Adaptive finite element analysis and modelling of solids and structures. Findings, problems and trends

Abstract

SUMMARYThe main purpose of this paper is two-fold: first, to present a critical review of available errorcontrolled adaptive finite element methods-with absolute global and goal-oriented error estimatesfor approximate solutions of a given mathematical model and the model error of the mathematical model considered and its dimensions, enhanced with our own recent results in fracture mechanics. The second related important matter is which physical/mathematical models-including various boundary layers and other disturbances-and which accuracy of the finite element solutions in which norm are prescribed and necessary for the safety and reliability of a certain engineering problem, e.g. in structural engineering, of course depending on materials, systems and the type as well as the magnitude of prescribed loadings (i.e. actions in a factorized concept) acting on a specific structure. It is hard to understand why the current codes for structural engineering design in Europe-especially the European Standards (Eurocodes) and the relevant German DIN-Codes (DIN: Deutsches Institut für Normung e. V.)-do not lay out any rules or bounds for the prescribed accuracy of the computation of material and structural resistance, i.e. the validity of mechanical modelling and the accuracy of associated numerical methods, usually today the finite element method. It can be observed that behind the existing codes there still lies the old thinking of, e.g., classical elastic beam theory of so-called second order for stability problems and the corresponding elastic potential for plates and shells in buckling. Modern flexible finite element modelling by deriving theories in variational form for thin-walled beams, plates and shells from rather general 3D theories, for both linear and non-linear problems, capturing inelastic deformations and layer effects by material and dimensional expansions are not addressed in the Eurocodes for structural engineering. The necessary consequence is that joint work of computational mechanics specialists and structural engineers, working in national and international code committees, is absolutely necessary, in order to avoid a lot of costly troubles and dangers of today by improving and updating the existing codes. THE GAP BETWEEN COMPUTABLE ERROR BOUNDS AND THE REQUIREMENTS OF TECHNICAL CODESThe goal of reliable and efficient engineering analysis is the proof of material and structural resistance against all possible loadings (actions) with sufficient safety at minimal cost, guaranteeing full usability within lifetime. Codes, standards and technical rules sufficiently support the engineering design of most technical components in all engineering fields. But requirements and instructions for adequate mechanical modelling and the prescribed accuracy of static and dynamic analysis are not included. The thinking behind these codes is obviously still influenced by classical beam and plate models for which 'exact' analytical or numerical solutions by series are possible under a couple of restricting conditions, bu...

On the adaptive use of the Taylor assumption in computational homogenization of thin metal sheets

“…Most importantly, all these different subscale modeling assumptions translate to the macroscale as macroscale model errors. Methods for estimating the model error have been developed by, for instance, Stein et al [6], Oden and Vemaganti [7], Braack and Ern [8] and Larsson and Runesson [9][10][11]. The purpose of this paper is to extend the results of Lillbacka et al [5] and of Larsson and Runesson [10] with the purpose to investigate the possibility of carrying out fully model-adaptive two-scale CMM computations under the specific assumption of macroscale plane stress.…”

Adaptive hpq finite element methods for the analysis of 3D-based models of complex structures. Part 1. Hierarchical modeling and approximations

“…Also the recent advances in the p-method, closely related to the hp approach, can be noticed here [78,79]. Some open or unresolved problems, as well as the current trends and prospects for the future within the hp and other adaptive methods are addressed in [26,69,83], for example.…”

A hierarchical h adaptivity methodology based on element subdivision

“…The adaptive process is based on the estimation of the error in the magnitude of interest Q(ees) using the recovery type error estimator given in the following expression for element e, as proposed for example in [29] and [30]: (17) to reduce the computational cost asociated to the evaluation of |J|.…”