【作者】 马成勇； 田志凌； 彭云； 何长红； 张晓牧； 肖红军；

【Author】 Ma Chengyong Tian Zhiling Peng Yun He ChanghongZhang Xiaomu Xiao Hongjun (Central Iron and Steel Research Institute,Beijing, 100081)

【机构】 钢铁研究总院结构材料研究所；

【摘要】 超低碳贝氏体焊接材料的开发与研究是近年来国际上焊接领域研究的热点之一。作为一种全新的焊接材料,其研制过程中必然涉及大量多元非线性因素对材料性能的影响问题,而且影响ULCB焊接材料的众多影响因子之间又存在复杂的交互作用和非线性效应。基于人工智能和多元非线性分析的计算机辅助设计方法,在解决这类问题上具有独到之处,特别适合于研究对象复杂、机理不明或工作量太大而使理论分析难于进行的体系。本文采用基于判别分析、人工神经网络、分类图构造和遗传算法等技术开发的计算机辅助自建模系统,对获得的ULCB焊丝成分及力学性能数据进行处理,建立了ULCB焊丝熔敷金属成分及性能之间的定量模型。经试验验证,建立的神经网络定量模型用于预测ULCB焊丝熔敷金属的强度和塑性,获得了令人满意的效果,尤其是预测熔敷金属抗拉强度及断面收缩率具有很高的精确度。更多还原

【Abstract】 The development of ULCB ( Ultra low carbon bainitic) welding materials is one of the study hotspot in the welding field in recent years. Since it is a new material,the effects of a number of multi-variate nonlinear factors on material properties must be involoved. In addition, there are complex interactions and nonlinear effect among these factors. The CAD ( computer aid design) method based oh artificial intellegance and multivariate nonlinear analysis is suitable to resolve these problems, especially, the system in which the investigating objectives are complex,the mechanism is not clear or workload is too large to analyse theoretically. In this paper,a self-modelling system based on discriminant analysis, artificial neural networks, classification mapping and genetic algorithms is used to deal with the data on the compositions and properties of the ULCB welding material, and the qualitative model, geometrical model and quantitative model are established. The test results show that using the artificial neural network model to forecast the strength and ductility of the ULCB as-deposited metal, satisfied results can be obtained.更多还原