Deep learning-based method for characterizing the cutter runout phenomenon in micro milling

In-process geometrical parameters of cutter runout and cutting edges in micro milling are very difficult to be measured online due to the limitations of techniques and operation space. This article presents a deep learning-based method to identify the in-situ cutter runout parameters and cutting edges radii in micro milling without the requirement of complicated and nearly impossible online measurements. The great advancement lies in that the in-situ radii differences in relation to different cutting edges are separately characterized and identified, and then combined with the radial cutter runout model to study the micro milling mechanics. An implicit mapping relationship between the geometrical parameters (i.e. runout parameters and cutting edges radii) and the features of cutting forces is intelligently constructed just by machine learning the theoretical predictions rather than experimental results. The required training data in the intelligent method are predicted by the theoretical cutting force model under different combinations of runout parameters and cutting edges radii. Then, the peak value, valley value and phase width of the predicted cutting forces corresponding to every combination are extracted as crucial features, which are further utilized as inputs of a deep learning model for training the implicit mapping relationship. The runout parameters and radii of cutting edges in actual micro milling are finally identified from the trained relationship just based on the extracted features of measured cutting forces. A series of micro milling experiments conducted with different cutters verify the proposed theoretical method.