The analysis of diffraction patterns obtained from machined surfaces can be used to characterise the microgeometry and, consequently, the production process. Unfortunately, the analysis applied to real surfaces is rather complex. In this paper a new approach based on the technique of the neural networks is proposed to recognise turned surfaces. The diffraction images are acquired and the distributions of the light intensity are used as inputs to the networks. The ability of supervised networks of classifying the surfaces according to the received training is assessed. With unsupervised networks a classification of the surfaces in quality levels is performed. © 1995 CIRP.
File in questo prodotto:
Non ci sono file associati a questo prodotto.