Scholarlink

<p>Double-stage hot deformation tests were implemented to systematically reveal the flow characteristics and microstructure evolution of Ti55511 alloy with fully β phase. The double-stage hot deformation parameters cover wide ranges of strain rates (0.001&#xa0;s<sup>–1</sup>–0.1&#xa0;s<sup>–1</sup>), temperatures (1163–1223&#xa0;K), first-stage strains (0.3–0.9) and inter-stage holding times (0–120&#xa0;s). Experimental results show that the reloading yield stress significantly is lower than the yield stress in first-stage (stage-I) deformation. The main softening mechanisms, static recrystallization (SRX) and metadynamic recrystallization (mDRX), contribute to a decrease in the reloading yield stress in the second-stage (stage-II) deformation. When the inter-stage holding time exceeds 60&#xa0;s, the abnormal grain growth occurs, leading to an increased average grain size. A visco-plastic self-consistent (VPSC) model incorporating double-stage deformation parameters is presented. The model accurately reproduces the microstructure evolution during the double-stage hot deformation. However, its computational efficiency is limited. Therefore, by integrating experimental data with VPSC output, a novel model combining a particle swarm optimization (PSO) algorithm with a long short-term memory (LSTM) network (PSO-LSTM) is introduced to predict flow stress and microstructure evolution. The mean absolute error (<i>MAE</i>), correlation coefficient (<i>R</i><sup>2</sup>) and root-mean-square error (<i>RMSE</i>) values between experimental and predicted stresses of the PSO-LSTM model are 0.6252&#xa0;MPa, 0.9987 and 1.8637&#xa0;MPa, respectively. Additionally, the proposed PSO-LSTM model can accurately predict the average grain size evolution during the double-stage hot deformation.</p> Graphic Abstract <p></p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Modeling Flow Behaviors and Microstructure Evolution of Ti55511 Alloy During the Double-Stage Hot Deformation Process Utilizing Machine Learning Algorithm

  • 期刊论文

  • 发表日期：

    2025年6月24日

查看全文
  • Song Zhang,
  • Y. C. Lin,
  • Yun-Han Ling,
  • Dao-Guang He,
  • Ming-Song Chen,
  • Miao Wan,
  • Gui-Cheng Wu,
  • Ning-Fu Zeng,
  • Hui-Jie Zhang,
  • Majid Naseri

摘要

Double-stage hot deformation tests were implemented to systematically reveal the flow characteristics and microstructure evolution of Ti55511 alloy with fully β phase. The double-stage hot deformation parameters cover wide ranges of strain rates (0.001 s–1–0.1 s–1), temperatures (1163–1223 K), first-stage strains (0.3–0.9) and inter-stage holding times (0–120 s). Experimental results show that the reloading yield stress significantly is lower than the yield stress in first-stage (stage-I) deformation. The main softening mechanisms, static recrystallization (SRX) and metadynamic recrystallization (mDRX), contribute to a decrease in the reloading yield stress in the second-stage (stage-II) deformation. When the inter-stage holding time exceeds 60 s, the abnormal grain growth occurs, leading to an increased average grain size. A visco-plastic self-consistent (VPSC) model incorporating double-stage deformation parameters is presented. The model accurately reproduces the microstructure evolution during the double-stage hot deformation. However, its computational efficiency is limited. Therefore, by integrating experimental data with VPSC output, a novel model combining a particle swarm optimization (PSO) algorithm with a long short-term memory (LSTM) network (PSO-LSTM) is introduced to predict flow stress and microstructure evolution. The mean absolute error (MAE), correlation coefficient (R2) and root-mean-square error (RMSE) values between experimental and predicted stresses of the PSO-LSTM model are 0.6252 MPa, 0.9987 and 1.8637 MPa, respectively. Additionally, the proposed PSO-LSTM model can accurately predict the average grain size evolution during the double-stage hot deformation.

Graphic Abstract