基于改进天鹰优化算法优化LSTM的滚动轴承故障诊断方法

A rolling bearing fault diagnosis method based on improved Aquila optimization algorithm to optimize LSTM

针对天鹰优化(Aquila optimizer,AO)算法容易陷入局部最优,长短时记忆(long short-term memory,LSTM)神经网络容易受参数影响的问题,提出了一种基于改进天鹰优化(improved Aquila optimizer,IAO)算法优化LSTM神经网络的模型,并将其应用于滚动轴承的故障诊断中。首先,引入超立方策略优化了种群初始质量,设计自适应螺旋策略平衡了AO算法的全局搜索和局部搜索能力,并通过利用高斯变异策略增强了AO算法跳出局部最优的能力。然后,将所提IAO算法对LSTM的权值和阈值进行优化,构建了基于IAO-LSTM网络的滚动轴承故障诊断模型。最后,凯斯西储大学(Case Western Reserve University,CWRU)轴承数据集和帕德伯恩大学(Paderborn University,PU)轴承数据集的试验结果表明:与其他故障诊断模型相比,IAO优化后的LSTM模型的分类准确率更高,能有效识别滚动轴承的各种故障类型。

Here,aiming at problems of Aquila optimization(AO)algorithm being easy to fall into local optimization and long short-term memory(LSTM)neural network being easily affected by parameters,a model based on improved Aquila optimizer(IAO)algorithm for optimizing LSTM neural network was proposed and applied in fault diagnosis of rolling bearings.Firstly,the hypercube strategy was introduced to optimize the initial quality of a population,and the adaptive spiral strategy was designed to balance the global and local search abilities of AO algorithm.Gaussian mutation strategy was used to enhance the ability of AO algorithm for jumping out from local optimization.Then,the proposed IAO algorithm was used to optimize weights and thresholds of LSTM neural network for constructing a rolling bearing fault diagnosis model based on IAO algorithm-optimized LSTM neural network.Finally,simulation test results of Case Western Reserve University(CWRU)bearing dataset and Paderborn University(PU)bearing dataset showed that compared with other fault diagnosis models,IAO algorithm-optimized LSTM neural network model has higher classification accuracy and can effectively identify various fault types of rolling bearings.