基于机器学习的高强高导Cu基复合材料力-电性能统一预测模型研究  

Unified Prediction Model for the Mechanical and Electrical Properties of High Strength and High Conductivity Cu Matrix Composite Materials Based on Machine Learning Algorithms

作　　者：刘楠 郭青成 马麟趾 王嘉琦 LIU Nan;GUO Qingcheng;MA Linzhi;WANG Jiaqi(School of Materials Science and Engineering,Xi'an University of Technology,Xi'an 710048,China;Shaanxi Provincial Key Laboratory of Electrical Materials and Infiltration Technology,Xi'an 710048,China;Conductive Materials and Composite Technology Engineering Research Center of the Ministry ofEducation,Xi'an 710048,China)

机构地区：[1]西安理工大学材料科学与工程学院,陕西西安710048 [2]陕西省电工材料与熔渗技术重点实验室,陕西西安710048 [3]导电材料与复合技术教育部工程研究中心,陕西西安710048

出　　处：《铸造技术》2024年第1期44-49,共6页Foundry Technology

基　　金：国家自然科学基金(51834009,51974244,51605382);西安市科技计划项目(2021SFGX0004)。

摘　　要：颗粒增强铜基复合材料具有良好的力学、电学性能,但增强体特征参量与材料性能之间的定量关系难以量化确定。为建立Ti B和Ti B2陶瓷增强相与铜基复合材料力学与电学综合性能之间的映射关系,以求大幅提高铜基复合材料强度的同时,将其导电率降低在可接受范围内,提出了一种基于蚁群算法优化的BP神经网络铜基复合材料力-电性能统一预测模型(ACO-BP-Cu)。通过BP神经网络建立铜基复合材料性能与特征参数间关系,通过蚁群算法全局寻优确定BP神经网络模型结构。实验表明,ACO-BP-Cu模型能够根据Ti B和Ti B2陶瓷增强相特征参数有效预测铜基复合材料各项性能,且相对决策树、线性回归、K邻近法等9种回归算法准确率更高,稳定性更强。Particle-reinforced copper matrix composites exhibit good force-electric performance,but the quantitative relationship between the characteristic parameters of the reinforcement and the force-electric performance is difficult to quantify.To establish the relationship between the TiB and TiB2 reinforcement and the force-electric performance of copper matrix composites,greatly improve the strength of the copper matrix and control the change in conductivity within an acceptable range,a back propagation(BP)neural network and ant colony optimization based copper matrix composite performance prediction model(ACO-BP-Cu)was proposed.The relationship between the performance of the copper matrix composites and characteristic parameters was determined via a back propagation neural network,and the model structure was determined via global optimization of ant colony algorithms.The results show that the ACO-BP-Cu model can effectively predict the performance of copper matrix composites according to the characteristic parameters of TiB and TiB2,and have higher accuracy and stability compared with 9 regression algorithms including decision tree,linear regression,and K-nearest neighbor algorithms.

关 键 词：铜基复合材料 BP神经网络 蚁群算法 机器学习 导电率 

分 类 号：TB30[一般工业技术—材料科学与工程]