PROCESS MONITORING AND TOOL BREAKAGE RECOGNITION IN MILLING

The main aim of this thesis is to develop a robust, adaptive inprocess tool breakage detection algorithm. Tool breakage, unless recognized in time, can lead to irreparable damage to the workpiece and even the machine tool itself. In this thesis, two independent inprocess tool breakage detection algorithms have been developed using cutting force signals.

A detailed mechanistic model of milling forces has been developed. The influences of various transients on cutting forces in milling, such as presence of run-outs, entry, exit, and milling over slots and holes on cutting forces have been examined using this model.

A first order auto-regressive time series (AR(1)) model is shown to be sufficient to detect the tool breakages in milling operations. The method requires a trial cut in order to determine the residuals of the tool breakage thresholds.

An alternative tool breakage detection method, which does not require a trial cut, has been developed for variable mission, flexible manufacturing cells where the batch size can be as small as one work-piece. Transients are separated from the tool breakage by comparing the average force patterns produced by each event. It is shown that the radial throws have an average force pattern which is similar to that occurring during tool breakage. The influence of radial throws is different at different radial widths of cut. To this end, in-process identification of axial depth and radial width of cut has been developed using measured cutting forces. The identification method was used to adapt run-out thresholds to the changes in radial width of cut. Tool breakages, which produce larger deviations than the run-out thresholds, are detected using the identification, the threshold adaptation, and the average force pattern of tool breakage.

The methods proposed have been tested on-line on a vertical milling machine. Successful results were obtained in monitoring the changes in the milling process including tool breakage. The methods developed are practical and can be used for unmanned machine tool monitoring.