We introduce an extension to the Discrete Multiplier Method (DMM), called Minimal $\ell_2$ Norm Discrete Multiplier Method (MN-DMM), where conservative finite difference schemes for dynamical systems with multiple conserved quantities are constructed procedurally, instead of analytically as in the original DMM. For large dynamical systems with multiple conserved quantities, MN-DMM alleviates difficulties that can arise with the original DMM at constructing conservative schemes which satisfies the discrete multiplier conditions. In particular, MN-DMM utilizes the right Moore-Penrose pseudoinverse of the discrete multiplier matrix to solve an underdetermined least-square problem associated with the discrete multiplier conditions. We prove consistency and conservative properties of the MN-DMM schemes. We also introduce two variants - Mixed MN-DMM and MN-DMM using Singular Value Decomposition - and discuss their usage in practice. Moreover, numerical examples on various problems arising from the mathematical sciences are shown to demonstrate the wide applicability of MN-DMM and its relative ease of implementation compared to the original DMM.
翻译:我们引入了“分辨倍增法”(DMM)的扩展,称为“最小值 $@ ell_ 2$ Norm 分辨倍增法”(MN-DMM),其中对多节量的动态系统采用保守的有限差异办法,而不是像原DMM那样,按程序而不是按原DMM的分析来构建。对于具有多节量的大型动态系统,MN-DMM减轻了最初的DMM在构建满足离散倍增条件的保守方案时可能出现的困难。特别是,MN-DMM利用离散倍增倍矩阵的右面摩尔-Penrose假冒来解决与离散倍增量条件有关的最不确定的最低问题。我们证明了MNM-DM计划的一致性和保守性。我们还引入了两种变体――混合的MNM-DMM和MN-DMM,使用Singal值分立,并讨论其在实践中的使用情况。此外,关于数学产生的各种问题的数字实例显示MM的可广泛适用性及其与原DMMMM的相对容易执行。