热处理工艺模拟专用材料数据库的设计与实现  

Design and implementation of materials database for heat treatment process simulation

作　　者：张伦凤 王治涵 赵俊渝 安康 徐骏[1] 顾剑锋[1] Zhang Lunfeng;Wang Zhihan;Zhao Junyu;An Kang;Xu Jun;Gu Jianfeng(Institute of Materials Modification and Modeling,Shanghai Jiao Tong University,Shanghai 200240,China;Chongqing Gearbox Co.,Ltd.,Chongqing 402263,China)

机构地区：[1]上海交通大学材料改性与数值模拟研究所,上海200240 [2]重庆齿轮箱有限责任公司,重庆402263

出　　处：《金属热处理》2023年第9期247-252,共6页Heat Treatment of Metals

基　　金：国家重点研发计划(2018YFA0702905)。

摘　　要：材料参数是热处理工艺模拟中至关重要的数据支撑,但目前国内相关的材料数据库十分缺乏,少数现有数据库存在数据准确性低、完整性差、无法共享等问题,且仅通过化学成分区分材料,无法应对热处理工艺模拟对参数的需求。为此,设计出以化学成分与微观组织为主的数据结构体,自主开发了一款在线的专用材料数据库。该数据库针对热处理工艺模拟所需的材料参数特性,优化设计了数据存储结构。采用B/S架构设计实现了数据共享,同时提高了数据使用便捷性。另外,该数据库通过采用数据挖掘技术,引入了多元线性回归、贝叶斯线性回归、决策树和随机森林4种机器学习算法,建立了一种创新的数据提取机制。可通过对现有数据的应用分析确定有效的数据提取策略,进而获取当前所需的实际数据,初步解决了当前普遍存在的数据缺失问题,有力支撑了热处理工艺模拟的开展。Material parameters are the crucial data support in heat treatment process simulation.However,at present,there is a lack of relevant material databases in China,and a few existing databases have problems such as low data accuracy,poor integrity,and inability to share data,they only distinguish materials based on chemical composition,which cannot meet the parameter requirements of heat treatment process simulation.Therefore,a data structure focusing on chemical composition and microstructure was designed,and an online special material database was also independently developed.The database optimizes the data storage structure according to the characteristics of material parameters required for heat treatment process simulation.Adopting B/S architecture design realizes data sharing and improves the convenience of data use.Furthermore,by using data mining technology,the database introduces four machine learning algorithms:multivariable linear regression,Bayesian linear regression,decision tree,and random forest,and establishes an innovative data extraction mechanism.The effective data extraction strategy can be determined through the application analysis of existing data,and then the actual data requires at present can be obtained,which preliminarily solves the problem of data missing currently,and strongly supports the development of heat treatment process simulation.

关 键 词：材料数据库 热处理 机器学习 数据挖掘 

分 类 号：TG156.6[金属学及工艺—热处理]