针刺C/C复合材料结构-工艺-性能一体化软件平台  

Structure-process-performance integrated software platform for needled C/C composite

作　　者：成博 弓站朋 邓俊楷 白侠[2] 李艳[2] 张承双[2,3] CHENG Bo;GONG Zhanpeng;DENG Junkai;BAI Xia;LI Yan;ZHANG Chengshuang(State Key Laboratory for Mechanical Behavior of Materials,Xi'an Jiaotong University,Xi'an 710049,China;Xi'an Aerospace Composites Research Institute,Xi'an 710025,China;Shaanxi Key Laboratory of Aerospace Composites,Xi'an 710025,China)

机构地区：[1]西安交通大学金属材料强度国家重点实验室,西安710049 [2]西安航天复合材料研究所,西安710025 [3]陕西省航天复合材料重点实验室,西安710025

出　　处：《固体火箭技术》2022年第5期807-814,共8页Journal of Solid Rocket Technology

摘　　要：针刺C/C复合材料的纤维排布方式和针刺工艺对其性能具有重要影响,结合材料信息学开发了一套基于Web界面的针刺C/C复合材料结构-工艺-性能一体化软件平台,包含材料数据库和机器学习模型两个模块。将实验、工业生产或数值模拟得到的“结构-工艺-性能”的材料数据存入到已开发的数据库模块中进行有效存储和管理,并利用这些数据,基于神经网络回归算法训练机器学习模型,建立结构-工艺-性能之间的映射关系。针对目前实验数据昂贵且稀缺,采用高通量有限元数值模拟得到针刺C/C复合材料结构-工艺-性能(有效弹性力学参数)的材料参数数据,训练了预测力学性能的机器学习模型,并部署在软件平台中;通过Web用户界面调用机器学习模型,对特定结构和工艺参数条件下的针刺C/C复合材料的有效弹性力学参数进行预测。结果表明,预测值相对误差小于2%,验证了软件平台的功能和用途。利用这一软件平台,有望进一步基于实验数据和对应的机器学习模型对针刺C/C复合材料性能进行预测,从而为设计和优化针刺C/C复合材料的结构参数和工艺参数提供理论指导。The carbon fiber placement mode and needle-punching process of needle-punched C/C composite have an important influence on its mechanical properties.Combined with material information science,a set of structure-process-performance integrat-ed software platform for needle-punched C/C composites based on Web interface was developed,including two modules:material database and machine learning model.The"structure-process-performance"data from experiments,industrial production or numeri-cal simulation were stored in the developed database module for effective storage and management.By using these data,the machine learning model was trained based on neural network regression algorithm,and the mapping relationship among structure-process-per-formance was established.Aiming at current expensive and rare experimental data,high-throughput finite element numerical simula-tion was used to obtain the material parameter data of the structure-process-performance(effective elastic mechanics parameters)of needled C/C composites,and the machine learning model for predicting mechanical properties was trained and deployed in the soft-ware platform.The machine learning model was invoked through the Web user interface to predict the effective elastic mechanical parameters of needled C/C composites under specific structural and process parameters.The results show that the relative error of the predicted values is less than 2%,which verifies the function and application of the software platform.This software platform is expected to further predict material properties based on the experimental data and corresponding machine learning models,so as to provide theoretical guidance for designing and optimizing the structural parameters and process parameters of needled C/C compos-ites.

关 键 词：针刺C/C复合材料 材料信息学 软件平台 机器学习 

分 类 号：V258[一般工业技术—材料科学与工程] TP3111[航空宇航科学与技术—航空宇航制造工程]