Detecting catastrophic failure events in large-scale milling machines

Catastrophic failure in milling machines is a major concern for manufacturers employing these processes in the production of vital parts. Tool chipping or breakage can lead to machine breakdown, which is a costly consequence in today's highly demanding industry. This paper introduces a novel and practical concept for the detection of failure events in milling. Using the historic data of the machining process (a collection of average spindle power signals) the detection algorithm computes discrete probability distributions representing the power consumption profile along finite synchronous process segments. These distributions play a central role in identifying failure; an unexpected occurrence in the process. Using a combination of real data collected from a powerful industrial milling machine and failure disturbance simulations, concept testing results illustrate that the proposed algorithm is capable of promptly detecting catastrophic faults while keeping unnecessary interruptions to the machine operation to a minimum.