# Contribution au développement d’outils analytiques et numériques pour quantifier et qualifier la robustesse des structures

Abstract : Localized initial failures in constructions can sometimes be followed by disproportionate damage (collapse) spreading to the whole or the major part of a building. Since the partial and progressive collapse of the Ronnan Point tower (London, $1968$) caused by a gas explosion, the concept of robustness has been introduced in standards. Structural robustness is defined as the ability of a structure to withstand unforeseen events causing local damage like fire, explosion or impact, without suffering disproportionate collapse. This definition encourages engineers to include the concepts of initial damage (local failure) and disproportionate damage (global failure) in design procedures. The main objective of this PhD work is to develop a simulation tool in order to highlight the potential weakness in a structure when uncertain sollicitations (accidental events) and/or dimensional fault (design or realization) interfere with the standard predictions. The robustness is evaluated by an index varying from 0 (non-robust structure) to 1 (very robust structure) and is calculated from the initial and global failure probabilities. The proposed methodology is based on an event tree analysis summurizing all the distinct potential scenarios, from the initial damage to the collapse of the structure. The developed approach is applied to statically indeterminate unidimensional structures like beams and frame. The redundancy's consequence is that the break of one or several cross sections will not necessarily lead to the collapse of the whole system: the redistribution of the internal efforts allows the remaining undamaged parts of the structure to support the external (applied) loading. The methodology is illustrated by some examples of clamped-clamped beam and frame, loaded with punctual forces. The cross sections are supposed to have an elastic behaviour until the formation of plastic hinges (local failure). Two types of probabilistic laws, Gaussian and Log-normal, are tested by the developed approach and by Monte-Carlo simulations. The chosen random variables can be either independent or correlated. The resulting complete event tree contains all the exclusive paths from an localised damage to the global failure, without intersection between branches stemming from the same node. This specific property allows to evaluate the robustness indexes of the structure with the ratio between the local and global probabilities, according to each scenario. The analysis of the event tree and of the robustness indexes allows to highlight the potential brittleness which could cause a generalized collapse of the structure with respect to accidents or malicious acts. The developed methodology provides an effective tool of simulation and diagnostic, both in the design phase and in the rehabilitation one, useful to the reinforcement of existing or future buildings and to ensure the safety of people and surrounding structures.
Keywords :
Document type :
Theses

Cited literature [119 references]

https://tel.archives-ouvertes.fr/tel-01923391
Contributor : Abes Star :  Contact
Submitted on : Thursday, November 15, 2018 - 11:25:12 AM
Last modification on : Thursday, March 5, 2020 - 5:39:08 PM
Long-term archiving on: : Saturday, February 16, 2019 - 1:47:09 PM

### File

2018PA100060.pdf
Version validated by the jury (STAR)

### Identifiers

• HAL Id : tel-01923391, version 1

### Citation

El Hadji Boubacar Seck. Contribution au développement d’outils analytiques et numériques pour quantifier et qualifier la robustesse des structures. Génie civil. Université de Nanterre - Paris X, 2018. Français. ⟨NNT : 2018PA100060⟩. ⟨tel-01923391⟩

Record views