A Graphical Causality based Modelling Approach for Condition Monitoring

The wide use of electromechanical systems in critical applications motivates the need for condition monitoring (CM) of such systems. Signal-based methods require simulating all possible malfunction conditions of a system which in reality not possible. Model-based CM is a promising solution and provides a cost-effective and appropriate approach for simulation of all possible operating conditions and can be used for early fault detection and diagnosis. Many models have been utilized to simulate the behaviour of electric machines; however, most of these simulations are based on abstract numerical models rather than a structured illustration of the system. To overcome these problems, a Bond Graph (BG) model with qualitative simulation has been used in this thesis. BGs are efficient at modelling the dynamics of system behaviour based on physical structure, causality and mathematics.

Qualitative simulation (QS) represents semantic knowledge concerning the performance of a particular system for qualitative reasoning. QS could be applied with minimum knowledge of system variables and even with incomplete system model . Therefore, this study focuses on the investigation of QS procedures to develop a more effective and realistic approach to monitor an industrial machine. Significantly, the study has also developed a qualitative BG fault detection approach based on temporal causal graphs (TCGs), qualitative reasoning and, forward propagation. It mainly describes the dynamics of an AC induction motor (ACIM), which is commonly used in numerous industries. It also used to detect ACIM electrical faults (broken rotor bar and stator winding imbalance that commonly occur in ACIM). Further promotes the diagnostics performance of by introducing an algorithm for ACIM diagnosis, which is based the qualitative influence of the faults on the motor current.

In order to evaluate the proposed QS approach, to enhance the knowledge of the dynamic behaviour of ACIM, a BG model of the AC induction motor has been introduced. The developed BG model can simulate the motor behaviour under different conditions, including a healthy motor, a motor with two broken rotor bar levels (one and two broken bars), and a motor with two different levels of stator winding imbalance. The investigation was based on the motor current spectrum analysis using the FFT signal processing technique.

An experimental study investigates the effects of broken rotor bars and stator imbalance on the motor current. The BG model results and corresponding results from the experimental study have been in good agreement. Moreover, it has also been shown that these outcomes are agreed upon with outcomes reported in the literature.

The TCG and forward propagation results indicated that this approach could be used for the CM of ACIMs. It can detect the effects of broken rotor bars and stator imbalance on the whole system behaviour, showing that this developed QS approach is an efficient technique for extracting diagnostic information, ending up with accurate fault detection using TCG and qualitative reasoning.

The QS technique was validated based on a 20-SIM simulation of the ACIM BG model. The observed results show that a QS approach can accurately detect a broken rotor bar and stator imbalance faults.

The investigation continued by examining the qualitative influence of the seeded electrical faults on the motor current signatures. The results from the experimental study confirm that the BG model and qualitative influence give accurate diagnoses.

Comparison evaluation has been done to compare the graphical causality-based approach with work in the literature. The graphical causality-based approach represents an efficient and meaningful technique for simulating the dynamic system behavior. The diagnostic approach based on TCG is very effective for the detection of ACIM electrical faults. Moreover, it overcomes the limitations of some qualitative studies in the literature.