Analysis of Grinding Process Acoustic Emission by a Neural Network

The article describes how information about the surface produced by a grinding process can be estimated on–line from detected acoustic emission (AE) signals. This estimation is performed by an adaptive measurement system with a neural network structure. The system is used either in a learning or in an analysis mode of operation. During the learning process, the network obtains the complete AE spectrum as well as the surface autocorrelation functions. During the analysis mode of operation, the surface correlation functions are estimated on–line by nonparametric regression from the AE spectra.