基于蝗虫优化Bi-LSTM网络的电机轴承故障预测

Motor bearing fault prediction based on grasshopper optimized Bi-LSTM network

为了有效预测复杂工况下的电机轴承故障,提出一种利用蝗虫优化算法(GOA)优化双向长短时记忆网络(Bi-LSTM)的电机轴承故障预测方法。首先,对电机轴承的振动信号利用互补集合经验模态分解(CEEMD)算法进行分解,获得能够表征振动特征的多组固有模态分量(IMF),计算每组IMF分量的奇异值能量,并组成奇异值能量谱作为电机轴承的性能退化指标。其次,采用GOA对Bi-LSTM网络的多个超参数进行迭代寻优,提高模型的预测精度与收敛速度,从而获得一组最优的超参数组合。最后,利用优化后的Bi-LSTM网络实现电机轴承的故障预测。以开源的电机轴承振动信号进行测试实验,实验结果表明,相较于其他多种预测模型,所建立模型具有较高的预测精度同时还具有较强的鲁棒性,能够及时为检修工作提供理论支撑,具有一定的研究价值与工程意义。

In order to effectively predict motor bearing faults under complex operating conditions, a motor bearing fault prediction method using a grasshopper optimization algorithm(GOA) optimized bidirectional long short-term memory(Bi-LSTM) network is proposed.Firstly, the vibration signal of the motor bearing was decomposed by the complementary ensemble empirical mode decomposition(CEEMD) algorithm to obtain multiple sets of intrinsic mode functions(IMF) that can characterize the vibration, calculate the singular value energy of each IMF component, and form the singular value energy spectrum as the performance degradation index of the motor bearing. Secondly, GOA was used to iteratively seek the optimization of multiple hyperparameters of the Bi-LSTM network to improve the prediction accuracy and convergence speed of the model, so as to obtain an optimal set of hyperparameter combinations. Finally, the optimized Bi-LSTM network was used to realize the fault prediction of motor bearings. The experimental results show that compared with other prediction models, the model established in this paper has higher prediction accuracy and stronger robustness, which can provide theoretical support for maintenance work in time and has certain research value and engineering significance.