Estimation of cutting temperature using machine learning based on signal information received from power analyzer in vortex machining conditions

Abstract

Cutting temperature directly determines power consumption by influencing tool wear and workpiece quality. Especially for workpieces that cause high heat generation due to low thermal conductivity coefficients, such as Ti6Al4V, it is a challenge to control and determine the cooling method and cutting parameters accordingly. Pre-process modeling and estimation of the cutting temperature due to the cutting variables is necessary for effective process planning and machining efficiency. In this study, it is aimed to use various signal information received from the power analyzer as input features to effectively model and estimate by using the machine learning method the cutting temperature occurring in the turning using various cooling environments of Ti6Al4V workpieces. The features extracted from the four signal information were used both individually and by obtaining hybrid signal features where all the features are combined. While individual signals were used directly in establishing the models, hybrid signals were ranked using the feature ranking method according to six different importance levels. As a result of the study, it was determined that the cutting temperature could be predicted with high success by modeling it according to the features extracted from the signal information obtained from the power analyzer.