چكيده به لاتين
Considering the key role of induction motors in different fields of industry, designing a
comprehensive strategy for repair and maintenance of motors and early diagnosis of faults in order to prevent long term delay and reducing the cost of repair and maintenance would be scientifically and economically feasible. Intelligent techniques, among other techniques have proven special abilities. In this thesis, in order to increase the reliability of detection of faults such as outer ring of bearings, rotor shaft fracture, short circuit in stator winding, a method using neural data fusion is proposed. Due to the fact that neural network is a feature level fusion method, in thesis a feature level fusion has been utilize. After collecting data, different features are being extracted from three phase current and voltage signals. the extracted features are being fed to the feature selection unit which is comprised of a decision tree. When the efficient features of each phase are being selected, they are being sent to the neural system so thet the exsitance and type of the faults are being detected.
In order to increase the reliability of detection of the above mentioned faults, two types of fusion are being proposed. The first method fuses the features of the three phase current signals and the other method fuses the features of the three phase current and voltage signals simultaneously. The performance of the proposed fault detection system has been studied using an induction motor coupled with an Electro-pump under normal condition, in the pretense of unbalanced disturbance of power suplay and pump dryruning. In order to detect the phase which the short circuit fault has been occurred, the fusion of the feathers of current signals are being used. The results indicate that the reliability of fault diagnosis and detection of the faulty phase is improved.
