A NEURAL-NETWORK PROCESS MODEL FOR ABRASIVE FLOW MACHINING OPERATIONS

This paper describes the development of a predictive process modeling system for the abrasive flow machining (AFM) process. This process is used for polishing and surface removal of workpieces with an internal flow path. The core of the process modeling system is a set of neural network models that predicts surface finish and dimensional change. Th ese neural network models are then paired with a heuristic search algo rithm to select sets of machine setup parameters for the AFM process. The heuristic search is specifically designed to avoid allowing the ne ural networks to extrapolate. The completed system was validated using several test pieces, and the results were very promising. The system is currently planned for implementation into the production process. T he system has the potential to significantly reduce the development ti me for new applications of the process and can also be used to suggest alternative machine setup parameters when certain media types are una vailable.