基于神经网络的无刷直流电机传感器故障检测及容错系统

Brushless DC motor sensor fault detection and fault-tolerant system based on neural network

针对传统容错控制系统只能检测传感器单相或两相出现故障且检测及容错算法相对复杂的缺点,提出了一种基于神经网络的故障霍尔传感器故障检测方法与基于无位置传感器系统的容错控制系统,利用神经网络分类功能对多种故障类型,如换相延迟、换相提前、单相故障等更多故障类别进行诊断。通过仿真与测试平台实验验证,所提出的基于神经网络的传感器故障检测及容错控制系统能显著降低霍尔传感器故障对电机转速的影响,并使电机在霍尔传感器故障时能够稳定运行。

For traditional fault-tolerant control systems that can only detect single-phase or two-phase faults in sensors, the detection and fault-tolerance algorithms are relatively complex, this paper presents a fault detection method based on neural network and fault-tolerance based on sensorless systems. The control system uses neural network classification to diagnose multiple failure types such as commutation delay, commutation advance and single-phase failure. Through simulation and test platform experiments, the sensor fault detection and fault-tolerant control system based on neural network proposed in this paper can significantly reduce the effect of Hall sensor fault on the motor speed, and make the motor run stably when the sensor fails.