Generating instead of Programming Diagnostics

The complexity of car subsystems and their interaction is growing. This fact, together with the existence of many variants, enforces a trade-off between the quality and coverage of the respective diagnostics and the efforts spent on their development. What is needed is a comprehensive, general (system-independent) and systematic approach to the generation of diagnostics and its automation. This has to be grounded on an appropriate representation of the underlying knowledge about the respective technologies. Libraries containing behavior models of vehicle components capture an essential portion of this knowledge. Models of car subsystems can be automatically composed from such library elements and provide the basis for the automated generation, instead of programming, of diagnostics. The grounding in an explicit model results in an improvement and guarantees of fault coverage and, hence, safety. As of today, the technology is mature enough for introduction into industrial work processes. We explain the foundations of the model-based solution and demonstrate its applicability using a case study on automated generation of on-board diagnostics for the comfort system of a VW Polo with four control units connected via a CAN-Bus. We argue that the representation of expertise in terms of models makes it available – independently of location, time, and people to computer-based support to various other work processes, such as FMEA, diagnosability analysis, sensor placement, test generation und workshop diagnosis.