Online Detection System Of Grinding Performance Of Microcrystalline Corundum Grinding Wheel Based On Multi-sensor Fusion

As a common precision and ultra-precision machining process,grinding is the key to determine the processing accuracy and the surface quality of workpiece.In the actual production,it occupies an increasingly important position.Faced with small batch,multi-species,high-quality producing targets,the requirements of selecting grinding wheel and designing grinding process are getting higher and higher,which is closely related to the detection of wheel grinding performance.However,due to the complexity of grinding and the diversity of grinding performance evaluation,a single sensor can not meet the requirements of online detection.Online detection based on multi-sensor fusion has become a trend of intelligent processing production.Based on the test hardwares,including a laser micrometer,a multi-component dynamometer,an acoustic emission sensor and a power meter,a multi-sensor information fusion virtual-instrument system was developed by means of Lab VIEW’s development platform and MATLAB toolkit.It completes the automatic processing and the real multifaceted online detection of wheel grinding performance,that is,the online identification of wheel wear state,and the online forecast of grinding workpiece surface roughness and grinding ratio.In this paper,an artificial neural network information fusion algorithm(PCA-ANN)based on principal component analysis(PCA)is applied in the multi-sensor fusion online detection system.The principal component analysis is reduced by the correlation of the extracted features,and the expression of the information is completed by the artificial neural network.The algorithm reduces the information redundancy of the extracted features,accelerates the running speed of the system,and improves accuracy of the online detection.For the specific development of the online detection system,the frequency domain and time-frequency domain of the signal are analyzed by means of FFT spectrum,power spectrum and wavelet packet analysis.Then the specified time-frequency domain of the signal is selected according to the analysis result.The characteristics of the signal related to the grinding performance were obtained,and then the feature fusion was performed by PCA-ANN method to complete the information expression.Based on the principal component analysis,feature extraction,grinding performance representation and information fusion algorithm,this paper completes the specific programming and development of the whole system.Sample experiments of hardened 20CrMnTi ground by microcrystalline corundum grinding wheel are used to analyze the extracted features and the grinding performance of the wheel.The influence of the delay of the power signal in the grinding process is reduced,the guiding function of the extracted feature to the grinding process is proposed,and the validity of the selected feature is verified.Through testing and debugging of the system,it is proved that the PCA-RBF method does better than PCA-BP method in this field.The designed system can achieve high-precision online identification of wheel wear status,high-precision online prediction of grinding surface roughness and grinding ratio.