Big Data Analytics in Process Safety Management (PSM)

Size: px
Start display at page:

Download "Big Data Analytics in Process Safety Management (PSM)"

Transcription

1 Big Data Analytics in Process Safety Management (PSM) CCPS Asia Pacific Regional Technical Steering Committee (TSC) Meeting 2 nd Oct 2018, Singapore Sudhakar Kabirdoss, PE Global Process Safety, Micron Technology Singapore. Source: IBM,

2 Key Objectives for Industries Manufacturing Operations Business Growth Productivity Risk Management R & D Marketing

3 Big Data - Size - Speed - Complexity - Uncertainty 1 KB = 10 3 byte 1 MB = 10 6 byte 1 GB = 10 9 byte 1 TB = byte 1 PB = byte 1EB = byte 1 ZB = byte 1 YB = byte

4 Data Growth in Industries (IoT) Source: CISCO

5 Data Domain Knowledge Value Meaningful Result Application

6 Competitive Advantage Analytics Decision Optimization Predictive Modeling What is the best decision? What will happen next? Prescriptive Advanced Analytics Forecasting What if these trends continue? Predictive - Descriptive - Diagnostic - Predictive - Prescriptive Basic Statistical Analysis Reporting with Early Warning Dynamic Reporting Ad Hoc Reporting Why is this happening? What actions are needed? Where exactly are the problems? How many, how often, where? Diagnostic Descriptive Basic Analytics Reporting Basic Reporting What happened? Data Information Intelligence Decision Support Decision Guidance

7 Big Data Applications in other industries Customer Service Sentiment analysis Customer category Brand Perception Supply chain Optimization Product distribution Forecasting Database marketing Financial risk management Healthcare Drug delivery Personalized medications Disease Diagnosis Banking & Finance Risk Management Fraud Detection Forecasting Process monitoring Pattern detection Manufacturing Process Performance Optimization Yield Improvements Equipment Performance Asset utilization Fraud detection

8 Business Understanding [Process Safety policy, metrics, standards, guidelines, eqpt. spec,etc] Data Understanding [parameters, type of operations, mode, source, etc.] Typical approach in data analytics Deployment Data Data Preparation [Impute, clean-up, formatting etc] Modeling [Classification, Regression, Neural] Evaluation [Assessment, Validation, Testing]

9 Unstructured Structured Supervised Learning Historian Data Data Type Incident database Process Parameters Alarms Event logs Equipment monitoring data and so on Equipment Inspection data ITPM data (SAP, CMMS) Design Data SOPs, P&IDs, PFD, HMB, Plot Plan, Layout Incident Investigation Reports Shift communications PSM Audit Reports Photos, images, videos Equipment Inspection Reports Unsupervised Learning PHA Reports

10 Tools Logos, names and brands cited herein are the property of their respective owners and more

11 Process Safety Pyramid Predictive modeling, Prevention Statistical analysis, extrapolation Alerts, Hazard communications Standard/Adhoc reports API 754 Process Safety Pyramid

12 Benefits Real Time Risk Evaluation Optimal Maintenance Schedule Asset Management Resource Allocations Visualization Dashboards

13 Case Study- Pump Failure Prediction Variable Variable Name Data Type x1 date NUM x2 vibration NUM x3 weather NOMINAL x4 noise NOMINAL x5 remote_start BINARY x6 bearing_temp NUM x7 seal_oil_pressure NUM x29 operator_skill_level NOMINAL

14 Case Study- Pump Failure Prediction (contd.) screenshot

15 Case Study- Pump Failure Prediction (contd.) Confusion Matrix

16 Case Study- Pump Failure Prediction (contd.)

17 Incident Prediction Future works in Process Safety Equipment/Instrument Failure Prediction Dynamic Risk matrix Text analysis to complement with data analysis

18 Challenges Business requirements Data availability Data collection Data quality Discipline integration

19 How much data is being analyzed now? Questions? Thank you!