Fault Detection Method Based on Improved Dynamic Partial Least Square Method
In order to solve the problem that the characteristic signals of incipient faults of rolling bearings are weak and that in actual industrial operation,the data often have temporal correlation,which makes it more difficult to extract fault features in the incipient stages of rolling bearings,an improved dynamic partial least squares method based on the theory of depth decomposition for incipient faults detection of rolling bearings is proposed.Firstly,the time delay parameter is selected to form a dynamic augmentation matrix of the original data matrix,secondly,the order of deep decomposition is determined,then the deep decomposition theory is applied to obtain each subspace generated by the decomposition.Finally,the partial least squares(PLS)method is used to calculate the statistics and control limits of each subspace,and then compare the statistic of each subspace with its corresponding control limit to determine whether the system has a fault.The proposed DeepDPLS method greatly improves the incipient faults detection rate of rolling bearings compared with PLS method and its related methods.The fault detection rate of the DeepDPLS method proposed reaches 100％in the first-order decomposition compared to the DeepPLS method,the established model is more accurate and can detect incipient faults of rolling bearings earlier.The DeepDPLS method proposed is feasible and effective for incipient faults detection of rolling bearings by experiment and simulation.
rolling bearingincipient faults detectionpartial least squares methodmulti-order decompositiondynamic characteristics
NETL
NSTL
万方数据
