Steel sheet laser welding-optimization with neural networks

ABSTRACTThe employment of laser systems for completely automated welding processes needs an analysis of physical mecha- nisms by means of mathematical models to correlate the characteristic quality indexes of welding to working parameters.In this paper we describe the application of neural networks methods in order to correlate, for different kinds of steel, the form factor and the penetration depth to working parameters and chemical and thermal properties of working materials. The results are quite satisfactory even a direct search of the optimal process conditions is not possible.1. INTRODUCTION Optimization of laser welding or cutting means, generally, to select the variables or parameters affecting the process and to find analytical correlation between these parameters in order to produce an objective function of one or more van-ables to maximize or minimize. Usually this problem, when applied to technological processes, turns out to be analyticallynon-resoluble if the process parameters are not independent and if there are constraints to respect, in these cases it is only