【摘要】 概述了采矿、岩土、土木等工程材料本构模型辨识及其物理力学参数的反分析等研究 ;针对具体工程材料最佳本构模型的选择 (模型辨识 )及其物理力学参数确定问题 ,视工程对象 (岩土体 )为一复杂系统 ,施工中实测获得的工程对象随时间的响应 (如载荷、变形位移等宏观信息数据 )隐式地蕴含着工程材料的本构信息 ,即工程对象的应力 -应变本构关系和模型的物理力学参数。作为系统识别模拟的理想工具 ,神经网络能够学习模拟到工程材料的本构关系 ,这样将工程对象体的本构模型和参数识别为一“黑箱”,给定某一应变向量就可得到期望的输出应力向量 ,而无需写出显式参数的数学表达式 ,此即基于神经网络的隐式材料本构模型——智能本构模型。更多还原

【Abstract】 This paper reviews the identification of constitutive model of engineering materials and the parameters’ back analysis.Considering the determination of optimal model and corresponding physical parameters,the engineering rock mass is regarded as a complex system.The response corresponding to the varying time during the engineering practices implicitly incorporates the constitutive information of engineering materials,i.e.,the relationship between stresses and strains and the physical and mechanical parameters of constitutive model.As the ideal tool of system identification,neural network is capable of learning and mimicking the constitutive material relationship.Accordingly,the engineering material constitutive modeling and parameters identification can be regarded as a “black box”,given some vectors of strains,the desired stresses can be obtbined without the necessity of explicitly written out mathematical expressions.This is the so called implicit neural constitutive modeling,also named as intelligent constitutive modeling which potentially can be incorporated into a numerical procedure like FE code.更多还原