目的 通过机器学习方法预测三维热拉弯的回弹量,以解决传统解析方法在复杂工艺参数下计算繁琐、流程复杂等问题。方法 首先进行了TC4钛合金三维热拉弯试验,探究不同工艺参数对成形质量的影响,随后构建了三维热拉弯数值仿真模型,并验证了数值模型的准确性,然后通过参数化建模方法,获得了包括温度、水平弯曲角度、垂直弯曲角度、摩擦系数4个参数在内的2 430组高质量的训练数据集,最后构建了极限梯度提升（XGBoost）和随机森林（RF）2种机器学习预测模型,通过对预测结果的分析,证明预测模型的准确性,然后与试验数据进行对比,以此来验证模型的有效性与稳定性。结果 XGBoost算法的决定系数R2达到0.98,均方根误差（RMSE）为0.55,随机森林算法的决定系数R2为0.94,RMSE为0.63。结论 XGBoost算法表现出最高的预测精度,且XGBoost算法对三维拉弯回弹量的预测误差不超过3.4%。说明通过XGBoost算法可以实现多点三维热拉弯成形回弹变形的高精度预测。
Abstract
The work aims to employ machine learning methods to predict the springback amount in 3D hot stretch bending, addressing the computational complexity and intricate procedures of traditional analytical methods under complex process parameters. Firstly, a three-dimensional hot stretch bending experiment of TC4 titanium alloy was conducted to investigate the effect of different process parameters on forming quality. Subsequently, a 3D hot stretch bending numerical simulation model was established, and the accuracy of the numerical model was verified. Then, a parametric modeling approach was adopted to obtain a high-quality training dataset containing 2 430 samples with four parameters, including temperature, horizontal bending angle, vertical bending angle, and friction coefficient. Finally, two machine learning prediction models, eXtreme Gradient Boosting (XGBoost) and Random Forest (RF), were constructed. The prediction results were analyzed to demonstrate the accuracy of the prediction model and comparisons with experimental data were made to validate the effectiveness and stability of the model. The XGBoost algorithm achieved a coefficient of determination (R2) of 0.98 and a root mean square error (RMSE) of 0.55, while the random forest algorithm yielded an R2 of 0.94 and an RMSE of 0.63. Analysis indicates that the XGBoost algorithm exhibits the highest prediction accuracy, with its prediction error for 3D stretch bending springback not exceeding 3.4%. This demonstrates that the XGBoost algorithm can achieve high-precision prediction of springback deformation in multi-point 3D hot stretch bending forming.
关键词
机器学习 /
三维热拉弯 /
预测模型 /
TC4钛合金 /
极限梯度提升
Key words
machine learning /
3D hot stretch bending /
prediction model /
TC4 titanium alloy /
XGBoost
{{custom_sec.title}}
{{custom_sec.title}}
{{custom_sec.content}}
参考文献
[1] FU K N, ZHAO Z W, PENG H L, et al.Manufacturing of Ultra-Thin Large Titanium Alloy Tube Using the Novel Hot Gas Pressure-Bending Process[J]. Journal of Materials Processing Technology, 2024, 326: 118358.
[2] QIAO Z, ZHAI R X, TIAN L J, et al.Springback Prediction and Control of TC4 Titanium Alloy Electrically Assisted Stretch-Bending Based on Electro-Thermal- Mechanical Coupling Constitutive Model[J]. The International Journal of Advanced Manufacturing Technology, 2025, 139(3): 1451-1465.
[3] SHAO Y L, LIU Z Z, DING H C, et al.Characterization of the Deformation Behavior and Microstructure Evolution in Laser Bending of TC4 Titanium Alloy Heavy Plate[J]. Journal of Laser Applications, 2024, 36(2): 022001.
[4] DENG X W, HUI S X, YE W J, et al.Numerical Simulation Research on Hot Stretch Straightening of Thin-Walled T-Shaped Profile of Titanium Alloy Assisted by Resistance Heating[J]. The International Journal of Advanced Manufacturing Technology, 2022, 121(9): 6483-6495.
[5] YANG X M, DANG L M, WANG Y Q, et al.Springback Prediction of TC4 Titanium Alloy V-Bending under Hot Stamping Condition[J]. Journal of Central South University, 2020, 27(9): 2578-2591.
[6] DENG R, XUE S, HE Q Q, et al. Research on the Effect of Quasi-β Forging Deformation on the Structure and Properties of TC21 Titanium Alloy[J]. Journal of Materials Engineering and Performance, 2025, 34(15): 15970-15979.[LinkOut]
[7] 高嵩, 于鹏, 梁继才, 等. 型材三维拉弯成形数值模拟及回弹特性研究[J]. 锻压技术, 2018, 43(6): 53-59.
GAO S, YU P, LIANG J C, et al. Research on Numerical Simulation and Springback Characteristic for 3D Stretch-Bending Parts[J]. Forging & Stamping Technology, 2018, 43(6): 53-59.[维普]
[8] 高嵩, 吴宇航, 李奇涵, 等. 铝型材三维拉弯成形回弹预测的理论解析与数值模拟[J]. 锻压技术, 2021, 46(4): 143-149.
GAO S, WU Y H, LI Q H, et al.Theoretical Analysis and Numerical Simulation on Springback Prediction for 3D Stretch Bending of Aluminum Profile[J]. Forging & Stamping Technology, 2021, 46(4): 143-149.
[9] 高嵩, 于长春, 梁继才, 等. 铝型材多点三维拉压复合弯曲成形工艺[J]. 机械工程学报, 2019, 55(20): 152-159.
GAO S, YU C C, LIANG J C, et al.Multi-Points 3D Stretch-Press Bending Technology for Aluminum Profile[J]. Journal of Mechanical Engineering, 2019, 55(20): 152-159.
[10] 惠生猛, 毛晓博, 湛利华. 铝锂合金回弹预测的机器学习及有限元仿真与实验[J]. 中国机械工程, 2024, 35(12): 2114-2121.
HUI S M, MAO X B, ZHAN L H.Machine Learning and Finite Element Simulation and Experimentation for Springback Prediction of Al-Li Alloys[J]. China Mechanical Engineering, 2024, 35(12): 2114-2121.
[11] 刘贵彬, 解小琴, 王平, 等. 15CrMoG钢管弯曲回弹的改进PSO-BP评估[J]. 机械设计与制造, 2023(10): 130-133.
LIU G B, XIE X Q, WANG P, et al.Improved PSO-BP Evaluation of 15CrMoG Steel Pipe Bending[J]. Machinery Design & Manufacture, 2023(10): 130-133.
[12] 王睿乾, 潘林, 童志远, 等. 航空发动机导管CNC弯曲回弹智能预测及补偿系统开发与应用[J]. 航空制造技术, 2024, 67(1): 106-111.
WANG R Q, PAN L, TONG Z Y, et al.Development and Application of Intelligent Prediction and Compensation System for Springback of Aero-Engine Pipes during CNC Bending[J]. Aeronautical Manufacturing Technology, 2024, 67(1): 106-111.
[13] 梁继才, 李义, 高嵩, 等. 型材多点柔性拉弯成形回弹预测[J]. 吉林大学学报(工学版), 2017, 47(1): 185-190.
LIANG J C, LI Y, GAO S, et al.Springback Prediction for Multi-Point 3D Stretch Bending Profile[J]. Journal of Jilin University (Engineering and Technology Edition), 2017, 47(1): 185-190.
[14] 陈光耀, 李恒, 贺子芮, 等. 基于机器学习的管材弯曲回弹有效预测与补偿[J]. 中国机械工程, 2020, 31(22): 2745-2752.
CHEN G Y, LI H, HE Z R, et al.Effective Prediction and Compensation of Springbacks for Tube Bending Using Machine Learning Approach[J]. China Mechanical Engineering, 2020, 31(22): 2745-2752.
[15] ZHANG S Y, FU M Y, WANG Z L, et al.Springback Prediction Model and Its Compensation Method for the Variable Curvature Metal Tube Bending Forming[J]. The International Journal of Advanced Manufacturing Technology, 2021, 112(11): 3151-3165.
[16] 占少伟, 龚俊杰, 韦源源, 等. 基于DPSO-BP神经网络的V形自由折弯成形角度和回弹预测[J]. 锻压技术, 2023, 48(8): 151-157.
ZHAN S W, GONG J J, WEI Y Y, et al.Prediction on V-Shaped Free Bending Angle and Springback Based on DPSO-BP Neural Network[J]. Forging & Stamping Technology, 2023, 48(8): 151-157.
[17] 王鹏鹏, 程子詹, 凌强, 等. 基于深度学习神经网络的铝合金型材回弹预测[J]. 锻压技术, 2024, 49(7): 105-111.
WANG P P, CHENG Z Z, LING Q, et al.Springback Prediction of Aluminum Alloy Profile Based on Deep Learning Neural Network[J]. Forging & Stamping Technology, 2024, 49(7): 105-111.
[18] CU S Y, SUN Y, WU K.Machine Learning Methods for Springback Control in Roll Forming[J]. International Journal of Material Forming, 2024, 18(1): 7.
[19] GAO X S, ZHANG K, ZHANG Z T, et al.XGBoost-Based Thermal Error Prediction and Compensation of Ball Screws[J]. Proceedings of the Institution of Mechanical Engineers, Part B: Journal of Engineering Manufacture, 2024, 238(1/2): 151-163.
[20] 钱尼君, 秦智军, 刘华龙, 等. 基于极端梯度提升算法的钢丝绳拉伸疲劳寿命预测[J]. 失效分析与预防, 2025, 20(3): 169-178.
QIAN N J, QIN Z J, LIU H L, et al.Tensile Fatigue Life Prediction of Steel Wire Ropes Based on Extreme Gradient Boosting Algorithm[J]. Failure Analysis and Prevention, 2025, 20(3): 169-178.
[21] TAO H F, CHEN R, XUAN J P, et al.Prioritization Analysis and Compensation of Geometric Errors for Ultra-Precision Lathe Based on the Random Forest Methodology[J]. Precision Engineering, 2020, 61: 23-40.
[22] 刘震锋, 陈建政, 赵春云, 等. 基于随机森林的钢轨轮廓在线识别方法研究[J]. 机械, 2024, 51(10): 52-58.
LIU Z F, CHEN J Z, ZHAO C Y, et al.Online Recognition Method of Steel Rail Profile Based on Random Forest Algorithm[J]. Machinery, 2024, 51(10): 52-58.
[23] SUTOU A, WANG J F.Influence-Balanced XGBoost: Improving XGBoost for Imbalanced Data Using Influence Functions[J]. IEEE Access, 2024, 12: 193473-193486.
[24] CHOI D K.Data-Driven Materials Modeling with XGBoost Algorithm and Statistical Inference Analysis for Prediction of Fatigue Strength of Steels[J]. International Journal of Precision Engineering and Manufacturing, 2019, 20(1): 129-138.
[25] 王孟超, 武川, 孟亚飞, 等. 基于机器学习模具钢大型锻件力学性能预测[J]. 精密成形工程, 2024, 16(12): 198-208.
WANG M C, WU C, MENG Y F, et al.Prediction of Mechanical Properties of Large Forgings in Mold Steel Based on Machine Learning[J]. Journal of Netshape Forming Engineering, 2024, 16(12): 198-208.
基金
国家自然科学基金（51805045）; 吉林省科技发展计划（20240302115GX）; 吉林省博士研究生托举工程（BST202516）
PDF(8425 KB)
渝公网安备 50010702501719号 渝ICP备15012534号-4
