Performance for Assets · Ø Recurrent savings of € 500.000 / year on energy equivalent to 7.000...
Transcript of Performance for Assets · Ø Recurrent savings of € 500.000 / year on energy equivalent to 7.000...
PerformanceforAssetsTurningdataintoactionableintelligence
www.performanceforassets.com
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
Productionprocessescontinuouslygenerate
massiveamountsofdata
but...
“Onaverage,between
60%and73%ofalldatawithinanenterprise
goesunusedforanalytics”-Forrester
Moreover,traditionalconditionmonitoringsystemsgiveanarrowviewonprocessbehaviour
Today’s Industrial Challenges
Tobroadentheview,wecorrelatetheconditionmonitoringmeasurements
withtheoperationaldataandcentralize,filterandsynchronizeallavailableinformation
PLCCMS
Furthermoreweevaluateandintegrate
theotherunstructureddatasourcesthatcanpotentiallyberelevant
Today’s Industrial Challenges
Detectanomaly
Diagnose:Wheredoestheanomalycomefrom?
Prognose:Howdoesanomalyevolveandwhentotakeaction?
Actionable intelligence
for performance optimization & predictive maintenance
Our approach: Methodology
Immediate/onlineactionsSafety,reliability&continuity
Mid-termactionsPlanning,savings&performance
LongtermactionsProcedures,strategy&investments
Our approach: Actionable Intelligence
Combiningprocess/assetknowledgewithdataanalytics..
..isthekeytosuccess!
Machine learning: 3 approaches
Onlineconditionmonitoring
Onlineidentificationofanomaly
Offlinehumandiagnosis
Physicalmodels
Theoreticaloptimumcondition
Impossibletoincludeallhypothesis
Datadrivenmodels
Historicalprocessdata
Relativeviewonequipmentstatus
!
Datadrivenmodels
Our approach: Hybrid model
Onlineconditionmonitoring
Physicalmodels
Hybridmodels
Robustmodelsandoptimizedmachine-learning
bycentralizing,synchronizingandanalysingallassetdata!
Weuse+30provenanalytictoolssuchasthedecisiontreeto:
Rule 2
Rule 4 Rule 3
Rule 1
Our approach: Analytic tools
1. Analysethehistoricaldata
2. Automaticallyselectthekeyparametersthatrelatedtoadrift/deviation
3. Createrulesbycombiningdifferentparameters
4. Testtherobustness
Aftervalidation,therulescaneasilybecopiedtoallsame type equipment and are automaticallyadapted (auto-learning) to the different operatingconditionsofeachuniqueequipment
PREDICTIVE MAINTENANCE
PERFORMANCE OPTIMIZATION
COST REDUCTION & REVENUE INCREASE
€
€
Our goal: Providing actionable intelligence for…
ENERGY SAVINGS COMPENSATION OF LOSS OF KNOWLEDGE
2008 R&D Project
2009 Various industrial
pilot projects
2011 Implementation
on 52 wind turbines
2016 Application on steam turbines
and CNC machines
2017 Establishment of
P4A
Ø Startedin2008asanR&Dprojectfocusingonthedevelopmentofaremoteassetmanagementplatformforrotatingequipmentinindustry–needofintelligentmonitoringtoolforourO&Mproductionbasedcontractsintheconventionalindustry
Ø Healthmonitoringplatform,modelingbasedusingthelatestofdataminingtools
Ø Since2008~6million€wasinvestedtodeveloprobust,reliablemonitoringandmanagementtoolforrotatingequipment
Ø In2017,PerformanceforAssetswasestablishedtomarketthiscollaborativeplatformtoalargeraudience
Track Record
Projects
Case1:Improvedproductionuptime&safety
Howtoavoidreturningfailures
onacriticalmixerinachemicalplant
causinglongdowntimeandsafetyrisks?
Case 1: Improved production uptime & safety
Case1:Improvedproductionuptime&safety
Beforerepair
Afterrepair
Lookingatthedatawithtraditionalchartsandplots,itisnoteasytoseeifpatternshaveoccurredbeforethefailure
Thehealthindicatorchangesbehaviourassoonasthemixerentersinabnormalconditions(4monthsbeforethefailure!)
Failure
Normaloperationbeforethestartofthedegradation
Machinelearninghelpstoidentifywhichconditionsondatadefineahealthysituationoranupcomingfailure
Failure
Case1:Improvedproductionuptime&safety
Step1:DetectionofanomalyØ Useofhistoricaldatatoanalyseconditionsbefore/afterafailureandidentifykeyfactorsto
establishahealthindicator
Ø Unhealthyconditionscanbedetected4monthsbeforethefailure
Step2:DiagnosisØ Rootcauseanalysisbasedondataandournewhealthindicator
Step3:PrognosisØ Predictionofremaininglifetimebasedonnewhealthindicator
Step4:Intelligence-PredictiveMaintenanceØ Implementationofapredictivemaintenancetoolwithalarmthresholdstoprotectthe
equipmentfromupcomingfailures
Case 1: Improved production uptime & safety
Thebenefits
Ø Improvedproductionreliabilityandsafetythroughpredictivemaintenance
Ø Automatedalarmstoprotectequipment
Ø Easyimplementation
Ø Self-learningbigdatamodel
Case2:€500.000/yearenergysavings
Howtooptimizethesteamextraction
ofthesteamnetworkinourphosphateplant
andtherebyminimisetheenergybill?
Case2:€500.000/yearenergysavings
Operatordashboardforoptimizedsteamextraction
Extractfromthedecisiontree(model)createdbyP4Atoindicatethebestwaytooperatethesteamnetwork
Step1:DetectionofinefficienciesØ A+100.000tonsofsteam/yeargapinenergyefficiencywasrevealedthroughourprocess
analysisanddataminingapproach
Step2:DiagnosisØ Rootcauseanalysisthroughvariabilityexplorationofhistoricaldataandprocessthankstoour
analyticstool
Step3:PrognosisØ Wasteofsteamextractionof5tons/hour,resultingina€60/hourfinancialloss
Step4:Intelligence–PerformanceOptimizationØ Installdashboardtomonitorextractionflow(actualvsoptimum),improvecommunication
betweenvariousdepartmentsandenhancereportingpractices
Case2:€500.000/yearenergysavings
Thebenefits
Ø Increasedsteamextractionof5tons/hour
Ø SignificantcostreductionØ Bettermanagementandmonitoring
Ø Moresustainableoperations
Ø Implementedin<3months
TheROI
Ø Recurrentsavingsof€500.000/yearonenergyequivalentto7.000tonsofCO2peryearingasconsumption,ora15%reduction
Case2:€500.000/yearenergysavings
Case 2: Advanced turbomachinery monitoring
Howtogetanabsoluteview
ontheconditionandbehaviour
ofmyturbomachinery?
Case 2: Advanced turbomachinery monitoring
Elaborationofamechanicaldynamicmodelbasedonreverseengineering
Case 2: Advanced turbomachinery monitoring Fine-tuningofthedynamicmodelusingthebalancingmachineInfluencesdueto;
1. bearingdesign2. labyrinthseals3. impeller&turbineblades
Usingthisinputweconfigure
upto50auto-learningmodels/assetwhichwethenfollowonline!
Thebenefitsofourhybridmodel?
Ø Usingphysicalmodelsonlygivesanarrowviewontheassetbehaviour
Ø Bycorrelatingthephysicalmodelwith
historicalandreal-timeconditiondata
wearemonitoringtheassetfromallangles
Ø Ourrobustauto-learningmodelsallow
reliable,continuousconditionmonitoring
Case 2: Advanced turbomachinery monitoring
Dashboard
Asset info with alarms
Asset health status
Various trend analysis options
ASSET MANAGER: Production analysis
ASSET MANAGER: Gap analysis
ASSET MANAGER: Power VS. Torque
MAINTENANCE MANAGER: Alarm overview
MAINTENANCE MANAGER: Failure analysis
020406080100120140160
050100150200250300
nbofo
ccuren
ces[-]
duratio
n[h]
Failuredistribution[FL595]
SumofOccurrences SumofDuration
MAINTENANCE MANAGER: Prognostic
Conditions Prognostic
highwind 6months
Lowwind 12months
Highwindunderpowerreduction
8month
Services
Data value chain
Dataacquisition Datastream Storage Analytics Visualization
Ø PLC/DCSØ CMSØ ImagesØ TextØ VideosØ Sound
Ø QualityØ RelevancyØ Cyber
securityØ GovernanceØ Gemalto
Ø OnpremiseØ PrivatecloudØ Bigdata
Ø P4AModelsØ DataMaestro
(Analytics)Ø OpenSourceØ PhysicalØ Clustering
Ø DashboardsØ TrendingØ ChatbotØ Opensource
IBM
Wintell
Servicepackages
Approach:Processandbusinessunderstanding1. Identifyneedsandconcerns
2. Identifydata/informationalreadyavailable
3. DefineKPI’s
Targets:Presentpotentialimprovementsandassociatedsavings1. Offlineanalysisofhistoricaldataandreporting
2. Onlineimplementation
3. Additionalquickwinsmodelsusingonthesamedataintodifferentareas
Howtogetstarted?
FLASHANALYSIS
Deliverable:Reportwithopportunitiesforprocessoptimization/PDM
Project flow
Step 1: Process & Business Understanding Ø Interview with process engineers to define main Kpi(s) Ø Evaluation of P&IDs / PFDs Ø Evaluation of data availability / Key data identification
Step 2: Flash Analysis Ø Definition of key performance indicators (KPIs) Ø Performance & maintenance benchmarking Ø Assessment of potential improvements opportunities (cost savings, performance,
maintenance) Ø Evaluation of constraints or road blocks
Step 3: Workshops Ø Workshops with operators and plant staff for Ø Brainstorming root causes of variability and process improvement ideas Ø Objective to engage operators in the project Ø Ideas presented on a root cause tree
Step 4: Advanced Analytics Ø Preparation of historical data -> KPI calculation (flag possible data quality issues) Ø Analyse trends and assess long, mid-term (like seasonality) and “local” variability Ø Compare performance, failure rates, with best historical performance or industry
benchmarks Ø Define a realistic target considering production, availability and quality requirements Ø Quantify the performance gap and failure rates reduction Ø Estimate the potential cost savings
Project flow
Step 5: Model development & offline testing Ø Key parameter identification Ø Model development and evaluation Ø Model testing offline with independent/new data sets
Step 6: Online implementation
Ø Setup of prototype dashboard through Wintell web portal Ø Online implementation
Project flow
Do your assets need a performance boost? www.performanceforassets.com