In order to improve fault diagnosis rate of neural network on Hall sensors

an improved neural network model is designed.This model is composed of back propagation（BP） neural networks in parallel.The output of each BP neural networks is weighted and integrated by the algorithm to obtain a smaller error output.As a result

the improved neural network model is applied to the Hall sensor fault diagnosis system of the brushless DC motor（BLDCM）

and the Hall sensor fault tolerance processing is realized by using a position sensorless system.The simulation results show that the accuracy of improved neural network fault diagnosis is 96 %

which is higher than that of traditional BP neural network.The fault-tolerant control system can significantly reduce the impact of Hall sensor fault on motor rotating speed

so that the motor can be assured to stably operate when Hall sensor is failed.