Distortion-oriented welding path optimization based on elastic net method and genetic algorithm

The optimization of welding path is one of the most combinatorial problems in robotic welding. This paper is focused on the optimization for both productivity and quality in robotic welding. The productivity related scheduling problem is similar to the well-known traveling salesman problem (TSP). An optimization strategy for TSP based on the “elastic net method” (ENM) and artificial neural network is suggested in this paper. So for the most welding points which have less influence on the final distortion, they can be arranged in a better way that the welding robot can reach all the points and the total distance is the minimum. For the rest of the welding points, distortion and residual stress are the two main factors concerned. A thermo-mechanical model is developed to predict residual stress and distortion, and a welding sequence with minimum distortion has been found using a genetic algorithm (GA). These two solutions are integrated into the whole optimal welding path according to “nearest neighborhood criterion”. Finally, the optimization strategy is successfully carried out on a real product.