The seventh National Conference INCOSEIL 2013 System of

The seventh National Conference INCOSE_IL 2013 System of Systems "Sys. ML and Modelica Integration using FMI – a Case Study " Lev Greenberg (IBM), Alessandro Mignogna (ALES), Peter Aronsson (Wolfram-Mathcore), Daniel Wadler (IAI) 1 Sys. ML and Modelica Integration using FMISystem of Systems

Overview • • • SPRINT System Engineering Simulation Goals Current Practice – Simulation Techniques State of the Art (Example) FMI Approach FMI for Sys. ML FMI for Modelica Demonstration of Model Integration and Simulation Conclusion 2 Sys. ML and Modelica Integration using FMISystem of Systems

SPRINT This work is supported by the SPRINT (Project Number: 257909) EU 7 th framework programme grants. 3 Sys. ML and Modelica Integration using FMISystem of Systems

SE Simulation Goals • Promote cooperation between different engineering domains. • Leverage cyber-physical modeling to validate software by simulation with plant models. • Validate System Specifications and descriptions early and often. 4 Sys. ML and Modelica Integration using FMISystem of Systems

Current practice Simulation techniques Co-simulation Hosted simulation Desyre (hosting tool) -Heterogeneous models -Distributable -Sync/coord overhead -Heterogeneous models -Not distributed -Model export required -High performance Requires proprietary solution! 5 Sys. ML and Modelica Integration using FMISystem of Systems

State of the Art • IBM Rhapsody approaches – Import Simulink block into Sys. ML/Rhapsody – Sys. ML and Simulink integration using S-function Simulink RT workshop models Simulink • Sys. ML model Simulink models Simulink Rhapsody plug-in Sys. ML model S-function generation Simulink model Other vendors also provide similar proprietary solutions Roadblocks to Multi-Tool Integrated Simulation: • • • Current approaches are bound to specific proprietary tools Each tool bring it strengths, all-in-one approach doesn’t fit cyber-physical SE We need a flexible open simulation framework 6 Sys. ML and Modelica Integration using FMISystem of Systems

Functional Mock-up Interface (FMI) Approach Problems / Needs supplier 1 supplier 2 supplier 3 supplier 4 supplier 5 Component development by supplier Integration by OEM Many different simulation tools ? Solution supplier 1 supplier 2 supplier 3 supplier 4 supplier 5 tool 4 tool 5 Reuse of supplier models by OEM: DLL (model import) and/or tool 1 tool 2 tool 3 Tool coupling (co-simulation) Added Value supplier 1 Early validation of design Increased process efficiency and quality OEM FMI Protection of model IP of supplier ! OEM supplier 2 supplier 3 Blocwitz, Otter, et al, retrieved from: https: //trac. fmi-standard. org/export/700/branches/public/docs/Modelica 2011/The_Functional_Mockup_Interface. ppt 7 Sys. ML and Modelica Integration using FMISystem of Systems

FMI - Main Design Idea • FMI for Model Exchange: FMU Tool Solver Model Version 1. 0 released in January 2010 • FMI for Co-Simulation: – Reuses as much as possible from FMI for Model Exchange standard Tool FMU Model Solver – Version 1. 0 released in October 2010 Blocwitz, Otter, et al, retreived from: https: //trac. fmi-standard. org/export/700/branches/public/docs/Modelica 2011/The_Functional_Mockup_Interface. ppt 8 Sys. ML and Modelica Integration using FMISystem of Systems

XML schema (. xsd) defined by the FMI specification Blocwitz, Otter, et al, retreived from: https: //trac. fmi-standard. org/export/700/branches/public/docs/Modelica 2011/The_Functional_Mockup_Interface. ppt 9 Sys. ML and Modelica Integration using FMISystem of Systems

FMI Support in IBM Rhapsody Prototype of FMI plugin for IBM Rhapsody • Features: – Export Sys. ML block to FMI 1. 0 model-exchange – Support both state-chart and activity diagrams behavior FMI plugin steps Sys. ML element FMI element block FMU atomic input flowport scalar input discrete variable atomic output flowport scalar output discrete variable Code generation for Sys. ML block <<FMUParameter>> attribute scalar parameter variable Code generation for FMI wrapper “not annotated” attribute scalar Internal variable … constant attribute initial values start value of scalar variable XML Model description generation DLL compilation Archiving into FMU file • In progress: – Complex types (e. g. , structures) – Dealing with events – Units 10 Sys. ML and Modelica Integration using FMISystem of Systems

FMI Support in Wolfram System. Modeler Will be available in the next release: • Support for export of Modelica models to FMI 1. 0 for model-exchange. – Full support for all mandatory parts of the standard, including state events, time events, dynamic state selection, etc. Future work: • Support for the upcoming FMI 2. 0 standard. • Import of FMI for model-exchange to use FMUs in Modelica models. • Support for export to FMI for co-simulation. 11 Sys. ML and Modelica Integration using FMISystem of Systems

SE Use Case 1. User 1 creates Sys. ML model of overall system 2. User 2 creates UML models of software elements of vehicle and control station 3. User 3 creates modelica models for mechanical elements and control 4. FMUs created for behavioral models 5. Simulation Engineer cofigures Simulation using FMUs 6. Simulation Engineer performs Simulation verifying System behavior 12 Sys. ML and Modelica Integration using FMISystem of Systems

Model Overview 13 Sys. ML and Modelica Integration using FMISystem of Systems

14 Sys. ML and Modelica Integration using FMISystem of Systems

Conclusion 1. Test case performed on prototypes with significant modeling constraints 2. We succeeded in demonstrating multitool simulation using an Open Standard. 3. Open Standards: § § frees the user from "Tool Provider Politics" provides the user with flexibility of Modeling Tools and Simulation Platforms. 15 Sys. ML and Modelica Integration using FMISystem of Systems

Thanks/ תודה 16 Sys. ML and Modelica Integration using FMISystem of Systems