基于改进粒子群的采煤机齿轮箱故障诊断

Gearbox Fault Diagnosis of Shearer Based on Improved Particle Swarm Optimization

采煤机截割部摇臂齿轮箱承担着综采工作面截割部动力传动的重任,其故障与否直接影响采煤机正常工作。而传统的故障诊断方法-BP神经网络采用基于梯度下降的算法,存在容易陷入局部极小值、收敛速度慢等不足,这些不足严重影响了BP网络的应用。然而粒子群算法(PSO)有很好的全局收敛特性。因此,为了提高网络的性能,采用粒子群算法来优化BP神经网络,将改进的PSO引入神经网络的拓扑结构,用PSO的迭代代替BP中的梯度修正。结果表明:提出的改进方案可以有效地优化神经网络,提高其在采煤机齿轮箱故障诊断中的应用价值。

The shearer cutting unit rocker gear box bears the task of powertrain for the cutting unit of mechanized mining face,it＇s failure affect the normal operation of shearer directly.But the traditional fault diagnosis method-BP neural network is an algorithm which is based on gradient descent, it has some defects,such as easy to fall into local minima,slow convergence and so on, these defects affect the BP network seriously. However PSO has good global convergence properties. Therefore, in order to improve the performance of network,in this article PSO is improved PSO is introduced into the BP neural network PSO particle swarm iteration.The results show that ：The can optimize the neural network effectively,and improve fault diagnosis. chosen to optimize the BP neural network, the topology,instead of the gradient correction by improvement program proposed in this article it＇s application value in the shearer gearbox