Study of milling machining roughness prediction based on cutting force

Cutting parameters can to be properly adjusted to increase machining precision, to enhance cutting efficiency, and to reduce cost. These parameters are generally chosen according to the recommended data from the tool maker’s technical manual with the synergy of the operator’s experiences. However, inappropriate parameters selection might not only damage cutter and the machine tool, but also worsen the surface roughness of the workpiece. Researches on improving surface roughness has been studied over the years, for example, on the optimization of cutting conditions, vibration measurement and analysis, and cutting forces measurement and analysis, to name only a few. However, most of them must have sensors installed on the machine tool for data collection before the analysis. This will increase the cost. Furthermore, the measurement of cutting force needs to have dynamometer which is not only expensive but is also not appropriate to be installed on a small-sized machine tool. This research focused on establishing the relationship between the cutting force and the surface roughness for the workpiece machining in a small-sized machine tool. In order to calculate the cutting force, cutting force coefficients were obtained based on the cutting test on the other machine tool rather than the small-sized machine. Tapping test was also conducted to obtain the frequency response function of the machine. The cutting forces were calculated based on the cutting force coefficients and the frequency response function. The calculated cutting forces were then converted through the proposed methodologies into eleven characteristic values. Finally, two commonly used machine learning algorithms, multivariate linear regression and generalized regression neural network were adopted for regression analysis to establish the relationship between the cutting forces and the surface roughness. The results have shown that the prediction by generalized regression neural network is better than that of multivariate linear regression. The major contribution of this paper is that appropriate algorithms have been studied, compared and verified so that the relationship between cutting force and surface roughness can be established. This means that surface roughness can be estimated based on the calculated cutting forces incorporating the machine tool’s frequency response function.