Improvement in Fault Tolerant Capability of ST-DTC for Five-Phase Induction Motor using Neural Network

The performance of switching table-based Direct Torque Control (ST-DTC) depends on number and type of switching states. If there are greater number of switching states and are distributed uniformly in the space, then DTC can handle not only different types of loads but also it can be operated in smoother way during high and low speed operation. When DTC is considered for fault tolerant drive, there are uneven distributions of switching states. In this context, higher level inverter can be preferable as it gives greater number of switching states which are distributed nearly uniformly in the space. The artificial neural network (ANN)-based DTC has the capability to handle such situation in better way if the training data are properly prepared. In this paper, the improvement of fault tolerant capability of ST-DTC with three-level inverter (3-LI) and ANN-based DTC for a five-phase induction motor (5PIM) with one phase open (phase ‘a’) are compared. The result shows that the use of ANN for fault tolerant DTC reduces the torque and current ripple by 3% and 3.36% respectively. The 5PIM 3-LI gives an opportunity to use five-level torque comparator to handle transient and steady-state load separately. Moreover, with ANN-based DTC, the torque and current ripples are further reduced.