22 March 2000, T/prediction of sand erosion in process.doc

PREDICTION OF SAND EROSION IN PROCESSAND PIPE COMPONENTS

Asmund Huser and Oddmund Kvernvold

Det Norske Veritas, Veritsaveien 1, N-1322 Hvik, NORWAY

Published in:

Proc. 1st North American Conference on Multiphase Technology. Banff, Canada 1998.

Ed. J.P. Brill, G.A. Gregory. BHR group Conf. Series. Publ. No. 31 pp. 217-227

21 ABSTRACTPrediction of sand erosion in pipe systems is outlined presenting two procedures, one forgeneral complex geometries, and one for standard pipe components.

Firstly, a general procedure to calculate sand erosion in three-dimensional flow geometriesis outlined. This procedure is designed for the detailed assessment of sand erosion rates incomplex process components such as production choke valves, manifolds, diverter systems,intruding sand probes, multiphase sensors, etc. The procedure is based on the ComputationalFluid Dynamics (CFD) programme CFX4. DNV has, however, developed specific models tocalculate erosion based on general flow and particle calculations performed within CFX4.Validation examples show good comparison with experiments and an application exampledemonstrate that the procedure is applicable to visualise the erosion mechanisms, tooptimise inspection and maintenance routines as well as provide a valuable design tool.

Based on experimental findings and the above procedure for complex geometries, DNV hasalso developed a DNV Recommended Practice (RP O501) which can be applied to calculatesand erosion in typical pipe components: Pipe bends, blinded Tee bends, straight pipes,welds and reducers. The new RP provides valuable tools for dimensioning of piping systemsand components, optimisation of production, and inspection and maintenance planning. Theoil companies Statoil, Norsk Hydro, Saga Petroleum, Conoco, and partly Amoco havesupported the development of the new RP, and the RP is now applied in the Norwagiansector in the North Sea,

2 INTRODUCTIONSand erosion has become an important concern both for design of new field developmentsand prolongation of existing oil and gas fields. High cost of retrieving process componentsmakes it attractive to design against sand erosion, especially for sub-sea installations. Formany existing fields, erosion potential increases toward the end of the well life due todecreasing pressure, increasing GOR, and increasing sand production. For existing fields,

3inspection and maintenance intervals and methods can be obtained by detailed assessment ofthe erosion rates of the different components in system. In this assessment, the wholeproduction line has to be considered. The highest velocity in the production line is usuallythe pressure reduction choke valve, hence the design and material selection of this is crucial/3/. The second most erosion prone component may be a bend, a manifold or a flowmeter. Ifthe choke valve is manufactured in an erosion resistant material, the weakest point may alsobe the second point.

Design to prevent sand erosion is often done by ad-hoc methods that are independent of thesand production rate. One such method is the API RP 14E /2/, where an erosional velocityis calculated. This method may give conservative designs when little sand is produced,whereas non-conservative designs when sand is present. By the present approach, which isbased on extensive experiments and testing /3/4/5/6/, design tools are funded on anunderstanding of the important factors that influence erosion.

3 PREDICTION OF EROSION IN COMPLEX FLOW GEOMETRIESThe method applied to predict erosion rates in complex flow components, such as chokevalves, multi-phase flowmeters, manifolds, etc. is here outlined. A general methoddeveloped to perform erosion calculations in arbitrary geometries is developed based onflow and particle track calculations performed with a standard Computational FluidDynamics (CFD) package. In the present work the CFD programme CFX4 is applied. Theprediction method for sand erosion applies the results from flow solution and a particle trackcalculation to calculate the erosion rate on the internal surfaces of the geometry. Specificmodels have been developed to calculate erosion rates based on the results from CFDsimulations.

3.1 Method of analysisThe sand erosion analysis is performed in four steps; grid generation, flow solution, particletrack calculation and erosion rate calculation. The first three steps are performed with CFX4,whereas the models developed by DNV perform the fourth step. The approach applied in thefirst three steps and the DNV erosion model is outlined below.

The CFD programme CFX4 contains most of the features that are of importance for erosionproblems. The grid system is suited to model complex streamlined, or irregular flowdomains. With a multi-block grid system, optimised grid may also be created, makingcomputational times friendly. However, care must be taken when creating the grid. Thestandard k-e method is applied for the turbulence modelling, and a converged turbulencefield must be achieved in order to predict the correct particle movement. Only a grid that hassufficient resolution, orthogonality near the walls, and a reasonable grid expansion ensures aconverged turbulence field. If non-orthogonal grids are applied near the walls, a solutionmay be obtained, but this may give unreasonable values for k and e which may causeparticles to hit the wall with unreasonable speed or particles to be attracted to the walls. Thismay cause too high erosion rates at these locations.

During the flow calculations a steady state one-phase flowfield is produced. The flow maybe either incompressible, or compressible. By assuming mixture quantities for the flowparameters (such as velocity, density, viscosity, etc.), multi-phase flows are approximated.

When the flow- and turbulence- fields are converged, the particle tracks are solved on thesteady state flowfield. Up to 10 000 particles are released at arbitrary locations at the domain

4inlet, where the particle tracking routine within CFX4 is applied. The invoked forces actingon the particles are drag force, gravity force and pressure gradient force. Further, acorrection term to the particle velocity is included dependent of the turbulence level.Particles are assumed to be spherical, and the particle diameter is specified together with thesand density to give the particle size and weight. Constant particle diameter or a Gaussiandistribution of the size is implemented. The restitution coefficient is one for an idealreflection, and denotes the reduction factor of the wall-parallel velocity after the hit:

inout Euu |||| = (1)

Here u||out is the wall-parallel velocity after the hit, u||in is the wall parallel velocity before thehit, and E is the restitution coefficient. The particle velocity component normal to the wallkeeps its value and only changes sign after the hit. Typically a restitution coefficient of E =0.8 is applied. Equation (1) is the default formulation applied din CFX4. For applications toerosion it is assessed that the angle of attack, particle shape and size, material type as well asthe particle velocity will influence the particle exit angle and velocity. However, this is anarea of further research.

Erosion calculations are performed with a general method based on the situation shownbelow, where u is the hit velocity and a is the hit angle /1/:

a

u

WallParticle

The general equation for the erosion rate is written as follow /7/8/9/:

A

FmKuCE

w

pn

unitL r

a )(&=

(2)

Here EL is the erosion rate in mm/year, Cunit = 3.151010 is a converting factor from m/s tomm/year, K is a material constant, pm& (kg/s) is the massflow of sand that hit the area, A (m)is the size of the area exposed to erosion, rw (kg/m) is the wall material density, n is thevelocity exponent which is dependent of the material, and F(a) is a number between 0 and 1given by a functional relationship dependant of the material.

The erosion rate in some given small sub-area is found by the summation over all particlesthat hit within the defined area:

=

=

nhit

ii

ni

w

FuMAK

R1

6

)(10

ar

,(3)

where R (mm/kg total sand feed) is the erosion rate in the sub area, M is the total number ofparticles (total sand feed), and nhit is the number of hits in the sub-area. The materialdependant parameters have been determined from extensive laboratory tests of a number ofdifferent materials given in /3/. Parameters for two frequently used materials are listed inTable 1. The functional relationship, F(a), for two materials is given in Figure 1. It should

5be noted that the erosion resistance of different materials is varying, and care must be takenwhen selecting a good erosion resistant material. Parameters applied for tungsten carbid(WC) in Table 1 yield a good WC material with 6% cobalt (Type Sandvik DC-05).

Table 1 Recommended values for material constants for two typical materials to beapplied in Equation (3). From ref. /3/.

Material K n rw

Steel grade 2.0 10-9 2.6 7 800 kg/m

Tungsten carbide (WC) 1.1 10-10 2.3 15 250 kg/m

00.10.20.30.40.50.60.70.80.9

1

0 15 30 45 60 75 90

Impact angle - a

F( a

)

Brittle (WC)Ductile (Steel)

Figure 1 Function F(a ) for typical ductile and brittle materials /1/.

3.2 Validation casesTwo validation cases are presented indicating that the erosion rate may be predicted for bothsingle-phase and multi-phase fluid flow.

3.2.1 Bean chokeIn a bean choke, flow is going through acontraction as sketched to the right. Theexperimental investigations /4/ were reproducedfor carbon-dioxide gas at subsonic speed. Thefollowing flow and particle parameters areapplied:

Inlet velocity: 11.4 m/s Inlet pressure: 14.1 bar

4 5

I n l e t

O u t l e t

g

6 Inlet temperature: 36 C Fluid viscosity: 1.5 10-5 kg/(ms) Molecular weight: 44 Inlet ID: 54 mm Outlet ID: 20 mm Particle diameter: 0.25 mmMost sand particles hit the 45 contraction and bounces off and hit the second time insidethe smaller outlet pipe on the opposite side as the first hit. The maximum erosion rate on theoutlet pipe is obtained a small distance from the contraction. In comparison with theexperiment, a good agreement on the level of erosion is obtained (Figure 2). The restitutioncoefficient is E = 0.8 is applied this case. The restitution coefficient does influence theresults and may give an explanation of why the location of the maximum point is slightlyoff.

0

0.5

1

1.5

2

2.5

3

0 0.05 0.1 0.15 0.2 0.25 0.3

Distance (m)

Ero

sion

rat

e ( m

m/k

g sa

nd)

CFD calculation

Experiment

Figure 2 Erosion rate along bean. Comparison with experimental results /4/.

3.2.2 Pipe bendA 5D-pipe bend with a vertical inlet and horizontal outlet, as shown below in Figure 3, ismodelled and compared against experiments. The multi-phase parameters are converted tomixture parameters, which are applied in the model. Assuming incompressible flow,following parameters apply:

Inlet mixture velocity: 36.3 m/s Gas-to-liquid ratio (GLR): 14.5 m/m Mixture viscosity: 1.8 10-5 kg/(ms) Mixture density: 72.3 kg/m Pipe inner diameter, D: 26.5 mm Radius of curvature: 5D

7 Particle size: 0.25 mm

g

R=5D

Inlet

Outlet

Figure 3 Sketch of flow model (left) and particle tracks from simulations (right).

Excellent agreement is obtained when comparing the CFD results with the experiment(Figure 4), despite the approximation to single phase flow. The DNV RP for calculation oferosion in pipe bends /1/ gives a maximum erosion of 7.3 mm/kg sand in this case. Thisconfirms that the RP is on the conservative side (see paragraph 4). The DNV developedprogramme ERBEND is also seen to predict the erosion rate well. This programme employsa 2D model of particle paths to calculate erosion in pipe bends.

The effects of multi-phase flow on particle distribution and erosion patterns are not wellunderstood. The present results indicate that the erosion pattern may be distorted by themultiphase flow, explaining some of the differences between the experiment and the CFDresults in Figure 4. However, this should be an area of further research.

0

0.5

1

1.5

2

2.5

3

3.5

4

4.5

0 15 30 45 60 75 90

Degrees along bend

Ero

sion

rat

e ( m

m/k

g sa

nd)

CFD

Experiment

ERBEND calculation

Figure 4 Erosion rate along outer side of bend. Comparison with experiment /5/, and2D model DNV*ERBEND.

83.3 Application to process componentsThe production choke valve gives an example of application of the erosion routines tocomplex geometries. The Multiple Orifice Valve (MOV) in Figure 5 is modelled as shownin Figure 6 containing only the flow domain with a high-pressure part (dark) and a low-pressure part (light). The complex flow domain is shown by the solution in Figure 7 with theinlet pipe, the valve house with the internal turning fork, the two bean holes where the flowis accelerated and the low-pressure outlet spool. Turning the two orifice holes, so that theyare non-aligned, regulates the flow. The modelled choke presented here is partly open withthe bean holes non-aligned.

Particle tracks are shown in Figure 8 in order to get an understanding of the erosionmechanisms. At small opening angles, the particles are directed toward the sidewalls of theoutlet spool with a high velocity. This causes the erosion rate to reach a maximum at twolocations caused by the two bean holes as illustrated by the two maximum points which areshifted 180 in Figure 9.

A series of operating ranges have been modelled by changing the opening angle, pressuredrop, particle size, and outlet geometry. In Figure 10 are shown results from four caseswhere the opening angle (or opening percent) is varied. For the largest opening, a reducederosion rate is obtained. For all cases the erosion rate for both steel walls and WC walls isdetermined. From each case, the contours of constant erosion on the internal surfaces arefound and the point of maximum erosion rate determines the lifetime of the choke. Aconsolidation of all simulation cases is applied to develop correlation equations that areapplied to predict the maximum erosion rates for the relevant operating conditions.Operators can apply this to determine service and inspection intervals as well as optimaloperations on fields with a high sand production. The erosion rate is proportional to the sandproduction rate so that monitoring of the sand production is required to be able to follow theerosion development. On-line sand monitoring is recommended when the sand production ishigh in order to be able to spot sudden changes in the sand production rate.

Inlet

Straightoutlet spool

Turning fork seen from the side

Bean holeorifice

Figure 5 Typical Multiple Orifice Valve (MOV) with two bean holes shown in fullyopen position.

9Figure 6 Geometry model showing the flow domain of the Multiple Orifice Valve(MOV) choke applied in the CFD model. This valve is partly open with the bean holes

non-aligned.

Figure 7 Pressure contours, Pa (left) and velocity vectors, m/s (right) in symmetryplane of MOV choke at partly open valve.

Figure 8 Particle tracks in 3D model of MOV choke seen from two angles. Tracks areshaded by the particle speed showing the slow moving particles in the inlet and the

house (dark), and the accelerated light particles in the contraction.

10

Figure 9 Iso-contours of constant erosion rate on the folded out outlet spool wall. Thehorizontal axis is the angle around the spool, and the vertical axis is the flow direction.

To the left is sketched the profile of the outlet spool.

0.0001

0.001

0.01

0.1

1

10

100 150 200 250 300 350 400

Pipe axis outlet spool wall

Rel

. ero

sion

rat

e

20% open30% open45% open60% open

Figure 10 Maximum erosion rate on outlet spool wall for four openings of valve. Theprofile of the outlet spool is sketched below the pipe axis.

4 PREDICTION OF EROSION IN TYPICAL PIPE COMPONENTSProcedures for calculation of erosion rates in the pipe components: pipe bends, blinded Teebends, straight pipes, weld joints and reducers have been developed based on extensiveexperiments and CFD modelling cases. Results are consolidated in correlation equations foreach component type and can be applied to predict the maximum erosion rates for relevantoperating conditions. Correlations are of the form given in Eq. (3) where componentdependant coefficients and calculation procedures are incorporated to account for thevarious geometries. Due to spread in the underlying experimental and numerical results, thecorrelations are made on the conservative side. The complete set of correlation proceduresare outlined in a DNV Recommended Practice /1/ which is commercially available from

11

DNV. Also a hand-on PC computer programme is available. This can be applied for e.g.dimensioning of piping systems and to optimise production in an existing system.

Below is given an example demonstrating the use of the RP procedure for estimation of thepipe dimensions, D, for the following production conditions: Limiting component in pipe: 90 pipe bend Radius of curvature of bend: 2D Production rate: 1.8 106 Sm

3/day gas = 20 kg/s

Well head conditions: 120 bar and 85C Sand production: 0.5 g/s = 32 kg/day Particle diameter: 0.35 mm Sand particle density: 2600 kg/m3

Maximum erosion rate: 0.1 mm/yearThe physical properties for the HC mixture at the well head conditions will have to bedetermined from process simulations by appropriate programs; i.e. PROCESS, MAXISIM,HYSYS etc. Based on the physical properties given in Table 2, the mixture properties areobtained for relevant pipe diameters in Table 3. Based on the parameters in Table 3, theprocedure for erosion in a pipe bend can be applied for dimensioning of the pipe system. Inthe calculations a 90 bend with bend radius equal to 2D is used for all cases.

Based on the RP results in Table 3, a pipe with dimension 6" (D=0.15 m) should be selectedfor the given case. However, if sand content becomes lower, the production rate can beincreased, see Table 4.

Table 2 Physical properties obtained from process simulation programme.

Gas mass flow(kg/s)

Liquid massflow

(kg/s)

Gas density, rg

(kg/m3)

Liquid density,r l

(kg/m3)

Gas viscosity, g

(kg/ms)

Liquid viscosity,

l(kg/ms)

15 5 100 700 1.5 10-5 0.6 10-3

Table 3 Mixture properties and resulting erosion rate for relevant pipe diameters

Diameter

(m)

Superficial liquid

velocity, Vl

s

(m/s)

Superficial gas

velocity, Vg

s

(m/s)

Mix. velocity,V

m(m/s)

Mix density,rm

(kg/m3)

Mix.viscosity,

m

(kg/ms)

Erosion ratefrom RP

(mm/year)

0.125 0.6 12.2 12.8 127 4.5 10-5 0.5

0.15 0.4 8.5 8.9 127 4.5 10-5 0.1

0.175 0.3 6.3 6.6 127 4.5 10-5 0.03

12

Table 4 Recommended maximum production rates for different sand productionrates for a 6" pipe with 2D pipe bends in order not to exceed a maximum erosion rate

of 0.1 mm/year. The same pressure, temperature, HC composition as in the aboveanalysis is assumed.

Sand feed Recommended maximumproduction rate

Vl

sV

g

s

(g/s) (kg/d) (kg/s) (MSm3/d) (m/s) (m/s)

0.01 0.9 81 7.3 1.6 34

0.05 4.3 44.7 4.0 0.9 18.9

0.1 8.6 34.6 3.1 0.7 14.6

0.5 43.2 20 1.8 0.4 8.1

5 CONCLUSIONPrediction of sand erosion in pipe systems is outlined presenting two procedures, one forgeneral complex geometries, and one for standard pipe components.

A procedure for general three dimensional flow geometries is designed for the detailedassessment of sand erosion in complex process components such as production chokevalves, manifolds, blinded Tee bends, intruding sand probes, multi-phase meters etc. Thecalculation procedure employs the Computational Fluid Dynamics (CFD) programme CFXas a framework. The application of a general CFD programme gives the opportunity tomodel flow in general geometries. The flowfield and sand particle tracks from the CFDprogramme are applied to calculate the erosion rates on all internal surfaces. A correlationequation developed by DNV, based on material testing of several materials, is applied whencalculating the erosion rate as a function of the particle hit velocity and angle, and materialgrade. Validation examples show good comparison with experiments and an applicationexample demonstrate that the procedure is applicable to visualise the erosion mechanisms,as well as providing a valuable design tool.

By the aid of experimental findings, and the procedure for complex geometries, is developeda DNV Recommended Practice (RP O501) which can be applied to calculate sand erosion intypical process components: Pipe bends, blinded Tee bends, Straight pipes, welds andreducers.

The present procedures provides a valuable tool for dimensioning of piping systems andcomponents, optimisation of production, inspection and maintenance planning, as well asdesign of sand monitoring systems and choke valves.

13

6 REFERENCES

/1/ Det Norske Veritas Recommended Practice Erosive Wear in Piping Systems DNVRP O501, 1996.

/2/ American Petroleum Institute API Recommended Practice for the Design andInstallation of Offshore Production Platform Piping Systems API RP 14E section 2.5b1991.

/3/ K. Haugen, O. Kvernvold, A. Ronold and R. Sandberg Sand erosion of wear resistantmaterials: Erosion in choke valves Wear 186-187 (1995) 179-188

/4/ P.Jensen, O.Kvernvold & R.Sandberg 1991 An experimental investigation of theerosion characteristic in bean chokes VERITEC Report no.: 91-3059

/5/ O.Kvernvold & R. Sandberg 1993 CRDN 617: Production rate limits in two-phaseflow systems: Erosion in piping system for production of oil and gas DNV Report no.:93-3252.

/6/ A.K. Cuson & I.M.Hutchings Influence on erodent particle shape on the erosion onmild steel Proc. of the 6th Int. Conf. On Erosion by Liq. and Sol. Impact, Cavenish1983.

/7/ G.P. Tilly Erosion by impact of solid particles Treatise on Material Science andTechnology, Academic Press, New York, 1979.

/8/ E. Rask Tube erosion by ash impaction Wear 13, 301-315, 1969.

/9/ W.F. Adler Assessment of the state of knowledge pertaining to solid particle erosionFinal Rep. CR79-680, to the U.S. Army Research Office for contract DAAG29-77-C-0039, 1979.