Processing noisy cutting force data for reliable calibration of a ball-end milling force model

Machine tool vibrations are common in machining and often result in noisy cutting force measurement. These noisy measurement data compromise the calibration accuracy of empirical cutting force models if they are directly used to determine the empirical model parameters. In this paper, reliable calibration of a ball-end milling force model from noisy cutting force data is addressed. The pertinent section of a noisy cutting force signal measured from a model calibration test cut is first fitted to a polynomial function to reduce signal fluctuation. The empirical parameters of the cutting force model are then solved using an iterative two-stage numerical procedure. Two solution methods with distinct characteristics have been developed and implemented in the present work: the forward and the backward solution method. The former method is a direct solution method and applicable to cutting force data with relatively low levels of noise. The latter method, although an approximate solution method, is more tolerant to the increased noise magnitude in the force data. The applicability of these two solution methods for reliably determining valid empirical parameters of the ball-end milling force model from noisy cutting force data was demonstrated and evaluated.