transforma global network of sales and technical partners - cypher energy - pangolin associates...
TRANSCRIPT
Transform your data
into solutions
Who is PEPITe ?DATAmaestro Advanced Analytics Platform
Applications for Yield, OEE, productivity, quality, energy optimisation, predictive
maintenance
DEDICATED APPLICATIONS
Performance optimization based
on advanced analytics and
change management in
operations
Production performance
monitoring and optimization
in aquaculture (salmon
industry)
FindIT App
Performance & health care for
assets (wind turbines, steam
turbines, CNC, etc.)
Wintell App
SERVICES
OPTImaestro projects
A global network of sales and technical partners
- Cypher Energy- Pangolin Associates (Looking for pilots)
ITC Infotech (India)
Confipetrol AndinaPdM Xperts (Peru)
Energies (Col)
- Indea, Techflam, DNV (W. Europe)
Franz Inc. (Asia)
Otelo International (Mx)
Texas A&M University(USA)
PowerMar(US/CAN)
PROSPECTIVE
SAG (JPN)
BelgoBras
Sales partnerTechnical partner
La
st M
od
ified
11
/27
/20
17
4:0
9 P
M R
om
an
ce
Sta
nd
ard
Tim
eP
rinte
d
6McKinsey & Company
Industry 4.0 will impact operations through mainly four key disruptive
technologies
Disruptive
technologies
Industry 4.0
Advanced
analyticsBig data
Human
machine
interface
Additive
Manufacturing
and Advanced
Robotics
Industry 4.0
Capacity &
Quality Increase
Lean
Operations
Mass
Production
Source : Mckinsey & Company
PEPITE CORE BUSINESS
SOURCE: Electricity Consumers Resource Council estimated the cost of August 213 blackout in US between $4.5 and $8.2 billions
Type of project Impact
Optimize use of energy in exothermic processes
Reduce energy costs by 15%Chemicals
Predict and understand root causes of breaks in paper sheets
Reduce shutdowns and increase OEEof 5%Paper making
Use historical data to predict in real time the quality of the steel
Increase yield and reduce scrap by 5%Steel making
Forecast dynamic security of transmission grid
Avoid costly curtailment of loads or generations; in the worst case avoid black-outs (several billions1)
Electrical network
Collect data from hatcheries and provides analytics features to decrease malformation rates
Reduce malformation rates of fish by 20%Hatcheries
Predictive Maintenance project to enhance O&M services
Reduced unplanned down timeCost saving of 10% (lower insurance costs)Wind mills
Analyse drilling operation data to increase ROP
Faster drilling and less downtimes due to well head failure
E&P drilling operations
What is Artificial Intelligence ?
• Artificial Intelligence is a key technology to enable advanced analytics in a big data world
• Artificial Intelligence concepts were invented by Alan Turing; in 1950, he published a paper called “Computing machinery and intelligence.” founding the root of AI.
• There are two main technologies in AI :
Expert systems
Using a knowledge
base, computer uses
an inference engine to
take decision
Machine learning
Using historical data :
1. the computer learns
relationships
(knowledge)
2. the computer uses
what he has learnt to
take decisions
Evolution technology – a new world of opportunities
DMpack is a cheap portable device with embedded services for data collection, data storage, advanced analytics, dashboarding
IBM State of the art : 1956 DMpack State of the art : 2017
Pilot ProjectFrom identified opportunities, a pilot is
selected with clientAnalytics is applied to solve the case
(OPTImaestro)
How to start digital analytics journey ?
STEP 1Digital Data Diagnostic
STEP 2 : Proof of Value Project
STEP 3 : Analytics Capability Transfer
Digital Data DiagnosticWhere is the data ?
What are the data flows ?How is it organized ?
Install DMpack on-siteCan we stream data to the cloud ?
What are the savings opportunities ?How big are these opportunities ?
How to leverage opportunities with analytics ?
Client with Analytics Capability Pilot enables client training in advanced
analytics capabilitiesClient can implement analytics projects using
DATAmaestro technology
1-2 Months
2 – 6 months
SaaS access fee per user per year
(*) Indicative estimates; can vary depending on application complexity
OPTImaestro project frameworkSTEP 2 : OPTImaestro project to optimize plant operation
DATAmaestro Platform setup/configuration
Clo
ud
DATAmaestro Platform
DATAmaestro CollectorConnect, collect
DATAmaestro LakeStore, merge, resample
DATAmaestro AnalyticsVisualise, analyse, diagnose, predict
DATAmaestro DashboardsReal-time monitoring
Industrial/Manufacturing Process
Collectors, Lake, Analytics and Dashboards are web services that can be installed in any secured cloud infrastructure (such as AWS, Azure, etc.).PEPITe can also provide its own secure cloud service.
Cloud based setup
Cloud based infrastructure In order to leverage the power of big data infrastructure, DATAmaestro can be also integrated in 3rd party big data infrastructure.
Integration with Big Data infrastructure
Cloud based infrastructure DATAmaestro can be installed on customer premises in a few days with the DMpack.
On premises setup
OPC UA DB/Historian native API
DM
pac
k
Plant Site
Self service analytics with DATAmaestro access
DATAMaestro is flexible and can be adjusted to comply with clients policy on data access and security
IiIIoT
Performance for Assets
Turning data into actionable intelligence
www.performanceforassets.com
Introduction
Half of my expert-personnel will retire in the near future
Can processes be optimized
and how?
How do I ensure safe and reliable operation of aging equipment?
Where can I save energy?
Why does my machine not perform
as expected?
How can meantime between failures be
extended?
How to enhance the value of existing
products & services?
Today’s Industrial Challenges
Production processes continuously generate
massive amounts of data, but..
“On average, between 60% and 73% of all data
within an enterprise goes unused for analytics” - Forrester
Today’s Industrial Challenges
Moreover traditional condition monitoring systems give
a very narrow view on the behaviour of your asset / process
To broaden the view on process behaviour unlimited other (structured and unstructured) data sources are available
and potentially more relevant.
Detect anomaly: identification only
Diagnose: why does anomaly happen?
Prognose: how does anomaly evolve and when to take action?
Combining process / asset knowledge with data analysis is the key to success!
Actionable intelligence for performance optimization & predictive maintenance
Immediate actions in safety, continuity, reliability and recovering of performance
Mid-term actions in planned intervention, replacement planning, savings, increase of performance
Long term actions in improved procedures, investment strategy, cash out planning
Our approach: Methodology
Our approach: Hybrid model
Online condition monitoring with global alarms + human analysis
Physical models
+
Correlation of Historical process data
+
Additional benefitWhen applied to multiple similar machines (fleet), new knowledge is applied to avoid the same failure in these machines too
!The P4A Hybrid model
Machine learning ♻ Hybrid model combining approaches of models type 1,2 and/or 3 to optimize machine-learning and models qualityand put online with the current data of the specific machine
PREDICTIVE
MAINTENANCEPERFORMANCE
OPTIMIZATION
COST REDUCTION & REVENUE INCREASE
€
€
Our goal: Providing actionable intelligence for…
ENERGY SAVINGSCOMPENSATION OF
LOSS OF KNOWLEDGE
Case studies
Case: Improved production uptime & reliability
Goals: – Determine the tool wear and provide real-
time feedback to operator– Early detection of degradation: rail bearing,
ball screw
P4A solution– Instrumentation : currents, vibration,
temperatures– Realtime analysis
• Signal processing : phase detection, filtering, enveloping
• Data analytics : ISHM
Case: Improved production uptime & reliability
Operation optimisationSignal processing
Operation optimisationData analytics
Tool wear indicator
Raw motor current
Machining phase detector
Cumsum of ISHM distance
Cause identification
Predictive maintenanceVibration signature
– Mixers (predictive maintenance)
– Steam network (energy savings)
– Steam turbine (predictive maintenance & performance)
– CNC machine (predictive maintenance & performance)
– Wind turbines (predictive maintenance & performance)
– High voltage motor (predictive maintenance)
– Transformers (predictive maintenance )
Some projects
Do your assets need a performance boost?www.performanceforassets.com
creating value from data
Machine & process monitoring in the age of big data
18/04/2018
helps you create value from your data
6Years of Expertise
11People
Data Science
AI – Machine Learning
Prescriptive Analytics
Optimization algorithms
Data and Process Mining
Engineering
PID control
Vision and Sensors
Physics - Mechatronics
Advanced Process Control
Scope
Consulting and Services
End to end data projects
Software products
Custom made software
OUR SERVICES ARE USED FOR…
Process Improvement
Process control improvements to:
- resolve quality problems
- reduce cost
- improve stability and safety
Smart Machines
Software and sensor selection for:
- inspection robots
- sorting machines
- measurement devices
Advanced Analytics
Prescriptive Analytics for:
- knowledge discovery
- decision support
- root cause analysis
Anomaly Detection &
Analytics for Machine
Sensor Data & Log Files
ANOM LYYA
Traditional monitoring
2. Descriptive statistics
3. Dedicated sensors 4. KPI’s and dashboards
1. Hand-written rules
6
What changed?
More data is available- Sensors, software logs, contextual data,
lab measurements…- Easier to collect (IoT etc)- No storage / speed limits
Machine Learning and AI are mature techs
Some problems with traditional monitoring
• You can’t write rules for every issue:• there’s no time
• you don’t know what could go wrong
• Problems happen in the software execution / in the communication with someone else’s equipment.
• Not every issue is a vibration.
• Statistical analysis doesn’t always work (highly dynamical or complex signals)
Solution: Unsupervised Anomaly DetectionAI learns what’s normal by analysing (all) the data, no human input.
Unique capabilities
Yazzoom has built on top of open source Big Data technologies like Logstash, Kafka, Kibana and Elastic Search a powerful reusable and extendible technological solution with some unique capabilities:
Process mining: The ability to learn from the various heterogeneous data streams statistical models that describe how the machines, devices and systems have been used and which process steps were executed by those systems.
Context dependent anomaly detection: The ability to detect in real-time anomalies or abnormal process execution in both numerical data and log files, while taking into account the context in which the system operates.
Compared to traditional monitoring
Detect unknow problems, abnormal trends and patterns,
external sources
Detect problems that are not vibrations
Scale easily to more data, new sensors, changes, variable components or subsystems…
Detect issues in software execution in advanced machinery
Handle complex signals with better detection of
issues and less false alarms
Take context into account (machine state,
temperature, raw material properties, recipe)
Some highlights of AI-based anomaly detection- Compared to traditional systems:
- Can be used to determine where to add extra sensors / get extra data: the algorithms can pinpoint where they lack information to be accurate.
- Can be used for enhanced (cyber)security
1 minute to find the cause of an issue
Root cause analysis of downtime in 1min instead of 20.
1 hour early warning
Alarm raised before the end user noticed problems and contacted tech
support.
(Cyber)Security ?
Several recent Industrial (cyber)sabotage involved fooling traditional monitoring systems by feeding them fake sensor readings.
AI can detect that one sensor reading is not realistic given the value of all others.
Much harder to fool than traditional monitoring.
Who uses ?• Industrial production lines (continuous and
batch processes, robot-powered discrete
manufacturing)
• Medical imaging equipment
• Network and telecom infrastructure
• IoT-enabled utilities networks
• Power plant systems
• Consumer connected devices…
“Machine” users or manufacturers
• PoC
• Pilot
• Full Depolyment
…
Some misconceptions about unsupervised Anomaly Detection…
Doesn’t replace traditional approaches. It comes in complement.
Can use human input: learn faster, get feedback and become better.
Doesn’t need a supercomputer to run.
Doesn’t need to be in the cloud.
Big data, small project
PoC
Small subset of data, offline.
Goal: validate the presence of value in
the data
Typically 2-3 weeks work
Pilot
Small deployment, single system, basic
integrationGoal: refine
requirements, enhancemodels…
Typically 2-3 weeks work
Full Deployment
All systems monitoring, full
integration.
YANOMALY deployment model
Integrates with existing monitoring / data / IoT platforms / historians no big changes.
Edge or cloud.
Example Anomaly Dashboard in Browser
This view illustrates different types of anomalies discovered by process miningon log files
The user canzoom & filter & click-through tothe original log files
(More) Value inside machine data:
R&D
Service
Operations
• Gain pre-emptive knowledge of potential problems
• More efficient investigation and troubleshooting
• Reduce the mean time to repair of your service team
• Real-time detection and even prediction of anomalies
• Avoid down-time through automatic remediation actions
• Speed up root cause analysis for faster recovery
• Fraud detection
• Learning how machine/plant/utility is used
• Higher reliability through more realistic product testing
• Better product design thanks to knowledge of real usage
www.machine-analytics.com
+32 9 378 02 18