Bild 1 - Industrial information and control systems
Transcription
Bild 1 - Industrial information and control systems
Large scale MBSE – Lessons from SAAB Erik Herzog, Ph.D., SAAB Fellow - Systems Engineering 2011-04-04 INCOSE Large Scale MBSE Seminar The domain What I do for a living @ SAAB • SE Process and methodology development/harmonisation – towards ’one SAAB’ @ SAAB Aeronautics • • • • • SE Process and method development and management Process adaptation for large projects Introducing a System architecture and design methodology using SysML Methodology development in research projects Ph.D. student supervisor @ INCOSE • Ordförande INCOSE Sverige @ Home • • • Family management Golf Innebandy The world we live in Complex technologies Integrated into complex products Life cycle issues Change is constant Developed in complex organisations The challenge We need to do more with less resources while improving product quality How to optimise the SE capability? The paradox Whenever we add a new mathod/tool - the stengths of the tool will be used out of proportion When DOORS was introduced at SAAB traceability links was created in abundance Over use of tool capabilities may actually decrease the overall performance of an organisation Adding tool support must reduce overall workload! If we add something new then we have to stop doing something else • We can’t just stack up new methodologies, the old ones have to be adapted or retired Where we were Our ’old’ development process Stakeholder Requirements Definition System Requirements Analysis Define Functional Architecture Typ JAS System Documentation Define Physical Architecture PS PS HU HU PS PS CZ CZ PDD PDD PDD PDD PDD PDD Trace links SSS SSS PDD PDD PDD PDD SSDD SSDD PDD PDD PDD PDD DOORS SRS SRS PDD PDD PDD PDD Kod Kod PDD PDD PDD PDD EqSp EqSp PDD PDD PDD PDD Integrate Sequential process Document driven approach • PS PS SE SE Implement SDD SDD PDD PDD PDD PDD PS PS J2 J2 PS PS SA SA Complete one document prior to initiating work on the next one Fixed document structure Much focus on requirements (requirements in design documentation) Traceability is costly to implement and maintain Insufficient attention on describing what the system looks like (in a reader friendly format) Where we want to go 1 (2) Development process that emphasise parallel aspect of work Where we want to go 2 (2) Clear definition of system design System components Interfaces MBSE Concept of execution Specification Specification PDD PDD PDD PDD Design Design PDD PDD PDD PDD Trace links Documentation structure basedon system complexity WS • Specification Specification PDD PDD PDD PDD A/C Design Design PDD PDD PDD PDD Specification Specification PDD PDD PDD PDD Design Design PDD PDD PDD PDD MSS … Test Test PDD PDD PDD PDD NAV EWS Modell Modell … Dynamic depth Early focus on integration, verification, transit to use and validation System – Subsystem – Capabilitiy lifecycles WS level WS/AC concept design Equipment A Equipment and subsystems Equipment concept Equipment B Capability A Capability level Product design, realisation & integration Equipment verification Equipment Integration Equipment Acquisition Equipment concept Equipment Acquisition Equipment verification Equipment Integration Capability realisation Capability Capability concept design detailed design & integration Capability B Capability verification Capability realisation Capability Capability concept design detailed design & integration Capability C Overall capability definition set (first baseline) Capability verification Capability realisation Capability Capability concept design detailed design & integration Capability D Constraints implied by equipment A fixed Verification Capability verification Capability realisation Capability Capability concept design detailed design integration Constraints implied by equipment B fixed Capability verification SAAB MBSE vision MBSE validation validation in traditional approach Scoping the presentation Models (examples) Requirements Prototypemodel Black Box Designmodel Y=x+1 Realisationmodel X-2 Y+2 2x+5 transform Testmodel Realisation 10010100101001 00101010101001 0101010 Requirements containing model fragments expressing required system properties. Prototype model potentially executable, meeting high level functional requirements. May be used for code generation. Design model: Architectural and functional design at a suitable level of abstraction. Realisation model potentially executable, meeting requirements, structure corresponds to the realisation. Realisation executable object code (for verification in target system) Test model realises requirements based testing. May be used to verify the prototype and realisation models as well as the realisation. Model development intensity vs lifecycle Realisation models Design models Prototype models Capability realisation Capability Capability concept design detailed design & integration Capability verification Domain models vs system models Domain models typically provide the following benefits • • • • Rapid prototyping Early validation Code generation • Production code • Simulation Communication System models • • • • Communication Early validation Integration Verification Some experience from SAAB Where are we MBSE-wise? Skeldar System architecture in UML/SysML. Software generated from Systembuild and Rhapsody models Gripen NG (C/D) System architecture in SysML. Software generated from Simulink and Rhapsody/Bridgepoint models. User interface models generated using VAPS Physical models in Dymola MBSE challenges Integration across tool boundaries Training of third parties PDM system integration Variant management The legacy of the document paradigm Obsolescence Training of third parties The readers of our models/deliverables must understand what we are trying to convey • • We must limit our usage of the modelling language And present it in a reader friendly way Information must be in a format that is natural to interpret for the receiver • Once basic model semantics is understood Training must be adapted to each stakeholder Minimised SysML set Accept that some information will not be SysML formatted Integration across tool boundaries It migh be possible to integrate all tools in the development tool chain • • • Links Data exchange … Tight integration limits flexibility And is risky/costly over time The challenge is to find the right balance Information management Face it, our support systems are made for managing documents • • CM PDM The document paradigm is excellent for providing a coherent view on a design Sign-off routines work for documents • How to do it on models? Not all design information is suitable for SysML-ification • Keep traditional document format where appropriate Not all information in the model is intended for publication A good document generator (and model package structure) is a prerequisite for MBSE acceptance Model structure for documentation SSDD outline (MIL-STD-498) Document template 1. 1.1 1.2 1.3 SCOPE Identification System overview Document overview Package ”Scope” • Controlled files 1 1..* Package ”System-wide design decisions” • Controlled files 1 1..* 2 REFERENCED DOCUMENTS 3 SYSTEM-WIDE DESIGN DECISIONS 4 4.1 4.2 4.3 SYSTEM ARCHITECTURAL DESIGN System components Concept of execution Interface design Package ”System components” • Controlled files • Block Definition Diagrams • Picture + Description • Controlled files • Blocks • Name + Description • Tags 1 0..* 1..* 1..* 0..* 1..* 1..* 0..* 5 REQUIREMENTS TRACEABILITY 6 6.1 NOTES Terms and abbreviations Package ”Concept of Execution” • Controlled files • Use Case Diagrams • Use cases • Name + Description • Controlled files • Statechart Diagrams • Name + Picture + Description • Controlled files • Activity Diagrams • Name + Picture + Description • Controlled files 1 0..* 1..* 1..* 1..* 0..* 0..* 0..* 0..* 0..* 0..* 0..* Package ”Interface Design” • Internal Block Diagrams • Picture + Description • Controlled files • Parts • Picture + Description • Controlled files 1 1..* 1..* 0..* 1..* 1..* 0..* Package ”Notes” • Controlled files 1 Variant management and incremental development It is not one product, one release any more The system model need to support a baseline for subsystem development While the system model itself is evolving And doing so for multiple variants/configuration of the system Obsolescence Tool landscape is changing very fast No matter which tool you select there will soon be others that appear more competitive Tool providers are small and notorious for not meeting promises in their development plans What to do? • • Make an educated guess Consider the exit strategy up-front Steps towards MBSE One process that clearly communicates a common way of working • If different parts of the organisation have conflicting views on how work shall be performed then we will be forced to introduce MBSE many times. Clearly communicate what the processes shall deliver and in which format • Same approach for all projects applying MBSERestriktioner Don’t use all possibilities in the language! Training • Capture how you want the engineers to work, not multiple alternatives ways describing how they could potentially work Pilots and Mentors • Engage the engaged first and ensure that there is adequate support available for the engineers that apply the new methodology Think about documentation up front • • • All elements in the database need not end up in all formal documentation Elements that will not be used in any formal documentation should not be entered in the database Use MBSE where its strengths add value, not for everything just because it is possible What is next? Usage of architecture models for integration domain design and simulation models Real parametrics Capture of requirements, and requirement integration with models • Focus on the essential requirements Introduction of industrial strength formal methods