Characterization of Manufacturing Processes Based upon Acoustic Emission Analysis by Neural Networks

Summary This article describes the application of a Gaussian neural network for the monitoring of manufacturing processes. The network memory is formed by a self-organized learning process which is stimulated by a multi-component input comprised of acoustic emission (AE) signals and process parameters. By using a trained memory and a non-parametric, multi-dimensional regression the process parameters are associatively estimated from AE signals alone. In this article the on-line estimation of surface roughness during grinding, and the classification of tool sharpness in a drilling process are described. The dynamical forecasting of the AE signal generated in drilling is also demonstrated.