|
|
Prediction of hot deformation behavior of 2205 duplex stainless steel based on PSO-BP |
JI Ya-feng1, WANG Xiao-jun1, MENG Yuan2, WANG Hai-shen3, LIU Yu1, LI Xu4 |
1. School of Mechanical Engineering, Taiyuan University of Science and Technology, Taiyuan 030024, Shanxi, China; 2. Shanxi Information Industry Technology Research Institute Co., Ltd., Taiyuan 030012, Shanxi, China; 3. Qian′an Iron and Steel Company of Shougang Co., Ltd., Tangshan 064402, Hebei, China; 4. State Key Laboratory of Rolling and Automation, Northeastern University, Shenyang 110819, Liaoning, China |
|
|
Abstract Aiming at the problem that the constitutive equation established by traditional mathematical and statistical methods cannot accurately reflect the hot processing properties of materials, a hot deformation prediction model with machine learning combined with optimization algorithm was proposed. A single-pass hot compression test was conducted using a Gleeble-3800 thermal simulation test machine to study the rheological behavior of 2205 duplex stainless steel under the heating temperature of 900-1 100 ℃ and the strain rate of 0.01-10 s-1. The Arrhenius model considering strain correction and the improved BP neural network model based on particle swarm algorithm(PSO-BP) were constructed,and the predicted values of the PSO-BP model were used to draw the hot processing diagram based on dynamic material model (DMM). The prediction ability and stability of the two models were evaluated by calculating the mean square correlation coefficient(R2), root mean square error(RMSE) and average relative error(AARE). The results show that the PSO-BP model has better performance in predicting the flow properties of 2205 duplex stainless steel with R2, RMSE and AARE of 0.999 79, 1.138 7 and 1.43%, respectively, which is a 2% improvement in R2 and a 10.147 6 MPa and 4.979% reduction in RMSE and AARE, respectively, relative to the strain-compensated Arrhenius constitutive model. Moreover, the drawn hot processing map is in good agreement with the experimental hot processing map, and the optimal processing intervals are from 1 000 to 1 100 ℃ for processing temperature and 3.5 to 10 s-1 for strain rate. Therefore, the proposed PSO-BP model has strong reliability and applicability, can accurately describe the hot deformation characteristics of 2205 duplex stainless steel, and provide theoretical guidance and technical support for the forging, rolling and other processes of 2205 duplex stainless steel.
|
Received: 01 August 2022
|
|
|
|
[1] 宋志刚,丰涵,吴晓涵,等. 中国双相不锈钢的发展及研究进展[J].中国冶金,2022,32(6):2.(SONG Zhi-gang,FENG Han,WU Xiao-han,et al. Development and research progress of duplex stainless steel in China[J].China Metallurgy,2022,32(6):2.) [2] XUE R J,WANG X C,YANG Q et al. Grain size distribution characterization of aluminum with a particle swarm optimization neural network using laser ultrasonics[J].Applied Acoustics,2021,180(12):108125. [3] 于加学,孙杰,张殿华. 基于深度学习的热轧带钢头部厚度的命中预测[J].钢铁,2021,56(9):19.(YU Jia-xue, SUN Jie,ZHANG Dian-hua. Accurate prediction of head thickness of hot-rolled strip based on deep learning[J].Iron and Steel,2021,56(9):19.) [4] 王东城,徐扬欢,段伯伟,等. 数据驱动的热轧带钢边部线状缺陷智能预报模型[J].钢铁,2020,55(11):82.(WANG Dong-cheng,XU Yang-huan,DUAN Bo-wei et al. Data-driven intelligent prediction model of edge seam defects for hot rolling strip[J].Iron and Steel,2020,55(11):82.) [5] JI Y F,SONG L B,SUN J et al. Application of SVM and PCA-CS algorithms for prediction of strip crown in hot strip rolling[J].Journal of Central South University,2021,28(8):2333. [6] 汪雅婷,黎俊良,袁楷峰,等. 基于GA改进BP神经网络预测热变形流变应力模型的建立[J].材料工程,2022,50(6):170.(WANG Ya-ting,Ll Jun-liang,YUAN Kai-feng et al. Establishment of hot deformation flow stress prediction model based on GA improved BP neural network[J].Journal of Materials Engineering,2022,50(6):170.) [7] 任军帅,李欣琳,肖松涛,等. 新型Ti-Al-Zr-Nb-Mo-Si钛合金热变形行为及基于BP神经网络模型的本构关系研究[J].材料导报,2020,34(s1):283.(REN Jun-shuai,LI Xin-lin,XIAO Song-tao et al. Hot deformation behavior and constitutive relationship of Ti-Al-Zr-Nb-Mo-Si alloy based on artificial neural network model[J].Materials Reports,2020,34(s1):283.) [8] 何婵,邹德宁,赵洁,等. 2507超级双相不锈钢的流变应力本构关系及热加工图[J].金属热处理,2022,47(1):94.(HE Chan,ZOU De-ning,ZHAO Jie,et al. Flow stress constitutive relationship and processing map of super duplex stainless steel 2507[J].Heat Treatment of Metals,2022,47(1):94.) [9] 李华英,刘国祥,宋耀辉,等.含4.35%铜抗菌不锈钢的热变形行为[J].钢铁,2022,57(9):123.(LI Hua-ying,LIU Guo-xiang,SONG Yao-hui et al. Hot deformation behavior of antibacterial stainless steel containing 4.35% copper[J]. Iron and Steel,2022,57(9):123.) [10] 王庆娟,王钦仁,杜忠泽,等. 40Cr10Si2Mo钢的热变形模型及动态再结晶行为[J].钢铁,2021,56(11):112.(WANG Qing-juan,WANG Qin-ren,DU Zhong-ze et al.Hot deformation model and dynamic recrystallization behavior of 40Cr10Si2Mo steel[J]. Iron and Steel,2021,56(11):112.) [11] 马昕,许斯洋,周舸,等.Ni60Ti40形状记忆合金的热变形行为[J].中国冶金,2022,32(9):26.(MA Xin,XU Si-yang,ZHOU Ge, et al.Thermal deformation behavior of Ni60Ti40 shape memory alloy[J].China Metallurgy,2022,32(9):26.) [12] 万鹏,王克鲁,鲁世强,等. 基于应变补偿和PSO-BP神经网络的Ti-2.7Cu合金本构关系[J].材料工程,2019,47(4):113.(WAN Peng,WANG Ke-lu,LU Shi-qiang et al. Constitutive modeling of Ti-2.7Cu alloy based on strain compensation and PSO-BP neural network[J].Journal of Materials Engineering,2019,47(4):113.) [13] 姜炳春,卢立伟,文泽军,等. LZ92镁锂合金流变应力预测模型[J].材料热处理学报,2020,41(9):147.(JIANG Bing-chun,LU Li-wei,WEN Ze-jun et al. Flow stress prediction model of LZ92 magnesium lithium alloy[J].Transactions of Materials and Heat Treatment,2020,41(9):147.) [14] 刘书含,孙文强,石晓星,等. 基于BP神经网络的热风炉群煤气消耗量预测[J].中国冶金,2022,32(2):77.(LIU Shu-han,SUN Wen-qiang,SHI Xiao-xing et al. Prediction of gas consumption of a hot blast stove group based on BP neutral network[J].China Metallurgy,2022,32(2):77.) [15] 陈兵,唐晓垒,韩烬阳,等. 基于粒子群算法的拉矫机工艺参数优化设计[J].钢铁,2021,56(4):111.(CHEN Bing,TANG Xiao-lei,HAN Jin-yang, et al. Optimization design of tension leveler process parameters based on particle swarm optimization[J].Iron and Steel,2021,56(4):111.) [16] 岳远旺,温彤,刘澜涛,等. 基于BP神经网络预测的TC4热加工图[J].稀有金属,2014,38(4):567.(YUE Yuan-wang,WEN Tong,LIU Lan-tao, et al. Predicted processing map of TC4 titanium alloy based on BP neural network[J]. Chinese Journal of Rare Metals,2014,38(4):567.) |
[1] |
LI Shi-qian, HE Tao, DU Xiang-yang, HUO Yuan-ming, JIA Dong-sheng, LI Han-lin. High temperature flow stress prediction of LZ50 steel based on modified J-C and Z-A models[J]. Iron and Steel, 2023, 58(4): 148-156. |
[2] |
ZHANG Jing,WANG Cunyu,WANG Hui,CAO Chenxing,CAO Wenquan. Research on hot deformation behavior of austenite Fe30Mn9Al09C low density steel[J]. JOURNAL OF IRON AND STEEL RESEARCH , 2023, 35(4): 434-442. |
[3] |
HU Zhiqiang1,2, WANG Kaikun2, ZHAN Jianming1, SHI Hongqi1. Investigation on hot deformation behavior of new hot work die steel 5CrNiMoVNb[J]. JOURNAL OF IRON AND STEEL RESEARCH , 2023, 35(4): 443-453. |
[4] |
LI Wei-gang, LIU Wei-ji, XIE Lu, ZHAO Yun-tao. Rolling force prediction of hot rolled strip by graph convolutional networks[J]. Iron and Steel, 2023, 58(3): 89-96. |
[5] |
ZHENG Ce, LI Ying-ju, TAO Wen-zhe, LUO Tian-jiao, YANG Yuan-sheng. Hot deformation microstructure and behavior of silicon steel solidified under low voltage pulsed magnetic field[J]. Iron and Steel, 2023, 58(2): 159-167. |
[6] |
LI Hua-ying, LIU Guo-xiang, SONG Yao-hui, ZHAO Hai-quan, LI Juan, TIAN Ying-hao. Hot deformation behavior of antibacterial stainless steel containing 4.35% copper[J]. Iron and Steel, 2022, 57(9): 123-129. |
|
|
|
|