The Research Of On-line Prediction Technique For Surface Roughness After NC Turning Based On Neural Network

Besides the size, shape and location accuracy, the surface precision of mechanical parts is another important indicator used to measure their machining quality. As the main device of advanced machinery manufacturing industry, the CNC machine tool can control the size, shape and location accuracy much better than the traditional machine tool. But the obstacle of its further automation is still the stability control of the surface roughness in the cutting process, which is not resolved by the traditional machine tool.Based on the formation mechanism and evaluation system of surface roughness, its causes in the CNC turning process and the relationship between them and the cutting parameters are studied in this paper. It is found that the relative vibration between workpiece and tool caused by their interaction, resulting in relative offset, leads to the instability of the surface roughness of parts and the impact of the radial vibration on the surface roughness is distinctly higher than that of other directions' vibration. According to modern test theory of machining vibration signals, the acceleration signal in different CNC precision turning conditions is extracted with the piezoelectric acceleration sensor of the home-made equipment. Then, after the subsequent processing, it is input to the trained artificial neural network with the higher nonlinear mapping capability to predict the surface roughness.The results show that the surface roughness from the trained artificial neural network in different conditions is very close to that from the machining experiments. This establishes a new type of theory for online prediction of surface roughness during machining process. With the future NC system which can automatically adjust the cutting parameters to reach the required surface roughness, its stability control will be achieved. Thus, both the needs of machining efficiency and surface quality can be ensured.