碳钢焊条熔敷金属力学性能遗传神经网络预测.docVIP

碳钢焊条熔敷金属力学性能遗传神经网络预测 ---------------------------------------------------------------范文最新推荐------------------------------------------------------ 碳钢焊条熔敷金属力学性能遗传神经 网络预测 摘要 为了更好的反映焊条原材料成分与其熔敷金属力学性能之间的映射关系，本论文对E4303和E5015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。分别采用RBF、BP、GA-BP三种神经网络方法建立由焊条原材料成分预测焊条力学性能的神经网络模型。用这三种神经网络模型对训练样本以外的试验数据进行预测。结果表明用GA-BP结合的计算方法所建立的神经网络模型预测焊条力学性能的平均绝对误差最小。E4303的屈服强度的平均相对误差达到9.83%、抗拉强度的平均相对误差达到13.85%、延伸率的平均相对误差达到13.47%、冲击功的平均相对误差达到9.41%,E5015的屈服强度平均的相对误差达到8.69%、抗拉强度的平均相对误差达到6.41%、延伸率的平均相对误差达到11.25%、冲击功平均的相对误差达到8.35%。7068 1 / 9 关键词碳钢焊条 神经网络 力学性能 预测 遗传算 法 毕业设计说明书,论文,外文摘要 TitlePrediction Of The Deposited Metal Mechanical Properties of Carbon Steel Electrode Based on Genetic Neural Network Abstract 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 E4303and E5015 carbon steel electrode,Mechanical properties indexes of deposited metal including tensile strength，yield strength，elongation percentage，impacting energy ,Using the methods of RBF、BP、GA-BP neural network，three models for predicting electrode mechanical properties directly from primary material components are built ---------------------------------------------------------------范文最新推荐------------------------------------------------------ respectively,The models are used to predict the experiment data except training samples. Results show that the prediction average relative errors of the modle for predicting electrode mechanical properties directly from primary material components base on GA-BP is the least,And the prediction average relative errors of tensile strength of E4303 is9.83%，the yield strength is 13.85%， the elongation percentage is 13.47%，the impacting energy is 9.41%; the prediction average relative errors of tensile strength of E5015 is 8.69%，the yield strength is 6.41%， the elongation percent