Dissertation Defense: Model Based Fault Diagnosis For Automotive Functional Safety

Jiyu Zhang, PhD Candidate, Mechanical Engineering

All dates for this event occur in the past.

Center for Automotive Research Room 198
Center for Automotive Research Room 198
930 Kinnear Road
Columbus, OH 43212
United States

Committee Members

  • Professor Giorgio Rizzoni, Chair
  • Professor Vadim Utkin
  • Assistant Professor Vishnu Sundaresan
  • Assistant Professor David Hoelzle
  • Professor Bilin Aksun-Guvenc

Abstract

Functional safety is an important element for future automobile development. To ensure functional safety , the automotive industry has issued a functional safety standard - ISO26262 to standardize the functional safety requirements during different phases of a automobile's safety lifecycle. This dissertation proposes a model based approach for achieving automotive functional safety through model-based diagnosis. This approach begins with functional safety and diagnostic requirements definition through hazard analysis and risk assessment, and works through a design of a diagnostic strategy that leads to implementation of the algorithms to satisfy the functional safety goals. In particular, this dissertation introduces a systematic diagnostic strategy based on structural analysis to detect and isolate various faults in a complex system. The advantage of using structural analysis approach for FDI is that it can efficiently analyze the analytic redundancy of a system as well as systematically designing structured residual generators to satisfy the diagnostic requirements. This dissertation show the effectiveness of this approach in providing diagnostic solutions by presenting an application of the permanent magnet synchronous machine drive system in electrified vehicles, considering sensor fault detection and isolation. The diagnostic strategy is proven to be effective in detecting and isolating various sensor faults in a PMSM drive system. The design of this diagnostic approach can be integrated into the functional safety lifecycle to provide solutions to functional safety problems. This dissertation illustrates this concept by showing a case study of a safety case - the torque functional safety of pedal-by-wire systems. The case study starts with investigating the problem by hazard analysis and risk assessment. And then it uses fault modeling method to conduct an quantitative analysis on the effect of the faults. The fault modeling method can be used to assist hazard analysis and risk assessment in defining the risk level of each potential hazard, so as to help define the functional safety requirements. Then, the structural FDI approach is applied to analyze the diagnosability of various faults in a pedal-by-wire system and systematically design diagnostic tests toe detect and isolate pedal mechanical stiction fault and pedal sensor faults. Then some fault mitigation strategies are designed to mitigate the effect of these faults on torque functional safety. The structural analysis approach systematically generates candidate equation sets for designing residual generators, whose number may be considerable in a complex system. This dissertation proposes a novel approach for selecting residual generators to downsize the solution sets, considering feasibility of residual generators, diagnosability requirements, computation complexity while computing a residual as well as sensitivity and robustness of various residual generators. Based on these criteria, the optimal diagnostic test could be extracted from a large number of candidate equation sets to achieve the most desirable performance.