As the abnormal state data of rolling bearing under variable working conditions featured with high-dimensional fuzzy classification in the feature space encounters the extreme difficulty in partitioning the sub features of the abnormal state data,which increases the difficulty of bearing abnormal detection,a classification method of local tangent space for abnormal state of rolling bearing under variable working conditions is proposed.The local tangent space arrangement method is used to reduce the dimension of the rolling bearing data under variable working conditions,thus meeting the mapping conditions of the classification space in the local tangent space.The depth confidence network is applied to extract the abnormal features of the data through the training of the abnormal data.By using the nonlinear mapping function,the extracted features are input into the SVM classifier,and the two-dimensional feature matrix is mapped into the three-dimensional classification space,to which the hyperplane structure is added.Under the guidance of polynomial kernel function,the corresponding sub-feature classification region is found,and the abnormal state of rolling bearing under variable working conditions is detected according to the classification results.The experimental results show that the method has a high detection rate for abnormal conditions and takes less detection time for early abnormal points before and after adjusting the bearing load.
rolling bearing under variable working conditionlocal tangent space methoddata dimension reductiondeep belief networkSVM classifierabnormal state detection
国家科技期刊平台
NSTL
维普
万方数据
