The FE$^2$ method is a very flexible but computationally expensive tool for multiscale simulations. In conventional implementations, the microscopic displacements are iteratively solved for within each macroscopic iteration loop, although the macroscopic strains imposed as boundary conditions at the micro-scale only represent estimates. In order to reduce the number of expensive micro-scale iterations, the present contribution presents a monolithic FE$^2$ scheme, for which the displacements at the micro-scale and at the macro-scale are solved for in a common Newton-Raphson loop. In this case, the linear system of equations within each iteration is solved by static condensation, so that only very limited modifications to the conventional, staggered scheme are necessary. The proposed monolithic FE$^2$ algorithm is implemented into the commercial FE code Abaqus. Benchmark examples demonstrate that the monolithic scheme saves up to ~60% of computational costs.
翻译:FE$2美元的方法是一个非常灵活但计算成本昂贵的多尺度模拟工具。 在常规执行中,微粒变位在每个大型迭代圈内被迭代解决,尽管作为微尺度边界条件而强加的宏观变位株只代表估计数。为了减少昂贵的微尺度迭代次数,本贡献提供了一种单一的FE$2美元的办法,在微型和宏观变位上,在普通的牛顿-拉夫森循环中解决。在这种情况下,每个迭代内方程式的线性系统通过静态凝聚解决,因此只需要对常规的、交错的公式作非常有限的修改。提议的单项变位法在商业的FE代码Abaqus中实施。基准例子表明,单项变位办法节省了约60%的计算费用。