【摘要】 针对行星齿轮箱特征提取困难的问题，提出一种基于精细时移加权多尺度模糊熵（refined time-shift weighted multiscale fuzzy entropy，简称RTSWMFE）、改进监督型几何和统计保持流形嵌入（improved supervised geometryandstatistics-preservingmanifoldembedding，简称IS-GSE）和白骨顶优化算法支持向量机（coot optimizationalgorithmsupportvectormachine，简称COOT-SVM）的行星齿轮箱故障诊断方法。首先，利用RTSWMFE提取高维故障特征信息；其次，采用IS-GSE对高维特征进行降维，提取出敏感、低维的特征；最后，将低维特征输入COOT-SVM中进行识别分类。行星齿轮箱故障诊断实验结果表明：IS-GSE方法采用余弦相似度与欧式距离相结合的距离度量方式，并融入监督学习思想，降维效果较佳；COOT-SVM方法对经RTSWMFE和IS-GSE二次提取的故障特征识别精度达到100%。更多还原

【Abstract】 Considering the difficulty in extracting the features of planetary gearbox, a novel fault diagnosis method for planetary gearbox based on refined time-shift weighted multiscale fuzzy entropy(RTSWMFE), improved supervised geometry and statistics-preserving manifold embedding(IS-GSE) and coot optimization algorithm support vector machine(COOT-SVM) is proposed in this paper. High-dimensional fault feature information of planetary gear box is extracted using RTSWMFE, and the sensitive and low-dimensional features are extracted using IS-GSE. Low-dimensional features are input into COOT-SVM for recognition. The experimental analysis of planetary gearbox fault diagnosis shows that IS-GSE combines cosine similarity with Euclidean distance to construct distance matrix and integrates the idea of supervised learning, and the accuracy of fault feature recognition based on COOT-SVM combined RTSWMFE and IS-GSE is 100%.更多还原