Research On Data-driven Fault Diagnosis And Predictive Maintenance Of Machine Tool Parts

With the rapid development of science and technology,manufacturing is becoming increasingly important in the national economy.In the industrial production background of big data and intelligent manufacturing,maintenance and equipment are the core elements of intelligent manufacturing.How to use production data to solve failures and predict failures is very important.Gears and bearings are the most typical transmission components of rotating equipment.Working machines under high load and high speed for a long time are prone to failure.Rotating machinery failures caused by the failure of bearings and gears are not a few,and once they fail,they will trigger many chain reactions.Therefore,bearings and gears have always been the focus of research in machine tool fault diagnosis.On the basis of research on faults,how to do predictive maintenance is also an extension point.This thesis takes gears and bearings as research objects,and conducts fault diagnosis and predictive maintenance research.The main research contents are as follows:To study the characteristics of gear failure,which is a form of gear failure,and use a step-by-step method to preprocess the collected sound signals.Using the statistical method-principal component analysis method,the principal components whose original variables meet the conditions are used as feature samples,and the features are sent to Pca-LSSVM,Ga-Pca-LSSVM,Pso-Pca-LSSVM three improved support vector machine optimization models Diagnosis,the results verify the validity and accuracy of the model.Taking the bearing fault data set of Case Western Reserve University as the research object,the vibration signals of the base end,fan end and drive end bearings were selected for FFT,EMD and wavelet packet decomposition analysis.Decided to use wavelet packet decomposition method to extract fault feature vectors,and send the features to Gs-Pca-Lssvm,Ga-Pca-Lssvm,Pso-Pca-Lssvm three improved support vector machine models for fault diagnosis,and effectively realized the state type Recognition,high accuracy.Carrying out theoretical research on machine tool predictive maintenance,a diagnostic reasoning model based on data knowledge ontology is proposed forpredictive maintenance.According to the diagnostic reasoning of knowledge ontology,from extracting semantic knowledge to construct the corresponding maintenance diagnostic semantic knowledge base to using knowledge sharing mechanism to achieve the reason of fault reasoning and maintenance decision support.This theoretical study is supported by the semantics of data knowledge,and the whole life cycle is studied.The equipment state-symptom failure and mapping-symptom correlation algorithms are used to realize the mapping from machine tool feature state space to fault symptom space,deduction,induction,and traceable reasoning to finally achieve maintenance decision,to achieve the unification of dynamic diagnosis and static maintenance knowledge.