SSA-SqueezeNet在轴承损伤程度识别中的应用

Application of SSA-SqueezeNet in Bearing Damage Degree Identification

变负载工况下轴承振动信号呈现非线性、非平稳特性,轴承损伤程度复杂特征难以提取。本文提出一种基于麻雀搜索算法(sparrow search algorithm,SSA)结合轻量级卷积神经网络(SSASqueeze Net)的滚动轴承损伤程度识别方法。首先,该故障诊断模型利用SSA来自适应调节Squeeze Net网络中的超参数,使得网络结构在提升精度的同时降低波动;其次,将一维振动信号经过连续小波变换输入到优化后的网络中进行训练;最后,通过凯斯西储大学滚动轴承实验数据和XJTU-SY滚动轴承加速寿命试验数据集对所提方法进行验证。对比其他诊断方法,结果表明,优化后的网络模型能够准确地实现滚动轴承损伤程度和寿命状态的识别,具有较强的跨负载自适应能力和泛化性能。

Due to the non-linear and non-stationary characteristics of bearing vibration signals under variable load conditions,the complex features of bearing damage degree are difficult to extract.A rolling bearing damage degree identification method based on sparrow search algorithm(SSA)combined with lightweight convolutional neural network(SSA-SqueezeNet)was proposed.Firstly,the fault diagnosis method used SSA to adaptively adjust the hyperparameters in the SqueezeNet network so that the network structure reduced fluctuations while improving accuracy.Secondly,the rolling bearing one-dimensional vibration signal was input to the optimized network after continuous wavelet transform for training.Finally,the proposed method was validated by using Case Western Reserve University rolling bearing experimental data and XJTU-SY rolling bearing accelerated life test data.Compared with other diagnosis methods,the results show that the optimized network model can accurately identify the damage degree and life state of rolling bearings,and has strong cross-load adaptive capability and generalization performance.