Paul Bonner 28 September 2017 THE USE OF UNISIM AS A DIGITAL TWIN MODEL Inside Industrie 4.0 / IIoT applications UOP CPS Case Study
IIoT Introduction to Industrial Analytics 1 IIoT Solves OLD problems in new and innovative ways Leverages key developing technologies High bandwidth communications and massive low-cost storage in the cloud Low-cost pervasive sensor technology Advanced analytics (big data) technologies and machine learning Device inter-operability standards Changes the way of how and where work is done Remote collaboration (owner / vendor / OEM) Enables Center of Excellence Captures and applies knowledge Routine monitoring done by machines
IIoT Industrial Architecture 2 Inspire Service Platform Uniformance Insight Asset Sentinel PKI Manager Pulse Big Data Analytics Cognitive Learning Tableau, R, etc Cloud Historian Sentience Cloud Platform Digital Twin UniSim 3 rd Party Models 3 rd Party Apps INspire Partners 3 rd Party Cloud 3 rd Party Vendors INspire Partners Supply Chain Service Management Platforms Stranded Assets Asset Health and Diagnostic Data SDX Historian Secure Data Exchange LIMS ERP Plan SDX Historian SDX IIoT SDX IIoT IIoT IIoT DCS DCS MES IIoT IIoT DCS
The Power of The Digital Twin 3 Plant Availability & Optimization UOP Process Specialist OEM Partner Ecosystem Equipment Availability Utilizing Partnerships Digital Twin Customer SME Analytics Based Insight at Enterprise Level Cyber Secured Data Stream Decisions & Actions Disparate Data Across Enterprise Asset Data Sensors Process Data ERP, Historian, etc. Connected Plant UNCOVERING THE VALUE OF DATA TO DRIVE ACTION
The Digital Twin 4 Unifies existing data silos into a virtual entity Federates data across different applications to drive end-to-end integration Leverages process simulation technology beyond current scope of process design Utilizes Cloud to overcome maintainability issues and enables 3 rd party expertise CAPTURING THE COMPLEXITY OF AN ENTIRE REFINERY INTO A SCALABLE DATA MODEL
Process Analytics Overview 5 Run-time Analytics Unit / Site Additional Models / Rules Off-Line Analytics Unit / Site / Multi-Site Event Detection Model Deviation Detection Heuristic Trained Normal & Abnormal First Principles Statistical State estimation Visual Data Analytics Pattern search Value Search Combinations Cleanse / Filter Data Driven Analytics Multivariate statistical (PCA, PLS ) Black-box (Neural Nets ) Machine Learning Big Data Data vol. & variety (unstructured / text) Feature Selection / Extraction ML (Random Forest, SVM, Naïve Bayes ) Process Data Real-time & Historical (Small Data) Process Engineer Time Required / Skillsets Required Data Scientist
Where Industrial Analytics Meets Predictive Analytics Optimisation: Random Testing: Predictive Modeling: Statistical Modeling: What s the best action? What if we try this? What will happen next? What is the pattern? Prescriptive Analytics Predictive Analytics First Principles Modeling Uses known physics & chemistry Discovery/Alerts: Query/Drill Down: Where should we look? Why did it happen? Diagnostic Analytics Ad Hoc Rpt/Scorecards: Standard Reports: How many, when, where? What happened? Descriptive Analytics TDWI Boston - Techniques for Advanced Analytics
Big Data Analytics Myths 7 Big Data / Analytics replaces the need for process knowledge / engineering experience Data Analytics is a replacement for fundamental models You don t need GOOD data as long as you have a LOT of data You surely have enough data already to get value from big data analytics You don t need to change your work processes to get full value from big data All analytics software is about the same it is all shareware downloaded from the internet
Big Data / Analytics replaces the need for process knowledge 8 The manufacturing processes we operate are fundamentally deterministic Processes have 3 Levels of determinism - Process is governed by its fundamental Chemistry and Physics - The closed loop control system gives the same output in response to the same input - The each operator is trained to provide intervention to the process in the same way
Digital Twin Solution Architecture 9 Purity Production Energy Capacity Customer B Customer C Customer A Ongoing capture of plant (process/lab) data Proactive, ongoing dialogue and recommendations UOP expert review Gross Error Detection Data Conditioning Data Reconciliation Data Cleansing Visualisation Advanced Computation HADOOP data store Solution-specific calculations & models UOP expert reviews Customer D Shared Data Store with Common Metadata model enables: Big Data / ML platform for continuous innovation Templated solutions for scalable deployments
Solution Architecture & Flow Customer Site Data Cleansing Model Tuning Prediction Optimization Unisim Sentinel Unisim Unisim Unisim SDX PHD Mass balance, Reconciliation Parameter Estimation Complex SS Model Complex OPT Model Rendering Sentinel Scheduling, Data interchange & Data Model Hadoop Data Lake Analytics 10 File Number
Uniformance Asset Sentinel Mgt. Reporting & KPI s Engineering Reliability, Maintenance, & Process Uniformance Sentinel Asset Model Dashboards / Displays / Trends Work Requests Maint. Mgt Orchestration web Service Templates Attributes Calculations Rules Equipment Lib. Analytics Embedded User Defined UniSim Other Notifications (E-mail & Alerts) Event Detection (Symptoms/Faults) Calc Results Alerts & Data Historian Real-time / Historical Data Operator DCS / PLC Smart Instruments Historian Visual Inspections Vibration Monitoring Other Applications 11 File Number
Some Details 12 Number of UniSIM flowsheets used Typically 3 x number of Units plus 3 Mass balance, Reconciliation and Parameter estimation flowsheets per unit Complex Steady State and Complex Optimization. How does UniSIM and Uniformance Asset Sentinel work together? UAS process the Raw input data through a series of steps, cleansing, correction, aggregation and lab data normalization. An orchestration web service added as a software extension to UAS triggers the execution of the multiple UniSim design flowsheets. The Orchestration services monitor the readiness of the aggregate input data structure, the completion status of any pre-requisite flowsheets and its tag-time driven execution schedule. The data is triggered to send to the USD flowsheet when conditions are met. The macro level progress of the flowsheet is monitored and the return codes reviewed. How are the results from UniSIM extracted and fed back to the user? The USD results are fed back to UAS via output staging tables. This data is then processed through calculation models as required and fault models to identify potential issues. UAS results (USD pass through and all other results) are exported in UFL format (structured text files) which are processed into a data lake structure for visualization using a Tableau web served interface. The UFL files are also processed into PHD for use. What are the advantages to a UniSIM Power User? The main benefit of the integrated UAS USD solution is to regularly run the agreed set of simulations in consistent and unattended fashion at a higher frequency than could be achieve manually.
13 Example - UOP CPS Connected Performance Services
Delivering the Connected Plant with CPS Solving Key Customer Challenges CPS Architecture Unplanned Downtime Process issues Equipment failures Underperforming Assets Sub-optimal operations Performance vs peers Human Capital Challenges Knowledge gaps Operational excellence Energy and Emissions Emission standards Energy reduction DATA COLLECTION Proactive, ongoing dialogue and recommendations. UOP expert review Ongoing capture of plant (process/lab) data Gross Error Detection Data Conditioning Data Reconciliation DATA CLEANSING Purity Production Energy Capacity VISUALIZATION ADVANCED COMPUTATION CPS solution-specific calculations and models UOP expert reviews Analyze plant performance to reveal full potential through a cloud-based service Around-the-clock monitoring of plant data and rigorous simulations Provides on-going, operational recommendations to close performance gaps Leveraging UOP Process Models & longstanding experience in operational support and troubleshooting Customer value of $0.30-$0.50/bbl in refining & $10-$20/MT in Petrochemicals Customer Site Secure UOP Cloud
Cloud-Enabled Portfolio of Services Customers Reach and Sustain Full Potential KPI & Operating Envelopes Pre-empt process issues with embedded root cause analysis Operate closer to constraints Drive best operating practices Reliability Improve turn-around planning by managing asset lifecycle Gain new reliability insights through data analytics Process Optimization Maximize profitability through changing process and economic conditions Evaluate new opportunities with whatif analysis Unit-specific Benchmarking Improve operations through actionable unit advice Set realistic, achievable targets Utilities Management Enhance heat recovery, fuel gas and H2 systems management Optimize supply and demand balance
CPS - Process Reliability Advisor MOVIE 16 What Is Behind It? Matches unit configuration Tuned UOP process models Technology specific fault models Embedded UOP knowledge, context, and background What Does It Do? Monitoring of unit performance Constraint limitations Event detection & mitigation Knowledge transfer UOP 7432A-16
CPS - Process Optimization Advisor MOVIE Advisory Service to Maximize Profits as Conditions Change What Is Behind It? Rigorous simulation to match unit configuration Tuned UOP process models SQP Optimizer driving an economic objective function Cloud hosted & Maintained by UOP What Does It Do? Maximize process economics against constraints Provide operational recommendations
CPS - Impact of Optimizing an Aromatics Complex FEED RATE PARA- XYLENE BENZENE LIGHT NAPHTHA PROFIT CHANGE ($US million/yr) KEY ACTIONS Feed A 2% 1% 2% 4% +3.5 Lower Reflux - Xylene & Raffinate Columns Reduce Tatoray Conversion Enables Increased Feedrate Base Case Typical UOP aromatics complex Unconstrained feed Optimization Problem Setup Maximize obj. function ($): Products (PX + Bz + By-Products) Feed Utilities Feed B - 2% 8% 4% +5.0 Feed C 5% 5% 4% 5% +9.5 Assumptions: $380 Naphtha to PX spread, $180 Naphtha to Bz. Complex size 600KMTA Naphtha Splitter Increase C6s to Plat Enables Feed Increase to Top of Complex Debottleneck to Rebalance PIX Loop Full Complex under-utilized Increase feedrate until constraint met
Honeywell Connected Plant Solutions from UOP 19 Process Reliability Advisor CCR PlatformingTM Unit UnicrackingTM Unit UOP FCC Unit UOP Aromatics Complex OleflexTM Unit UOP Russell Gas Plants Process Optimization Advisor UOP Naphtha Complex UOP Aromatics Complex