QUANTITATIVE BOND GRAPH APPROACH FOR TROUBLESHOOTING ELECTRICAL SYSTEMS

Abstract

A fault occurs in a system when actual behavior of the system is different from its expected behavior. Detection of faults has been a challenging task from the very beginning. With the evolution of science, many miracles came into being in the field of engineering. Especially fields like electronics that contain discrete electrical components like resistors, inductor, capacitors and diodes etc laid foundation for the modern developed complex mechatronics systems. These small electrical components got so much importance because these are the building blocks of huge and complex systems like jets, control systems, industrial boilers and vehicles etc. Fault detection and isolation (FDI), a subfield of control engineering, concerns with monitoring of a system, i-e fault detection, identifying the type of fault and isolating the detected fault. A single component failure in electrical circuit may lead to the collapse of entire system. It is desirable to have timely fault detection, so that the fault can be dealt with to minimize further damage. The knowledge of fault severity is of great importance while dealing with the fault, so the estimated value of faulty parameter is needed as well. Manual decisions are often time consuming and costly so a fault diagnosis system should be automated to provide the fault detection alarm as soon as possible. The automated system should be capable of detecting the fault, identifying the faulty component and the estimate of faulty value. In this work fault detection and isolation of some electrical systems containing components like resistors, inductors, capacitors, diodes and transistors is performed using Bond Graph modeling approach, with the notion of Analytical Redundancy Relations (ARRs). Bond graph is a modeling technique used for modeling of the complex systems. ARRs are constraints of the system’s model expressed using known variables (sensors readings, known inputs, parameters etc). These relations are called redundancy relations because the behavior of these relations is observed over time for checking the consistency of the system. Bond graph modeling when used with the concept of ARRs is known as Quantitative Bond Graph approach. Fault isolation is done using the fault signatures of the parameters from fault signature matrix (FSM). And a parameter estimation technique based on the ARRs is applied on the fault candidate to obtain its estimated value. Parameter estimation is done using non linear least squares.