基于改进ALIF与FA-BP的滚动轴承故障诊断

Rolling bearing fault diagnosis based on improved ALIF and FA-BP

针对滚动轴承的故障信号存在大量噪声信号和滚动轴承故障的准确诊断等问题,提出一种基于改进自适应迭代滤波算法与萤火虫算法优化BP神经网络相结合的故障诊断新方法。首先采用自适应迭代滤波算法对故障信号进行分解得到若干个内禀模态函数,再进行奇异值分解,绘制差分谱曲线并选择重构信号,对其进行二次降噪;然后通过萤火虫算法寻找BP神经网络的最佳参数,建立FA-BP故障诊断模型,提取降噪后的内禀模态函数中心频率形成特征矩阵,输入故障诊断模型;最后应用于美国凯斯西储大学的轴承数据进行检测,准确率达99.4%,诊断时间为3.18 s。该方法与BP神经网络、萤火虫算法网络、遗传算法网络、遗传算法优化BP神经网络的诊断模型相比,大大提高了诊断效率并具有较高准确率。

In allusion to the problem that the fault signal of rolling bearing has a large number of noise signals and influences the diagnostic efficiency of rolling bearing fault,a new fault diagnosis method based on the combination of improved ALIF(adaptive iterative filtering algorithm)and the BPNN(BP neural network)optimized by FA(firefly algorithm)is proposed.The ALIF is used to decompose the fault signal to obtain several intrinsic mode functions.The SVD(singular value decomposition)is conducted to draw the difference spectrum curve and select the reconstructed signal,so as to carry out the secondary noise reduction.The FA is used to find the best parameters of BPNN,and the FA-BP fault diagnosis model is established.The center frequency of the intrinsic mode function after noise reduction is extracted to form the feature matrix,so as to input into the fault diagnosis model.The bearing data of Case Western Reserve University in the United States is used for detection,with the accuracy rate of 99.4%and the diagnosis time of 3.18 s.In comparison with the diagnostic models of BPNN,FA network,genetic algorithm network and genetic algorithm optimized by BPNN,this method can greatly improve the diagnostic efficiency and has high accuracy.