Optimizing process parameters of in-situ laser assisted cutting of glass–ceramic by applying hybrid machine learning models

Abstract

Glass-ceramic is an advanced optical material. However, its high hardness and brittleness pose significant challenges to achieving high-quality surfaces during the machining process. Currently, in-situ laser assisted diamond cutting technology has emerged as an effective method for machining hard and brittle materials. To achieve the most favorable machining effect, this study utilized a range of methods to determine the optimal process parameters. The experiments were initially designed using the Taguchi method and response surface methodology. The influence of laser power, cutting depth, spindle speed, and feed speed on the surface quality of glass–ceramic was investigated through analytical techniques including variance analysis and signal-to-noise ratio analysis. Subsequently, a hybrid machine learning model was developed to predict the surface roughness of glass–ceramic, including synthetic minority over-sampling, stepwise regression, bagging, and artificial neural network technologies. The results indicated that varying parameter combinations during in-situ laser-assisted diamond cutting had a significant impact on surface roughness. The contribution rates of laser power, cutting depth, spindle speed, and feed speed to surface roughness were determined as 32.40 %, 23.15 %, 11.63 %, and 22.18 %, respectively. Moreover, a hybrid machine learning prediction model for glass–ceramic surface roughness achieved an R2 value of 0.9737 on the testing set with a mean absolute error of 4.1858. Subsequently, an improved quantum genetic algorithm was adopted to determine the optimal process parameters, achieving a smooth surface quality of Sa 14.022 nm with laser power of 12 W, cutting depth of 2 μm, feed speed of 1 mm/min, and spindle speed of 2274 rpm. The error between the surface roughness obtained from the verification experiment and the optimized result was below 8 %.