Research On The Key Technology Of Fault Diagnosis And Health Management Of Mine Hoist Transmission System

As a key channel connecting the underground production system and the ground,the mine hoist will directly affect the safety of the miner and the capacity of the mine's production if it fails.Therefore,it is of great significance and great economic value to realize the fault diagnosis and health management of mine hoists.In recent years,with the continuous development of signal analysis methods,data mining,and artificial intelligence technologies,data-driven fault diagnosis methods have received widespread attention,and now they have become one of the main research directions for implementing intelligent fault diagnosis.In this paper,the mine hoist drive system is taken as the research object.By analyzing the current research status of data-driven fault diagnosis methods at home and abroad,the fault diagnosis methods based on the three key components: the inverter,bearing and drive motor in the drive system are studied.Due to the complicated structure of the hoist drive system,it is necessary to further analyze and dig out the failure cause after identifying the fault status of the components.The data-driven fault diagnosis method is mainly aimed at different parts of different types of equipment.Under different life cycles and working conditions,the fault status can be effectively identified,but the underlying causes of the failure have not been found.Ontology-based fault diagnosis can model fault diagnosis knowledge at a macro level and dig deep-seated causes of failures.Therefore,it is necessary to study the combination of data-driven fault diagnosis and ontology-based fault diagnosis to achieve the pattern recognition.The whole process of fault knowledge reasoning and diagnosis.Research on the above issues,the main contents include:(1)The fault diagnosis method of NPC three-level inverter based on adaptive electric period division and random forest is studied.Based on the circuit and fault analysis of the NPC three-level inverter,an Adaptive Electrical Period Partition(AEPP)algorithm that mainly depends on the characteristics of the output current signal is proposed.Use the Maximal Overlap Discrete Wavelet Transform(MODWT)to divide the three-phase current signals with varying periods into electrical periods,calculate 11 kinds of statistical parameters and construct the original statistical feature set,combines low frequency components normalized using the Park's Vector Modulus(PVM).An NPC three-level inverter open-circuit fault diagnosis model based on random forest classifier is constructed,using the imulation analysis data and the inverter fault experimental platform data for experimental analysis.The experimental results show that the proposed fault diagnosis model has better diagnostic performance for variable-period three-phase current fault signals,and can achieve ideal fault diagnosis accuracy under the condition of variable speed of the motor.(2)A fault diagnosis method for elevator bearings based on sensitive feature selection and maximum local boundary criteria is studied.Study the bearing signal analysis method based on Dual-Tree Complex Wavelet Packet Transform(DTCWPT),use DTCWPT to process the original vibration signal,perform single-node reconstruction on the terminal node,and obtain reconstructed signals in different frequency ranges.The statistical parameters of the reconstructed signal and its Hilbert envelope spectrum are calculated to construct the original feature set.In order to build feature subsets with the sensitive features that can effectively reflect bearing faults from high-dimensional original feature sets,a Sensitive Feature Selection by Feature Clustering and Correlation Coefficient between Features(FSFCC)method is proposed.FSFCC)to quantify the fault state sensitivity of statistical features,and select sensitive features to build feature subsets.In order to remove the high-dimensional feature concentration redundancy and interference features,reduce the computational complexity,and improve the data separability,a feature localization method based on the Maximum Local Margin Criterion(MLMC)is proposed.Low-dimensional representation of high-dimensional feature sets,and improve the discrimination performance of feature sets.Based on the Support Vector Machine(SVM)classifier,combined with FSFCC and MLMC,the OFS-FSFCC-MLMC-SVM bearing fault diagnosis model is constructed.The bearing fault data of Case Western Reserve University in the United States and the SQI of Spectra Quest in the US-MFS mechanical failure comprehensive simulation test bed bearing failure data for failure experimental analysis under different operating conditions.The experimental results show that the FSFCC method can effectively select sensitive features,and the MLMC method can reduce redundancy and interference features,improve the discrimination performance of the feature set,use Both methods can significantly improve the accuracy of the diagnostic model for fault diagnosis under different operating conditions.(3)The fault diagnosis of hoist bearing based on sensitive feature transfer learning is studied.In view of the two limitations of data-driven at present,(1)the majority of the fault diagnosis model based on data driven,is between the training data set and test data set,with the same distribution under the assumption of constructing the industrial scenario,test data of variable working condition of equipment and training data distribution differences,can cause difficult to achieve ideal performance of fault diagnosis.(2)Due to the diversity of variable working conditions and faults in the actual industrial scene,it is difficult to obtain a large number of labeled training samples in the actual fault state,which to some extent limits the application of intelligent fault diagnosis method in the actual industrial scene.Therefore,based on the study of Transfer Component Analysis(TCA),an Modified Transfer Component Analysis(MTCA)is proposed,which increased the consideration of the difference of conditional distribution between data in different domains and the optimization goal of minimizing the divergence in data classes,so as to improve the domain adaptability and enhance the discrimination performance.On the basis of the bearing vibration signal analysis method based on DTCWPT and fault sensitive feature selection method FSFCC,the bearing fault diagnosis model ofs-fsfcc-mtca-svm was constructed by combining the MTCA method.The experimental results show that the proposed MTCA method can effectively reduce the distribution differences between different domains,improve the adaptability of domains and enhance the discriminant performance,so that the fault diagnosis performance of the fault model can be achieved.(4)Study on fault diagnosis of hoist drive motor based on intra-class feature transfer learning and multi-source information fusion.The EEMD-based analysis method of motor fault vibration signals and stator current signals is studied.The effective IMF component is selected,and the envelope spectrum and marginal spectrum components are combined to extract statistical features to construct the original feature set.Based on the study of Stratified Transfer Learning(STL),an improved feature transfer learning method is proposed.The drive motor fault diagnosis models OFS-MSTL-SVM and OFS-MSTL-RF were constructed based on the SVM classifier and random forest classifier respectively.The SQI-MFS mechanical fault comprehensive simulation test bench of Spectra Quest Company was used to drive the motor fault vibration signal and current signal.The experimental analysis shows that the proposed method can effectively improve the accuracy of fault diagnosis under different working conditions of the drive motor.However,the performance of different fault diagnosis models using different source signals for fault diagnosis is different and may be one-sided.So put forward a drive motor fault diagnosis framework based on DS evidence fusion theory,using DS evidence theory for decision-level fusion based on OFS-MSTL-SVM and OFS-MSTL-RF models.The fusion fault diagnosis results show that the proposed multi-source information fusion framework can further improve the accuracy of fault diagnosis.(5)The research based on the combination of data-driven fault diagnosis and ontology-based fault diagnosis,realize the whole process from pattern recognition to fault knowledge reasoning and diagnosis.Based on the analysis and summary of the fault diagnosis knowledge of the elevator drive system,a fault ontology knowledge base is constructed,and the Neo4 j graph database is used to visually display and store the fault ontology knowledge base.The semantic mapping method is used to associate the data-driven fault state recognition results with the fault phenomenon instances in the fault ontology knowledge base,realizing the advantages complementary of the two methods,and digging into the underlying cause of the failure.Finally,based on the above work,a fault diagnosis and health management system for the mine elevator drive system was designed and developed.The system contains four modules: system operation management,historical equipment failure information,fault status identification for hoist drive system based on data-driven,Fault cause analysis and system health management based on fault diagnosis knowledge map.