基于增强型HBA算法优化SVM模型的机械故障诊断方法

Mechanical Fault Diagnosis Method Based on Enhanced Honey Badger Algorithm Optimizing Support Vector Machine Model

摘要:

机械设备振动信号的复杂性及其复杂的环境因素导致其故障的准确诊断难度较高。为此，提出一种增强型蜜獾算法优化支持向量机模型SVM的故障诊断算法EHBA-SVM。为了提高蜜獾算法对目标问题的搜索精度，设计Logistic混沌映射种群初始化提升初始解的多样性，引入黄金正弦策略解决算法全局搜索阶段的种群同质化问题，以增强全局搜索能力，同时设计自适应多向Levy飞行机制兼顾算法的全局搜索精度和收敛速度，以避免迭代后期易于陷入局部最优的不足。利用增强型HBA算法搜索支持向量机SVM模型的超参数组合(c,σ)，构建适用于机械故障诊断且泛化能力更强的EHBA-SVM模型。最后，通过旋转机械振动故障模拟实验平台进行实验分析，提取信号时域特征表征不同设备故障状态，同时利用主成分分析法对高维特征进行特征降维，减少特征冗余与噪声干扰。结果表明，与同类模型相比，EHBA-SVM模型在故障类型诊断与分类上具有更高的准确率和诊断效率，为复杂工况环境下的故障诊断提供了一种可行方案。

Abstract:

Because of the complexity of mechanical vibration signals and complex environmental factors, it is difficult to accurately diagnose faults. Therefore, a fault diagnosis algorithm EHBA-SVM based on enhanced Honey Badger algorithm optimizing support vector machine SVM model is proposed. In order to improve the search accuracy of Honey Badger optimization algorithm for target problems, Logistic chaotic mapping population initialization was designed to improve the diversity of initial solutions. The golden sine strategy was introduced to solve the problem of population homogeneity in the global search stage of the algorithm to enhance the global search capability. And an adaptive multi-directional Levy flight mechanism was designed to take into account the global search accuracy and convergence speed of the algorithm. Therefore, it can avoid the deficiency of local optimization in the later iteration. The enhanced HBA algorithm is used to search the hyperparameter combination (c,σ) of SVM model, and the EHBA-SVM model is constructed which is suitable for mechanical fault diagnosis and has better generalization ability. Finally, the rotating machinery vibration fault simulation experiment platform was used for experimental analysis, and the signal time-domain features were extracted to characterize the fault states of different equipment. Meanwhile, the principal component analysis method PCA was used to reduce the feature dimension of high-dimensional features to reduce the feature redundancy and noise interference. The results show that the EHBA-SVM model has higher accuracy and efficiency in fault diagnosis and classification compared with similar models,which provides a feasible scheme for fault diagnosis under complex working conditions.