Classification of laser welds by acoustic signature

Abstract

The application of a backpropagation neural network to the classification of acoustical signals emanating from the laser welding process is discussed. The investigations which are discussed demonstrate that, at least in a relatively simple setting, the backpropagation network is capable of determining whether or not a laser weld has achieved full or partial penetration from its acoustical signature. This result is seen as having important implications for future developments in monitoring and control of these processes.<>