BP神经网络预测船用钢焊接接头力学性能研究

Research on BP neural network prediction of marine steel welded joints mechanical properties

采用不同成分母材和焊丝进行焊接工艺试验,研究母材成分、焊材成分、热输入和焊接位置等参数对焊接接头力学性能的影响,为进一步提升模型预测精度,通过遗传算法对BP神经网络进行优化,将优化权值和阈值赋值给BP神经网络进行建模,预测结果表明,优化模型稳定性好,提高了预测的精度和泛化能力,为焊接接头力学性能预测提供了借鉴,对船用钢和焊材冶金成分设计具有借鉴意义。

Welding process experiments were conducted using different compositions of base materials and welding wires to study effects of parameters such as base material composition,welding material composition,heat input,and welding position on welded joints mechanical properties.To improve prediction accuracy of model,genetic algorithm is used to optimize BP neural network.Assign optimization weights and thresholds to BP neural network for modeling,Prediction results indicate that optimized model has good stability,improved prediction accuracy and generalization ability,provided reference for predicting welded joints mechanical properties,and has reference significance for metallurgical composition design of marine steel and welding materials.