基于数据驱动的磁电复合材料性能研究  

Study on the performance of magnetoelectric composites materials based on data-driven

作　　者：王欢[1] 文建彪 李瑨哲 WANG Huan;WEN Jianbiao;LI Jinzhe(College of Civil Engineering,Xiangtan University,Xiangtan 411105,China;School of Mechanical Engineering and Mechanics,Xiangtan University,Xiangtan 411105,China)

机构地区：[1]湘潭大学土木工程学院,湖南湘潭411105 [2]湘潭大学机械工程与力学学院,湖南湘潭411105

出　　处：《湘潭大学学报(自然科学版)》2025年第1期97-107,共11页Journal of Xiangtan University(Natural Science Edition)

基　　金：湘潭大学博士启动基金(KZ08049)。

摘　　要：如何准确且快速预测磁电系数是磁电多功能器件优化设计领域的一大难题.针对这一问题,该文提出了一种新的遗传算法(GA)优化的反向传播(BP)神经网络的预测磁电系数模型.该模型弥补了BP神经网络模型中经验化计算隐含层神经元个数与手动设定模型最佳训练轮次的不足,有效地提高了GA优化后的BP神经网络的预测精度.结果表明,GA优化后的BP神经网络模型的预测指标R 2高达98.49%,比普通BP模型的预测指标R 2高2.37%,同时该模型的误差更低,故GA优化后的BP预测模型在预测磁电系数上有显著的优越性与更高的精确度.另外在GA-BP模型的基础上引入SHAP模型,充分解释输入参数对磁电系数的影响程度.本研究为更快速、简洁地预测磁电系数,优化磁电复合材料设计提供了有效的支持.How to predict the magnetoelectric(ME)coefficient accurately and quickly is a difficult problem in the field of optimization design for magnetoelectric multifunctional devices.To solve this problem,a new model of predicting ME coefficient based on BP neural network optimized by genetic algorithm(GA)is proposed in this paper.This model makes up for the shortcomings of empirically calculating the number of hidden layer neurons and setting the best training epochs of the BP model,and effectively improves the prediction accuracy of BP neural network.The results show that the prediction factor R 2 of the GA-optimized BP neural network model is 98.49%,which is 2.37%higher than this factor value of ordinary BP model,and the errors of GA-optimized BP neural network model is lower.It means the higher accuracy of the GA-optimized BP prediction model in predicting the ME coefficient.In addition,SHAP model is introduced on the basis of GA-BP model to fully explain the influence of input parameters on the ME coefficient.This study provides effective support for predicting the ME coefficient and optimizing the design of ME composite materials more quickly and concisely.

关 键 词：磁电系数 磁电复合材料 GA优化 BP神经网络 SHAP模型 

分 类 号：TB34[一般工业技术—材料科学与工程]