E4303焊条力学性能模糊神经网络智能预测  

Intelligent Prediction of E4303 Electrode Mechanical Properties Based on Fuzzy Neural Network

作　　者：黄俊[1] 徐越兰[1] 

机构地区：[1]南京理工大学材料科学与工程学院,江苏南京210094

出　　处：《南京理工大学学报》2011年第2期219-223,共5页Journal of Nanjing University of Science and Technology

摘　　要：为了获得反映焊条原材料成分与其熔敷金属力学性能之间映射关系的预测模型,该文对E4303碳钢焊条进行配方设计和堆焊试验,测定其熔敷金属的抗拉强度、屈服强度、延伸率、冲击功4项力学性能指标。采用自适应模糊神经网络方法建立了直接由焊条原材料成分预测焊条力学性能的模糊神经网络模型。用该模糊神经网络模型对训练样本以外的试验数据进行预测。结果表明,抗拉强度和屈服强度的预测平均相对误差在5%以内,延伸率指标预测平均绝对误差仅为0.021,冲击功指标预测效果与BP网络相比有明显改善,说明该模糊神经网络预测模型能够直接根据焊条原材料成分较准确地预测其熔敷金属的力学性能。To acquire a prediction model reflecting the relationship between primary materials formula and the deposited metal mechanical properties of electrodes,formula design and resurfacing welding experiments are made on E4303 carbon steel electrode.Mechanical properties indexes of deposited metal including tensile strength,yield strength,elongation percentage,impacting works are also measured.Using the method of adaptive fuzzy neural network,a model for predicting electrode mechanical properties directly from primary material components is built.The model is used to predict the experiment data except training samples.Results show that the prediction average relative errors of tensile strength and yield strength are all below 5%,the prediction average absolute error of elongation percentage is only 0.021,the predicting effect of impacting works is improved compared with that using the BP network.This fuzzy neural network prediction model can accurately predict the deposited metal mechanical properties directly from primary material components.

关 键 词：碳钢焊条 模糊神经网络 力学性能 智能预测 

分 类 号：TG407[金属学及工艺—焊接]