点击次数：

论文类型：期刊论文

发表时间：2024-04-26

发表刊物：International Journal of Manufacturing Research

卷号：19

期号：1

页面范围：98-118

ISSN号：1750-0591

关键字：friction stir welding; FSW; temperature; BP neural network; genetic algorithm; particle swarm algorithm

摘要：The thickness of launch vehicle fuel tank reaches 18 mm, and is welded using friction stir welding (FSW) technology. The increase in thickness will directly affect the temperature distribution of welded joint, thereby affecting the welding quality. Temperature measurement experiments of FSW 18 mm thick 2219 aluminium alloy were conducted. Based on experimental data, a prediction model of peak temperature in Nugget Zone based on BP and improved GA-BP neural network was built. The results showed that the GA-BP neural network had higher prediction accuracy. Subsequently, a temperature prediction model of peak temperature both on advancing and retreating sides was established based on GA-BP and PSO-BP neural network. The results indicated that PSO-BP model showed better performance to realise dual-objective prediction. The temperature prediction model achieves accurate prediction of the temperature of FSW 2219 aluminium alloy thick plate, providing reference for the control of welding temperature of the fuel tank.