More comprehensive and flexible process models have been developed, adapted and tested for a new intelligent machining control scheme. These models provide on-line estimates of tool wear and tool wear rate, and off-line evaluations of different constraints according to process capacity and part quality requirements.

In the first part of this thesis, different modelling techniques, including statistical and neural network techniques have been applied to build constraint variable models for cutting power, noise level, chip merit mark and surface roughness. Once the constraint models are defined, a rule-based system guides their applications according to the particular requirements and process set-up available. Experimental validations and comparisons of the models has been carried out and demonstrated the advantage of using the neural networks over other techniques.

Modelling for on-line process optimization involves tool wear and tool wear rate estimation. In the second part of the thesis, analytical functions have been derived from the theory of contact mechanics, allowing exceptionally good estimates of compressive and shear stresses in the chip-tool and workpiece-tool interfaces. Such interfacial stress modelling served as a baseground for parameter estimation of tool wear and tool wear rate. These models have been tested successfully by simulation on several alloys including SAE 1112, SAE 4135 and other soft steels, and using carbide and HSS tools. The experimental data for the simulation have been provided by extensive process monitoring.

Demonstration of the entire approach for IMS has indicated great promise for future development and application to intelligent machining control.