CNC Machine Table Feed System Fault Diagnosis

CNC machine is the main equipment of modern industrial production, especially inthe processing of complex structure, large-scale and high precision parts, CNC machinetools play an irreplaceable role. But CNC machine tools are usually in high-speed,variable load and reciprocating impact of work conditions, work long times of CNCmachine tools can lead to failure, especially in some mechanical parts such as screw,bearing, guide rail and so on. Research on CNC machine tools to carry out faultdiagnosis can timely discovery machine failure and identify hidden faults, so as toimprove the reliability of the machine, push CNC machine fault diagnosis technologyconsists of regular maintenance to repair and maintenance of real-time changes in orderto reduce maintenance costs and create greater economic efficiency.This paper studies the common fault forms of CNC machine tools and its failuremechanism, and based on BP neural network designs the fault diagnosis system of CNCmachine table feed system. The system mainly includes fault type and mechanismanalysis, experimental program design, data acquisition system hardware and softwaredesign, signal analysis and feature extraction and fault diagnosis based on neuralnetwork model design and so on. The focused research is the signal processingtechnology including signal preprocessing, feature extraction and feature selectiontechnology and the design of two level fault diagnosis model and implementation and soon.First of all, the common fault forms and mechanism of CNC machine tool is studied,and more frequent failure of the mechanical parts are the focused research. According tothis, designed an experimental program, including the selection and setting fault pieces,measuring point selection, sensor selection and installation, as well as specificexperimental design process.Secondly, study the technology of data acquisition and data acquisition system isdesigned, including hardware system design and software design. Hardware design is onthe basis of NI-PXI chose data acquisition platform, data acquisition card, thecorresponding cable and regulate device, and its parameters have been set; Software system design is mainly based on LabVIEW and MATLAB platform to design a dataacquisition module, data analysis module and database management module, and thencompile the program.Again, study the technology of data processing, and the data processing is dividedinto three big step. As a first step, the collected data signal preprocessing, including theremoval of singularities zero mean processing and signal processing; The second step,the pre-processed signal in time domain, frequency domain analysis and waveletanalysis, and extract the corresponding time-frequency characteristic value; The thirdstep, to extract the time-frequency characteristic value for further selection andextraction, including characteristic value the preliminary selection and feature extractionbased on kernel principal component analysis of two parts, eventually get used to thecharacteristics of the fault diagnosis value.Finally, the two level fault diagnosis model of CNC machine table feed systemwhich is based on the BP neural network are established. The first level is the totalnetwork, which is used to diagnose the faults of different parts; The second level issub-networks, which is used to diagnose the same parts of different fault, divided intorolling bearing and ball screw networks. Two level fault diagnosis model realizes thefault of preliminary discrimination and the details of fault diagnosis function.