Asset Optimization for the Process Industries: From Data to Insights to Actions June 2017 Willie Chan, Ashok Rao, Bill Mock Aspen Technology, Inc. 1
AspenTech Disclaimer Aspen Technology may provide information regarding possible future product developments including new products, product features, product interfaces, integration, design, architecture, etc. that may be represented as product roadmaps or product visions. Any such information is for discussion purposes only and does not constitute a commitment by Aspen Technology to do or deliver anything in these product roadmaps or otherwise. Any such commitment must be explicitly set forth in a written contract between the customer and Aspen Technology, executed by an authorized officer of each company. 2
AspenTech recognizes George Stephanopoulos Founding Chair, AspenTech Academy Established in 2012 to Accelerate application of research innovations in Process Engineering software Promote the use of Process Systems Engineering software in universities 3
Evolution of Process Systems Engineering Unit Process Plant & Sites 70 s 80 s & 90 s 00 s to Today ENGINEERING - DESIGN & OPTIMIZE THE ASSET MAINTENANCE - MAINTAIN THE ASSET MANUFACTURING - OPERATE THE ASSET SUPPLY CHAIN - OPTIMIZE THE SUPPLY CHAIN 4
Technology Trends The Industrial Internet of Things (IIOT) is changing the world as we know it! By 2020 - over 7 billion Internet-connected industrial devices * Gartner 2015 Market Trends Connected industrial devices will proliferate worldwide Analytics Cloud Embeddable HW/SW Sensors and equipment will become increasingly smarter Cybersecurity Industrial information systems will move to the cloud Networking Connected devices will generate massive amounts of condition data Manufacturers will look to analytics to provide deep process insight 5 http://www.gartner.com/newsroom/id/3165317
Example: Wealth of Sensor Data in Plant Historians 6
Example: P&IDs and Equipment Datasheets 7
Example: 3D Plant Model 8
Example : Maintenance Data 9
Example: Textual / Unstructured data 10
Asset Optimization A comprehensive, holistic approach to driving the highest possible financial return over the entire asset lifecycle What if you could leverage the rich wealth of asset data to Run to the limits of performance Operate Asset Lifecycle Maintain Maximize uptime with actionable insights Push the boundaries of what s possible 11 Design
Asset Optimization A comprehensive, holistic approach to driving the highest possible financial return over the entire asset lifecycle What if you could leverage the rich wealth of asset data to Run to the limits of performance Operate Push the boundaries of what s possible Smart Sensors Big Data Mobile Connectivity Cloud Computing High Fidelity Modeling AI & Machine Learning Advanced Visualization Maintain Maximize uptime with actionable insights 12 Design
Solution Vision Data Sources Operational Data Maintenance Data Design Data Other sources Aggregate / Clean / Consolidate Time Series Data Events Asset / Material Info Textual Context Insights Asset Optimization Engine Models Workflows Messaging and Actions Operations Guidance Maintenance Guidance Design Guidance Capture Organize Implement Predict and Prescribe Rules Data sources that span time series, maintenance records and textual content like operating manuals Insights by bringing together design, operations and maintenance domains, and automation of complex analyses and workflows Guidance with an objective to minimize plant lifecycle costs subject to the level of risk the organization can tolerate 13
Enabling Technologies Operational Data Data Sources Aggregate / Clean / Consolidate Insights Messaging and Actions Closed Loop Control Data Historian LIMS Sensors Maintenance Data Sensor Needs Enterprise Asset Management Asset Breakdown Data Operator Records Cognitive Computing Analytics Engine Process Modeling Spare Parts Design Data Datasheets Other Sources P&IDs 3D Modeling Big Data Ecosystem Machine Learning Knowledge Graphs Natural Language Processing Visualization Molecular Modeling Ensemble Modeling Prescriptive Guidance Work Orders Design improvements literature PC / Mobile VR AR Capture Organize Predict and Prescribe Implement 14
Data Science Pattern Discovery, Pattern Search, and Process Data Visualization Process Sensors Pattern Discovery Compact Data Visualization Pattern Search 15
Precursors detected by RCA Historian Time series Data Science Root Cause Analysis (RCA) KPI exhibits undesirable spikes. RCA to identify precursors/causes Bayesian Network RCA Model (Machine Learning) M03 M01 M04 Precursor predictive lag: 80 min M02 Feature Extraction (Data Science) Predictive Capabilities (Data Science) - Structure learning - Conditional probabilities - Causal links connect nodes Real-time Data Warning! High probability of KPI spike @ time=xxx 16
Empirical Modeling Multivariate Analysis Data Conditioning Anomaly Detection Key Contributions Operational Guidance Optimization of Operation 17 PLS / PCA Modeling isolates the most important data dimensions
Molecular Characterization From Analytical Data to First-Principle Molecule-based Process Models Whole Crude Data ~ 20 points Molecular Distributions Distilled Cuts Data ~ 100 points GC-MS Data ~ 1000 points MC Model FT-ICR-MS Data ~ 50,00 points Molecule-based Process Model 18
Operations Analytics Combining Data Science, 1 st Principle & Empirical Modeling for Equipment Analytics Connect plant measurements to simulation model Deploy column model and monitor plant operation Connected historian as historical and real-time plant data source Auto tune model with historical plant data Display measured and calculated values (predictions) Monitor operation using real-time data Real time dashboard 19
Maintenance Analytics Training of Agents from Maintenance and Historical Data to Monitor Plant Operation Sensor Data Smart Trained Agents Asset Monitoring and Failure Prediction Asset Work History 20
Areas for Future Research & Development Operational Data Data Sourcing Sourcesand Transmission Aggregate / Clean / Consolidate Cognitive Reasoning Computing Insights and Visualization Messaging and Actions Closed Loop Control Data Historian Cognitive Computing Maintenance Data Enterprise Asset Management Design Data Datasheets Other Sources LIMS Asset Breakdown Data P&IDs Sensors Operator Records 3D Modeling literature Prescriptive Engine Knowledge Graphs Natural Language Processing Ensemble modeling Model Calibration Prescriptive Engine Big Data Ecosystem Machine Learning Molecular modeling Cognitive Computing Machine Learning Knowledge Graphs Natural Language Processing PC / Mobile Cloud and Edge Analytics Cybersecurity Analytics Engine Software Defined Storage Augmented Reality Virtual Reality Mobile Devices Visualization Sensor Needs Knowledge Graphs Spare Parts Work Orders Design improvements Ensemble modeling Capture Organize Predict and Prescribe Implement VR Process Modeling Molecular Modeling Ensemble Modeling Prescriptive Guidance AR 21
Empowering the Next Generation of Process Engineers Engineering Modeling Optimization Data Science Machine Learning Analytics 22