Surface roughness prediction for Ti-48Al-2Cr-2Nb micro-milling based on 1DCNN-LSTM neural network

Surface roughness is the main index to measure the surface quality of micro machined parts. In order to improve the accuracy of surface roughness prediction in micro milling, a deep neural network prediction model based on one-dimensional convolution-long short-term memory (1DCNN-LSTM) is proposed. Using the efficient data processing mechanism of one-dimensional convolution network and the accurate prediction ability of long-term and short-term memory network, the problems of batch sequence data processing, sample key feature learning and small sample data surface roughness prediction are effectively solved. Taking spindle speed, feed speed, milling depth and micro milling cutter spiral angle as control variables, the prediction model of micro milling surface roughness is trained and verified by experimental data. The results show that compared with the traditional machine learning model, the average prediction error of 1DCNN-LSTM neural network is only 5.9%, which verifies the high-precision prediction performance of the model based on small sample data, and provides a new method for the prediction of micro-milling surface roughness.