On-Board Fault Identification in an Automobile Fully-Active Suspension System

Abstract

The problem of on-board fault detection and Identification in a broad-bandwidth, hydraulic, fully-active suspension system is considered. The system is large-scale, non-linear, as well as stochastic. The fault detection and identification unit is based on a properly selected partial system model and an appropriate fault detection and identification methodology. Two designs are considered based on the Nearest Neighbor and a novel robust geometric methodology, respectively. Their performance characteristics are assessed via numerical simulations. While the Nearest Neighbor based design is effective with faults of fixed size, only the robust geometric design is shown to be capable of identifying faults of variable size.