The identification of parameters for visco-plastic models via finite-element methods and gradient methods
The identification of parameters for visco-plastic models via finite-element methods and gradient methods
复制标题
通过有限元方法和梯度方法识别粘塑性模型的参数
DOI：
10.1088/0965-0393/2/3a/013
复制
发表时间：
1994
影响因子：
1.8
通讯作者：
E. Stein
中科院分区：
文献类型：
--
作者：
R. Mahnken;E. Stein
在这项工作中，我们提出了一个统一的战略识别的粘塑性模型的材料参数，从测试数据的复杂结构。为了考虑相关的不均匀变形和应力，使用有限元法。最小二乘类型的目标函数通过基于梯度评估的方法来最小化，例如SQP方法或由于Bertsekas的投影算法。详细说明了灵敏度分析，即目标函数梯度的确定。最后给出了一个递推公式。在数值例子中，我们比较了基于梯度的方法与进化方法的齐次问题。关于非齐次问题，我们讨论的结果，由于Steck的材料法。
In this work we present a unified strategy for identification of material parameters of visco-plastic models from test data of complex structures. For consideration of the associated inhomogeneous deformations and stresses the finite-element method is used. The objective function of least-squares type is minimized by a method based on gradient evaluations, such as an SQP method or a projection algorithm due to Bertsekas. The sensitivity analysis, i.e. the determination of the gradient of the objective function, is explained in detail. As a result a recursion formula is obtained. In the numerical examples we compare gradient-based methods with evolutionary methods for homogeneous problems. Concerning inhomogeneous problems we discuss the results obtained for a material law due to Steck.