On-line sensor fault detection, isolation, and accommodation in automotive engines

Abstract

This paper describes the hybrid solution, based on artificial neural networks (ANNs), and the production rule adopted in the realization of an instrument fault detection, isolation, and accommodation scheme for automotive applications. Details on ANN architectures and training are given together with diagnostic and dynamic performance of the scheme

Accelerated Artificial Neural Networks on FPGA for Fault Detection in Automotive Systems

“…The authors describe an offline software-based ANN for detecting sensor faults in engines. In [13], the authors present an evaluation of the Instrument Fault-Detection, Isolation, and Accommodation (IFDIA) scheme and present a proof-of-concept scheme on a DSP platform. Though their results show improved sensitivity to faults, software execution presents a bottleneck for larger ANN networks, since parallelism is not exploited.…”

Precedent-Free Fault Isolation in a Diesel Engine EGR Valve System

“…Franchek et al [19] employed a nonlinear model identified by a process called system probing which used the system frequency response to identify significant regressors. Some researchers [20,21] have successfully applied dynamic neural networks (NNs) to detect air path faults while others [5,22] used physical models to model the dominant phenomena and NNs to model secondary, highly nonlinear phenomena.…”

“…Many researchers have utilized detailed physicsbased models (see [9,10,23,31], for example), but such models require extensive a priori knowledge and their simplicity often limits their detection of more complex faults. For this reason, other researchers have pursued semi-physical [5] or datadriven approaches [1,19,20] with varying degrees of accuracy and computational burden. The GSMMS approach discussed in this section is a dynamic, data-driven modeling approach and will underlie the EGR FDD strategy presented later.…”

Research and Application of Failure Detection Method Based on Closed-loop Systems

“…The goal was to generate several symptoms indicating the difference between nominal and faulty status. On-line sensor fault detection, isolation and accommodation in automotive engines have been studied by Capriglione, [5][6][7]. Their paper described the hybrid solution, based on artificial neural networks (ANN), but their methods used ANN just as classifiers rather than dynamic models.…”