基于IHA-TPE-LightGBM融合模型的NiTi基形状记忆合金相变温度预测方法-中国材料进展

文章信息/Info

Title:: Prediction Method for Phase Transition Temperature of NiTi Based Shape Memory Alloy Based on IHA-TPE-LightGBM Fusion Model

Author(s):: LI Jun; XU Liang; CHEN Xiaoran; College of Computer and Control Engineering, Northeastern Forestry University, Harbin 150040, China

Keywords:: NiTi based alloy; genetic algorithm; simulated annealing algorithm; feature selection; tree-structured Parzen estimator; LightGBM

摘要:: 提出了一种基于IHA-TPE-LightGBM的融合模型预测NiTi基形状记忆合金的相变温度(Tp)的方法。融合遗传算法与模拟退火算法形成改进混合算法(improved hybrid algorithm, IHA)，筛选影响Tp的特征，减少特征冗余并优化模型性能；利用非标准贝叶斯优化算法(treestructured Parzen estimator, TPE)优化最佳模型的超参数，提升模型的精度。结果表明，提出的温度预测模型IHA-TPE-LightGBM的R2评价指标为0.92，验证了该方法的有效性。该研究方法有助于开发新型NiTi基形状记忆合金，可以加快未来高性能弹热材料的发现。

Abstract:: This article proposes a fusion model IHA-TPE-LightGBM to predict the phase transformation temperature (Tp) of NiTi based shape memory alloys. The improved hybrid algorithm (IHA) is formed by integrating genetic algorithm and simulated annealing algorithm, it is used to screen features that affect Tp, finally reducing feature redundancy and optimizing model performance. Using non-standard Bayesian optimization algorithm (tree structured Parzen estimator, TPE) to optimize the hyperparameters of the optimal model and improve the accuracy of the model. The results indicate that the R2 evaluation index of the temperature prediction model IHA-TPE-LightGBM is 0.92, which verifies the effectiveness of the method. This method will contribute to the development of new NiTi based shape memory alloys, which can accelerate the discovery of high-performance elastocaloric materials in the future.