Master Thesis - Past Projects - Abstract

Fault handling architecture in embedded automotive systems

ID: LITH-IDA-EX--07/065--SE

An embedded system is a computer system designed to perform a dedicated function, in our case to control a heavy duty diesel engine. It is important that such a system, especially when implemented in an automotive application, is dependable. This means that the system shall be robust against software and hardware related faults, since faults may cause the vehicle to stop non deliberately or even worse, crash. To make a system robust against faults it is common to use Fault Detection and Isolation (FDI) combined with Fault Tolerant Control (FTC). This means that faulty components in the system are detected by the embedded computer system and the computer system contains logic to isolate the faulty behviour of the components. In addition to this FDI part, the computer system also contains FTC software to control the hardware system in the presence of faults i.e. mask the faults to the user of the vehicle. FTC often includes performing accommodation actions to change the operating mode of the system from nominal mode to some fault-tolerant operating mode. In addition to performing accommodation actions to mask faults to the user it is also important to avoid software related problems when a fault prevents the computer system from operating in normal mode. In this thesis a fault handling architecture for an Engine Management System (EMS) to a Scania diesel engine is presented. The EMS software is partitioned into two parts, application and central service software. To prevent software architectural problems from occurring a new architecture for the application is presented, this architecture partitions the application into three layers; raw data, derivation and control. Architectural rules are implemented to restrict the development of each layer. Moreover Scanias present method to propagate fault status information throughout the EMS is evaluated and improvements are suggested. This thesis also includes a decision machinery to choose optimal accommodation actions in case of the occurrence of faults in the system. This decision machinery is implemented in the central service software. Finally a case study is performed where an EMS is built with the architecture and the decision machinery developed in this thesis. This EMS is used to control a model of a diesel engine implemented in Matlab Simulink.

Keywords: Fault Handling, Bayesian Decision Theory, Engine Control

Author(s): Fredrik Johansson

Contact: Simin Nadjm-Tehrani