Induction motors are the most common motors for converting electrical energy to mechanical energy . 60% of produced energy is consumed by Induction motors. They have the major industrial applications, and therefore the maintenance of induction motors is so prominent. Fault diagnosis contributes to maintenance programming and reducing of plant shutdowns. Motor-current-signature analysi (MCSA) is one of the most applicable methods for fault detection and isolation of induction motors. Signal processing techniques are presented to analyse stator current. In recent years discrete wavelet transform is mainly employed to detect rotor faults, while this transform has some limitations such as: aliasing, oscillations, and shift variance. These shortcomings can have some bad impacts on fault detection and identification. For example, there may be a fault but discrete wavelet transform cannot detect it or perhaps produces fault alarms. The dual-tree complex wavelet transform is a recent enhancement which solves the discrete wavelet shortcomings. This transform has some important properties such as nearly shift invariance and aliasing reduction. Dual-tree complex wavelet transform is naturally parallelized for efficient hardware implementation. A new technique based on dual-tree complex wavelet transform is proposed in this thesis. The procedure has the following steps: current, envelope and speed is sampled from electrical motor model based on winding function method, signal is decomposed to approximation and detail coefficients then useful features are selected in time and frequency domain. Broken rotor bars and broken end ring are identified in different loads. Noise, temperature fluctuations and voltage changes are considered in signal sampling. Artificial neural networks and support vector machines are used to monitor the condition of motor. Finally, key words Dual-Tree Complex Wavelet Transform, Motor-Current-Signature Analysis, Broken Rotor Bars, Fault Diagnosis , ltr"
