Research On Fault Identification Method Of CNC Machine Tools For Health Status Diagnosis

As the key technology equipment in modern manufacturing,CNC machine tool has become the mother of modern industry with its high precision and high efficiency.The outage events caused by machine tool faults are bound to inevitably bring huge economic losses,while due to the difficulty in diagnosis,the maintenance system of periodic maintenance and replacement is widely adopted for the mechanical parts of CNC machine tools,which causes great waste of human and material resources,so it is very important to assess the health status and diagnose the faults.This paper takes the common fault components of CNC machine as the research object,such as the servo system,rolling bearing and the variable speed gearbox,on the basis of modern theory and method of machine learning,the fault mechanism analysis,fault feature extraction,fault pattern classification and health status assessment are studied in depth.A fault-based triangular fuzzy multi-attribute fault diagnosis method based on Bayesian network is proposed for the fault of servo system of machine tool spindle.The fault extraction method based on manifold learning is presented for the gearbox fault of CNC machine.In order to solve the problem of health status recognition of rolling bearings in CNC,a classification method for rolling bearing faults based on deep convolutional neural network is proposed.Finally,based on the above results,a model of fault identification system for CNC machine health status diagnosis is established.The main work of the paper is as follows:(1)Aiming at the characteristic that the attribute weight of fault information of CNC machine tool servo system is unknown,a gray relation interval triangular fuzzy multi-attribute fault diagnosis method based on Bayesian network is proposed in the paper,through the analysis of the servo system mechanism.The fault information probability matrix is obtained.The algorithm of interval trigonometric fuzzy multi-attribute is proposed for finding out the cause of the fault of the servo system of the CNC machine tool spindle.The gray correlation degree of the reference sequence and the comparison sequence is solved in this method to obtain the gray correlation degree of the interval number representation and sort.The effectiveness of the proposed method is verified by simulation experiments.It is determined that the overheating of the spindle motor is the main cause of the failure of the test servo system.(2)In view of the nonlinear coupling fault characteristics of the gearbox,a supervised Laplacian feature learning algorithm for gearbox faults is proposed based on the manifold learning algorithm principle.By combining the consistency of local neighborhood information and class label information,the intrinsic geometry embedded in high-dimensional nonlinear dataset can be effectively utilized.Then the algorithm is applied to the fault feature extraction of the gearbox in CNC machine tool to solve the fault mode classification task.A new method of gearbox fault feature extraction based on manifold learning is proposed.The method extracts the intrinsic manifold features in the high-dimensional fault data by directly learning the data,largely retains the overall geometric structure information embedded in the signal,and guides the clustering of the training data by using the category label information.The complex modal space is transformed into a low-dimensional feature space,which makes it easy to implement pattern classification and fault diagnosis.The gearbox failure test experiment is carried out.Compared with the traditional feature extraction methods PCA,LDA and Laplacian Eigenmaps,the new method has better classification ability and significantly improves the classification performance of fault pattern recognition,thus verifying the feasibility and effectiveness of the proposed method.(3)Taking the rolling bearing of CNC machine tool as the diagnosis object,a classification method of bearing fault based on deep convolutional neural network is proposed combined with the deep learning theory.By utilizing the strong feature learning ability of deep convolutional neural networks,this method can extract effective features from the original vibration signals and classify the health status of rolling bearings.The test results of the test sample set and the numerical control machine tool rolling bearing data set show that the method has achieved good fault classification effect.It is proved to be an effective fault classification method.(4)The CNC machine tool fault identification system for health status diagnosis is established.The system uses cRIO data acquisition controller to collect the vibration signal and noise signal of each research object.Using Labview and Matlab hybrid programming technology,the processes for data acquisition,data analysis and time-frequency characteristics extraction are realized.Finally,the proposed fault identification method is used to machine tool spindle servo system,variable speed gear and rolling bearing fault testing the final diagnosis results were obtained by the diagnostic methods proposed by each research object in CNC.The diagnosis results verified that the proposed system model is effective.