Model-based Diagnosis Tutorial PHM-E 12

Transcription

1 Model-based Diagnosis Tutorial PHM-E 12 Peter Struss Tech. Univ. of Munich Univ. College Cork OCC M Software GmbH struss@in.tum.de - 1

2 Outline 1 Introduction: Model-based Systems 2 Component-oriented (Qualitative) Modeling 3 Consistency-based Diagnosis 4 (4 Test Generation) 5 Other Tasks during the Product Lifecycle 6 Summary and Perspectives - 2

3 Knowledge-based Systems Model-based systems are knowledge-based systems are not simply Systems based on knowledge Problem solver but Systems grounding their solution on a knowledge base Knowledge base - 3

4 Knowledge Base and Problem Solver Provlem solver: A usually task-specific, possibly domain-independent algorithm which can process the represented knowledge Problem solver Knowledge base: an explicit declarative formal representation of knoweldge about a certain domain and/or class of tasks Knowledge base - 4

5 Knowledge Base and Problem Solver : What and How Separate: How (a certain class of problems is solved) Problem solver What (is the individual problem to be solved) Knowledge base Advantages: Transparent, maintanable, extensible solution Re-use of (parts of) knowledge base and problem solver - 5

6 For instance: diagnosis Observations: When braking with ABS, car is yawing to the right, and brake pedal feels harder than normally Yawing : - under-braking at left side - over-braking at right side hydraulic unit harder brake pedal front left wheel rear right wheel underbraked overbraked - 6

7 Diagnosis: What and How Knowledge about the subject How is it structured? How does it work? Knowledge about - Structure - Componenten behavior hydraulic unit harder brake pedal Diagnosis Algorithm From knowledge about the subject and observations of the system behavior infer diagnosis hypotheses underbraked front left wheel rear right wheel overbraked - 7

8 Diagnosis? OBS Task: Determine, based on a set of observations: What`s going on in the system? - 8

9 Model-based Diagnosis MODEL? OBS Task: Determine system models that are consistent with the observations - 9

10 Iterative Diagnosis Model Revision MODEL OBS Task: Incremental modification of fault hypotheses Initial hypothesis: correctly designed device - 10

11 Key Idea: Generation of Diagnosis Systems Library: Component Models Generic Diagnosis Test Generation Generic Engine FMEA Engine Engine User: specifies structure only Diagnosis system is generated automatically Component Behavior Structure System Model The same for FMEA, Testing,... CAD Data Specific Diagnosis Specific Specific Test System FMEA Generation Generation - 11

12 Generic Diagnosis Algorithm?! - Simple Example Models explicit conceptual and structural model Behavior models Compositional ( context-free ) (possibly) qualitative models cmd p sensors Mech. pump drive f - 12

13 Simple Example Diagnostic Reasoning Inconsistent partial models: conflict combined evidence from several conflicts Pump or pressure sensor or container defect OR mechanical drive and flow sensor logical theory: consistency-based diagnosis - 13

14 Demonstrator: Turbo Charger System On-board detection and localization of faults related to black smoke under realistic conditions (e.g. sensors) with model-based techniques from Artificial Intelligence - 14

15 Project Vehicle Model Based Diagnosis (2/97-1/99) Università degli Studi di Torino - 15

16 Demonstrator Vehicle Set-Up MAC 2 serial line ETK ECU VS100 RAZ R - 16

17 Demonstrator Car (Volvo) with RAZ R - 17

18 Demonstrator Car (Volvo) Switchboard for Fault Injection - 18

19 Demonstrator Turbo Control Subsystem Scenario 1 Leakage Scenario 3 Boost pressure Scenario 2 Air flow sensor - 19

20 VMBD Demonstrator: Leakage in Air Intake - 20