Morphological Analysis and Intelligent Optimization of the Fusion Zone in Electrofusion Welding of PE Pipes

Abstract Taking dn400 large-diameter PE electrofusion fittings as the research object, a theoretical model of electrofusion welding was established and experimental verification was completed. Based on this, the influence of welding parameters and wire feeding methods on the morphology of the fusion zone and welding performance was analyzed. A welding database was established using finite element simulation calculations and Latin hypercube sampling (LHS). Random forest (RF) was used to distinguish the fusion state under different welding parameters and wire feeding methods, and BP neural network (BPNN) was used to predict welding performance. The particle swarm (PSO) algorithm was used to optimize the morphology of the fusion zone and welding performance. The results showed: compared with uniform wire distribution, non-uniform wire distribution has a better effect on controlling and improving the morphology and performance indicators of the fusion zone. The RF classification model and BPNN both achieved an accuracy of over 96% in distinguishing the fusion state and predicting welding performance. The PSO optimization algorithm can significantly improve the morphology of the fusion zone, with all welding performance indicators improving by more than 8%.