曾青南,王卓健,柴向东.基于GWO-SVR的LPBF TA15钛合金尺寸精度模型预测研究[J].精密成形工程,2024,16(11):117-125.
ZENG Qingnan,WANG Zhuojian,CHAI Xiangdong.Model Prediction of Dimensional Accuracy of LPBF TA15 Titanium Alloy Based on GWO-SVR[J].Journal of Netshape Forming Engineering,2024,16(11):117-125. |
| 基于GWO-SVR的LPBF TA15钛合金尺寸精度模型预测研究 |
| Model Prediction of Dimensional Accuracy of LPBF TA15 Titanium Alloy Based on GWO-SVR |
| 投稿时间：2024-05-20 |
| DOI：10.3969/j.issn.1674-6457.2024.11.014 |
| 中文关键词: 激光粉末床熔化 TA15钛合金 灰狼算法 支持向量回归 尺寸精度 |
| 英文关键词: laser powder bed fusion TA15 titanium alloy Gray Wolf algorithm support vector regression dimensional accuracy |
| 基金项目:航空发动机与燃气轮机基础科学中心重点项目(P2022-DB-Ⅳ-003-001) |
|
| 摘要点击次数: 128 |
| 全文下载次数: 10 |
| 中文摘要: |
| 目的 由于激光粉末床熔化(LPBF)成形过程中熔池存在复杂的热膨胀和收缩行为，导致LPBF成形TA15钛合金零件尺寸与工艺参数之间存在复杂的非线性关系。为了有效预测TA15合金零件的尺寸精度，提出了一种群体智能算法结合支持向量回归(SVR)模型的方法。方法 以激光功率、扫描速度、扫描间距和层厚为实验变量，以成形零件X、Y、Z 3个方向的尺寸精度作为响应。通过正交实验设计生成数据集，建立以LPBF工艺参数为输入、TA15合金尺寸精度为输出的SVR模型。利用灰狼算法对SVR模型的超参数进行优化，并采用统计评价指标与SVR、PSO-SVR模型进行对比。结果 通过粒子群(PSO)算法和灰狼(GWO)算法优化的SVR模型能够有效预测LPBF成形TA15合金不同方向上的尺寸精度;其中GWO-SVR模型的预测精度和效率最高，在X方向尺寸精度预测模型上的决定系数(R2)、平均绝对相对误差(AARE)和均方根误差(RMSE)依次为0.917、19.7%、0.027，在Y方向尺寸精度预测模型上的R2、AARE和RMSE依次为0.906、9.7%、0.021，在Z方向尺寸精度预测模型上的R2、AARE和RMSE依次为0.911、10.3%、0.019。结论 采用GWO优化的SVR模型，其预测性能和计算效率显著提高，研究结果可为增材制造TA15合金的尺寸精度预测提供数据支持和理论参考。 |
| 英文摘要: |
| The complex thermal expansion and contraction behavior of the melt pool during laser powder bed fusion (LPBF) forming results in a complex nonlinear relationship between the dimensions of LPBF formed TA15 titanium alloy parts and the process parameters. In order to effectively predict the dimensional accuracy of TA15 alloy parts, the work aims to propose a method that combines a population intelligence algorithm with a support vector regression (SVR) model. Laser power, scanning speed, scanning pitch and layer thickness were used as experimental variables, and the dimensional accuracy of the formed parts in X, Y and Z directions was used as the response. The data set was generated by orthogonal experimental design to establish the SVR model with LPBF process parameters as input and TA15 alloy dimensional accuracy as output. The hyperparameters of the SVR model were optimized by the Gray Wolf algorithm and compared with the SVR and PSO-SVR models based on statistical evaluation metrics. The SVR model optimized by Particle Swarm (PSO) algorithm and Gray Wolf (GWO) algorithm could effectively predict the dimensional accuracy of LPBF formed TA15 alloy in different directions. Among them, the GWO-SVR model had the highest prediction accuracy and efficiency and the coefficient of determination (R2), average absolute relative error (AARE), and root mean square error (RMSE) of the dimensional accuracy prediction model in X direction were 0.917, 19.7%, and 0.027 in order. The R2, AARE, and RMSE in Y direction were 0.906, 9.7%, and 0.021. The R2, AARE, and RMSE in Z direction were 0.911, 10.3%, and 0.019. The predictive performance and computational efficiency of the SVR model optimized with GWO are significantly improved, and the results of the study can provide data support and theoretical reference for the prediction of dimensional accuracy of additively manufactured TA15 alloy. |
|
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