| Summary: | The induction motor is without doubt the most widely used form of electric power device in any modem power system, giving power to millions of homes, farms and factories all over the world. The early detection of developing machine faults can therefore be vital if the costs of lost production arising from motor failures are to be avoided. Traditional induction machine condition monitoring techniques usually involve the use of sensors embedded in the machine to measure, for example, temperature or vibration. There has also been considerable interest in detecting winding and other machine faults by examination of terminal current waveforms using data gathered under steady-state operating condition and may involve the calculation of quantities such as input power or negative sequence components. Recent trends in condition monitoring also include the detection of machine faults using data acquired during speed transients. A different approach for machine condition monitoring and fault identification using terminal and rotor speed data is presented in this thesis. In this method, a stochastic search is carried out to estimate values of machine parameters which give the best possible match between the performance of the faulty experimental machine and its mathematical model, thus identifying both the location and nature of the winding fault. The performance of three stochastic search methods, Simulated Annealing (SA), Particle Swarm Optimization (PSO) and Bacterial Foraging Optimization (BFO), ABSTRACT when used for condition monitoring of the stator and rotor windings of a three-phase Induction Motor (IM), is evaluated in this thesis. The proposed condition monitoring technique uses time domain terminal voltage and current data in conjunction with the optimization algorithm to indicate the presence of a winding fault and provide information about its nature and location. The technique is demonstrated using experimental data from a laboratory general machine set. In order to verify the proposed fault detection technique, experimental tests were carried out simulating different potential induction motor winding faults under different operation conditions. In this method, stator currents are calculated using an ABC-abc induction motor model and compared to the actual measured currents to produce a set of current errors that are integrated then summed to give an overall error function. Fault identification is carried out by adjusting the model parameters off-line, using the stochastic search method to minimize the error. The new set of model parameters then defines the nature and location of the fault. Unlike many other methods, the new stochastic search based approach does not require any expert prior knowledge of the type of fault or its location; both are identified as an integral part of the optimisation process.
|
|---|
