volume II, Bruges 29 September.

28. Canadian Standards Association, 1999. Z662:1999, Oil andgas pipeline systems.

Sample

copy


Sample

copy


BLASTING AND CONSTRUCTION activity nearexisting pipelines is a common occurrence which can

result from new construction or expansion of existingfacilities, such as mines and quarries, near the pipeline.Pipeline operators have a need to ensure that these activitiesdo not compromise the integrity of their pipelines.Unfortunately, the answer is not as simple as maintaininga specific distance from the pipeline. Factors such as theenergy of the blast, local geology, and the integrity of thepipeline must factor into this assessment.

Measurement of the effect of blasting on the pipeline ispossible using strain gauges and/or accelerometers placeddirectly on the pipe. Although direct measurement ispossible, it is not always practical or cost effective. In manyinstances, such as the construction of a new line in the sameright-of-way, the location of the blast is constantly changing,and direct measurement in these instances would result inexposing a good portion of the pipeline. For this reasonoperators must depend on indirect measurements such asvibration monitoring to infer the stress on the pipe fromblasting.

There has been considerable research into the effects ofblasting vibrations on structures over the years.Unfortunately, for the pipeline industry, much of thisresearch has been focused on the effects of blasting onabove-ground structures. A resource that brings much ofthis research together can be found in the book by Dowding


*Author’s contact detailstel: +1 614 410 1605email: [email protected]

Blasting and constructionvibrations near existingpipelines: what are theappropriate levels?

by Robert B Francini*1 and William Nik Baltz2

1 Kiefner and Associates, Inc., Worthington, OH, USA2 El Paso Natural Gas, Colorado Springs, CO, USA

CONSTRUCTION AND industrial processes such as mining and quarry blasting, or pile driving, neara pipeline create vibrations that will result in stress on the pipeline which is added to the normal

operating stress on the line. The obvious way to determine this stress is by exposing the pipeline, installingstrain gauges, reburying the line, and measures its response to the event. This is very costly and not apractical approach in most cases. A common method that is used for monitoring these activities is tomeasure the vibration of the ground above the pipeline: the question then becomes, how do thesevibrations relate to the stress on the pipe? The answer to this question is very important both to the pipelineand the construction/process operators, as it will determine what the allowable vibration levels are.

This paper presents the results of a recent project where four pipelines located within a coalfield productionarea were instrumented with strain gauges and the stress on the lines measured, along with the particlevelocity in the soil above the lines during a large blast. This data will be compared with blasting data generatedby the Bureau of Mines and Esparza. The paper then looks at methods for estimating the stress to see howthey compare with the data. Finally, guidelines are presented for determining acceptable vibration levelsover a pipeline based on these results and an integrity analysis of the pipeline.

Sample

copy


on construction vibrations [1]. The former United StatesDepartment of the Interior Bureau of Mines had conducteda large amount of research on the effects of blasting onstructures. The reports of most interest to pipeline operatorsare the Bureau of Mines reports on surface mine blastingnear presÄï:surized transmission pipelines [2], the effectsof millisecond-delay intervals on vibrations [3], and blastingvibration effects on structures [4]. In addition, the PipelineResearch Committee of the American Gas Association(AGA) sponsored two research efforts on pipeline responseto blasting [5, 6]. As will bee seen, the range of applicabilityof this research is quite different.

We will focus first on the results of a project to measure thestress in four active pipelines resulting from blasting toremove the overburden at a nearby surface mine. Theseresults will be compared with existing methods for analyzingblasting stresses. Finally, we will see how these results canbe used to relate surface vibration to the stress on thepipeline and offer recommendations on the appropriatevibration levels on a pipeline based on allowable stresslevels.

NomenclatureE = Young’s modulus for steel (30 x 106psi)n = ratio of the energy in a charge to the

equivalent amount of ANFO (ammoniumnitrate and fuel oil, dimensionless)

PPV = peak particle velocity (ips)R = distance from the pipe to the blast (ft)t = wall thickness of the pipe (in)W = weight of charge (either total or charge per

delay) (lb)e

long= longitudinal strain (in/in)

ehoop

= hoop strain (in/in)s

Esp= estimate of blasting stress using equation

developed by Esparza (psi)s

long= longitudinal stress (psi)

shoop

= hoop stress (psi)s

max= estimate of the maximum blasting stress

(psi)s

max_upper= upper bound estimate of maximum blasting

stress (psi)n = Poisson’s ratio for steel (= 0.3, dimensionless)

eniLenilepiPepyt

retemaiD)mm/ni(

llaWssenkciht)mm/ni(

edarG tliubraeYPOAM/POM)aPM/isp(

1rebmuN tcudorP 912/526.8 2.8/223.0 B 2591 9.21/078,1

2rebmuN tcudorP 423/57.21 5.9/573.0 25X 9991 9.41/061,2

3rebmuN saglarutaN 955/22 5.9/573.0 06X 5002 8.5/548

Table 1. Pipelines within the rights-of-way near the coal mine blasting.

Fig.1. Location of pipelines in relation to blast site.

Sample

copy


BackgroundA coal-mining company was expanding its mining operationat a facility in Wyoming. Two operators had pipelines in thevicinity that were affected by this expansion. One had twoproducts pipelines: an 8-in, 0.322-in wall thickness, GradeB, pipeline, and a 12-in, 0.375-in wall thickness, X52pipeline. The other operator had a 22-inch, 0.375-in wallthickness, X60 natural gas pipeline. A complete descriptionof these pipelines is summarized in Table 1. The expansionresulted in the relocation of a 2-mile portion of the 22-inpipeline. The 8-in line was isolated and temporarily idled.Even with the relocation, the final blasting came within

500ft of the 22-in gas pipeline and a slightly furtherdistance from the 12-in product pipeline. Figure 1 shows asketch provided by the mine showing the locations of thesepipelines in relation to the blast site.

Site instrumentationFour sites were selected to be instrumented based on theinformation supplied by the mine. Two of these sites wereon the 12-in pipeline and two sites were on the 22-inpipeline. These sites were selected because they were locatedclosest to the blast site. One site for each pipeline was

Hole 3

El Paso Denver to Rock Springs Main Line 22-inch

approximately 250 ft

78 ft.

92 ft.

Hole 4

ConocoPhillips Pioneer 12-inch

Instrumentation shed

Hole 1Hole 2

Fig.2. Site layout for monitoring mine blasting.

Keep side as vertical as possible

Strain gauge

Toward Mine

Side View

End View

Accelerometer

Fig.3. Diagram of bellhole configuration and instrumentation locations.

Sample

copy


located on the straight section of pipe to the east of the fieldbends in each pipeline, while the second site was located ona straight section near the field bend on each pipeline todetermine if the bend would affect the results. Figure 2shows a sketch of the site layout along with the location ofthe instrumentation holes. The test holes were designed tominimize the disturbance of the soil around the pipe.

Figure3 shows the design of the bellhole along with thelocation of the strain gauges on the pipe. A biaxial straingauge was laid on the pipe at the top dead centre (TDC) andone was laid 90o from TDC on the side of the pipe towardthe blast. The biaxial gauge was oriented with one gaugealong the axis of the pipe and one gauge in the hoopdirection. This was the same configuration used in Ref.2.

Fig.4. Strain gauge installation on 2-in products pipeline at Bellhole 2.

Fig.5a. Longitudinal strain measuredat top dead centre of 12-in product

pipeline, Bellhole 1. Top graph is rawdata in Volts; middle graph shows the

frequency spectra of the data in therange of 0-50Hz; bottom graph is

the final data in min/in of strain afterfiltering the data using a 2-50Hz

bandpass filter.

Fig.5b. Circumferential strainmeasured at top dead centre of 12-in

product pipeline, Bellhole 1. Graphsas in Fig.5a.

Sample

copy


Figure 4 shows the final strain gauge installation at Bellhole2 on the 12-in products line. Note that the gauges werecovered with a protective tape before they were buried.

Since the soil around the pipe was disturbed by digging theholes, the seismometers were placed on the groundapproximately 20ft from the strain gauge locations. Theaccelerometers were oriented with the radial direction

being directed in a line from the strain gauges to the nearestblast location.

ResultsThree strain gauges were not working on the day of the test.Since a full set of gauges was working on both pipelines and

rotarepO eloH noitatneirO noitacoL

)ni/ni(niartS

mumixaM muminiM

tcudorphcni-21 1 lanidutignol CDT 0.22 5.81-

tcudorphcni-21 1 laitnerefmucric CDT 3.62 8.92-

tcudorphcni-21 1 lanidutignol tsalbgnicaf 1.12 3.91-

tcudorphcni-21 1 laitnerefmucric tsalbgnicaf 0.22 0.92-

tcudorphcni-21 2 lanidutignol CDT 9.82 8.62-

tcudorphcni-21 2 laitnerefmucric CDT 2.12 1.41-

tcudorphcni-21 2 lanidutignol tsalbgnicaf 8.21 2.31-

tcudorphcni-21 2 laitnerefmucric tsalbgnicaf A/N A/N

saglarutanhcni-22 3 lanidutignol CDT 8.41 6.91-

saglarutanhcni-22 3 laitnerefmucric CDT 3.72 2.74-

saglarutanhcni-22 3 lanidutignol tsalbgnicaf 4.51 8.11-

saglarutanhcni-22 3 laitnerefmucric tsalbgnicaf 5.61 9.83-

saglarutanhcni-22 4 lanidutignol CDT 8.91 6.42-

saglarutanhcni-22 4 laitnerefmucric CDT 7.51 1.23-

saglarutanhcni-22 4 lanidutignol tsalbgnicaf A/N A/N

saglarutanhcni-22 4 laitnerefmucric tsalbgnicaf A/N A/N

epiP-lleBeloh

noitatneirO noitacoL

ssertslautcA)aPM/isp(

evitavresnoC s ssert)aPM/isp(

mumixaM muminiM mumixaM muminiM

tcudorpni-21

1 lanidutignol CDT 0.5/817 2.3-/754- 8.6/689 2.6-/609-

1mucric -

laitnerefCDT 8.4/307 0.6-/568- 5.7/480,1 0.8-/761,1-

1 lanidutignol tsalbgnicaf 2.4/806 9.4-/907- 3.6/219 4.6-/429-

1mucric -

laitnereftsalbgnicaf 1.5/937 6.6-/269- 4.6/439 9.7-/741,1-

2 lanidutignol CDT 0.6/578 2.6-/309- 0.8/361,1 1.7-/320,1-

2mucric -

laitnerefCDT 3.4/326 8.3-/645- 8.6/489 0.5-/037-

2 lanidutignol tsalbgnicaf A/N A/N A/N A/N

2mucric -

laitnereftsalbgnicaf A/N A/N A/N A/N

epiP-lleBeloh

noitatneirO noitacoL

ssertslautcA)aPM/isp(

evitavresnoC s ssert)aPM/isp(

mumixaM muminiM mumixaM muminiM

ni-22

larutan

sag

3 lanidutignol CDT 7.2/093 9.3-/275- 2.5/957 7.7-/411,1-

3-mucric

laitnerefCDT 8.5/848 2.01-/084,1- 2.7/740,1 1.21-/157,1-

3 lanidutignol tsalbgnicaf 7.3/035 0.3-/924- 6.4/076 3.5-/577-

3-mucric

laitnereftsalbgnicaf 4.3/594 7.8-/262,1- 8.4/796 6.9-/893,1-

4 lanidutignol CDT 4.4/446 1.5-/647- 6.5/908 8.7-/821,1-

4-mucric

laitnerefCDT 4.3/784 6.7-/801,1- 9.4/317 0.9-/303,1-

4 lanidutignol tsalbgnicaf A/N A/N A/N A/N

4-mucric

laitnereftsalbgnicaf A/N A/N A/N A/N

Table 2. Summary of maximumstrains measured on 12-in productand 22-in natural gas pipelines.

Table 3a. Summary of maximumand minimum stresses calculatedusing Equns 1 and 2 for 12-inpipeline.

Table 3b. Summary of maximumand minimum stresses calculatedusing Equns 1 and 2 for 22-inpipeline.

Sample

copy


the gauges that were working in Bellholes 2 and 4 gaveresults that were in the same range as the remaining gaugesin the holes this was not thought to be a significant issue.

Typical strain gauge results are shown in Figs 5a and 5b.There are three plots in each figure: the top graph representsthe raw data from the test, in Volts. The middle graphshows the frequency spectra of the raw data in the range of2-50Hz. The bottom graph shows the measured strain inMUin/in after the data had been filtered using a 2-50Hz

band-pass filter. The band-pass filter was used to eliminatethe 60-Hz harmonic noise that was introduced by thegenerator used to power the equipment. In addition, therewas a low-frequency periodic signal superimposed on thedata whose source we were unable to resolve. The excitationfrom the blast was contained within this 2-50Hz range sono data were lost doing this. The measured strain neverexceeded the range of +30/-50 MUin/in: Table 2summarizes the maximum and minimum strain measuredat each location.

-3,000

-2,500

-2,000

-1,500

-1,000

-500

0

500

1,000

1,500

0 1 2 3 4 5 6

Time, sec.

Stre

ss, p

si

-1,500

-1,000

-500

0

500

1,000

1,500

2,000

2,500

3,000LongitudinalCircumferential

Fig.6. Stress measured at top deadcentre of 12-in product pipeline,

Bellhole 1.

Fig.7. Ground velocity at the surface measured at Bellhole 1 of the 12-in products pipeline. The top graphs are the radialvelocity and frequency spectra, the middle graphs are the vertical velocity and frequency spectra, and the bottom graphs arethe transverse velocity and frequency spectra.

Sample

copy


Since the strain is biaxial, the following formulas are usedto convert strain to stress:

σν

ε νεlong long hoop

E=

−+

1 2( ) (1)

σν

ε νεhoop hoop long

E=

−+

1 2( ) (2)

in which the nomenclature is given above.

The data were converted on a point-by-point basis usingEquns 1 and 2. Typical stress results are shown in Fig.6 (thearrows point to the appropriate axis). Tables 3a and 3bsummarize the maximum and minimum stress, and alsocontain a conservative stress estimate based on Ref.2. Thisvalue for the stress was calculated using the absolutemaximum (or minimum) values of the strain from eachrecording of the longitudinal and hoop strain in Equns 1and 2. The actual stress ranged from +900 to –1,500psi(+6.2 to –10.3MPa) and the conservative estimate rangesfrom +1,200 to –1,800psi (+8.3 to –12.4MPa).

The result from the seismometer located near Bellhole 1 isshown in Fig.7, and Table 4 summarizes the maximumground velocities measured at each bellhole.

DiscussionThe mine reported that it had used 522,167lbs of explosivefor the shot and the maximum charge per delay was2,400lbs. We estimate that Bellholes 2 and 3 wereapproximately 500ft from the nearest blast, and Bellholes1 and 2 were 550ft from the nearest blast.

Data from blasting have been found to correlate with scaleddistance (SD), the normal definition of which is given as:

SDR

W= (3)

where

R is the distance from the blast, andW is the charge per delay.

The stress on the pipe is shown in Table 4. As described

above, the conservative estimate of stress uses the maximum(minimum) values for the longitudinal and circumferentialstrain from each gauge to calculate the stress. We plottedthe absolute maximum value of stress measured for eachlocation along with the data from the Siskind [2] andEsparza [6] reports as a function of scaled distance, in Fig.8.The results from the mine blast compares very well with thedata from Siskind [2]; the Esparza [6] data are shown forreference only and are from a small charge per delay closein blasting which does not correspond to the case at themine blasting. The regression line shown in Fig.9 is basedon the Siskind data [2]. The curve for this fit is:

smax

= 19,580 SD-1.089 (4)

smax_upper

= 32,767 SD-0.933 (5)(upper 90% confidence level)

where smax

and smax_upper

are estimates of the maximum stressand upper-bound maximum stress that result from blasting.

The R2 value for the curve fit is 0.85. Since it seems toencompass all of the data, we recommend the upper 90%confidence level (Equn 5) as an estimate of the stressresulting from mine or quarry blasting under similarconditions. The stress is assumed to be the same in thehoop and axial directions.

The maximum stress is compared with the peak particlevelocity (PPV) in Fig.9. Once again, the curve fit is based onthe Siskind [2] data for the same reasons presented above.The curve fit is:

smax

= 302.9 PPV + 201.7 (6)

smax_upper

= 365.6 PPV + 446.0 (7)(upper 90% confidence level)

The value of R2 for Equns 6 and 7 is 0.82. Strictly theintercept for both equations should be zero since therewould be no stress at a PPV of 0ips. This did not result ina conservative upper bound, so Equns 6 and 7 were usedwith the understanding that there will be no stress at 0ipsPPV. We recommend using the upper 95% confidencelevel for estimating the stress resulting from future blastingunder similar conditions since this is the confidence levelthat is typically used in pipeline risk management.

The peak particle velocity (PPV) is compared with thescaled distance in Fig.10 along with the Siskind [2] and

rotarepO noitacoLyticolevlaidaR

)s/mm/spi(yticolevlacitreV

)s/mm/spi(

esrevsnarTyticolev)s/mm/spi(

tcudorpni-21

1elohlleB 36/84.2 001/29.3 45/21.2

2elohlleB 96/27.2 08/61.3 25/40.2

saglarutanni-22

3elohlleB 001/29.3 411/84.4 631/63.5

4elohlleB 97/21.3 07/67.2 78/44.3

Table 4. Summary of the maximumground velocity during mine blastmeasured near the instrumentationBellholes on 12-in and 22-inpipelines.

Sample

copy


Esparza [6] data. The Blasters’ Handbook [7] presents fourequations that can be used to estimate the PPV as afunction of scaled distance if there is no local informationon this relationship. These equations are:

PPV = 24.2 x SD-1.6 lower bound typical data (8)

PPV = 160 x SD-1.6 typical data (9)

PPV = 242 x SD-1.6 upper bound typical data (10)

PPV = 605 x SD-1.6 high response (11)

Equations 8 through 11 are included in Fig.10. The datafor this particular mine blast lie between the typical dataand upper-bound typical data curves. It should be notedthat the Blasters’ Handbook recommends that it is preferableto use local data if they are available, since PPV data can varywidely. These data can be obtained by performing a

regression on PPV versus scaled-distance data for a series oflocal shots at various distances from the shot.

Recommendations foracceptable vibration levels

These results can be used to determine acceptable vibrationslevels for blasting near pipelines. The first step of theprocess is to determine the maximum acceptable stress onthe pipeline. The pipeline integrity data should be checkedto ensure that there are no defects in the vicinity of theblasting. If there are defects the maximum acceptable stresslevel should be determined based on the predicted failurestress for the defects. If there are no defects, the acceptablestress can be based on the applicable code for the pipelineor an engineering analysis of the pipeline. Once theacceptable stress level has been determined, Equn 5 can be

1

10

100

1,000

10,000

100,000

1.0 10.0 100.0 1,000.0

Scaled distance, ft/lb0.5

Max

imum

stre

ss, p

si

Siskind DataEsparza DataHole 1Hole 2Hole 3Hole 4

Regression Line

90% Confidence Level

0

500

1,000

1,500

2,000

2,500

3,000

3,500

4,000

0.00 1.00 2.00 3.00 4.00 5.00 6.00 7.00 8.00 9.00 10.00

Peak particle velocity, ips

Max

imum

stre

ss, p

si

Siskind DataEsparza DataHole 1Hole 2Hole 3Hole 4

Regression Line

95% Confidence Level

Fig.8. Mine blast data for maximumstress versus scaled distance compared

with data from the Siskind and Esparzareports.

Fig.9. Mine blast maximum stressversus peak particle velocity

compared with data from the Siskindand Esparza reports.Sam

ple co

py


0.100

1.000

10.000

100.000

1.0 10.0 100.0 1,000.0

Scaled distance, ft/lb0.5

Gro

und

vitr

atio

n, ip

s

Siskind dataEsparza dataHole 1Hole 2Hole 3Hole 4Lower bound for typical responseTypical responseUpper bound for typical responseHigh response

1

10

100

1,000

10,000

100,000

1.0 10.0 100.0 1,000.0

Scaled distance (ft/lbs0.5)

Stre

ss, p

si

Siskind dataEsparza dataHoles 1-4Upper 90 %Regression lineEsparza Equation

Fig.10. Mine blast data formaximum peak particle velocityversus scaled distance comparedwith data from the Siskind andEsparza reports.

Fig.11. Mine blast data formaximum stress versus scaleddistance compared with data fromthe Siskind and Esparza reports,including the Esparza equation.

used to determine the smallest acceptable scaled distance,and this is then used in the lower-bound PPV given byEqun 8.

This procedure gives a very-conservative PPV limit since itis based on the lower-bound PPV equation. If a localrelationship for PPV is known, or if it is possible to carryout some test shots to determine this relationship, the localPPV-scaled distance should be used.

Equation 7 could be used to determine PPV directly, butthe main drawback to this approach is that it does not allowthe use of local PPV data.

As an example of this method, let us assume that a quarrywants to expand its operation near a natural gas pipeline.The pipeline is Grade X60 operating at 72% of SMYS(specified minimum yield stress). The operator hasdetermined that there are no defects in the pipeline in the

section affected by the quarry blasting. The operator hasalso determined that it is acceptable for the stress in thepipeline to go as high as 90% of SMYS based on anengineering analysis of its pipeline. This means that thestress on the pipeline as a result of the blasting can go ashigh as 10,800psi (74.5MPa). This value is used in Equn 4to determine the smallest-acceptable scaled distance fromthe pipeline. The result is 3.28ft/lb0.5 (0.67m/kg0.5). Finally,the scaled distance is used in Equn 8 to determine themaximum-acceptable PPV, which is seen to be 3.6ips(91.4mm/s).

As stated above, this is an extremely-conservative value. Onthe other hand, if it has been determined that therelationship between PPV and scaled distance is better-approximated by the typical data Equn 9, this acceptablePPV level increases to 23.9ips (607mm/s). Although thissecond vibration level seems very high, Siskind [2] reportsa PPV level higher than this value where the stress on the

Sample

copy


pipe was lower than our allowable limit. This shot did notresult in damage to the pipeline. If Equn 7 is used, the resultis a PPV of 28.3ips (719mm/sec). These calculations showthat, based on stress alone, pipelines in good condition canwithstand very high vibration levels from blasting.

The vibration levels calculated above are considerablyhigher than the value of 2ips (50mm/s) that is normallyquoted as the maximum allowable PPV for blasting. Thisnumber is based on the response of above-ground structuresto blasting, and is based on the threshold point where theplaster in a dwelling will begin to crack [1, 4]. Our analysisshows that a pipeline in good condition can withstandmuch higher vibration levels without damage. However,due to the “perceived” notion that blasting can result indamage to pipelines, and geotechnical factors that we havenot addressed, we think that levels above 10ips (250mm/s) would be unacceptable to most operators and regulators.Geotechnical factors that should be considered includesoil saturation, the potential for the earth to slump into thepipe and the rockiness of the soil. Therefore we wouldrecommend this as an upper limit for PPV during blastingnear a pipeline providing it is supported by an analysisbased on the condition of the pipeline. We have found thatlevels of 5-10ips (125-250mm/s) are usually acceptable toblasters.

Up until this point we have not considered the frequencyspectra of the shot. Studies have shown that the frequencycontent of the blast can have a profound effect on thestructures response due to resonance [1]. Neither theSiskind [2] nor Esparza [6] studies considered the frequencyresponse of the pipeline, and these effects can be consideredto be built-into the analysis. However, in the case ofvibratory equipment operating near a pipeline, the frequencyresponse of the pipe should be considered.

A note on EsparzaEsparza [5] developed an equation for the stress in buriedpipe resulting from blasting:

σ = 8 882 5

0 77. ( ).

.nW

Et RE (12)

in which the nomenclature is given above.

This equation is based on a series of experiments carriedout by Esparza [5, 6] which were more representative of

blasting that would occur during construction. They usedsmaller charges that were set closer to the pipe. Siskind [2]notes that these results do not necessarily agree with theresults for mine blasting in his report. Figure 11 is a repeatof Fig.8 with Equn 12 included; the data Esparza used togenerate Equn 12 are not included in this plot. We can seethat for small scaled distances, Equn 12 predicts muchhigher stresses than those measured by Siskind [2], andunder-predicts the stress for larger scaled distances. Siskind[2] notes that the waveforms recorded in Ref.5 are morerepresentative of a shock pulse than the vibratory motionrecorded during their mine blasting tests. Assuming thatthis is the case, then Equn 12 should be used in place ofEqun 5 in construction situations where the charges areplaced much closer to the pipeline than would typicallyoccur during mining and quarrying operations.

AcknowledgmentsThe authors would like to thank El Paso Natural Gas,ConocoPhillips Pipe Line Company, and the Black ButteCoal Company for funding this project. We would also liketo thank Robert Gertler without who’s over 40 years ofexperience with strain gauging this project would not havebeen possible.

References1. C.H.Dowding, 2000. Construction vibrations. ISBN 0-

9644313-1-9.2. D.E.Siskind, M.E.Stagg, J.E.Wiegand, and D.L.Schulz, 1994.

Surface mine blasting near pressurized transmission pipelines,RI Number 9523, Bureau of Mines, Denver, CO, USA.

3. J.W.Kopp and D.E.Siskind, 1986. Effects of millisecond-delay intervals on vibration and airblast from surface coalmine blasting. RI Number 9026, Bureau of Mines, Denver,CO, USA.

4. H.R.Nicholls, C.F.Johnson, and W.I.Duvall, 1971. Blastingvibrations and their effects on structures. BuMines B 656,Bureau of Mines, Denver, CO, USA.

5. E.D.Esparza, P.S.Westine, and A.B.Wenzel, 1981. Pipelineresponse to buried explosive detonations, Vols I and II. PRCICatalog No. L51406, PRCI, Arlington, VA, USA.

6. E.D.Esparza, 1991. Pipeline response to blasting in rock.PRCI Catalog No. L51661, PRCI, Arlington, VA, USA.

7. R.B.Hopler, Ed., 2000. Blasters’ handbook: 17th Edn. Int.Soc. Explosives Engineers, Cleveland, OH, Chap. 38.

Sample

copy


The driversThe utilization of natural gas on a large scale in centralChile began in 1997 with the completion of the Gas Andespipeline to transport natural gas from Argentina to Santiago.A second pipeline was soon extended to the coast. Duringthe period 1995 to 2000, the initial focus of natural gasdistribution companies in Chile was centred on theimplementation of business infrastructure, which includedthe construction of networks of both polyethylene andsteel pipelines. After the start of distribution activities, thefocus changed to the operation and maintenance processes.

Some companies foresaw pipeline integrity management asa potentially-useful tool to achieve that change of focus. Infact, in 2004, one natural gas distribution company inChile began developing a distribution integrity-managementsystem for its relatively-new network.

In parallel the Regulator, the Superintendencia deElectricidad y Combustibles (SEC), was thinking along thesame lines and, in 2006, started working on a modification

of the gas regulations, which contemplated pipeline anddistribution integrity-management sections.

Unfortunately in February, 2007, a serious accidentoccurred in the historic centre of Valparaíso, which causedfour fatalities, the collapse and burning of five buildings,and important damage to 36 heritage buildings. Thisaccident prompted SEC to bring forward the part of the D-PIM rule requiring an immediate assessment of regulatorycompliance and a risk analysis for each LDC and electricitydistribution company. This preliminary D-PIM requirementwas named PEIRS (Plan de Evaluación de Integridad deRedes Subterráneas: the Underground pipeline integrityevaluation plan).

The Regulator required the LDCs to conscientiously analyzetheir systems, detect relevant areas from a risk point of view,and take action to improve and/or remedy potentialshortcomings.

Industry reactionIndustry reaction was diverse. On the one hand somecompanies saw this requirement as an opportunity toimprove their processes and to align efforts oriented to afuture D-PIM regulation. Other companies reacted

Author’s contact details:tel: +56 2 428 0800email: [email protected]

Natural gas distribution integritymanagement in Chile: a newway of doing things

by Enrique Acuña CDandilion Ingeniería Ltda, Santiago, Chile

A DISTRIBUTION pipeline integrity-management (D-PIM) system is currently being implemented inChile. A serious commitment from both the Regulator and industry players (LDCs) is facilitating the

development of a comprehensive management system, initially focused on communications to educate theorganization, then quality control related to the preparation and re-engineering of procedures, and re-education of personnel to develop a new way of doing things. This new direction is focused on the thirdstep of process and risk management.

The Chilean Regulator instructed companies to use the ASME B31.8S standard as a guide for their D-PIMactivities, which resulted in a very useful framework for conceptualizing and developing most of the D-PIMactivities. The most important discovery was the realization that the involvement of a well-educated D-PIMmanagement control group was a prerequisite for the organization.

Sample

copy


defensively, assigning limited resources to accomplish theminimal legal requirement, with no real impact on theirmanagement processes.

PEIRS applicationFor those companies that decided to go beyond theminimum authority requirements, local consultants wererequested to develop a complete model to implementPEIRS for their distribution systems.

The first step was to establish a ‘Baseline Regulatorydocument’, the basis of which included all applicablecodes, regulations, and company procedures. This activityrequired an important effort.

The second step involved the arduous task of collecting allthe data required order to demonstrate compliance notonly with all applicable regulations but also with the LDC’sown technical specifications, drawing standards, andoperating and maintenance procedures.

In parallel, and using the information from the Regulatorycompliance analysis, the risk team worked on the applicationof a ‘relative-risk analysis model’, a proprietary developmentmodel based on both the ‘subject matter experts’ (SME)and ‘indexed model’ approach. A relative index model [1]was used in this analysis, and used dynamic segmentationto assign risk attributes to each pipeline segment. Thisbecame feasible with the integration of the risk modeldatabase and GIS (geographical-information system). Thisintegration also made possible the geographicalrepresentation of relative risk, which was very useful in thevalidation and analysis of the results.

Finally, a set of recommendations was developed related torisk reduction and the correction of any shortfalls relatedto Regulatory compliance and/or company technicalspecifications, maintenance, and operating procedures.These recommendations were structured as follows:

• Normalization plan: to repair defects and correctsituations related to Regulatory or company standardcompliance shortfalls.

• Prevention plan: to prepare for and optimize controland response to threats not related to networktechnical or safety conditions.

• Inspection plan: to ensure the collection of all relevantand necessary information about actual pipelineconditions.

• Mitigation plan: to minimize consequences ofpotential incidents.

The recommendations of each plan where classified

according to the ANSI/ASME B31.8S document Managingsystem integrity of gas pipelines.

PEIRS findingsThree major distribution companies in Chile voluntarilyimplemented a rigorous model application for their PEIRSstudies. The findings were very similar:

• An understanding that D-PIM is an importantconcept for reducing risks and guaranteeing anextended life of distribution systems.

• Difficulties existed in obtaining effective supportfrom internal service areas for process improvement,organizational structure, and personnelcompetencies management.

• Some difficulties were encountered regarding theavailability of data.

• In some cases poor performance measurement andprocess management were found.

• When present, quality control systems were focusedon formal process compliance and not on processeffectiveness.

• Processes were not developed and managed basedon a formal risk approach.

• System design and construction were very safe,exceeding the minimum ANSI B31.8 requirements,particularly for steel pipelines operating at 10 bar(145 psig) or above.

• Third-party damage was the main threat. Currentefforts should be complemented with better third-party co-ordination where possible, and education.

• Lack of authorities’ agreement and co-ordinationmakes it difficult to implement a comprehensiveprevention programme.

D-PIM regulationIt is expected that a new gas safety regulation will bepublished shortly. This regulation will include a dedicatedsection on gas network integrity management systems(Sistema de Gestión de Integridad de Redes – SGIR). TheRegulator considers that companies should establish arobust management system. The idea is to implement, ona priority basis, an integrated system focused on riskmanagement,

SGIR will make reference to the US’ 49 CFR Part O Pipelineintegrity management and ANSI/ASME B31.8S Managing

Sample

copy


system integrity of gas pipelines. All ANSI/ASME B31.8Smanagement tools (communications, quality control,management of change, and performance plans) will beapplicable to both transmission and distribution systems.The principal difference for distribution systems will be inintegrity-management plans, particularly in the risk-analysisand inspection processes. Both of these issues were tackledduring PEIRS development phase.

Previous SGIR experienceOne pioneer LDC in Chile decided in 2004 to implementD-PIM, now referred to as SGIR. It took five years toachieve full SGIR implementation for all the operation andmaintenance processes, which were recently completed. Atthe beginning, the effort was focused on studying andanalyzing the new US gas pipeline integrity regulations andhow apply them to high-pressure distribution systems (steelpipelines operating at 35 bar), which are not essentiallydifferent to transmission pipelines.

The first step involved an educational process to explainand convince top management of the potential value of anSGIR. When management understood the risk conceptsand the benefits of a more formal approach to riskmanagement, regulatory compliance, operating procedures,and maintenance, it agreed that each of the distributionsystems should be managed with this structured approach.

Finally a ‘company commitment letter’, signed by thegeneral manager, was issued to each person in theorganization, asking their support for the implementationof the SGIR process. This initial process took nearly a yearto complete.

Local consultants assisted with the SGIR implementationand initially the efforts were focused on the communicationand quality control plan. Not only the engineering andO&M personnel, but the entire organization, needed to be

educated in SGIR concepts. Incident-prevention and -mitigation processes were given priority for re-engineering.

After approximately one year of development work theteam realized that most of the concepts used for the high-pressure network could be used for the low-pressuredistribution systems, particularly those involved withincident-prevention procedures. From that moment on,SGIR implementation was extended to all distributionsystems.

It was also realized that it would be necessary to assign themanagement of the system to a specific area of theorganization. A group of engineers, who were originallyfully focused on planning and administrative activities inthe area of operations, were assigned the task of managingthe process. Nevertheless, efforts to educate this group inthe technical concepts were necessary. They were involvedfrom the beginning in re-engineering procedures, whichfacilitated future process management and improvement.

Finally, after five years of work, the company is nowrunning the SGIR on its own, improving not only technicalprocesses but also financial and human resourceperformance.

The futureCompanies which implement an SGIR, will be wellpositioned to comply with the new regulation. Without anSGIR, companies potentially face a difficult scenario, astime is needed, firstly to educate not only senior managementbut also operating personnel, secondly to get realcommitment from the organization, and thirdly to preparea unit to manage the system.

It is not clear yet, how the Regulator will enforce andmeasure SGIR implementation, so some uncertainties stillexist.

Fig.1. PEIRS flow diagram.

Regulatory Base Line

Relative RiskAnalysis

Normalization Plan

Inspection Plan

Prevention & Mitigation Plans

Cicle Begining

PEIRS CICLE

ONE TIME EVALUATION PERMANENT CICLE

INFORMATION

Regulatory Base Line

Relative RiskAnalysis

Normalization Plan

Inspection Plan

Prevention & Mitigation Plans

Cicle Begining

PEIRS CICLE

ONE TIME EVALUATION PERMANENT CICLE

INFORMATION

Sample

copy


ConclusionsIt has been demonstrated that organizational commitmentat all levels is a prerequisite to successfully develop,implement, and run an SGIR. This is only possible if acomprehensive educational process is initially carried outat the organization’s own intrinsic velocity, and neitherfaster or slower.

SGIR is a new structured way of undertaking regularactivities; the difference is that these activities now arefocused by risk and managed in a rigorous way. Consultants’participation is necessary to contribute with the creationmethodology, planning, technical advice and – most

importantly – to be an independent mediator in internalnegotiations.

An educational effort with local authorities not directlyinvolved in safety regulations is necessary to obtain theircommitment and support for the company’s damage-prevention programmes. The Regulator will also potentiallyneed to develop a management plan to oversee compliance.

Reference1. W Kent Muhlbauer, 2004. Pipeline risk management manual,

3rd Edn. Gulf Professional Publishing.

Sample

copy


THE WORLD IS short of educated people. In thepetroleum industry, there is a particular problem.

Many bright and energetic people were drawn into theindustry in the past, in the USA particularly in the 1960sand 1970s, in Europe in the 1970s, in Asia and Australiain the 1980s. Many of those people are soon to retire: oneestimate has it that half the industry will retire in the nextdecade.

Secondly, there are extraordinary new challenges toaccomplish developments less expensively, more quickly,more safely, and in a way more sensitive to the human andnatural environments. We cannot be content with simplycontinuing to do what we have done in the past. Some ofthese pressures come from governments, some fromfinancial institutions, and some from the wider community,particularly from people concerned about questions suchas climate change.

Thirdly, we live in an international environment. Qualifiedpeople move freely between one country and another, andit is easier for them to do so that it ever has been in the past.There remain some constraints of language and culture,but every day there are fewer of them. If someone doesn’tlike where he is working, he moves on.

The same situation exists elsewhere. Different industriesare competing to attract talent with better pay and benefitsto feed their development. Before the economic crisis, andat the time of high oil prices greater than $100/barrel, thebanking sector paid better to attract graduates, amongthem engineers.

RecruitmentGetting recruitment right is difficult, and everyone worriesabout it. Universities are very much concerned about it,both for students and for academic staff. Academic staff areincreasingly difficult to recruit as it is hard to competeinternationally, and quality is suffering. Recruitment ofstudents is also difficult: they want to come to university,but choice of where to go and what to study is difficult andexposed to all kinds of pressures, for instance fromgovernments that want to use universities to secure socialgoals.

Developing extraordinary talentby Professor Andrew Palmer*1 and Andrew Ngiam2

1 Centre for Offshore Research and Engineering, National University of Singapore,Singapore

2 INTECSEA, Singapore

THIS PAPER explores two areas: how to bring talented people into the industry, and how we developtheir talents and knowledge. Part of that is recruitment: to put it more crudely, how does an

organization get more than its fair share of the best people? Specifics are more illuminating than generalphilosophy, and we shall rely on anecdote and example much more than on sociological research

This paper was originally presented at the 5th Asian Pipeline Conferenceheld in Kota Kinabalu, Malaysia, on 28-29 October 2009.

*Author’s contact details:tel: +65 6516 4601email: [email protected]

Sample

copy


An anecdote about a recruitment failure. The daughter ofone of the authors wanted to work in banking, and sheapplied to an investment bank, a household name. Therewere 2000 applicants, an elaborate interviewing andscreening process, and she was one of eight who wereselected. The bank put the group through a trainingprogramme, which was estimated to cost £40,000 per head.After four months the bank assigned each of them to adepartment. Within three-and-a-half years, every single oneof the eight had left the bank, and five out of the eight leftthe banking industry altogether. Without going into thereasons – and they would obviously be different for eachindividual – that was an extraordinarily wasteful andexpensive process for the bank – think of just reading 2000applications – and all the bank got out of it was a few man-years’ of work.

In the oil industry, many experienced engineers wereforced out in the 1980s and 1990s when oil was for a timeat rock-bottom prices below $15/barrel. No effort wasmade to attract people to join or to stay, and it led to a hugeshortage of experienced mid-level engineers when the oilprice hit $150/barrel a couple of years ago.

It is illuminating to talk to students and recent graduates inthe industry. They are very diverse and come from manydifferent backgrounds, and what each is searching for isdifferent. They want challenge, meaning, and content, aswe all do. Money is a signal, but not the primary goal. Manylook forward to working overseas to see the world, and theability to provide that experience is one of the assets the oilindustry has.

Some years ago an excellent graduate student came toCambridge from the University of Michigan. She had notheard about the petroleum industry as an undergraduate,and originally she wanted to go into environmental research.She learned about the offshore industry in the course of anMSc project on seabed trenching for a contractor, and itappealed to her as a career. She was made offers by twointernational majors, call them X and Y. Y offered moremoney than X, but she joined X rather than Y, because Xmade the job sound more interesting.

She emphasized the people she met. There can only havebeen a handful of individuals in each company, but whatthose individuals said was central to her choice: she did notsay anything about glossy brochures. It is noticeable thatsome companies engage their top people both in recruitmentand in keeping close to their recruits when they arrive. Ifyou go to a Schlumberger recruitment conference, forexample, you meet individuals very high up in management.

i:Many students want the opportunity to travel, and maychoose to travel after they graduate rather than joining acompany immediately and settling down, even if not ‘settlingdown to the long littleness of life’. This is particularly so forstudents who have been continuously in the education

process since they were five years old, and have not takena ‘gap year’ between school and university. They value thepossibility of taking reasonable vacations. By comparison,things like pension plans, health insurance, and sportsclubs carry no weight: they become more significant onlymuch later.

How do we find the best people? Another anecdote. Manyyears ago one of us worked for a consulting engineeringcompany in Holland. It had never recruited new graduates,but it was expanding. One day the engineering managerdecided that he wanted to recruit two graduates. TechnicalUniversity Delft was just down the road, the best engineeringand science university in the country (and one of the bestin Europe). The Professor of Offshore Engineering had alist of the students who were going to graduate a monthlater, and went down the list. ‘This is the best’, he said, ‘Youdon’t want that one. That one only wants to work in thethird world. That is the second best’. Three of us interviewed,we recruited two of the students, and both turned out verywell. That had been good for the company, good for Delft,and good for the students. We were looking for the verybest, and the professor had known them for years, and waslikely to be a good judge. We did not let the professorchoose whom we were going to employ, but we did let himadvise us whom to interview. The best: ‘one does not rejectan opportunity to join an elite force’ (the words of KimPhilby, a British spy who defected to the Soviet Union, butnever mind!).

Of course we are not only looking for students. We oughtnot to disdain the possibilities of bright people from otherindustries. There is a story about the development of radarin the UK in the Second World War. Trained scientistswere in short supply, and there was direction of labour. Theradar team were asked what kind of people they wanted.“Send us some more zoologists”, they replied, becausemany of the leaders of the successful radar effort had comefrom biology.

People are looking for something challenging andworthwhile and interesting. They are not prepared to putup with boring work on a promise that it will turn morechallenging in a few years’ time. That is a constant problemin engineering: companies recruit good people but fail tomotivate them, and they leave

The question does rate an uncomfortable thought: if thereis boring work that somebody has to do, perhaps you oughtnot to recruit the brightest: the second and third classmight do better. It is disquietingly reminiscent of AldousHuxley’s Brave New World, where people were born andeducated as alphas, betas, gammas, deltas, and epsilons,and Huxley has one of them remark ‘I’m glad I’m a beta’.But a better response would be to restructure the work,perhaps with the aid of the vast IT resources that are nowavailable, so that it becomes more challenging and lessboring.

Sample

copy


Career development

How do we develop talent?

Only too often, a company recruits a good person, and thenneglects him: ‘him’ does of course include ‘her’. Peoplehave extraordinary talents, but that talent can remainhidden or fall into disuse. If people are given demandingtasks, they will respond.

Again and again one comes across people who have becometired. They ask less of themselves, and fall back intoroutines. A conversation with a middle-aged engineer whowas talking about reeling pipelines is an example. He saidthat the diameter/thickness ratio ought to be less than 22,whatever the diameter. “But that can’t be right”, one of theauthors responded (and of course it is not right). “If you sayso”, he replied, “but that‘s what people do”. That attitudewas bad for everyone around him, and certainly bad for hisproject and for the operator who was paying for it.

In the internet age, an employee may have no long-termloyalty to the company he is working for. That is especiallytrue of talented people who are willing to travel, becausethe whole world is open to them. They – both the youngand the not-so-young – are looking for interesting jobs witha reasonable pay package. It is essential for companies tomake their workplaces “a fun place to work”. The ITindustry in Silicon Valley is a leader in doing this.

Technology can help, though it is not the whole answer.Electronic communication gives us immediate access to aworld of information; there is no excuse for not using thebest. One example is design codes. It would be naive tothink that we can do without codes; they pay an importantrole is the transmission of knowledge, experience, and thenotion of ‘good practice’. It would be equally naive to thinkthat codes are perfect. They are an imperfect product ofimperfect human beings, and reminds one of the saying ofa famous pipeline engineer, Otto von Bismarck, whoseremark applies forcefully to codes: “If your respect the law,or like eating sausage, you have not watched either beingmade”. It has been argued elsewhere that codes could bewritten in a way that is more user-friendly and less likely tolead to mistakes, though that drew a thought-provokingresponse from a consultant, who said that if codes werebetter written there would be nothing for consultants todo.

Still, one finds people who don’t keep up. The best seriesof pipeline codes, at least for underwater pipelines, arethose produced by DNV. The first was in 1976, and newerversions with very substantial changes came out in 1981,1996, 2001, and 2007. It is not uncommon to review

reports, and to find that the writers had used the 1976code, which is long obsolete and belongs in the waste-paperbasket. The up-to-date codes are instantly available, andthere is no good reason not to use them.

Formal training is part of education on the job, and can bethe most efficient way of bringing people into contact withnew ideas and filling in gaps in their previous training.There are many training opportunities, but some companiesignore them. Engineering seems to spend less on trainingthan other technologies like medicine and law.

Another thing that can be done is to give people widerexperience. If someone spends his whole working life witha design consultant, he will not understand the differentpressures and priorities that a construction contractor has,and that lack of a broader understanding will affect hisdesign. If someone spends his life with a contractor, he willnot understand the priorities of a government departmentor a certifying authority, and that too will have a bad effect(as well as making it more likely for him to become tired andbored).

That can be resolved by making it a part of everyone’s careerdevelopment for him to spend time in an organization ina different part of the industry and in different countries,preferably accomplishing real work over a serious period oftime rather than simply watching. A design engineer mightspend two years working in the field. A regulator mightspend two years with a designer, a research engineer in auniversity might spend time with a contractor, and so on.Everyone would benefit.

Some years ago an engineer in a company that we had bettercall a major operator asked a consulting company for ideasabout how we could all work together more efficiently. Theconsultant spent a lot of time on a reply, and wrote anidealistic response about exchanging people, collaboratingin training, sharing financial information, carrying outresearch together, and so on. Two weeks later there was ameeting with the operator’s engineer about something else.He did not mention the reply, but at the end of the meetinghe was asked if he had had time to read it. “Never mindthat,” he answered “what are your hourly rates?” Deeplyannoying though that answer was at the time, it points toanother truth, that we all have to operate in the real world,and that economics and finance are important.

Finally, a company should have a plan for progression. Anemployee should see a path ahead, either a technical or amanagement path depending on individual aptitude,expectations, and wishes. A talented person with the rightspirit should come to a senior management positionsomewhere between 30 and 40 years of age.

Sample

copy


Sample

copy


THE LATEST PIPELINE system designed and installedby StatoilHydro is Langeled, a 44-in diameter 1200-km

long subsea pipeline system with three different designpressure levels. Clearly, there is a great saving potential in


*Author’s contact detailstel: +47 959 99 161email: [email protected]

Designing offshore pipeline safetysystems utilising flow andpressure in multi-design-pressure pipeline systems

by Gjertrud Elisabeth Hausken*, Jørn Yngve Stokke, and Steinar Berland

StatoilHydro, Forus, Norway

THE NORWEGIAN Continental Shelf (NCS) has been a main arena for development of subsea pipelinetechnology over the last 25 years. The pipeline infrastructure in the North Sea is well developed and

new field developments are often tied -in to existing pipeline systems [3].

Codes traditionally require a pipeline system to be designed with a uniform design pressure. However, dueto the pressure drop when transporting gas in a very long pipeline, it is possible to operate multi-design-pressure systems. The pipeline integrity is ensured by limiting the inventory and local maximum allowablepressure in the pipeline using inlet and outlet pressure measurements in a safety-instrumented system (SIS).Any blockage in the pipeline could represent a demand on the safety system. This concept was planned tobe used in the new Gjøa development when connecting the 130-km long rich gas pipeline to the existing450-km long FLAGS pipeline system. However, a risk assessment detected a new risk parameter: theformation of a hydrate and subsequent blockage of the pipeline. In theory, the hydrate could form in anypart of the pipeline. Therefore, the pipeline outlet pressure could not be used in an SIS to control pipelineinventory. The export pressure at Gjøa would therefore be limited to the FLAGS’ pipeline code. Availablepressure drop over the Gjøa pipeline was hence limited and a large diameter was necessary.

Various alternatives were investigated: using signals from neighbouring installations, subsea remote-operated valves, subsea pressure sensors, and even a riser platform. These solutions gave high risk, reducedavailability, or high operating and/or capital expense. A new idea of introducing flow measurement in theSIS was proposed. Hydraulic simulations showed that when the parameters of flow, temperature, andpressure – all located at the offshore installation – were used, a downstream blockage could be detectedearly. This enabled the topside export pressure to be increased, and thereby reduced the pipeline diameterrequired. Flow measurement in SISs has not been used previously on the NCS.

This paper describes the principles of designing a pipeline safety system, including flow measurement, witha focus on the hydraulic simulation and designing the safety system. Emphasis will be put on improvementsin transportation efficiency, cost reductions, and operational issues.

such systems considering the reduced availability of steelon the market and the associated cost. The concept isespecially suitable for new, long gas pipelines, but also fornew tie-ins to existing infrastructure.

When connecting to an existing pipeline system, theoperational envelope becomes very complex. The transientconditions in the pipeline system are dependent on aminimum of two sources, control systems, gas compositions,safety systems, communication paths, etc. The 42-indiameter, 700-km long, Åsgard Transport is an example of

Sample

copy


such a system where eight sources export gas into the samemain pipeline.

The Gjøa tie-in to FLAGS was therefore initially not a newdesign challenge. However, when considering the significantprobability of hydrate formation in the downstream FLAGSpipeline, the system became special.

In previous designs, StatoilHydro had analysed full blockageby hydrate, but found it to be so unlikely that it was not abasis for design. Due to poorly-functioning dewpoint controlat one of the existing facilities and the fact that the operatorhad experience with hydrate blockage in large gas pipelines,it was agreed to include the hydrate case as a design basis forthis development.

The Tampen Link pipeline was the first pipeline that tied-in to FLAGS downstream of Brent B, and thereby includedthe hydrate case in its design. The pipeline connects theStatfjord A, B, C, and Gullfaks C platform to the British

pipeline system. The solution was based on a purely staticconsideration: the limit of the safety system was restrictedto 10% above the design pressure of FLAGS, whichrepresented the lowest design pressure in the system. Thisdesign is referred to as the base case for the Gjøa PipelineProtection System (PPS) in this paper.

For Gjøa, the new pipeline was significantly longer and thecost of a larger-diameter pipeline was considerable. TheGjøa source was also more ‘independent’ compared to theStatfjord system, where the sources are close to each other.The possibility of introducing a flow-based safety systemwas therefore promising.

This paper will outline the hydraulic simulations that werecarried out to prove the concept of a flow-based pressure-protection system. Focus will also be put on the operationalaspects, code requirements, and design of instrumentedsafety systems which all depend on the hydraulic behaviourof the gas export network.

Fig.1. The Norwegian subsea gas pipeline system.Sample

copy


NomenclatureDP = design pressureEV = emergency shutdown valveESD = emergency shutdown systemFLAGS = Far North Liquid Associated Gas SystemID = internal diameterIEC = International Electrotechnical CommissionMIP = maximum incidental pressureMOP = maximum operating pressureMSL = mean sea levelNCS = Norwegian Continental ShelfPCS = pressure-control systemPFD = probability of failure on demandPLEM = pipeline end modulePPC = pipeline pressure-control systemPPS = pipeline protection systemPRS = pressure regulation systemPSD = process shutdownPSV = process safety valvePT = pressure transmitterSIL = safety integrity levelSIS = safety instrumented systemTP = test pressure1oo2 = 1 out of 2 voting

Pipeline system descriptionThe new Gjøa development is located in the North Seaapproximately 90km west of the Norwegian shore. Thewater depth is typically 370m, and production is plannedto start up in 2010.

The Gjøa subsea configuration comprises four four-slotsubsea templates. There will also be a subsea tieback fromthe neighbouring Vega field 50km from the Gjøa platform,comprising and additional three four-slot templates. Theoil will be transported to the Norwegian refinery atMongstad, while the rich gas will be transported through anew 130-km long pipeline before entering the existingFLAGS pipeline system. The FLAGS pipeline is a 450-km,long 36-in diameter, pipeline from the Brent B platform tothe St Fergus terminal in Scotland.

In the future, the FLAGS pipeline system can be suppliedwith gas from up to six sources that may pressurize it abovethe maximum allowable according to design codes. TheTampen Link system pipeline that connects the Statfjordplatforms through a hot-tap is the latest development, andstarted gas delivery in 2007. Gjøa will connect to a pre-installed tee. The pipeline network is shown in Fig.3.

The basis for the hydrate case is that when gas from theGjøa or Statfjord platforms meets the gas in FLAGS,hydrate might form and cause a full blockage of the FLAGSpipeline. The fact that the Gjøa gas pipeline should also bedesigned to meet a back pressure equal to FLAGS’ designpressure meant that the export pressure at Gjøa 130kmNW of the tie-in would exceed the FLAGS design pressure.

Figure 4 gives an overview of the relevant pressure limits at,respectively, the Gjøa installation, the Gjøa pipeline, andthe FLAGS pipeline. From the figure, it can be seen that theFLAGS pipeline is the pressure bottleneck in the system;the Tampen Link system has the same design pressure asGjøa. The MIP (the maximum incidental pressure) according

Fig.2. Overview of the Gjøa field development.

Sample

copy


to the code is 10% above design pressure for these systems:consequently, for the Gjøa and the Tampen Link, the codesin Refs 1 and 4 allow for 189bar MIP at an unwantedincident. The FLAGS pipeline is designed to the codes inRef 2 and the MIP is 154.2bar; note that all pressures arereferenced to mean sea level (MSL).

Various design solutions were evaluated. The concept ofpipelines divided into sections of different design pressureis based on information being sent from the location of ablockage to the pressure source location for shut-down.Considerations was given to locating a remotely-operatedvalve systems (typically a high-integrity pressure-protectionsystem – HIPPS) at the interface between the Gjøa pipelineand FLAGS, with an umbilical tie-back to one of the closestplatforms. The solution was rejected for several reasons:tieing-back an umbilical was costly, and the suggestionmeant that the Gjøa production would be dependent onshut-downs at the other platform and the risk of

unintentional valve closure with subsequent productionshut-down at Gjøa. Subsea refurbishment of both valvesand critical pressure sensors located in harsh environmentwere also arguments against this solution.

The initial basis of the Gjøa development was therefore torestrict the export pressure to the allowable pressure inaccordance with the relevant code. Consequently, the Gjøasafety system was restricted to 10% above the FLAGSdesign pressure, which meant that only 11bar (includinghydrostatic head) was available for pressure drop over thecheck valve, the flexible riser, and 130km of pipeline. Therequired pipeline diameter was naturally 20% larger thanthe initial estimate when hydrate was not a basis for design.

Having a system with a compressible fluid, rich gas, arelatively-long pipeline should give some possibilities forlifting the set points of the safety system.

Brent A

SFC

SF

SFB

Brent B

NLGP

Statpipe

x

Tampen Link

Kårstø

SPUR

FLAGS 130 km

St.Fergus

Gjøa

Brent A

SFC

SFA

SFB

Brent B

NLGP

Statpipe

x

FLAGS

36”/450 km

Intrafield

Tampen Link

Kårstø

SPUR

Vega

Gjøa 130 km

St.Fergus

Gjøa

Brent A

SFC

SF

SFB

Brent B

NLGP

Statpipe

x

Tampen Link

Kårstø

SPUR

FLAGS 130 km

St.Fergus

Gjøa

Brent A

SFC

SFA

SFB

Brent B

NLGP

Statpipe

x

FLAGS

36”/450 km

Intrafield

Tampen Link

Kårstø

SPUR

Vega

Gjøa 130 km

St.Fergus

Gjøa

Fig.3. The FLAGS pipeline network.

Fig.4. Pressure limits accordingto the design code.

Sample

copy


Hydraulic analysisThe concept of flow-based PPS is that by using flowmeasurement in combination with export temperatureand pressure, abnormal combinations of these parametersgive time for a shut-down. The advantage of this design isthat the pipeline’s internal diameter can be reduced incomparison to having a fixed-pressure shutdown set point.In consequence, the Gjøa pipeline diameter was reducedfrom 720mm to 680mm.

The basis for the PPS set points are the maximum normalsteady-state export pressures as given in Fig.5. The x-axisrepresents the flow rate and y-axis the pressure topside atthe Gjøa offshore installation. The black line representsthe maximum operating pressure at Gjøa when there ismaximum packing in the pipeline system. The maximumpacking is based on the FLAGS pipeline being operated atthe design pressure of 140.2bar. In normal operation, thepipeline will be operated at a margin below design pressure,in the green-shaded area in the figure.

It should be noted that the pipeline leak-detection systemwill continuously model the gas system covering alloperational scenarios. In theory, these simulations couldalso be used as a safety system and shut down the operationwhen �detecting an abnormal incident. However, as a safetysystem has rigid requirements to the level of safety, a leak-detection system that is based on ultrasonic fiscal meteringcannot be used as a safeguard. As described later in thispaper, a safety system should always be separated from thepressure-control system (PCS).

The safety system developed for Gjøa does not continuouslymodel the operational conditions, but is based on a steady-state simulation representing the maximum operationalpressure. As export pressures above this level indicates anunwanted incident, shut-down will occur. From Fig.5,there is a 4-bar margin, the red-dotted area, before the PPSinitiates valve closure (the red line) with subsequentproduction shut-down. The margin is added to account forinaccuracy in measurements, input data, and simulations.

138

140

142

144

146

148

150

152

154

156

158

1 3 5 7 9 11 13 15 17

Flow (MSm³/d)

)grab( erusserP

Flow based PPS set point

Gjøa max normal Export pressure

Base case PPS set point

Fig.5. The basic concept offlow-based PPS.

138

140

142

144

146

148

150

152

154

156

158

1 3 5 7 9 11 13 15 17

Flow (MSm³/d)

)grab( erusserP

Flow based PPS set point

Gjøa max normal Export pressure

Base case PPS set point

Fig.6. Modified flow-basedPPS for Gjøa.

Sample

copy


This concept applies for all flow scenarios. However as thebase-case shutdown pressure, which is marginally below theFLAGS MIP, lies above the flow-based set points for lowflow rates, it was found sufficient to run flow-based setpoints only when necessary. That is, with reference to Fig.4,for flow set point pressures below the MIP, the base-case setpoint applies. Figure 6 illustrates that the PPS set point isconstant and equal to the base-case set point at low flows.Note that the margin to the maximum normal exportpressure, the red-dotted area, is more than 10 bar at lowflows.

In order to further prove the concept, extensive transienthydraulic evaluations were carried out. The combinationof various parameters would lead to a considerable numberof simulation cases. The first approach was therefore to listall the parameters and evaluate their relevance.

Some of the parameters were tie-in pressure, pipelinesystem packing, check-valve functionality, flow regime,control system of the other platforms, location of hydrateplug, the timeframe for total blockage, gas composition andtemperature, friction loss in pipeline components, etc.

The transient behaviour of the system was the mostchallenging, including variation in production rate andfunctionality of the control system at the variousinstallations, start-up and shut-down scenarios, sequenceof incidents, etc. The analyses that have influence on safetyare discussed in the following paragraphs, and theparameters having an impact on the system’s availability arefurther detailed in the section on safety system robustness.

The case matrix was gradually reduced by investigating theprobability of simultaneous incidents and initial simulationsthat excluded the parameter. For instance, the probabilitythat the PCS fails at the Gjøa platform at the same time as

failure occurred at another installation and a hydrate plugwas blocking the line, was ruled out as a design case.

Initial simulations were carried out with both rapid andslow build-ups of the hydrate plug. This parameter did nothave a significant influence on the results and also, fromexperience, a hydrate plug once initiated is known rapidlyto form and cause full blockage. Simulations with varioushydrate plug locations showed that a location justdownstream of the inlet to FLAGS caused the highestpressure. This might be an obvious finding as the gas iscontained in a minimum pipeline volume.

It was also found that the worst case occurred with maximumline packing and the maximum production rate of33.2Mm³/d, as given by the FLAGS operator. There couldbe three check valves in the system: upstream of the tie-into FLAGS, and two check valves located respectively in theTampen Link and the Gjøa PLEM at the tie-in to FLAGS,as shown in Fig.7.

The check valve in FLAGS was ruled out as the volumecontributed only 0.4 % of the total pipeline volume. Forthe same reason, the hydrate plug formation would causea nearly instant shut-down at Brent B, and the flow fromBrent B was therefore set to zero as a worst case.

Therefore, the steady-state flow regime considered was themaximum FLAGS capacity of 33.2Mm3/d divided betweenthe Gjøa and the Statfjord platforms only. Figure 6 showsthat only Gjøa flows above 14Mm3/d are relevant: the high-and low-flow cases were run with 14 and 17Mm3/drespectively, the latter being the maximum capacity ofGjøa.

The distribution of the remaining volumes – around 16and 19Sm³/d at the Statfjord platforms respectively – were

Tampen Link /SFLL

St F

ergu

sB

rent

B

“high case” : 17 MSm³/d

“low case” : 14 MSm³/d

Hydrate plug

“high case” : 16.2 MSm³/d

“low case” : 19.2 MSm³/d

In FLAGS:33.2 MSm

~ 0 MSm³/d

Gjøa

Gjøa check valve (CV)

Tampen Link

check valve (CV)

FLAGS

check valve (CV)

Tampen Link /SFLL

St F

ergu

sB

rent

B

“high case” : 17 MSm³/d

“low case” : 14 MSm³/d

Hydrate plug

“high case” : 16.2 MSm³/d

“low case” : 19.2 MSm³/d

In FLAGS:33.2 MSm

~ 0 MSm³/d

Gjøa

Gjøa check valve (CV)

Tampen Link

check valve (CV)

FLAGS

check valve (CV)

Fig.7. Check-valvelocation and flow scenarios.

Sample

copy


investigated in various simulations. It was concluded thatthe control systems on the various platforms reduced theflow accordingly. In such a scenario, the flow from Statfjordinto the Tampen Link pipeline only slightly increased thebuilt-up pressure in the timeframe of a Gjøa shutdown.This was later verified by simulations.

The tie-in pressure to FLAGS was set to 140.2bar, aparameter given by the FLAGS operator. The remainingparameters were put into a matrix shown in Table 1.

Note that the export temperature will have direct influenceon the export pressure and will give an offset from the curvein Fig.5. Simulations with varying export temperature werehowever not necessary to prove the concept, the temperatureparameter being included in the PPS logic.

Simulation cases and resultsThe results for high flow cases are given in Table 2, fromwhich it can be seen that it takes typically between 20 and27 minutes to reach shutdown pressure at Gjøa. From thesimulations, the maximum pressure is reached in theFLAGS pipeline about 5-10 minutes later. At the diameterinvestigated in the study, the worst case C gives a margin of2.3bar. The results clearly show that case A, the high-flowcase, is the more conservative in this sequence of events.

Figure 8 illustrates the case where both the subsea checkvalves are not functioning and there is maximum productionat the Gjøa fields.

The production from Gjøa runs at steady state until thehydrate blockage occurs after 45 minutes. The pressure at

the Gjøa PLEM by FLAGS (the pink line) increasesimmediately, and the flow through the Gjøa PLEM spool(the red line) decreases. As there are no check valvesworking between the Tampen Link and the Gjøa pipelines,the flow thereafter oscillates when gas is flowing betweenthe two systems. The Gjøa export pressure (the black curve)increases gradually until the PPS set point is reached, 4barabove the steady-state export pressure. After the Gjøashutdown, it takes approximately 8 minutes before thepressure at the Gjøa PLEM (equal to the FLAGS hot-tappressure) reaches its maximum. The system thereafter startssettling out and, due to the high export temperature, thepressure falls.

The absolute maximum pressure case is when the Gjøacheck valve is not functioning and the Tampen Link checkvalve prevents any back flow from Gjøa, Case C, showin inFig.9. The flow through the Gjøa PLEM representing theintersection between the Tampen Link and the Gjøapipeline is now only negative: that is, gas from the TampenLink flows into Gjøa through the non-functional checkvalve but there is no ‘positive’ gas flow from Gjøa into theTampen Link system. The peak pressure at the Gjøa PLEM(equal to the FLAGS hot-tap pressure) is the highest pressurereached for the eight cases, and is 2.3bar below the MIP forthe selected diameter.

All the previous simulations have been based on twounplanned incidents: hydrate blocking the pipeline andsubsequent failure of the Gjøa control system. Varioussimulations were carried out with multiple subsequentincidents.

In one case, the Gjøa compressor tripped due to productionupsets only 10 minutes after the hydrate blocked the

esaC noitpircseD

)wolfwol/hgih(AesaC liafsevlavkcehchtoB

)wolfwol/hgih(BesaC sliafevlavkcehckniLnepmaT

)wolfwol/hgih(CesaC sliafevlavkcehcaojG

)wolfwol/hgih(DesaC yltcecrrocnoitcnufsevlavkcehchtoB

esaCPIMSGALFwolebnigraM

)rab(aojGtanwodtuhsotemiT

)snim(

)hgih(A 2.3 22

)hgih(B 4.3 62

)hgih(C 3.2 22

)hgih(D 4.2 52

)wol(A 9.4 02

)wol(B 3.5 72

)wol(C 7.4 02

)wol(D 3.5 62

Table 1. Simulation cases.

Table 2. Simulation results.

Sample

copy


FLAGS pipeline. Gjøa thereafter restarts production againstthe blocked outlet. The other sources have also been fillingthe pipeline volume in the meantime. The case assumes nocontact with the Brent B platform where a shut-down hasalready occurred.

None of the cases that were run caused a pressure inFLAGS that exceeded the Case C high. There are anendless number of combinations that could have beenevaluated, but the probability of these situations arisingmust also be considered. The probability of even the casewhere Gjøa restarts against a blocked outlet was so low thatfurther transient safety simulations were not carried outwhen analysing the safety aspect. Case C was therebychosen for the design of the Gjøa pipeline.

Safety system robustnessAlthough the results from the hydraulic analysis wereconvincing, it was a concern that such a system would causeunnecessary shut-downs and would not be suited forimplementation at an offshore installation. When

investigating various operational upsets, the logic was furtherdeveloped. The rapid changes in flow in combination withthe slow pressure variations in a large pipeline systemwould not be handled by the pure steady-state approach.

The system was modelled using HYSYS simulations withthe steady-state-based regression logic in the control/safetysystem. The modelling showed unwanted shutdowns whenthe production rate was reduced. At reduced flow, thepressure would remain high in the pipeline system andreduce slowly, whereas the PPS set point would be quicklyreduced and an unnecessary shutdown would occur. Atincreased flows, the PPS set point would be higher than theactual operational pressure and would consequently notrepresent a trip scenario.

The problem was solved by including a simple functionusing the pressure gradient: firstly, in the scenario ofreduced flow, the pressure is reducing, i.e. there is anegative gradient. As there is no safety concern in a systemwhere the pressure is reducing, the PPS was set above theactual operating pressure with a 4-bar margin. In the caseof a pressure increase, the flow-based regression line would

140

142

144

146

148

150

152

154

156

158

160

0 60 120 180 240

Time (hrs)

Pres

sure

(bar

a)

-90

-60

-30

0

30

60

90

120

150

180

210

Mas

s flo

w (k

g/s)

Gjøa PLEM flow

Gjøa export rate

Gjøa PLEM pressure

Gjøa export pressure

140

142

144

146

148

150

152

154

156

158

160

0 60 120 180

Time (hrs)

Pres

sure

(bar

a)

-90

-60

-30

0

30

60

90

120

150

180

210

Mas

s flo

w (k

g/s)

Gjøa PLEM flow

Gjøa export rate

Gjøa PLEM pressure

Gjøa export pressure

Fig.8. Failure of both check valves,high flow from Gjøa (case A).

Fig.9. Gjøa check-valve failure: highflow Gjøa (case C).

Sample

copy


be active. The logic can be described by Equn 1:

PPSset

= F (flow, temperature, pressure) with dp > 0PPS

set = Pressure + 4 with dp < 0

(1)

where flow, temperature and pressure are measured topside.This is illustrated in Fig.10, in which the solid brown linegives the Gjøa operating pressure when varying the flowfrom 17Mm3/d to 10Mm3/d and back to 17Mm3/d after20 minutes. The operating pressure drops gradually whenreducing the flow, and thereafter increases with the flow.The dotted green line represents the original flow-basedPPS set point: the PPS set point ÿ:line crosses the Gjøaoperating pressure line, and consequently an unwantedshutdown occurs.

The dashed blue line represents the flow- and gradient-

based PPS setpoint. For reducing pressure, the set point is4 bar above the actual pipeline pressure, and hence anunwanted shutdown is avoided. For increased flow, the setpoint follows the steady-state-based regression as normal.

Sensitivities in parameters such as export temperature,pressure drop over the pipeline end structures, risers andpipeline friction loss, and gas composition were investigated.For this pipeline system where almost 20% of the totalpressure drop comes from the pipeline end modules (PLEM),these parameters have been thoroughly investigated in thedetailed design. The check valve in the downstream structureis the governing influence.

During the detailed design of the flow-based PPS, all theseparameters are taken into account. The chosen diameterwith corresponding margin to the FLAGS MIP allows forinaccuracies in the regression line caused by modelling of

Fig.10. Illustration of logic thatimproves system availability.

Fig.11. Simplified sketch of flow-based PPS.

Sample

copy


the system, and inaccuracies in flow, temperature andsensors, response time for valves, accuracies in data transfer,etc.

The simulations presented in this paper, carried out for anID of 660mm, proved that the worst case (Case C) gave amaximum pressure safely below the target. An ID of 680mmID was chosen to allow for inaccuracies and the fact thatthis concept has never before been used in a pipelinesystem.

Design of safetyinstrumented systems

Generally, a process-control system (PCS) is used to maintainnormal operation of a transport system. In addition, therecould be a requirement for a safety system totally segregatedfrom the PCS. The safety system will shut down thetransport system in case of erroneous operation by thePCS.

Traditionally, process safety systems comprise onemechanical safety barrier (such as a process safety valve,PSV, with flare system) and one electronic safety barrier(such as a process shutdown system, PSD). Such a design isnot feasible for large gas pipelines with sections withdifferent design pressures. The flare system would beenormous and the environmental aspects of flaring suchlarge gas volumes are unacceptable as well as generatinghigh capex and opex.

For this reason, such large gas pipelines on the NCS haveelectronic safety systems. The existing PPSs are based onprogrammable safety systems similar to those used for PSDand emergency shutdown (ESD) systems on offshoreinstallations worldwide. This is also used for Gjøa.

The acceptance criteria for the Gjøa PPS are given asfollows:

• Installations where human life may be exposed toan accidental leak or rupture:

• Annual probability of over-pressuring above the testpressure shall be less than 1 x 10-5.

• Annual probability of over-pressuring above theallowable pressure as stated in the relevant pressurevessel code shall be less than 1 x 10-3.

For Gjøa, the test pressure could be reached without anypressure safety system present, and hence 1 x 10-5 applies.

In order to define the required safety level of the PPS, it isnecessary to define how often there is a situation where thePPS is needed, defined as the demand rate of the PPS. Thedemand rate for Gjøa involves blockage of the pipeline,

which could occur both as a consequence of hydrate plugor blockage at the outlet plant onshore. Regarding thelatter case, the FLAGS operator has found that this is notthe governing case for design. The frequency of valveclosure is high; however, due to the large pipeline volume,the time before rupture pressure has been reached is verylong. Although the onshore plant has a high frequency ofblockage, the duration of each incident is very short incomparison. Consequently the governing case for thisdesign is the hydrate plug which, conservative estimatesfrom the operator have concluded, might once every 100years or 1 x 10-2.

All pressure sources which can deliver pressures above theFLAGS MIP must therefore in sum have a probability offailure on demand (PFD) of 1 x 10-3. The Gjøa contributionis given as 1 x 10-4. A PFD of 1 x 10-4 is only valid for thosesafety systems that prevent pressure to below the acceptableaccording to the FLAGS system design code (MIP). TheGjøa PSVs and PSD have set points that are too high andcan therefore not be considered as a safety layer for theFLAGS system. The Gjøa PSVs and PSD are designed toprotect the Gjøa topsides equipment and not thedownstream pipeline.

Simulations show that it will take a minimum of 22minutes before hydrates can block the FLAGS pipeline andfor the pressure to exceed the FLAGS MIP. For this reasonthere is time for both human intervention and/or thepipeline pressure control (PPC) system to take preventiveaction. The PPC system could be part of an overpressureprotection function if this system is independent of PPS. Inaddition:

• the control system must not be a part of the demandon the PPS;

• the maximum risk reduction (PFD) shall be lessthan 10-1 if the PPC is not built according to a safetysystem.

Human intervention may, in some cases, prevent a build-up of pressure above the design pressure. Provided certaincriteria are met, this barrier to overpressure can reduce thetechnical requirements for the PPS systems. Theserequirements are:

• there must be continuous control room operatormonitoring of the pipeline;

• the operator must have the capacity and capabilityof responding correctly to a pressure build-up in thepipeline;

• clear procedures must be established and followed;• there must be redundancy (back-up) in all human

and technical elements of the human interventionchain, i.e. in the control room, at the pressuresource, etc.

The amount by which the acceptance criteria as given maybe altered depends on the total time available for detection

Sample

copy


and response by the human intervention chain, and isgiven by:

( ( ))

R R for t

R R t for t

R R for t

= ≤ ≤

= + − ≤ ≤

= ≥

0

0

0

0 1

1090

111 1 12

100 12

(2)

where:

R0 is the original annual acceptance criterion (10-3 and

10-5 for MIP and TP, respectively)R is the new annual acceptance criterion for the PPS

systems when human intervention is includedthe variable t is the available response time (in hours)

until the MIP or TP is reached, following a pressurebuild-up event.

At Gjøa the topside process control system (PCS) will betotally segregated from the PPS, but will contain the samesoftware functions for protecting the FLAGS pipeline.Pressurisation from design pressure to MIP takessignificantly less than 1 hour; for this reason, the Gjøa PPSPFD requirement is set to10-3, while the control systemcontributes 10-1, making a total of 10-4, in accordance withthe total PFD requirement. Human intervention is nottaken into account.

For StatoilHydro, if the required PFD is equal to or higherthan 10-3, the total overpressure protection system shallconsist of dual independent safety layers. The Gjøa PPS hasa PFD requirement of 10-3, hence it is a dual independentsafety system. For this reason the Gjøa PPS system consistsof two independent loops called PPS-1 and PPS-2. Bothloops are duplicated where necessary in accordance withreliability and availability requirements. All pressure andtemperature sensors have a triple-redundant configurationboth for excellent availability, and so that one of the threesensors can be taken out of service and tested withoutdegrading the safety level of the total safety system. Theprocessing units are dual in each loop. Both loops will beSIL2 certified in accordance with IEC 61508 and IEC61511 [5, 6].

Figure 11 gives an overview of the patented PPS solution.Each PPS closes two safety-shutoff valves, and one valve iscommon for both PPS. Multiple valves are necessary inorder to reach the required safety level. These safety valvesare primarily used for other safety functions than PPS.With an added PPS solenoid the valves are also part of thePPS safety function, and the PPS remains independent ofother safety systems. The flow meter function will lift the setpoint at the most by 2-3bars above the normal PPS,corresponding to 20% of the total pressure drop.

With a failure of this part of the safety system, the system

will reduce production and return to the base-case setpoints but, since the control system is based on the samelogic, there will be no shutdown. The PPS will then discardthe flowmeter function and continue with a fixed, lower-pressure, set point.

The flowmeter element is a venturi type. In reliabilitycalculations, this is regarded as a pipe with normal processconnections. Due to the minor effect of an undetected ordetected failure in the flowmeter part of the PPS, there isno need for dual flowmeters. However, gas export availabilitygives a requirement for redundancy of pressure andtemperature sensors in the flowmeter.

An ISS will reduce the availability of the system. For thisreason the ‘unwanted shutdown’ failure mode of the PPSwas carefully investigated during design. Requirementsconsidered included:

• the possibility to test, repair and maintain the PPSwithout shutdown of gas export;

• blocking, override, and by-pass of the PPS shall notbe possible;

• any blocking/override/by-pass required for testingpurposes shall be registered in the alarm log;

• system degradation shutdown timers shall be used;• hydraulic behaviour will be employed to delay

shutdown.

In order to ensure high availability of the safety system, itis important to evaluate the safety system contribution’s toexport availability. Based on the reliability and availabilityrequirements, various software programming can be chosen.In the Gjøa PPS, the following have been implemented inaddition to the necessary vendor protocols:

• System degradation timers which allows the PPS tocontinue to operate, typically, for 72 hours if afailure is detected and the reliability is degraded,but the system is still functioning.

• Voting logic that switches to 1oo2 voting in case anyone of the three sensors is out of order. Voting logicwould even allow operation with one sensor out ofthree when monitored, although reliability isreduced.

• Flow algorithms which calculate flow based onpressure differential on the venture, and temperaturemeasurements.

There is also a fiscal flowmeter at the Gjøa topside measuringthe gas export flow rate, although this meter is not designedaccording to the safety codes Ref. 5 and 6; for this reason,the PPS cannot use the fiscal meter. However, the accuracyof the fiscal flowmeter will be an excellent verification forthe PPS flowmeter and can be compared on operatorscreens.

Sample

copy


SummaryHydraulic analysis proved that by using flow measurementsin combination with export temperatures and pressures,abnormal operation can be detected in due time for ashutdown.

For gas transport systems with sections of different designpressures, an excellent understanding of the pipelinesystem’s hydraulic transient behaviour in combinationwith using programmable safety instrumented systems gives:

• The necessary safety level against excessive internalpressures.

• Optimal reliability and availability similar to apipeline with a uniform design pressure.

• Minimum pipeline diameter.

• An excellent operational envelope with line-packingpossibilities.

When using programmable safety instrumented systems asprimary and secondary pressure protection, care must betaken to ensure simple, conservative, and testableprogramming for the PPS application.

The internal diameter of the new 130-km long gas exportpipeline from Gjøa was reduced from 720mm to 680mmas a result of the use of flow as an integral part of the PPS.This has given significant savings in the field developmentcosts, and has contributed to good HMS practice since thepipeline has less inventory in case of a gas leakage.

References1. DNV, 2000. Offshore standard OS-F101: Submarine pipeline

systems. January.2. ASME, Gas transmission and distribution piping systems:

ASME B31.8, and BSI, Code of practice for pipelines,BS8019 (PD8010-2).

3. M.Aamodt and G.Staurland, 2004. Designing offshorepipeline systems divided into sections of different designpressures. IPC04-0217.

4. ISO, Petroleum and natural gas – pipeline transportationsystems. ISO 13623.

5. IEC 61511, Functional safety – safety instrumented systemsfor the process industry sector.

6. IEC 61508, Functional safety of electric/electronic/programmable safety-related systems.

Sample

copy


THERE IS A CLEAR trend towards large-diameterpipelines and higher design pressures resulting in

larger wall thicknesses. Also crossings of sensitive publicworks or near sensitive buildings may require an increasedwall thickness. With a given pipeline diameter, an increasedwall thickness leads to a higher bending stiffness that canbe outside the range of today’s experience. From experiencein Netherlands with large-diameter pipelines, it becameclear that a better insight is needed.

In Ref.1 it was shown that there are high soil reactions atthe head of the pipe string in the curved section of theborehole, which may lead to damage of the coating or toborehole penetration and stuck pipe in the case of large-diameter pipelines. In Ref.2, design guidelines for thedesign bending radius for HDD were developed. It wasshown that pipe stiffness in relation to soil properties is animportant factor in this, and the requirements in the DCAguidelines have been modified because of this (Ref.3).

A further study was done to investigate the influence ofuplift and downlift during the pull back operation.

ModelIn Ref.1 the ABAQUS finite-element code was used andthe pipeline string and drill pipe were bent from initiallystraight into the shape of the borehole. The pull-back issimulated then by a series of prescribed displacements.Interaction between the soil and the pipe is by using springelements, and the gap between the pipeline wall and theborehole wall is taken into account. It was found that thepeak soil reaction is when the head of the pipeline from thestraight section enters the curved section. With furtherdisplacement in the curved section the peak soil reactiondecreases.

To study the influence of uplift and downlift, the ANSYSStructural 11.0 finite-element code was used for modellingthe pipe string and the borehole. In this model, the pipestring is above ground and lies at a vertical angle equal tothe entrance angle of the HDD. The pipe is pushed into theborehole (a check if there is a difference between push or

*Author’s contact details:tel: +31 5 0521 2190email: [email protected]

Horizontal directional drilling:the influence of uplift anddownlift during the pull-backoperation

by F Podbevsek1, H J Brink2, and J Spiekhout*2

1 Infinite Simulation Systems BV, Groningen, Netherlands2 NV Nederlandse Gasunie, Gronoingen, Netherlands

IN EARLIER WORK it was shown that, especially for large-diameter pipelines, the stiffness of the pipe inrelation to the soil properties is an important factor in avoiding problems related to the bending radius.

Also it was shown that there are high soil reactions at the head of the pipe string in the curved section ofthe borehole. In this paper, the influence of creating uplift or downlift when ballasting the pipe string duringthe pull back operation is studied.

Sample

copy


pull did not show significant differences in relation to soilreaction). From the calculations it was clear that in thecurved borehole section soil peak reaction is high at thehead of the pipeline as in the previous work [1].

In the “theoretical” example as modelled in ANSYS, theparameters were as follows:

pipe diameter: 48in (1219mm)wall thickness: 23mmborehole diameter: 1422mmdepth: 25mradius: 1000D = 1219m

mid-section length: 200mentry/exit angle: 11.4o

Model details:

• pipe meshed with 8-node shell elements (SHELL281)• contact between pipe and borehole with contact

elements (CONTA174) using the nodes of the shellelements

• rigid borehole meshed with higher-order elements(TARGE170)

The material behaviour is assumed to be linear elastic. The

Fig.1. The HDD pull-back operation HDD

Fig.2. Plot showing the starting situation of the analysis. The pipe is at an angle outside the borehole and is supported by arigid target surface.

Sample

copy


Fig.3. Element plot showing the borehole and the rigid target surface which lies at an angle above the ground level, andsupports the pipe when it enters the borehole.

Fig.4. Tail end of the pipe where an MPC contact is created to apply the prescribed displacement to the pipe.

Sample

copy


non-linearities come from the large displacement effectsthat are included in the analyses and the non-linear(frictionless) contact that is created between the pipelineouter surface and the borehole.

AnalysisDifferent analyses were performed to simulate the effect ofuplift and downlift on the pipe string in the borehole.Contact stiffness was varied to check for differences, and inthe model itself the contact algorithm (Augmented Lagrange,Normal Lagrange) and analysis options for dealing withinstabilities (numerical stabilization, slow dynamics) werealso varied to study the effect on the results. For instance,with a larger stiffness value convergence problems are morelikely to occur. The calculations showed a rather slow rateof convergence and were time consuming. This can beexplained from the dimensional properties of the steelpipeline string that behaves more like a flexible sausage dueto its relatively large length.

The following basic cases were analysed:

Case Uplift/ Contact Stiffnessdownlift

1 Downlift 20% 0.012 Downlift 100% 0.013 Downlift 100% 0.14 Uplift 20% 0.0015 Uplift 20% 0.01

ResultsDuring the pull-back operation, high pressures occur atlocations where the pipe ends are in contact with thecurved borehole. One has to keep in mind that these

locations are also known as stress singularities in the finite-element model. Stress singularities are local effects and themaximum value of the stress peak is thus not accurate.

Comparison between cases 1 and 2

• Increasing downlift results in a larger contact areabetween the pipe bottom and the borehole.

• Contact pressure increases with increasing downlift.


• No significant difference in contact size.• Increasing the contact stiffness with one order of

magnitude almost doubles the contact pressure.


• Good agreement in contact size. • Increasing the contact stiffness with one order of

magnitude almost doubles the contact pressure.


• A comparable distribution of contact properties(gap, penetration, and pressure) along the pipe. Inthe 20% downlift situation, higher contact pressureat the edges of the middle contacting zone. The 20%uplift situation shows higher contact pressure at theedges of the outer contacting zones. This seems tobe contradictory, but can be explained by the gapthat is present between top of the pipe and the topof the borehole that has to be bridged in the case ofdownlift.

ConclusionsDuring the pull-back operation high pressures occur at

Figs 5 and 6. Examples of calculation results: uplift 20%, stiffness 0.001, and downlift 100%, stiffness 0.1.

Sample

copy


locations where the pipe ends are in contact with thecurved borehole. There is no significant difference incontact pressures in the curved section at the pipe end forthe case of 20% uplift or 20% downlift�; however, it lookslike the 20% download situation is somewhat better thanthe 20% uplift situation in view of the soil reactions at thehead of the pipe.

References1. J.P.Pruiksma, H.J.Brink, H.M.G.Kruse, and J. Spiekhout,

2009. Analytical solution for the soil reaction force at the

head of the pipeline during the pull back operation ofhorizontal directional drilling. Pipeline TechnologyConference, Ostend, Belgium.

2. H.J.Brink, H.M.G.Kruse, H.Lübbers, H.J.A.M.Hergarden,and J.Spiekhout, 2008. Design guidelines for the bendingradius for large diameter HDD. The Journal of PipelineEngineering, 6, 4.

3. DCA guidelines, rev. 3, May 2007.

Sample

copy


Sample

copy

journal of pipeline engineering - pipemag.com - dec09 for... · leigh fletcher, miab technology pty...

Documents