彭志伟,武川,王园园,等.基于机器学习与有限元相结合的Ti6554合金热挤压棒材组织均匀性调控[J].精密成形工程,2024,16(8):91-101.
PENG Zhiwei,WU Chuan,WANG Yuanyuan,et al.Microstructure Uniformity Regulation of Ti6554 Hot-extrusion Bar Based on Finite Element Simulation Combined with Machine Learning[J].Journal of Netshape Forming Engineering,2024,16(8):91-101. |
| 基于机器学习与有限元相结合的Ti6554合金热挤压棒材组织均匀性调控 |
| Microstructure Uniformity Regulation of Ti6554 Hot-extrusion Bar Based on Finite Element Simulation Combined with Machine Learning |
| 投稿时间：2024-02-03 |
| DOI：10.3969/j.issn.1674-6457.2024.08.011 |
| 中文关键词: Ti6554钛合金 热挤压 随机森林 遗传算法 均匀性调控 |
| 英文关键词: Ti6554 alloy hot-extrusion random forest genetic algorithm uniformity regulation |
| 基金项目:国家自然科学基金(52075386);天津市自然科学基金多投入重点项目(22JCZDJC00650);中国博士后科学基金第67项研究基金(2020M672309);陕西省高性能精密成形技术与装备重点实验室项目(PETE2019KF02) |
|
| 摘要点击次数: 244 |
| 全文下载次数: 41 |
| 中文摘要: |
| 目的 寻求一种快速且高效的方法，以获得微观组织分布均匀的Ti6554钛合金棒材制备工艺参数。方法 通过实验验证了Ti6554钛合金棒材热挤压有限元模型的准确性，并基于有限元模拟获取的晶粒尺寸数据，构建以工艺参数为输入特征量、以组织均匀度为输出的机器学习数据集。随后，采用随机森林回归(RFR)算法构建Ti6554钛合金棒材的组织均匀度预测模型，同时采用粒子群(PSO)算法对模型进行优化，以提高预测精度。在此基础上，利用粒子群优化随机森林回归(PSO-RFR)模型映射的数据，建立了关于挤压速度v、挤压温度 、挤压比 和组织均匀度ΔDDRX的多项式方程，并利用遗传算法(GA)在搜索域进行目标搜索，最终确定最佳工艺组合。结果 与RFR算法相比，PSO-RFR算法构建的预测模型有更好的预测精度，采用GA算法获得的工艺组合(v=47.9 、 =989 ℃、g=9.06)制备棒材时，棒材微观组织晶粒细化、分布均匀性好。结论 相较于实验或者有限元的传统方法，该调控方法不仅节约了时间，还提高了效率，同时也为热挤压棒材的组织均匀性调控提供了新的思路。 |
| 英文摘要: |
| The work aims to find a fast and efficient way to obtain process parameters for the preparation of Ti6554 titanium alloy bars with uniform microstructure distribution. The accuracy of the finite element model for hot extrusion of Ti6554 titanium alloy bars was verified through experiments, and a machine learning dataset with process parameters as input feature variables and microstructure uniformity as output was constructed based on the grain size data obtained from finite element simulation. Subsequently, the random forest regression (RFR) algorithm was used to construct the microstructure uniformity prediction model of Ti6554 titanium alloy bars, and the particle swarm (PSO) algorithm was also used to optimize the model to improve the prediction accuracy. On this basis, polynomial equations for extrusion speed (v), extrusion temperature (t0), extrusion ratio (γ), and microstructure uniformity (ΔDDRX) were established by the data mapped by the particle swarm optimized random forest regression (PSO-RFR) model, and the genetic algorithm (GA) was used to perform the target search in the search domain to finally determine the optimal process combination. The results demonstrated that, compared with the RFR algorithm, the prediction model constructed by the PSO-RFR algorithm had better prediction accuracy, and when the process combination (v=47.9 mm/s, t0=989 ℃, and γ=9.06) obtained by GA algorithm was used to prepare bars, the microstructure of the bars was refined and the distribution was uniform. Compared with the traditional methods of experiment or finite element, this regulation control method not only saves time and improves efficiency, but also provides a new idea for the regulating the microstructural uniformity of hot-extrusion bars. |
|
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