Abstracts
Résumé
Dans cet article, on décrit un essai réalisé sous conditions contrôlées en laboratoire concernant l'initiation et le démarrage de la rupture par submersion d'une digue de revanche homogène et en moraine, un matériel utilisé dans les ouvrages réels, de granulométrie étendue et sans cohésion.
Les résultats obtenus sur cette digue expérimentale ont montré que le mécanisme de formation et d'érosion de brèche est très différent du mécanisme global et moyen de développement et de progression de la brèche adopté dans les modèles courants de calcul des ruptures de barrages. L'essai souligne l'importance de la turbulence et des pertes de charge locales dans certaines zones spécifiques de l'écoulement et met en évidence le rôle de l'instabilité des parois latérales de la brèche dont la base est sapée continuellement.
L'essai a été stoppé à un moment précis, soit lorsque l'accroissement rapide du débit de rupture était confirmé, afin de préserver les caractéristiques géométriques précises de la brèche en cours de formation. Les données ainsi recueillies sur la rupture de la digue d'essai sont utiles à tout développeur voulant valider sa compréhension physique de l'érosion de brèche ou un outil de calcul visant à reproduire ce mécanisme. Le texte présente aussi une analyse comparative effectuée par un outil informatique de prévision, BRECHE, et la confrontation des résultats produits selon cinq méthodes publiées durant les dernières décennies et reconnues.
Une appréciation des résultats obtenus par prévision et des nombreuses données issues de l'essai réalisé sur le modèle en laboratoire est présentée et discutée.
Mots-clés:
- Rupture de barrage,
- brèche de barrage en terre,
- érosion de brèche,
- prévision,
- crue de rupture,
- digue en terre submergée,
- modèle réduit,
- digue fusible
Abstract
It has been recognized for several decades that the reliability of forecasts for flooding downstream from a dam failure depends not only on the local topography but also on the failure mode of the dam, specifically on the breach outflow hydrograph. It is therefore of primary importance that the mechanism of formation of the breach is anticipated with an acceptable degree of accuracy. Several approaches to model this phenomenon have been reported in the literature, but the results from these models do not appear to be reliable when the predicted values are compared with those deduced from observations carried out on real or simulated dam failures. Each failure of a dam generates a particular form of hydrograph that will have a specific impact on the affected population. The most frequent cause of failure of earthfill and rockfill dams is by overtopping of the crest, although in rockfill dams initial piping failure followed by partial collapse of the crest resulting in breach initiation has been observed. Once initiation of a breach has occurred, continuous development until complete failure occurs is virtually assured if sufficient energy from the reservoir volume is available to sustain the breaching process. For this reason, breach formation is recognized as one of the essential factors in the modeling of a dambreak.
To anticipate the real hazard from dam overtopping, a predictive model for breach initiation and development would be extremely useful. Currently, the majority of dam safety studies assume a standard scenario for the breach formation, development and final failure based on some norms established as a result of regression analyses on some actual failures. Modeling of the breach and therefore prediction of the outflow hydrograph do not take into account specific local data on the construction techniques and soil properties, mainly because currently available methods that attempt to incorporate these factors still do not provide reliable results. To contribute to the understanding of the mechanism of breach formation and to determine the details of the process, a test of erosion failure was performed and documented at the Hydrodynamics Laboratory of the École Polytechnique de Montreal. This test, coupled with the subsequent development of a numerical model incorporating the interactions of the hydraulics-erosion process, has formed the basis of a doctoral dissertation, the principal objective being to reproduce as accurately as possible the mechanisms involved in the formation of a breach (ZERROUK, 2004; ZERROUK et MARCHE, 2004).
This paper describes a test carried out under controlled laboratory conditions on a homogeneous dam composed of moraine, a material used in the construction of real dams with an extended non-cohesive granulometry. The test simulated the initiation and the start of the failure by dike overtopping known as a " fuse plug " i.e. a dike that, during an event of exceptional rain or extreme inflow, would be used to limit the volume retained in the reservoir. The test, carried out on a laboratory scale but extrapolated to prototype dimensions without similarity considerations, nevertheless allowed identification of the various mechanisms that intervene in the breaching process.
The results obtained on this experimental dike showed that the mechanism of formation and erosion of a breach is quite different from the mechanisms assumed in standard methods of computation of the rupture. Current models give undue weight to a single failure mechanism, independent of lateral slope instability, surface erosion or shear stress energy. The laboratory test however indicated the importance of turbulence and erosion in certain specific zones of the flow and highlighted the role of sidewall instability at the base of the breach, which erodes continuously leads to lateral instability.
Currently, engineers attempt to predict breach outflow based on the formation of the breach using physical properties of the component material such as the angle of friction, the slope angle of the breach channel and a factor for the rate of growth of breach width. The present authors recognize major deficiencies in this approach such as the use of assumptions for the breach geometry, the progression of the rupture, the use of material transport expressions based only on shear stress considerations (FREAD, 1984c) or the use of empirical 'adjustment' factors influencing the erosion of compacted cohesive soils (SINGH et SCARLATOS, 1989).
The data acquired from the experimental dike were used in a data-processing tool for comparative analysis called BREACH, in which five methods for predicting the formation of a breach by overtopping of dams composed of granular material were programmed (ZERROUK, 2004; ZERROUK et MARCHE, 2001). Starting with the same data obtained from the laboratory test, each of the models was applied and the results discussed with respect to whether the principal factor for breach evolution that is presumed predominant for the breaching process in the model, corresponds in fact to the mechanism observed during the test.
The analysis highlights the factors, ignored in the current models, which may in fact explain the mediocre results obtained for the predictions. It indicates that the characterization of the breach geometry is the major uncertainty in these models, which result in peak outflows and corresponding times never being simultaneously predicted correctly.
Keywords:
- Dambreak,
- earth-fill dam breach,
- breach erosion,
- forecasting,
- flood of failure,
- overtopped earthen dike,
- hydraulic testing flume,
- fuse plug