Research On Fault Diagnosis Of Rolling Bearing Based On Multi-sensor Information Fusion

With the social progress of science and technology, large,complex,difficult precise parts need to be produced in energy,transportation,national defense and other fields related to people's livelihood, therefore higher requirements was put forward on CNC machine tools accuracy,reliability and longevity.In the high-speed, high acceleration, heavy loads and other non-operating conditions, vibration, shock, deformation and other factors often make a significant impact on the CNC machines tool, which prone to be malfunctioned or damaged. On the one hand, the quality of the workpieces were affected, on the other hand reduce the reliability of CNC machine tools, shorten life or even scrapped. Rolling bearings are the most important and fragile parts in the machine tool feeding system. About 30% mechanical failures are caused by rolling bearings, so there are meaningful and economic profit on the study of rolling bearings on the relationship between status information and malfunction. Traditional single sensor is susceptible to outside interference that will lead to failure and misjudgments. so a single sensor accuracy rate is low for rollingr bearings test. Based on the above reasons, this multi-sensor information fusion technology in the diagnosis of CNC machine tool technology and method rolling bearing failure. Multi-Sensor Information Fusion's basic principle take full advantage of multi-sensor by combing each sensors' space and time redundancy or complementary information basing on certain criteria to abstract consistent explanation and description, and its purpose is exporting more useful information, and eventually it will improve the diagnostic accuracy and effectiveness and eliminate the limitations of the individual sensors. The advantages of multi-sensor data fusion is prominently demonstrated in the information redundancy tolerance,fault tolerance, complementarities and low cost.This article designs a set of virtual instrument detection system for rolling bearing condition monitoring and fault diagnosis in the numerical control machine tool feeding system on LabVIEW software as the development platform. By analyzing the characteristics of rolling bearing fault signals, use acceleration sensor and current sensor detects rolling element bearing vibration signals and current signals, after pretreatment by NI ELVIS into the PCI-6251 data acquisition card. While combing database technology, LabVIEW and Microsoft Access database are connected to achieve functions to storage,query,add,modify and delete data. Using powerful hardware compatibility and Matlab signal processing capacity, LabVIEW can use Matlab Wavelet Denoising and wavelet analysis; rolling bearing fault signals is obtained by the demodulation first layer of the details of signal of wavelet of Hilbert transform, and carred out spectrum analysis through the fast Fourier transform. theoretical calculations can give us the frequency of roller bearing failure. Centering the failure frequency and regarding -2.5Hz the 2.5Hz bandwidth as eigenvalues, eigenvalues the inner and outer rings and roller bodies are calculated by the normalization processing. The value of the acceleration sensors and current sensors as the feature vectors of eigenvalues, then improved the result of malfunction information by BP neural network algorithm, finally, I carry out the network training and simulation in order to achieve high efficiency, high reliability fault diagnosis system.