stator turn-to-turn fault estimation of induction motor by using probabilistic neural network

Induction machines are extensively used in industry due to the wide demand and diverse applications. managing dealing with various faults, accurately detecting the fault and its severity as one of the biggest challenges will have a significant impact on the induction machine health and the quality of system operation. ignoring the faults will cause irreparable damage to the electrical machine and then to the industrial complex. knowing about exact fault conditions is the most basic issue in dealing with fault management. in this paper, turn to turn fault as one of the major problems of induction machines is discussed. for this purpose first, the fault is evaluated by negative sequences current, and second, a mechanism is used to distinguish between the source imbalance fault and the turn-to-turn fault. with the help of the information obtained from the faulty machine and two layers of the probabilistic neural network, the number of the turn-to-turn fault will be estimated. the simulation was performed under normal conditions as well as under fault conditions for a specified number of turn-to-turn faults. this method is tested for non-training data with different common ranges and a number of turn-to-turn faults. neural network output results are compared with the simulation in matlab, which shows the correct training and high accuracy of the proposed method to detect the number of stator faults.