论文部分内容阅读
针对单一特征在进行故障诊断时准确率不高的问题,提出了一种基于自组织神经网络(SOM)的滚动轴承状态评估方法。该方法首先从原始振动信号中提取出多特征数据,运用主成分分析(PCA)方法对多特征数据进行预处理,采用SOM进行网络训练,构建多特征数据的融合模型,输出竞争神经元层的权值矢量;然后,计算每一个样本到竞争神经元层权值矢量的最小欧氏距离,输出最终的融合指标;最后,通过比较待检测样本与正常样本的最小欧氏距离的差异来判断轴承的状态。将该方法应用于滚动轴承状态评估,试验结果表明:融合指标比单一指标对早期故障更加敏感、更加稳健;同时,融合指标能够定量地描述轴承状态的劣化过程。
Aiming at the problem that the accuracy of a single feature is not high enough for fault diagnosis, a self-organizing neural network (SOM) based rolling bearing condition evaluation method is proposed. Firstly, the multi-feature data is extracted from the original vibration signal. The PCA method is used to preprocess the multi-feature data and the SOM is used to train the network. A fusion model of multi-feature data is constructed and the output layer of competing neurons Then, the minimum Euclidean distance between each sample and the weight vector of competing neurons is calculated and the final fusion index is output. Finally, the difference between the minimum Euclidean distance of the sample to be tested and the normal sample is judged to determine the bearing status. The method is applied to the state estimation of rolling bearings. The experimental results show that the fusion index is more sensitive and robust than the single index for early failure. Meanwhile, the fusion index can quantitatively describe the degradation process of bearing status.