Bond Graph Modeling Procedures for Fault Detection and Isolation of Complex Flow Processes

Abstract

Many models for fault detection and isolation (FDI) will contain multi-port energy storage and flow-dependencies due to convection. Converting these models into qualitative ones leads to ambiguities that can be eliminated by analysis of boundary conditions. In this study the example of an automotive cooling system is used to demonstrate how these issues can be handled. Furthermore, time constants in the models need to correspond to the bandwidth of the data acquisition system. In order to achieve this, parameters leading to irrelevant time constants can be eliminated by assigning relatively small and large parameters to be zero or infinite respectively. However, this may result in (i) discontinuities and (ii) zero order causal paths ("algebraic loops"). Compensation of qualitative effects instead of reversal is shown to be an adequate solution for these problems.