Research on surface roughness prediction method based on composite penalty regression model

Abstract

With the development of manufacturing technology, traditional empirical formulas are hard to deal with the relationships between complex manufacturing techniques and surface roughness. Current data fitting-based modeling methods often do not consider the influence of various factors on surface roughness and lack data cleaning capability. This paper proposes a method to reduce the dimension of data features, which includes the Lasso model to determine the correlation degree of processing parameters and roughness, and solves the possible sparse coefficient relationship between processing parameters and roughness. The ridge regression is also introduced to predict the workpiece surface roughness. The results show that compared with the existing prediction model, this prediction method has high accuracy when given a small amount of training data.