统一格式的显式与隐式任意混合异步算法

AN ARBITRARILY MIXED EXPLICIT-IMPLICIT ASYNCHRONOUS INTEGRATION ALGORITHM BASED ON UNIFORM DISCRETIZATION FORMAT

动力学问题的有限元分析需要在每一时步求解系统信息,相对于静力学问题,其计算量要大得多.因而,提高计算效率,节省计算工作量是动力学求解方法研究的主要内容.该文针对大型复杂动力学系统的高效求解问题,提出了一种基于Newmark离散格式的显式、隐式任意混合异步算法,根据整体系统不同局部的物理力学特性和求解精度要求,在空间域及时间域内对动力学系统方程进行多尺度求解.该方法根据显式、隐式算法固有的信息传递机制,采取动态的可变边界处理方法,避免了异步边界上的误差积累;并通过对整体系统能量平衡的校验,动态地确定和修正仿真计算时步,可以有效地预防不稳定性的产生和发展.数值算例表明:该算法能在保持较高的计算精度的同时,极大地降低计算资源消耗,因而具有一定的实用价值.

Dynamical finite element method requires solving system information at each time step, and the computational effort is much larger than solving the static ones. Thus, to improve computational efficiency and save computational effort is one the of the main research content in dynamics. The present paper introduces an arbitrarily mixed explicit-implicit asynchronous integration algorithm based on uniform Newmark discretization format, for the efficiently solving of the large and complex dynamic systems. The overall dynamical system can be partitioned into different parts according to the physical and mechanical properties, as well as the requirements of solution accuracy, and the system equation can be solved in multi-scale both at the space domain and time domain. According to the inherent message passing mechanisms of the explicit and implicit algorithm, a variable boundary treatment method was adopted to avoid the accumulation of errors at the asynchronous boundary. The simulation time steps were dynamically determined and corrected according to the energy balance checking, which can effectively prevent the emergence and development of the instability. Numerical example shows that the proposed algorithm can greatly reduce the consumption of computing resources while maintaining high accuracy, thus it has a high practical value.