Date of Award
8-2011
Document Type
Thesis
Degree Name
Master of Science (MS)
Legacy Department
Mechanical Engineering
Committee Chair/Advisor
Pisu, Pierluigi
Committee Member
Vahidi , Ardalan
Committee Member
Miller , Richard
Abstract
Recent advancement in the field of automotive industry is largely due to the addition of electrical and electronic equipment; some of the safety features in the vehicle and advanced driver assistance systems are examples of such equipment. Furthermore, safe operation of the vehicle is highly dependent on the electrical power generation and storage system (EPGS). Therefore, to ensure optimal operation of this system, a reliable diagnosis of the system is essential. However, as the complexity of the electrical systems has increased, the identification of a malfunction has become an increasingly difficult task to handle.
In the current work, a model-based diagnostic approach for the EPGS system is formulated using the residual generation and adaptive threshold method. The EPGS system comprises an alternator and a battery. Since the focus of the current work in on the vehicle alternator subsystem of the EPGS system, a mathematical model of the alternator subsystem based on the physics of the processes involved is derived. This model is characterized by time-varying nonlinear ordinary differential equations. To simplify the diagnosis scheme development, an equivalent linear time invariant model based on the behavior of the input/output of the alternator is presented. Afterwards, three typical faults for a vehicle alternator, namely belt slipping fault, open diode fault and voltage regulator fault, are modeled and injected into the model separately to observe the effectiveness of the adaptive threshold-based fault diagnosis scheme for fault detection and isolation (FDI). The proposed adaptive threshold scheme for the EPGS system has proven to be more sensitive and more robust than previously presented diagnostic schemes for the same system as available in the literature.
In addition to the classical adaptive threshold method, a novel general methodology is presented for the derivation of adaptive thresholds in the case of linear time varying-parameter systems, and Gaussian distributed linear parameter systems. The high order of the threshold dynamics in general is the main drawback of this approach. To overcome this problem, order reduction methods can be used. In this thesis, we explore two approximations, namely the steady state threshold and a first order threshold approximation. The study shows that these approximations are effective in detection and isolation of faults, however, a false alarm rate is introduced.
Moreover, the qualitative modeling of the equivalent system via stochastic automaton is also investigated, and a new approach for the evaluation of the transition probabilities based on the Divergence Theorem is proposed.
Recommended Citation
Hashemi, Ali, "MODEL-BASED SYSTEM FAULT DIAGNOSIS UTILIZING ADAPTIVE THRESHOLD WITH APPLICATION TO AUTOMOTIVE ELECTRICAL SYSTEMS" (2011). All Theses. 1168.
https://open.clemson.edu/all_theses/1168