Research On State Recognition Method Of Driving Motor Of Mine Belt Conveyor Based On Multi-source Information Fusion

With the continuous innovation and development of the scientific and technological level,the monitoring and diagnosis of the operation status of large-scale mine equipment has been paid more and more attention.The drive motor is an important part of the drive system of the belt conveyor and other equipment.The safe and efficient operation of the motor is related to the reliability and stability of the mine production.Motor failure will not only lead to the destruction of related equipment,but also cause the stagnation of mine production,resulting in unpredictable human and material loss.Therefore,it is of great significance and practical value to monitor the running state of the belt drive motor and to identify and warn the fault condition.Traditional motor fault diagnosis is often analyzed for a single signal,and there is a one-sidedness in signal acquisition.The comprehensive composition of the motor should be considered.In the feature processing,feature selection and state recognition methods are needed to obtain a highly targeted feature subset.In the information fusion,the recognition results of the signals trained by the multi-classifier need to be fused,and the fusion results should be displayed in a visual interface.Carry out research work on the above issues.(1)In this thesis,the fault characteristics of motor vibration signal and stator current signal under different operating conditions and the signal analysis method based on empirical mode decomposition are studied.A vibration and current signal analysis method based on complete mean empirical mode decomposition is proposed to effectively solve the "modal aliasing" problem in EMD decomposition and eliminate false components.The selection of the IMF is determined by calculating the correlation coefficient between the eigenmode component and the original signal.Combining the envelope spectrum and the marginal spectral components,eleven statistical features including time domain and frequency domain features are calculated to construct a high-dimensional original feature set.(2)Aiming at the high-dimensional original feature set of signal samples,a feature selection method FSMDA based on random forest(RF)average accuracy reduction is proposed.The method uses the random forest model to train and test the features,and uses the difference of the out-of-feature error rate before and after the noise interference to describe the feature importance as the basis for feature selection.Combining Linear Local Tangent Space Alignment(LLTSA)with Extreme Learning Machine(ELM),Fuzzy C-Means(FCM),random forest and other classifiers,A state recognition model based on the vibration signal of the drive motor and the stator current signal is constructed,and the test bench data is used for verification.(3)This thesis analyzes the problems existing in traditional information fusion,and proposes a two-level information fusion model based on optimized D-S evidence theory.The first-level fusion focuses on the results of multiple state recognition models of the same type of signal,and then the fusion recognition results of different types of signals are subjected to secondary fusion.The Jousselme distance is used to measure the similarity between the original distances,and the recognition result of the random forest model is retained as the original evidence,which avoids the problems caused by the highly conflicting evidence.(4)By analyzing the needs of mine operators for the monitoring and diagnosis of belt drive motor,the state-of-the-art identification and intelligent decision-making system of mine belt drive motor based on.Net platform was designed and developed.Combining the state monitoring model with the multi-source information fusion result,the function of the traditional threshold warning monitoring system is improved,and the mine operator is provided with a fast and efficient equipment processing system,effectively improve the efficiency of field work,with strong scalability.The experimental results show that the feature selection method FSMDA proposed in this thesis can effectively select the features with higher importance to the state recognition model.The constructed state recognition model CEEMD-FSMDA-LLTSA-ELM/FCM/RF has a good adaptability and significantly improves the accuracy of recognition of the same working condition and variable working condition of the motor.Combined with multi-source information fusion results,the mine belt drive motor state recognition and intelligent decision-making system can effectively describe the fault type and has strong operability.