Smart Manufacturing Systems Design and Analysis Dr. Sudarsan Rachuri Program Manager Smart Manufacturing Systems Design and Analysis Systems Integration Division Engineering Laboratory NIST sudarsan@nist.gov Presentation to the Federal Big Data Working Group February 2 2015 1
Outline Smart Manufacturing System Design and Analysis Program objective and focus Program Thrusts and Projects Projects Reference Architecture Modeling Methodology Predictive Analytics System Performance Assurance Standards for SMSDA Collaboration Opportunities 2
Smart Manufacturing Systems Design and Analysis Performance of SMS Models and standards SysML, BPMN, Modelica Modeling methodology for Smart Manufacturing Systems Reference Architecture for Smart Manufacturing Systems Performance metrics and Standards ASTM E60.13 Performance Assurance for Smart Manufacturing Systems Standards and Measurement Science for Smart Manufacturing Systems Design and Analysis Standards Reference Architecture OAGi standards ISA 95 Performance metrics and Standards STEP-NC, MTConnect, PMML Real-Time Data Analytics for Smart 3 Manufacturing Systems
Projects Short description Reference Architecture Provides a common vocabulary and taxonomy, a common (architectural) vision, and modularization and the complementary context. - Composability Modeling Methodology The new technical idea is to bring together the conceptual modeling, information modeling, and behavior modeling paradigms - Compositionality Predictive Analytics Big data analytics for manufacturing System Performance Assurance Extending quality assurance to performance metrics, including tools for verification and validation. An architecture is said to be composable with respect to a specified property if the system integration will not invalidate this property, once the property has been established at the subsystem level. A highly composable system provides recombinant components that can be selected and assembled in various combinations to satisfy specific user requirements 4 Compositionality - The property of the system that the whole can be understood by understanding the parts and how they are combined
Reference Architecture ISA 95 ISA-95, Image courtesy: Automation world
Project: Modeling methodology for Smart Manufacturing Systems (MMSMS) OBJECTIVE Develop modeling methodology and tools to predict, assess, optimize, and control the performance of smart manufacturing systems to result in a significant increase in process efficiency by FY 2018. The new technical idea is to bring together the conceptual modeling, information modeling, and behavior modeling paradigms 6
Model of Manufacturing System Predict Physical Artifacts Raw materials Components M1 Physical Products Artifacts Co-products By-products Waste Auxiliary materials Information Artifacts Mi Mn Information Artifacts Resources Resources Energy Water Emissions Control Mi Machine i Σ 7 Pi Process i
Project Three: Performance Assurance for Smart Manufacturing Systems (PASMS) OBJECTIVE To develop metrics and assessment methods that will assure the performance of dynamic production systems with a predictable degree of reliability by 2018. Extending quality assurance to performance metrics, including tools for verification and validation. 8
Performance Assurance for Smart Manufacturing Systems (PASMS) 5% decrease in batch cycle time 10% improvement in machine reliability 10% reduction in water consumption 5% reduction in energy costs Source: www.ge-ip.com
Project Four: Real-Time Data Analytics for Smart Manufacturing Systems (DASMS) OBJECTIVE Develop standards, methods, and protocols for data analytics to enable real-time diagnostics and prognostics that will significantly increase the efficiency of dynamic production systems by FY18. Big data analytics for manufacturing 10
A lot of Media and Business coverage of Big Data Analytics Data is a core asset Those that adopted data-driven decision making achieved productivity that was 5 to 6 percent higher than could be explained by other factors, including how much the April 23, 2011 When There s No Such Thing as Too Much Information By STEVE LOHR companies invested in technology Brynjolfsson, Erik, Lorin Hitt, and Heekyung Kim, Strength in Numbers: How Does Data-Driven Decision Making Affect Firm Performance? (April 2011) The high-performing organization of the future will be one that places great value on data and analytical exploration (The Economist Intelligence Unit, In Search of Insight and Foresight: Getting more out of big data 2013, p.15). Data Representation Computational Complexity Statistical and machine learning techniques Data sampling, cleaning Human in the loop Up to 50% reduction in product development and assembly cost Up to 7% reduction in working capital Companies that gain a competitive edge with analytics can be found at all levels of technological sophistication The evidence is clear, Data Driven Decisions tend to be better decisions. In Sector after Sector, companies that embrace this fact will pull away from rivals Big Data Management Revolution, Harvard Business Review
What is Big Data? 12
Evolution of Analytics 13
Manufacturing Big Data Analytics ROI of Big Data Analytics in Manufacturing Lifecycle Data availability and quality Distributed Computing Infrastructure Bring Manufacturing Science and Data Science closer 14
Manufacturing Data analytics - A Big Picture Report McKinsey Global InstituteBig data: The next frontier for innovation, competition, and productivity May 2011 byjames Manyika, Michael Chui, Brad Brown, Jacques Bughin, Richard Dobbs, Charles Roxburgh, Angela Hung Byers 15
Data Analytics Across Manufacturing Life Cycle DESIGN Supply Chain MANUFACTURING Supply ANALYTICS Chain Post Use Use 16
Impact of Data Analytics Across Life Cycle Report McKinsey Global InstituteBig data: The next frontier for innovation, competition, and productivity May 2011 byjames Manyika, Michael Chui, Brad Brown, Jacques Bughin, Richard Dobbs, Charles Roxburgh, Angela Hung Byers 17
Huge Amount of Stored Data in Manufacturing : But converting Data to information assets is a challenge Report McKinsey Global InstituteBig data: The next frontier for innovation, competition, and productivity May 2011 byjames Manyika, Michael Chui, Brad Brown, Jacques Bughin, Richard Dobbs, Charles Roxburgh, Angela Hung Byers 18
We need to achieve data compression and timeliness across Layers Data Compression Kilobytes/second Processing Decision Business and User Goals Manufacturing Business Intelligence (web, desktop, mobile apps), Dynamic produc@on system, Opera@ons Protocols/standards Real @me/ On @me Years - > Months - > Weeks à Days Filters Gigabytes à Terabytes Filters Petabytes à Exabytes Intelligent data reduc@on Integra@on Rules Engine, Distributed and real- @me compu@ng, Apps, APIs, Web Services Megabytes Protocols/standards Analy@cs Predic@ve Models, Algorithms, Analy@cs engine, Model composi@on, Uncertainty quan@fica@on Protocols/standards Data Structured, mul@- structured, Streaming, DAQ, Data pre- processing *, Descrip@ve analy@cs Closed loop system Timeliness Hoursà Minutes Seconds or less Manufacturing Execu@on System, Manufacturing Opera@ons Management SCADA, PLC, HMI, DCS * Extrac@on, cleaning, annota@on SCADA Supervisory Control and Data Acquisi@on PLC Programmable Logic Controller HMI Human Machine Interface 19
We need standards and protocols Close Loop Description Required Standards and Technology Model (Output Layer) Opt. (Time) Model * Opt. (Cost) Model ** Simulation Model *** Model Classification Model Tuning and validation Connected Post Processing Problem Classification Problem (Analytics Layer) Connected Data (Input Layer) Asset Management Problem a Meta-Data Data 1 Agility Problem b Meta-Data Data 2 Sustainability Problem c Data Analytics (Data Mining) Meta-Data Data 3 Problem Commonalities Machine Learning/Training Function Data Classification Data of Data (meta-data) Data Preprocessing Real Shop Machine A Robot A Machine B Sensor A
Predictive Analytics Workflow Standards and protocols for this information flow Data visualization Data extraction/data stream Raw input s Input validation Data Preprocessing Predictive Model Data Postprocessing Prediction Decision Storage/ Decision Processing Outliers, missing values, invalid values Normalize, Discretize, Filter etc. Scaling, Decision, Scores etc. Standardize the predictive models Model definition Model Composition Model chaining Sender Receiver Standard Interface Standard ensures compatibility Receiver Sender Protocol Standard define both the transmitter and receiver function at the same time.
Big Data Analytics Solution Application Layer * SM: Smart Manufacturing * HDFS: Hadoop Distributed File System * SME: Small & Medium Enterprise * CAPP: Computer Aided Process Planning * FDC: Fault Detection & Classification System * YMS: Yield Management System * SVM: Support Vector Machin CAPP, MES, FDC, YMS, Integration Layer Model Life Cycle Creation à Deployment à In-Use à Disposal Life Cycle Control Duration Control, Uncertainty Resolution Feedback Control Analytics Modeling Layer Statistics Approach Machine Learning Approach R, Neural Network, SVM, Decision Tree, Big Data Infrastructure Layer R Hive, Hadoop, HDFS, MapReduce, Static Data Process Plan (STEP-NC), Production Plan, Master, Data Layer Dynamic Data Monitoring (MTConnect), Metrology, Defect, Manufacturing Process Shop Floor Layer Blue letters: under development 22
Data Analytics Past, Present and Future Data Volume Current DA Reduce the information overload Can we get same level of insights with less data? Relevant and Useful Data Past Present Future 23
Standards SDOs and Industry Consortia Standards activities (good vehicle for industry interactions) ASME ASTM IEEE OMG, MESA, ISA, MTConnect, AMT, AIAG OAGi, SME, ASME, DMG Exploring OPC Foundation, AutomationML 24
Performance Metrics and Standards ASTM E60.13 MESA Metrics SCOR Cloud-based Manufacturing and Supply chain Services TOSCA OAGIS Performa nce Model OMG DSME, QVT, ODL Machine Model M2M OPC UA onem2m CPS for Smart factory Model Repository for Composition Workflow Model OAGIS Process and Behavior Model Cloudbased Analytic s Services B2MML Analytics Model CRIPS- DM PMML PFA V&V Model ASME V&V M2M OPC UA onem2m CPS for Smart factory Optimiza tion Model AMPL GAMS OPL Cloud Infrastructur e for Data Hadoop Hive Resource ISO 15531 Process Plan ISO 10303-240 STEP-NC Macro Micro CPS for Smart Manufacturing System Manufacturing Data Exchange NC Program ISO 6983 Monitoring OPC UA MTConnect PMI ISO 1101, ASME Y14.5, 14.41 Product Data ISO 10303 Usecase Scenario for SMSDA Agility Sustainability Asset Utilization
POTENTIAL COLLABORATION Industry Use case modeling Reference architecture, performance metrics and assurance, modeling methodology, and big data analytics Standards for smart manufacturing use cases Verification and validation Data driven models and uncertainty quantification Participating in standards development activities (OAGi, OMG, DMG, ASTM E60.13, ASME V&V, MTConnect, MESA) Research collaboration With our academic partners 26
Acknowledgements Members of SMSDA Program 27