Title: Reliable engineering computations of large infrastructures

Authors: Achintya Haldar

Addresses: Department of Civil Engineering & Engineering Mechanics, University of Arizona, Tucson, AZ 85721, USA

Abstract: Engineering computational power and improved mathematical platforms have advanced exponentially in recent years. The presence of uncertainty must also be considered in such formulation. Simulation-based algorithms have been routinely used to address uncertainty-related problems. However, they have limited application to study the realistic behaviour of large infrastructure, and the computations could be very tedious. Incorporation of uncertainties in the context of reliable computations, applied to three different classes of problems, is discussed in this paper. The first part of the paper presents the incorporation of uncertainty in a finite element based computational formulation, denoted as the stochastic finite element method. However, it becomes very inefficient when applied to time-domain dynamic problems. The concept can be improved further by different mathematical schemes. They are discussed in the second part. In the third part, sophisticated computational schemes are integrated with noise-contaminated measured response information to extract features of practical interest in the context of structural health assessment.

Keywords: reliable engineering computations; reliability evaluation; implicit limit state function; stochastic FEM; finite element method; large infrastructures; nonlinear behaviour; structural health assessment; system identification; noise-contaminated measured responses; Kalman filters; uncertainty; risk assessment; three-storey steel frame structures; earthquakes; seismic loading.

DOI: 10.1504/IJRS.2015.072714

International Journal of Reliability and Safety, 2015 Vol.9 No.2/3, pp.92 - 111

Received: 17 Oct 2014
Accepted: 15 Apr 2015

Published online: 27 Oct 2015 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article