Multi-Modal Diagnostics
Researcher
- Matthew Schwall
Description
On-board vehicle fault detection and isolation (FDI) has received increased attention as market and safety demands have pushed for improved automotive maintainability and reparability. In conjunction with growing demands for better system performance and reliability, there are also strong constraints on the additional cost of such systems. This dictates that diagnostic systems should require minimal additional hardware such as sensors and computational power, and also have low complexity and development cost.
Model-based methods are the preferred means of achieving these goals, as they use analytical redundancy to reduce costly physical redundancy by comparing the measured behavior of the vehicle with the expected behavior as predicted by a model. The difference between the predicted and the measured value of a variable is called a residual. This research focuses on how multiple residuals can be used to reach an optimal diagnosis.
Our research utilizes a Bayesian network framework that enables the use of multiple system models with varying confidence levels, ontologies and assumption sets. Therefore, models created during the design process can be integrated. In addition, this structure facilitates the inclusion of all available information including the state of the vehicle and customer complaint information. The result is a methodology capable of generating low-cost diagnostic models early in the design cycle. For more information on the framework and motivation for using a Bayesian network structure, see the paper below from the 2001 IMECE.
In order to accurately determine fault probabilities, a hybrid dynamic Bayesian network is currently being used. Details on this system can be found below in the paper presented at the 2002 ACC. In this structure, continuous residuals are used as evidence directly in the network, and this paper discusses options for representing their probability distributions. The Bayesian network is used to model the temporal behavior of the faults, and the assumptions necessary to do this are analyzed. The diagnostic method is demonstrated on a car’s handling system and experimental results are presented.
Publications
Sponsors
» 105 reads