基于SSA-CMBE和PO-ELM的滚动轴承故障诊断方法

The Rolling Bearing Fault Diagnosis Method Based on SSA-CMBE and PO-ELM

摘要: 针对轴承故障特征易受外部噪声影响，进而导致诊断效果不佳的问题，提出了一种基于奇异谱分析(SSA)-复合多尺度气泡熵(CMBE)以及由美洲狮算法(PO)优化的极限学习机(ELM)的滚动轴承故障诊断方法。首先，通过SSA对滚动轴承信号进行奇异值分解(SVD)，得到多个本征模态分量(IMF)，计算各个分量的皮尔逊相关系数，筛选出相关系数较大的IMF分量进行重构并计算各分量的复合多尺度气泡熵作为特征向量。然后通过美洲狮算法优化ELM的权值和阈值，得到基于PO-ELM的诊断模型。最后将得到的特征向量导入PO-ELM诊断模型，对故障特征进行识别。实验表明，PO-ELM的故障识别准确率达到99.17%，相较于其他方法，该方法在诊断滚动轴承故障中具有更高的识别精度和有效性。

Abstract: To address the challenge of accurately and efficiently diagnosing faults in rolling bearings due to complex fault conditions and significant noise interference, a fault diagnosis method based on Singular Spectrum Analysis (SSA)-Composite Multi-Scale Bubble Entropy (CMBE) and Extreme Learning Machine (ELM) optimized by the Puma Optimization (PO) algorithm is proposed. Firstly, Singular Value Decomposition (SVD) is applied to the rolling bearing signal using SSA to obtain multiple Intrinsic Mode Functions (IMFs). Pearson correlation coefficients for these IMFs are calculated, and those with higher correlation coefficients are selected for reconstruction. The Composite Multi-Scale Bubble Entropy of each IMF is then computed as the feature vector. Next, the PO algorithm is used to optimize the weights and thresholds of the ELM, resulting in a PO-ELM-based diagnostic model. Finally, the obtained feature vectors are input into the PO-ELM diagnostic model to identify fault characteristics. Experimental results show that the fault recognition accuracy of PO-ELM reaches 99.17%. Compared to other methods, this approach demonstrates higher accuracy and effectiveness in diagnosing rolling bearing faults.