Sparse self-stress matrices for the finite element force method

From MaRDI portal





The application of finite element algebraic force method produces a system \(NF=P ,\) where \(N\) denotes the equilibrium matrix of order \(n\times m\), \(F\) is the unknown generalized force vector, and \(P\) is the external nodal load vector. The solution \(F\) is obtained as \(F=B_o P+B_X X,\) where \(B_o (m\times n)\) and \(B_X (m\times r)\) are self-stress matrices which satisfy \(NB_o=I\) and \(NB_X=0\), \(r=m-n\), and \(X\) is the redundant force vector obtained from \(B_X^T f B_X X=-B_X^T f B_oP ,\) where \(f\) is the unassembled system flexibility matrix and \(T\) denotes transposition. Clearly, \(B_o\) and \(B_X\) are not unique, and there are several ways to get such a pair. This paper presents an algorithm of obtaining \((B_X^{LU})^T\) from the \(LU\) factorization of \(N\), such that it satisfies the equation \(NB_X=0\) and yields a new \(B_X^G\) which also satisfies this equation. The algorithm performs convenient column operations, similar to those in the \(LU\) factorization, and by forming groups of redundant forces it improves the sparseness. Also, \((B_X^G)T\) has a banded structure. Two examples illustrate the application of the algorithm. The authors report that the actual application of the algorithm produces important computational savings, when compared with the application of another algorithm proposed by the second author in 1979.




Cited in
(18)


Describes a project that uses

Uses Software






This page was built for publication: Sparse self-stress matrices for the finite element force method

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2712710)