The ability to predict, measure and control the geometric accuracy of the machining process has become of significant importance to industry in its pursuit of successful competition in today's market place. This thesis describes the development and integration of the different subsystems required for the implementation of geometric adaptive control in bar turning. These subsystems include modeling of the machining system, on-line measurement of the workpiece geometry, tool actuation and predictive control design.

A mechanistic dynamic model of the machining system in bar turning is developed. The model accounts for the rather complex geometry of a multi-edged cutting tool and considers the dynamics of both the tool and workpiece. Also, the simulation of machining dynamics is significantly enhanced by incorporating a mechanistic ploughing force model which accounts for process damping. The model enables the simulation of the process for a wide range of cutting conditions and the prediction of cutting forces, workpiece accuracy and machined surface topography. The model predictions are verified using experimental results of turning hardened steel using ceramic inserts.

A novel non-contact sensor is then developed for on-line high speed monitoring of the workpiece diameter while using a cutting fluid. The measurement system employs three ultrasonic transducers and provides an absolute diameter measurement. An algorithm based on multi-probe measurements is developed to process the radius data and simultaneously compensate for sensor misalignment by tracking the center-line of the workpiece. The system is incorporated into a CNC lathe and provides an accuracy of ±5 μm within a working range of 20 mm. The accuracy and repeatability of the measuring system are tested experimentally under realistic cutting conditions. The applicability of the measurement system to provide complete on-line assessment of workpiece geometry is also demonstrated by evaluating the workpiece geometrical tolerances. The influence of different parameters such as fitting objective functions, size of data sets, and data conditioning on such strategy are investigated.

A model-based predictive controller is designed and implemented to minimize the form error of the workpiece in bar turning. The mechanistic model of the machining system is simplified and reformulated into the state space to permit design and real-time implementation of the control system. The measurement system is set as close as possil.ie to the cutting edge (at a distance of 0.27 mm). An optimal regulator with a Kalman Filter is designed. The on-line measurement of the workpiece diameter is used through the Kalman filter to update the model predictions with unmodeled process disturbances such as tool wear and thermal deflections of the tooling system. The system is implemented on a CNC lathe retrofitted with an Open Architecture Controller. The output of the geometric controller is used to govern the position of the tool tip in real-time by commanding the same servo motor which is driven by the preprogrammed G-code. The control system is tested experimentally under different cutting conditions and is found to provide a significant improvement of more than 90% in workpiece geometric accuracy.