Establishment of the optimal predictive model of die steel SKD 11 micromilling via nanofluid (graphene)/ultrasonic atomizing minimum quantity lubrication

This study is mainly utilized nanofluid (graphene) and ultrasonic atomization system to process SKD11 mold steel in micromilling to improve the quality of processed products, find the optimal quality characteristics, and construct an accurate prediction model. In this study, Taguchi's robust design is adopted. The L18(2¹×3⁷) orthogonal table is used to find the optimal parameter combinations for each quality characteristic, and the micromilling force and micromilling temperature are used as characteristic indicators. The control variables are the average thickness of nanographene, weight percent concentration, ultrasonic atomization amount, spindle speed, feed rate, air pressure, nozzle angle, nozzle distance. Also, a back propagation neural network (BPNN) is used to predict the micromilling processing mode. Taguchi method is used to optimize the hyper parameters in the neural network to improve the accuracy of the prediction model, reduce the input of the number of training samples, and then build a neural network prediction model that can accurately predict the quality characteristics of micromilling force and micromilling temperature. The prediction results show that the milling force parameter value error between the micromilling force model prediction and the single target optimization is 0.55%. The error between the micromilling temperature model prediction and the single-target optimization of the micromilling temperature value is 5.90%.