In order to investigate the effect of longitudinal-torsional ultrasonic vibration on the surface topography of ceramic materials grinding,taking ZrO2 ceramics as the research object and the grinding surface roughness as the evaluation index,multiple linear regression analysis method was used to establish the materials surface roughness fitting model of ordinary grinding (OG) and longitudinal-torsional ultrasonic grinding (L-TUG) through the orthogonal test,and the primary-secondary order and the influence degree of the process parameters on the surface roughness were studied.At the same time,BP neural network prediction model was used to optimize the surface roughness of L-TUG.The results show that:in L-TUG,the spindle speed has the greatest influence on the roughness value,the ultrasonic energy has the least influence; in OG,the grinding depth has the greatest influence on roughness value,while the spindle speed has the least influence.The prediction error of the BP neural network model is within the range of 1.070% to 9.396%,and the surface quality obtained by the optimal grinding parameter combination is the best,and the intelligent prediction to L-TUG surface roughness value with high accuracy can be realized.