predictive maintenance on engine failures
TRANSCRIPT
GE01_DT 409510395_Wind Speed QI-109 GE01_DT Cooling Fan-711.Feed Rate
1-8.Net VolumeCoal Motor Load
02F100.TOT.EV
03LBA32CT001-2
DC.SJ.ITLoad.PWR
TI-145 FR2001
TI-178 GE04_OS
FT9001 FT9001
FR5001 AF_NOISE
DC.SJ.PUE TI-102 DC.Zero DY-108 DC.SJ.C1.Z3.R3.PDU1.PF GE01_A_DT
FIC-144 02F100 fasttag
FI-151 0_ENG_AUX_STS
GE05_Energy C1:14AT5AC03.Air Flow FeedBin.Cmt
Boiler Cold Reheat Pressure
B737_FG117 DC.TimeLoad
D-110.Tank Pressure.PVGE04_DT QI-121 GE03_V_WIN
DC.Rk07R DC.Srv06R
GE04_Energy TI-121 FT9001
FAC.OAK.Power-Kh-Val.PV
DY-131DC.SJ.PUE
fic1001.C
GE02_OT
GE01_DT
02F102.1HRAVG BGT001 PI-111 facility_output
DM-05:BW.R DC.SJ.C1.Z1.R1.Rk06.S2.O03.PWR QI-111 FinalProductBin.On
94:GRDIDX.ProdID Boiler-209.Fuel Gas Flow fic1001.C FR50011-
8.Net Volume Coal Motor Load 02F100.TOT.EV 3LBA32CT001-2
FI-101 bf5e1d1d-39c9-4b5b-b3d3-c2ce05fa3a26 DM-05:BW.R AT401
0_CLR_FINAL_OUT_B_TMP F506_E990 339511775_Clear Sky Global Horiz GE01_DT
FI-111
GE01_CON AlarmTest.Input.Float32.1
D-110.Tank Pressure.PV Boiler Feed Pump #1
Boiler-209.Fuel Gas Flow DC.Srv01R 94:GRDIDX.Tr igger AC09.Power
403511195_Wind Speed
DC.C2Z1.Pwr.Ripple GE01_A_DT
1-16.Net Volume CB1992_MS 0_CMP_FLOW_TOTAL GE02_Energy
FeedBin.Cmt
DC.Zone1.Number
DailyTriggerFrqPrbCost_ER
AlarmTest.Input.Float32.1 AQUA2-TI-201.PV DC.SJ.SiteRealTimeITLoad.PWRFT9001
FT9001
DC.Srv01R Boiler-125.Fuel Gas Volume
Anacortes Refinery.Alkylation.Asset Problems B210_FG005.KPIExcursion
FT9001
fasttag
AT401
Crude Furnace
Draft Pressure: -0.5 WC
Stack Temp: 316 °F
Oxygen: 2.5%
Firebox Temp: 860 °F
Outlet Temp: 840 °F
Cold Oil Velocity: 6 ft/sec
Weather Conditions
Relative Humidity: 34%
Current Temp: 85 °F
High: 92 °F
Low: 57 °F
Wind: 8 mph/N
$
Total Production
Energy Efficiency
Downtime
Real-time Decision
Support
Business/Operation
Intelligence
Act:
Score,
Visualize
Deploy Apps,
Services &
Visualizations
Measure
Preparation Modeling
Feature &
Algorithm
Selection
Model
Testing &
Validation
Operationalization
Models
Visualizations
Ingest
Profile
Explore
Visualize
Transform
Cleanse
Denormalize
Advanced Analytics Journey
Today’s Basic Analytics
DescriptiveAnalytics
• Point-by-point querying
DiagnosticAnalytics
PredictiveAnalytics
PrescriptiveAnalytics
Analytics Maturity
• What happened? • Provides info about
past problems fleet-wide
• Why/When did it happen?
• Provides insight and visibility into what can be improved where
• What will happen? • Provides predictions
(foresight) that lower maintenance costs, optimize efficiency and productivity
• How can we make it happen?
• Provides recommendations for the best course of action to achieve desired outcomes; based on predictive analytics
Descriptive &
Diagnostic Analytics
Predictive &
Prescriptive AnalyticsBasic Analytics
Data Scientist
Interact directly with data
Built-in to SQL Server
Data Developer/DBAManage data and
analytics together
Relational Data
Analytic Library
T-SQL Interface
Extensibility
?R
R Integration
010010
100100
010101
Microsoft Azure
Machine Learning Marketplace
New R scripts
010010
100100
010101
010010
100100
010101
010010
100100
010101
010010
100100
010101
010010
100100
010101
Intelligent analytics across realmsEmbed R, Python or Azure ML on-premises or cloud
REGIONAL SEMINARS 2015 8
Scenario
We have 100 engines sending various
sensor data like rpm, burner fuel/air
ratio, pressure at fan inlet and 20
other measurements with
configuration settings for each
engines. The average life span of an
engine is about 206 cycles but it
varies widely from 140 to 360 cycles.
We want to predict the failure of
these engine ahead of time.
Azure Cortana Intelligence Suite
Action
People
Automated Systems
Apps
Web
Mobile
Bots
Intelligence
Dashboards &
Visualizations
Cortana
Bot
Framework
Cognitive
Services
Power BI
Information
Management
Event Hubs
Data Catalog
Data Factory
Machine Learning
and Analytics
HDInsight
(Hadoop and
Spark)
Stream Analytics
Intelligence
Data Lake
Analytics
Machine
Learning
Big Data Stores
SQL Data
Warehouse
Data Lake Store
Data Sources
Apps
Sensors and devices
Data
Predictions as Future Data (to PI 2015)
INGESTPREPAREDATA SOURCES
On Premise
Predictive Maintenance on Engine FailuresOn Premise with PI Integrator and SQL Server 2016 Enterprise and R Server
Power BI
ANALYZE PUBLISH CONSUME
SQL Server 2016
Enterprise
R Server
PI Infrastructure
Predictions as Future Data (to PI 2015)
Azure SQL Data Warehouse
INGESTPREPAREDATA SOURCES
On Premise
Predictive Maintenance on Engine FailuresMicrosoft Azure: Cortana Intelligence
Machine Learning
Power BI
ANALYZE PUBLISH CONSUME
Cortana
Web/LOB Dashboards
Azure SQL Data Warehouse
SQL Server 2016 Enterprise
R Services
Asset Framework (AF) data from PI System
13
Engine 1 Engine 100
PI Integrator for Cortana Intelligence
Publish PI AF data to Cortana Intelligence Stores
PI Integrator allows you to push “Analytics Ready” data directly to Cortana Intelligence
Azure Machine Learning Model
Understanding our Data
Failure Points of Engines Strong Correlation among sensors
Using Principal Component Analysis
PC1 shows a strong Variance Value of PC1 has strong correlation on RUL
Scoring the Predictive Model
Prediction versus actual remaining life – Using PC1 as our predictor, the model appearsto be more concentrated and accurate as remaining life approaches to zero