A data-driven method for prediction of surface roughness with consideration of milling tool wear

Surface roughness is a critical parameter to evaluate the quality and performance of machined parts. However, in the machining process of hard materials, their poor machinability could lead to severe tool wear, which would make it difficult to determine the surface roughness precisely due to the nonlinear and time-varying characteristics. In this paper, a data-driven framework for surface roughness prediction is proposed by monitoring the state of tool wear. Firstly, multi-domain features are extracted from cutting force signals as indicators of tool states, and the complex mapping relationship between these features and tool wear is constructed based on the designed convolutional neural network (CNN) model. Then, cutting parameters combined with the monitored tool wear are fed into the trained artificial neural network (ANN) model for surface roughness estimation. Finally, a series of milling tests are conducted to verify the performance of the established method, and it is shown that the presented method enables us to reliably evaluate surface roughness by comparing the prediction results obtained by the measured tool wear and the monitored tool wear respectively.