基于GA-BP神经网络模型的滚揉机轴承的故障诊断研究

Research on Fault Diagnosis of Tumbler Machine Bearings Based on GA-BP Neural Network Model

摘要: 针对滚揉机轴承在运行过程中易发生故障，且传统BP神经网络模型在故障诊断中易陷入局部最优、收敛速度慢及诊断精度受限的问题，本文提出了一种基于遗传算法(GA)优化BP神经网络的故障诊断方法。通过搭建滚揉机轴承振动信号采集平台，采集了内圈故障、外圈故障、滚动体故障及正常状态四类工况下的振动数据。对原始信号进行低通滤波与Min-Max归一化处理后，提取时域与频域特征共72维，并采用主成分分析(PCA)将特征维度降至前11个主成分，累计方差贡献率为90.94%。利用遗传算法对BP神经网络的初始权值、阈值及关键超参数进行全局优化，构建GA-BP故障诊断模型。实验结果表明，该模型在测试集上的整体诊断准确率为98.29%，其中对内圈故障、外圈故障和正常状态的召回率分别达到96.8%、100%和100%，对滚动体故障的召回率为96.7%。与PSO-BP和WOA-BP模型相比，GA-BP在准确率、召回率及F1分数(0.967)上均具有优势，验证了该方法在滚揉机轴承故障诊断中的有效性与稳定性。

Abstract: Aiming at the problem that the rolling machine bearing is prone to failure during operation, and the traditional BP neural network model is easy to fall into local optimum, slow convergence speed and limited diagnostic accuracy in fault diagnosis, this paper proposes a fault diagnosis method based on genetic algorithm (GA) to optimize BP neural network. By building a tumbling machine bearing vibration signal acquisition platform, the vibration data under four working conditions of inner ring fault, outer ring fault, rolling element fault and normal state are collected. After low-pass filtering and Min-Max normalization of the original signal, a total of 72 dimensions of time domain and frequency domain features were extracted, and principal component analysis (PCA) was used to reduce the feature dimension to the top 11 principal components, with a cumulative variance contribution rate of 90.94%. The genetic algorithm is used to optimize the initial weights, thresholds and key hyperparameters of BP neural network, and the GA-BP fault diagnosis model is constructed. The experimental results show that the overall diagnostic accuracy of the model on the test set is 93.16%. The recall rates for inner ring fault, outer ring fault and normal state are 96.8%, 100% and 100%, respectively, and the recall rate for rolling element fault is 96.7%. Compared with PSO-BP and WOA-BP models, GA-BP has advantages in accuracy, recall rate and F1 score (0.967), which verifies the effectiveness and stability of this method in bearing fault diagnosis of tumbling machine.