A distributed in-process supervision of milling based on signal processing machining model

This dissertation examines the problem of supervising milling operations. Milling is an important field of manufacturing, and supervision coupled with in situ sensors will provide the means by which in-process control can be implemented.;A model is developed by which the force signal in steady-state, multi-tooth face milling is fully described based on the force of each individual tooth. The single tooth force signal model has been heavily researched, so the multi-tooth model describes the composite signal in terms of what is known from the physical model for single tooth milling. Furthermore, there are important applications of this model in terms of observation of milling features and extraction of parameters of cut. Knowledge obtained from this analysis is then used to describe the events resulting from tool breakage and relate feature sizes to known physical parameters. A real-time computer algorithm has been implemented based on these principles and tested with milling data.;The supervision of a machine tool is modeled as a loosely coupled distributed computer system. This system integrates event detectors into a computationally extensible architecture. The modeling involves an object-oriented hierarchy for supervision functions temporally ordered through threads. In this research a computer numerical controller (CNC) was integrated with a workstation, a PC, and two digital signal processors (DSP). The CNC is normally an autonomous system. Therefore, an interface was developed that provides the CNC as a distributed system resource. DSPs are used to provide fast processing of sensor data to provide rapid recognition of milling events. The workstation ties the system together by means of an ethernet communications network and provides a user interface to the milling operation. An off-line data analysis tool is integrated into this system for testing and validation of algorithms.