System Architecture Virtual Integration (SAVI) Presentation to PDT Europe 2016

Transcription

1 System Architecture Virtual Integration (SAVI) to PDT Europe 2016 Greg Pollari, Rockwell Collins Nigel Shaw, Eurostep Limited

2 Agenda SAVI The problem SAVI The constium Two examples Conclusions Looking fward 2

3 Many systems integrated into one aircraft System complexity increasing Shared resources Complex interfaces Req / design errs 3

4 The impact of requirement/design errs is documented High-level Design RFP Response System Integration Checks High-level Req s in RFP PDR Req s Changes CDR Target Completion Trades Req s Defined Sys Design Detailed Design Sys Re-Design Sys Development Sys Integration V&V 70% errs 3.5% detected 1x cost Sources: NIST Planning rept 02-3, The Economic Impacts of Inadequate Infrastructure f Software Testing, May D. Galin, Software Quality Assurance: From They to Implementation, Pearson/Addison-Wesley (2004) B.W. Boehm, Software Engineering Economics, Prentice Hall (1981) INCOSE Systems Engineering Handbook, Version 3.2.2, % errs 80% detected x cost x (INCOSE 2011) COST GROWTH Slide source: D. Redman, GPDIS 2015 SCHEDULE DELAY 4

5 SAVI Approach System Architecture Virtual Integration Leverage MBSE best practices and tools SAVI developed with exemplar toolset seek to define tool characteristics, but not specific tool selection Reduce costs/development time through early and continuous model-based virtual integration Inter-domain and inter-model consistency checks Protect Intellectual Property (IP) Suppt definition/capture of incremental evidence f system safety analysis suppting certification approach Consistency checking of constituent models participating in integration is critical element of the SAVI concept 5

6 SAVI Virtual Integration V ision Requirements Engineering Predictive Sensitivity analysis f uncertainty Top-Level Verification Items Validated Confidence in implementation Acceptance Test System Design High-level ADL Keeping the system continuously integrated! System Test Hardware Architectural Design Software Architectural Design Detailed ADL Integration Test SW Int. Test HW Int. Test Component Hardware Design Component Software Design Specify - Code Interfaces Software Unit Test Hardware Unit Test Slide source: D. Redman, GPDIS 2015 Hardware Development Software Development generation of test cases updating models with actual data 6

7 SAVI Participants Full Members Airbus Boeing DoD Embraer GE Aviation Honeywell Rockwell Collins Siksky Liaison Members FAA NASA SEI Tool Vend Partners Adventium Labs Ansys (Esterel Technologies) Eurostep Limited 7

8 Two example challenges The specific case: to test consistency f a printed circuit card assembly The generic case: to compare models of the same related systems in different languages 13 separate sources Geometry: MCAD ECAD Excel Connects Logical Excel Signals SysML AADL Simulink The Sliding Mass Example System ica Both cases fit within a single Reposity and Exchange/Sharing capability

9 Root cause of the problem: different & Enterprises Languages & Values. Foundation process Extract content into a common view Manage model meta data and extracted content Process extracted data and apply checks Present extracted data and check results Languages Selection ling s Reposity Discovery & 9

10 Languages Selection ling s Reposity Discovery & Specific case find specific elements such as pad positions on circuit cards Generic case extract into the model of models 10

11 Reposity Languages s ling Selection Reposity Discovery & SAVI have created a specification f a Reposity and Layer Key issues are: Enabling access to extracted data while controlling access to the source models Allowing f cross enterprise sharing of models As an exemplar Eurostep has used ShareAspace to provide this functionality 11

12 Virtual Integration Languages s ling Selection Reposity Discovery & In the specific case we can bring together data extracted from all 13 source files to create a virtual integration that suppts testing geometric and logical consistency Do pads and connect positions match? Are the circuit board shapes consistent across ECAD and MCAD? Are the signals on the boards consistent with the interconnect tables and do they match between boards? The maj challenge in this process was to be sure how the different geometric spaces relate across MCAD, ECAD and connect definitions STEP standard expts used to enable the ECAD/MCAD comparison 12

13 Virtual Integration Languages Selection ling s Reposity Discovery & 13

14 Virtual Integration Languages Selection ling s Reposity Discovery & Tree view of extracted data sets allowing sources to be shown hidden Geometric view with all data impted into the MCAD assembly space 14

15 Consistency tests 1 Languages s ling Selection Reposity Discovery & comparing points between MCAD and ECAD profiles on the same board... checking board : MCAD : S checking board: ECAD : Both the MCAD & ECAD are the same sized shape... checking board : MCAD : S checking board: ECAD : Both the MCAD & ECAD are the same sized shape... checking board : MCAD : S checking board: ECAD : Both the MCAD & ECAD are the same shape but different sizes Total difference in X : 0 Total difference in Y : Q1: How well do the MCAD and ECAD boards match up?` 15

16 Consistency tests 2 Languages s ling Selection Reposity Discovery & Signals across boards and Publish/Subscribe table match Close position of Mating Markers shows alignment in plane of board Signals not consistent Q2: How well do connects align? Q3: Are the signals consistent? Hole positioning coming from MCAD 16

17 Discovery of equivalence Languages s ling Selection Reposity Discovery & In the generic case, need to identify where things should be consistent, then test if they are consistent Have all models in a single fm, i.e. the model of models Enables Application of a rule engine to find equivalences The user to identify equivalences and look f consistency Edit results from the rule engine Identify patterns that should match across models Apply rules to determine consistency SysML Simulink Use fuzzy comparison due to differences in names and conventions AADL ica 17

18 of s viewer Languages Selection ling s Reposity Discovery & Rule set applied to identify matching names across the four models 18

19 of s viewer Languages Selection ling s Reposity Discovery & Pick one proposed set to examine 19

20 of s viewer Languages Selection ling s Reposity Discovery & View its content 20

21 of s viewer Languages Selection ling s Reposity Discovery & Edit content of the discovered equivalent set Fuzzy matching has identified four occurrences of three different names used across models Plus one false match 21

22 of s viewer Languages Selection ling s Reposity Discovery & Set now only contains good matches Now can check associated properties, structures, etc. 22

23 Production scale models These are the same two pairs of models with different relationships and rules applied. 23

24 Conclusions The aim is to allow discovery of consistency issues much earlier than physical test In the general case this is a challenging problem Making progress on establishing both process and mechanisms The approach is feasible f specific domains Through the use of standard fmats, can resolve the spatial relationships and so perfm virtual integration and check consistency of integration SAVI is wking on: A Virtual Integration Process Methods f considering emergent model behaviour as well as static tests Specific capabilities such as safety and security across models 24

25 s capture our understanding of system and its components Separate models express decomposition and viewpoints Only by assuring ourselves that our models are consistent can we have confidence that subsequent analyses and their results can be trusted! How do we assure that our individual views and composed (virtual) perspectives are looking at one system? 25

26 As Based Systems Engineering becomes the nmal way of business, it will be even me imptant to minimise the risk due to inconsistency between models, both within and across enterprises Looking fward This problem is not going to go away! Website: savi.avsi.aero 26