Research On Wear State Detection Of Cemented Carbide Tool Based On IWOA-SVM

Tool wear will directly affect the quality of the workpiece,resulting in loss of economic benefits.In order to improve the accuracy of tool wear state detection,this paper studies the wear state detection of cemented carbide tools.The main contents of the text are as follows:(1)Determine the vibration signal as the wear state detection signal of cemented carbide tool.The tool wear mechanism and form are analyzed.The internal relationship between vibration signals and tool wear is explored.Combined with the tool wear process and practical application,the tool wear grade is established.The orthogonal experiment method is used to formulate the signal acquisition experiment plan.The vibration signal of the cemented carbide tool during the whole process is obtained,which provides data support for the subsequent research.(2)The vibration signal is analyzed from three dimensions:time domain,frequency domain and time-frequency domain.In the time domain,the dimensional and non-dimensional parameters of the vibration signal are analyzed.In the frequency domain,the Power Spectrum Estimation based on the AR model is used to analyze the frequency gravity and energy of specific frequency band.In the time-frequency domain,the EEMD based on adaptive thought is proposed to analyze the proportion of IMF energy of each order to the total energy.Through the analysis of these three dimensions,a total of 11 tool wear state detection features such as peak,skewness,frequency gravity and IMF energy ratio are extracted.(3)To improve the control parameter a in the Whale Optimization Algorithm(WOA),an Improved Whale Optimization Algorithm(IWOA)is proposed.Make it no longer adopt the strategy of linear decline,but adjust nonlinearly with the sigmoid curve.A one-to-one multi-class non-linear Support Vector Machine(SVM)is applied.IWOA is used to optimize the penalty factors and kernel parameters in the SVM.Finally,the IWOA-SVM model is established.The tool wear state detection features are divided into complete feature sets with all features and non-complete feature sets without time-frequency domain features.The results show that,compared with the model built with non-complete feature sets,the model built with complete feature sets increases the accuracy of tool wear state detection by more than 10%to 97.5%when the modeling time only increases by 5%.In addition,the WOA-SVM model,SGA-SVM model and BP model are compared with the IWOA-SVM model.The test results show that the detection accuracy of WOA-SVM model,SGA-SVM model and BP model are 96.2%,93.8%and 88.9%respectively,and the modeling time is 145.6s,164.2s and 263.4s respectively.The research results show that the IWOA-SVM model proposed in this paper not only has the highest accuracy,but has the shortest modeling time.It has good comprehensive performance.