In modern vehicles the architecture of electronics is growing more and more complex because both the number of electronic functions – e.g. implemented as software modules – as well as the level of networking between electronic control units (ECUs) is steadily increasing. This complexity leads to greater propagation of failure symptoms, and diagnosing the causes of failure becomes a new challenge.
Diagnostics aims at detecting failures such as defect sensors or faulty communication messages. It is subdivided into diagnosis algorithms on an ECU and algorithms running offboard, e.g. on a diagnostic tester. These algorithms have to complement each other in the best possible way. While in the past the diagnosis algorithm was developed late in the development process, nowadays there are efforts to start the development of such algorithms earlier – at least in parallel to developing a new feature itself. This would allow developers to verify the diagnosis algorithms in early design stages.
The following questions often arise when diagnosis algorithms are implemented: Does the developed diagnosis algorithm detect all the relevant failures correctly? Are the expected diagnostic trouble codes stored in the ECU’s error manager and is the expected substitute reaction of the system correct? Which part of the diagnosis algorithms is to run onboard and which part is to run offboard? If such questions are answered early in the development process, the function developers can be given feedback and design flaws can be detected.
This paper describes a method for verifying and testing the diagnosis algorithms by means of a PC-based simulation. A model of the system is simulated with failures such as faulty signal values. Setting up such a simulation on a PC – instead of waiting for the production hardware – enables each function and diagnostics developer to verify their implementations earlier and more easily.
As an example, this paper describes an intelligent light system including diagnosis algorithms at Daimler AG that was modeled in dSPACE’s TargetLink and SystemDesk. Failures are inserted into the system, which is then simulated.
Thus, this paper shows a possible method to verify diagnosis algorithms of electronic systems in early design stages and to use this knowledge for the design process.