| With the development of intelligent train operation control system,higher quality requirements are put forward for the switching and electrical equipment of its power system.Railway relay is one of the important components in train power control system,and its operation reliability is an important guarantee of train safe operation.It is a long-term process for railway relay to degenerate from normal state to failure state.In-depth analysis of the specific operation state changes in the degradation process and establishment of an effective operation state identification model can discover the degradation degree of relay performance in time and prevent the failure of train system.In this thesis,the operation state identification of railway relay is studied based on datadriven method,and related performance degradation parameters are extracted through relay electrical life test.On the basis of feature selection by correlation analysis and importance analysis,contact degradation sequence was characterized,and PSO-ELM identification model was established to identify different operating states of relays.Firstly,the failure mechanism of railway relay contact system is analyzed based on its basic structure and on-off process.After that,according to the test requirements,the relay electrical life test platform was built,with 3RH211-2KF40 relay as the test object and Labview as the host computer to realize the effective control and online monitoring of the relay on-off process.Based on the collected operation data,eight characteristic parameters that can characterize the relay contact performance degradation information were calculated.Secondly,in order to improve the recognition accuracy of the operating state recognition model,feature selection processing is carried out for the eight extracted parameters.Due to the large number of degradation features extracted with nonlinear characteristics,the optimal feature subset is obtained by combining correlation analysis and importance analysis.On this basis,the optimal feature collection is combined with Mahalanobis distance to characterize relay contact degradation degree sequence,and the contact degradation interval is divided according to its degradation trend.The standard deviations in different intervals and at the turning point were analyzed to verify the rationality of the division and lay a foundation for the subsequent state identification.Finally,the identification method of railway relay operation state based on PSO-ELM identification model is studied.The recognition accuracy of different input features in the same recognition model and the same input features in different recognition models are compared.The results show that in the study of operation state identification of railway relays,when the optimal collection is applied to the PSO-ELM classifier model,different operation states of relays can be identified more effectively. |
