Big & Fast Data Analytics Event Analytics for Production Surveillance and Machine Management Michael O Connell, PhD Chief Data Scientist TIBCO Copyright 2000-2014 TIBCO Software Inc.
Smart Energy Production Copyright 2000-2014 TIBCO Software Inc. 2
Telematics and Equipment Monitoring Auto and Truck Telematics Maintenance Safety Fuel Efficiency Scheduling and Routing Airplane Fleet Management Copyright 2000-2014 TIBCO Software Inc. 3
O&G Production & Equipment Surveillance Data Monitoring Motor temperature Motor vibration Current Intake pressure Intake temperature Flow Pump Components Electrical power cable Pump Intake Protector ESP motor Pump monitoring unit Copyright 2000-2014 TIBCO Software Inc. 4
Energy Operations Map Analytics and Reporting Environmental, Health and Safety Upstream E&P Midstream Downstream Refining & Supply Oilfield Services Exploration : Drilling, Seismic Well Engineering Decline/Type Curves and EUR Portfolio Management Production Surveillance Production Optimization & NPT Analysis Drilling and Equipment Inventory Mgt Supply Chain Mgt Asset Mgt Process Optimization Performance, Risk & Trading Trading Analytics Trading Oversight Refining & Supply POS Asset and Management Syndicated Data Analytics Refinery Turnaround Netback Analysis Downstream Fuels & Marketing Order & Contract Gas Station Mgt Customer Asset Mgt, Reliability Drilling Optimization Hedging Risk Analytics Sales & Marketing Call Center Logistics, Transportation & Distribution Pipeline Transportation Optimization Terminal Mgt Fleet Management Supply Chain Management Finance, Property, HR and other supporting applications Copyright 2000-2014 TIBCO Software Inc. 5
Copyright 2000-2014 TIBCO Software Inc. 6 Event Analytics 3. Develop rules and models to anticipate problems and opportunities before they occur 7. Over time, user evaluates and optimizes triggers, analyses 2. Establish bounds of operation, areas of opportunity BIG DATA AT REST FAST DATA IN MOTION 6. User takes action - Intervention - Initiate test - Maintenance 1. Data collected on production process 4. Predictions on Event Stream trigger Spotfire analysis, notification 5. User evaluates event analytics in context to do Root Cause Analysis
Business Value of Pump Monitoring Downhole equipment reliability: revenue increase Condition-based maintenance: efficiency savings Warranty cost recovery Monitoring by exception; thousands of pumps Analysis Performance and pump health surveillance Real-time intervention Component faults and failure analysis Effects of operating conditions, suppliers Results Improvements in NPT: $1000/well/day observed Prioritization of engineering and retrofit Supplier involvement in system reliability Monitoring: Electric Submersible Pumps Copyright 2000-2014 2000-2013 TIBCO Software Inc.
Leading Indicators of Failure Leading Indicators of Component Failure Energy consumption increase, pressure and flow remain constant a pump is slipping Voltage, current variability increase a harmonics in power between VSD and ESP Motor temperature increase a break in insulation around wiring aelectrical short Motor temperature decrease, pressure increase a gas buildup downhole Motor temperature decrease Pressure increase/decrease Current increase/decrease Pressure slope change Combinations of variables......... Relationships with Flow Copyright 2000-2014 TIBCO Software Inc. 8
Copyright 2000-2014 TIBCO Software Inc. 9 Leading Indicators of Failure Variable Speed Drive Harmonics in Voltage, Current Electric Submersible Pump
Monitoring Variables over Time Location Change Variable moves up or down Slope Change Variable changes trend Variance Change Variable becomes more/less volatile Process Threshold Shewhart control chart Failure Model y (0/1) = f (X, b) + e; f = logistic regression, trees, svm, nnet,... Trend Analysis Combination of Rules CUSUM Analysis Statistical Analysis Statistical Process Control Machine Learning
Demonstration Spotfire Data Discovery on Equipment Failure Identify leading indicators for failure: rules and TERR models Real-time Monitoring of Sensor Data Monitor real-time data ; intervene when indicator is triggered Spotfire Root Cause Analysis Understand failure events Fleet monitoring Monitor and maintain equipment Predict time to service from sensor data TERR and Spotfire On-the-fly TERR analytics in Spotfire Real-time monitoring of production & equipment Equipment maintenance TERR and Spotfire Notes: data have been generated and flow sped up to produce faster failure events; 11
Copyright 2000-2014 TIBCO Software Inc. 12 Demonstration: ESP Analytics Identify leading indicators for ESP failure Trap for patterns in real-time data 09.00 11.00 13.00 09.00 11.00 13.00
Copyright 2000-2014 TIBCO Software Inc. 13 Event Analytics for ESP Management Add patterns in to Spotfire Event Analytics Real-Time monitoring workflow Step 1 Load TERR Models / Rules Step 2 Step 3 Stream & Score New Data Detect Issues & Trigger Spotfire Automation - Grab data prior to trigger - Populate RCA template - Save to Spotfire library - Email responsible engineer
Copyright 2000-2014 TIBCO Software Inc. 14 Event Analytics for ESP Management Upon event trigger, populate Spotfire RCA template; email responsible engineer
Copyright 2000-2014 TIBCO Software Inc. 15 Event Analytics for ESP Management Responsible engineer clicks URL to launch Spotfire RCA; diagnose issue
Copyright 2000-2013 TIBCO Software Inc. 16 Vehicle Monitoring & Maintenance
Copyright 2000-2014 TIBCO Software Inc. 17 Equipment Monitoring & Maintenance
Copyright 2000-2014 TIBCO Software Inc. 18 Production & Equipment Surveillance
Production Surveillance Copyright 2000-2014 TIBCO Software Inc. 19
Copyright 2000-2013 TIBCO Software Inc. 20 Decline Curve Analysis
Copyright 2000-2014 TIBCO Software Inc. Fraction Failing 21 Reliability Analysis Data Failure data from tickets/orders In-service data from equipment Models Weibull : 2-parameter Applications Single mode failure Wear and tear.05.03.02.01.005.003 Failure Analysis Bearing Cage Failure Data with Weibull ML Estimate and Pointwise 95% Confidence Intervals Weibull Probability Plot.001.0005.0003.0002.0001.00005.00003 etahat = 11792 betahat = 2.035 200 300 400 500 600 800 1000 1200 1500 Hours
Copyright 2000-2014 TIBCO Software Inc. 22 Reliability Analysis Data Failure data from tickets/orders In-service data from equipment Models Flexible smoother Applications Multi mode failure Birth defects Wear and tear Failure Analysis
Copyright 2000-2013 TIBCO Software Inc. 23 Reliability Analysis Spotfire allows 2 failure modes to be analyzed. With burn-in and wearout, the joint hazard function shows the bathtub shape.
Spotfire Event Analytics Components TIBCO Spotfire TIBCO Runtime for R Email Notification TIBCO Spotfire (desktop or web ) In-Database In-Memory Alert Via Spotfire Automation Services On demand Spotfire Server Complex Events Processing TIBCO Streambase TERR Enterprise Data SQL Server, Oracle, SAP, OSI-PI / Historians Enterprise Data SQL Server, Oracle, SAP, Streaming Data Copyright 2000-2014 TIBCO Software Inc. 24
Copyright 2000-2014 TIBCO Software Inc. 25 Spotfire Event Analytics Components
Copyright 2000-2013 TIBCO Software Inc. 26 Analytics that Change the Game
First to Insight, First to Action Thank you! Michael O Connell, PhD Chief Data Scientist TIBCO Fellow moconnell@tibco.com http://about.me/moconnell +1-919-7401560 Copyright 2000-2014 TIBCO Software Inc. Copyright 2000-2013 TIBCO Software Inc. 27