改进HHO优化集成学习的钢材疲劳强度研究
Improved Fatigue Strength of Steel with HHO Optimized Ensemble Learning
马思铭 1艾沛钰 1付敏 1谷志新1
作者信息
- 1. 东北林业大学计算机与控制工程学院,黑龙江 哈尔滨 150040
- 折叠
摘要
钢材疲劳强度是机械部件的设计与失效分析中所需的重要信息,在实际工程中,疲劳载荷作用下的部件往往会出现裂纹甚至断裂,因此疲劳强度的准确预测尤为重要.针对传统S-N曲线计算周期长等问题,采用Stacking集成机器学习方法对钢材疲劳强度进行研究,并采用哈里斯鹰优化算法提高模型准确率,同时引入Piecewise映射、准反向学习,鲸鱼捕食策略共同改进哈里斯鹰算法(POW-HHO),帮助算法跳出局部最优解,提高收敛精度.通过构建的正向预测模型与POW-HHO算法相结合,进行钢材疲劳的逆向设计,采用Null Importance方法进行特征选择以提高设计效率,最终逆向设计结果对钢材疲劳强度的研究具有一定指导意义.实验结果表明,上述模型表现良好,具有正向预测与逆向设计的能力,且准确率较高.
Abstract
The fatigue strength of steel is an important information required in the design and failure analysis of mechanical components.In actual engineering,the components under fatigue load often crack or even fracture,so the accurate prediction of fatigue strength is especially important.In response to the long calculation period of traditional S-N curves and other issues,the fatigue strength of steel is studied by the Stacking ensemble learning method,and the Harris Hawk optimization algorithm is used to improve the model accuracy,while Piecewise mapping,quasi-inverse learning,and whale predation strategy are introduced together to improve Harris Hawk algorithm(POW-HHO),which helps the algorithm jump out of the local optimal solution and improve the convergence accuracy.By combining the constructed forward prediction model with the POW-HHO algorithm,the inverse design of steel fatigue is carried out,and the Null Importance method is used for feature selection to improve the design efficiency,and the final inverse design results are of certain guiding significance for the study of steel fatigue strength.The experimental re-sults show that the model performs well and has the ability of forward prediction and inverse design with a high accu-racy rate.
关键词
钢材疲劳强度/哈里斯鹰算法/集成学习/预测/逆向设计Key words
Fatigue strength of steel/HHO/Ensemble learning/Prediction/Inverse design引用本文复制引用
基金项目
国家自然科学基金资助项目(51975114)
出版年
2024
NETL
NSTL
万方数据