TY - JOUR
T1 - hp-version finite elements for the space-time domain
AU - Peters, D. A.
AU - Izadpanah, A. P.
PY - 1988/3
Y1 - 1988/3
N2 - A bilinear formulation of elasto-dynamics is offered which includes, as a special case, "Hamilton's law of varying action". However, the more general bilinear formulation has several advantages over Hamilton's law. First, it admits a larger class of initial-value and boundary-value problems. Second, in its variational form, it offers physical insight into the so-called "trailing terms" of Hamilton's law. Third, numerical applications (i.e., finite elements in time) can be proven to be convergent under correct application of the bilinear formulation, whereas they can be demonstrated to diverge for specific problems under Hamilton's law. Fourth, the bilinear formulation offers automatic convergence of the "natural" velocity end conditions; while these must be constrained in present applications of Hamilton's law. Fifth, the bilinear formulation can be implemented in terms of a Larange multiplier that gives an order of magnitude improvement in the convergence of velocity. This implies that, in this form, the method is a hybrid finite-element approach.
AB - A bilinear formulation of elasto-dynamics is offered which includes, as a special case, "Hamilton's law of varying action". However, the more general bilinear formulation has several advantages over Hamilton's law. First, it admits a larger class of initial-value and boundary-value problems. Second, in its variational form, it offers physical insight into the so-called "trailing terms" of Hamilton's law. Third, numerical applications (i.e., finite elements in time) can be proven to be convergent under correct application of the bilinear formulation, whereas they can be demonstrated to diverge for specific problems under Hamilton's law. Fourth, the bilinear formulation offers automatic convergence of the "natural" velocity end conditions; while these must be constrained in present applications of Hamilton's law. Fifth, the bilinear formulation can be implemented in terms of a Larange multiplier that gives an order of magnitude improvement in the convergence of velocity. This implies that, in this form, the method is a hybrid finite-element approach.
UR - https://www.scopus.com/pages/publications/0003022338
U2 - 10.1007/BF00317056
DO - 10.1007/BF00317056
M3 - Article
AN - SCOPUS:0003022338
SN - 0178-7675
VL - 3
SP - 73
EP - 88
JO - Computational Mechanics
JF - Computational Mechanics
IS - 2
ER -