基于CEEMDAN与PSO-SVM的滚动轴承故障诊断

    Rolling Bearing Fault Diagnosis Based on CEEMDAN and PSO-SVM

    • 摘要: 滚动轴承的故障状态识别研究对于保障旋转机械的正常运转和生产安全,提高设备可靠性,以及减少经济损失具有重要意义。针对滚动轴承故障状态识别问题,提出了基于自适应噪声完备集合经验模态分解(CEEMDAN)和粒子群优化支持向量机(PSO-SVM)的滚动轴承故障诊断方法。首先利用CEEMDAN对轴承振动信号进行分解,获得多个本征模函数(IMF)及残余项;然后选择适当的IMF,提取功率谱熵、能量熵、近似熵、样本熵和模糊熵共5个信号熵指标,构建故障特征集;最后结合PSO-SVM实现故障状态识别。为验证模型的有效性和可靠性,针对SKF6203轴承正常、外圈故障、内圈故障及滚动体故障,ER-16K轴承正常、外圈故障、内圈故障、滚动体故障和内圈-外圈复合故障,进行了故障诊断实验,对信号熵特征与故障信号的相关性进行了分析,并通过对比实验验证了算法的优越性。结果表明:CEEMDAN能够有效实现轴承信号的预处理,提高信噪比;5个熵指标所构成的故障特征集能够有效表达轴承故障的本质;该故障诊断方法在凯斯西储大学轴承数据集和东南大学轴承数据集上的故障状态识别准确率分别为94%和92.8%,且准确率波动较小。CEEMDAN,熵特征和PSO-SVM的结合,实现了对旋转机械滚动轴承故障的有效识别,该故障诊断方法可以为选煤厂众多旋转机械的可靠运行提供有力保障,推动选煤厂智能化建设的进程。

       

      Abstract: Research on fault state identification of rolling bearings is of critical significance for ensuring normal operation and production safety of rotating machinery, enhancing equipment reliability, and mitigating economic losses. Aiming at the problem of rolling bearing fault state identification, a diagnostic method based on complete ensemble empirical mode decomposition with adaptive noise (CEEMDAN) and particle swarm optimization-support vector machine (PSO-SVM) is proposed. First, CEEMDAN is used to decompose the bearing vibration signal to obtain multiple Intrinsic Mode Functions (IMF) and a residual term. Then, the appropriate IMF is selected to extract five signal entropy indexes including power spectral entropy, energy entropy, approximate entropy, sample entropy, and fuzzy entropy, to construct the fault feature set; and finally combine with the PSO-SVM torealize the fault state identification. To verify the effectiveness and reliability of the model, fault diagnosis experiments were conducted for SKF6203 bearings with normal, outer-ring fault, inner-ring fault and rolling-body fault, and for ER-16K bearings with normal, outer-ring fault, inner-ring fault, rolling-body fault and inner-ring-outer-ring composite fault, the correlation between signal entropy features and fault signals was analyzed, and the superiority of the algorithm was verified through comparative tests. The results show that CEEMDAN can effectively realize the preprocessing of bearing signals and improve the signal-to-noise ratio; the fault feature set composed of five entropy indexes can effectively represent the essence of bearing faults; the accuracy of this fault diagnosis method for fault state identification on the Case Western Reserve University bearing dataset and the Southeastern University bearing dataset was 94% and 92.8%, and with small fluctuation in accuracy. The combination of CEEMDAN, entropy features, and PSO-SVM achieves effective identification of rolling bearing faults in rotating machinery, the fault diagnosis method can provide a strong guarantee for the reliable operation of numerous rotating machines in coal preparation plants and promote the intelligent construction process of coal preparation plants.

       

    /

    返回文章
    返回