张子琪,周祥曼,郑事成,等.基于神经网络的电弧增材制造铝合金力学性能预测[J].精密成形工程,2024,16(1):43-51.
ZHANG Ziqi,ZHOU Xiangman,ZHENG Shicheng,et al.Neural Network-based Prediction of Mechanical Properties of Wire Arc Additively Manufactured Aluminum Alloys[J].Journal of Netshape Forming Engineering,2024,16(1):43-51. |
| 基于神经网络的电弧增材制造铝合金力学性能预测 |
| Neural Network-based Prediction of Mechanical Properties of Wire Arc Additively Manufactured Aluminum Alloys |
| 投稿时间：2023-10-10 |
| DOI：10.3969/j.issn.1674-6457.2024.01.005 |
| 中文关键词: 6061铝合金 TiC增强的6061铝合金 BP神经网络 粒子群算法 遗传算法 力学性能 |
| 英文关键词: 6061 aluminum alloy TiC-reinforced 6061 aluminum alloy BP neural network particle swarm algorithm genetic algorithm mechanical properties |
| 基金项目:国家自然科学基金(51705287);湖北省教育厅科研计划(D20211203) |
|
| 摘要点击次数: 1500 |
| 全文下载次数: 661 |
| 中文摘要: |
| 目的 预测不同工艺参数下电弧增材制造铝合金的力学性能。方法 通过实验建立了电弧增材制造6061铝合金及TiC增强6061铝合金力学性能的数据集，并建立了一种以焊接电流、焊接速度、脉冲频率、TiC颗粒含量为输入，以屈服强度和抗拉强度为输出的神经网预测模型，对比了反向传播神经网络(BP)、粒子群算法优化BP神经网络(PSO-BP)、遗传算法优化BP神经网络(GA-BP)3种预测模型的精度。结果 与BP模型和PSO-BP模型相比，GA-BP预测模型具有更好的预测精度。其中，GA-BP模型预测6061铝合金屈服强度最佳结果的相关系数(R)为0.965，决定系数(R2)为0.93，平均绝对误差(Mean Absolute Error，MAE)为2.35，均方根误差(Root Mean Square Error，RMSE)为2.67;预测TiC增强的6061铝合金抗拉强度最佳结果的R=1，R2高达0.99，MAE为0.46，RMSE为0.49，GA-BP具有良好的预测精度。结论 BP、PSO-BP、GA-BP 3种神经网络模型可以用来预测电弧增材制造铝合金的力学性能，GA-BP模型比其他2种模型的预测精度更优。与传统的实验方法相比，用神经网络模型预测电弧增材制造铝合金力学性能的速度更快，成本更低。 |
| 英文摘要: |
| The work aims to predict the mechanical properties of aluminum alloy produced by wire arc additive manufacturing under different process parameters. In this paper, a data set of mechanical properties of wire arc additively manufactured 6061 aluminum alloy and TiC-reinforced 6061 aluminum alloy was experimentally established. A neural network prediction model was established with welding current, welding speed, pulse frequency, and TiC particle content as inputs, and yield strength and tensile strength as outputs. The accuracy of three prediction models:backpropagation neural network (BP), particle swarm optimization BP neural network (PSO-BP), and genetic algorithm optimization BP neural network (GA-BP) were compared. The results indicated that the GA-BP prediction model had better prediction accuracy than the BP model and the PSO-BP model. Among them, the optimal relationship number R for predicting the yield strength of 6061 aluminum alloy using the GA-BP model was 0.965, with a determination coefficient R2 of 0.93, mean absolute error (MAE) of 2.35, and root mean square error (RMSE) of 2.67; The best result for predicting the tensile strength of TiC reinforced 6061 aluminum alloy was R=1, with R2 as high as 0.99, MAE as 0.46, and RMSE as 0.49. GA-BP had good prediction accuracy. In conclusion, BP, PSO-BP, and GA-BP neural network models can predict the mechanical properties of aluminum alloy produced by wire arc additive manufacturing, and the GA-BP model has better prediction accuracy than the other two. Compared to traditional experimental methods, the method of using neural network models to predict the mechanical properties of wire arc additive manufacturing aluminum alloys is faster and less costly |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
|
|
|
渝公网安备 50010702501719号