Fadi Hachem
In the past, the safety margin for dynamic water pressure loads in steel-lined pressure tunnels and shafts was considered as acceptable by using conventional design safety factors. Due to high peak energy demands, existing plants are operating nowadays under rough conditions to regulate the discharge and power with relatively fast and repeated opening and closing of turbines and pumps. The economic and social costs due to production losses, when these water conveying structures are emptied for investigations and repairs, are considerable. Furthermore, the failure of pressure tunnels and shafts may produce catastrophic landslides and debris flows. An extensive literature review showed that the existing design methods have been based on the idea of keeping the allowable stress in steel liner below yielding threshold. These methods use also some rules for construction details and tolerances which minimize the risk of formation of high local concentrated stresses. Since the beginning of use of very high-strength steel liners in new hydro plants, the actual design methods and safety assessment have become inappropriate. This type of steel has a high risk of brittle failure and fatigue. Therefore, an enhancement of the existing theoretical design model for steel-lined pressure tunnels and shafts is necessary. Generally applicable approaches for estimating the quasi-static, which means without FluidˆStructure Interaction (FSI) and frequency-dependent water-hammer, wave speed in steel-lined pressure tunnels have been analyzed. The external constraints and assumptions of these approaches are discussed in detail and the reformulated formulas are then compared to commonly used expressions. For thin steel liners and weak rock mass modulus, Jaeger's and Parmakian's relationships overestimate the water-hammer velocity by approximately 3 – 4.5 %, while in Halliwell's formula this overestimation reaches 7.5 %. The quasi-static wave speed is significantly influenced by the state of the backfill concrete and the near-field rock zone (cracked or uncracked). In the case when these two layers are cracked, the quasi-static wave speed is overestimated in between 1% and 8% compared to uncracked concrete and near-field rock layers. Depending on the stiffness of steel liner and penstock, the FSI leads to significant difference in wave speeds values. As a first step, a fluid-structure interaction model is proposed as a basis for the development of new design criteria which consider fracture mechanics to access the response of high-strength steel liners. The effect of the backfill concrete and the surrounding rock mass has been mechanically modeled by a spring, a dashpot, and a lumped additional mass. The quadratic dispersion equation which results from FSI model, has been solved in the frequency domain through a numerical example. In this example and compared to the quasi-static case, the FSI approach results up to 13% higher wave speed values in the high-frequency range (higher than 600 Hz) and up to 150% lower values for frequencies between 150 and 300 Hz. In the intermediate frequency range (between 80 and 800 Hz), the precursor mode has a cut-off frequency which depends on the longitudinal distribution of the stiffness of the liner. The first acoustic mode begins to propagate at a frequency near 525 Hz. This cut-off frequency depends on the radial stiffness of the steel liner. For practical applications, the aforementioned wave speed differences in the quasi-static and FSI cases can be tolerated because of the uncertainty in the estimation of the rock mass characteristics and the presence of air in the water. The dynamic pressures obtained from classical water-hammer theory are not overly affected by such differences in wave speed while the FSI may lead to higher extreme dynamic pressures with higher frequencies. The influence of local drop of wall stiffness of pressurized waterways on the pressure wave speed and wave dissipation during transients was investigated experimentally. The weak reaches are resulting from local deterioration of the backfill concrete and the rock mass surrounding the steel liner. The change of wave speed generated by the weakening of the radial liner supports creates reflection boundaries for the incident pressure waves. A new signal processing procedure to identify the presence of these weak reaches has been proposed and validated by physical experiment tests. During water-hammer events, pressure and vibration records have been acquired at the both ends of a multi-reach steel test pipe. The weak reaches are simulated by replacing the steel reaches with Aluminum and PVC materials. The acquired data have been assessed using, amongst others, the Fourier Transform, wavelet decomposition, and cross-correlation techniques. The developed new monitoring method shows that wave speed and wave dissipation ratio are good indicators of the presence of local and large changes in stiffness. This method is also able to locate the weakness of stiffness along the test pipe when one PVC reach is used. When steep front wave have been generated inside the test pipe, it was possible to locate the position of the weak reach boundaries with a maximum relative mean error of 5.9 %. The severity of the local stiffness change has been also estimated with a maximum relative mean error of 20.6 %. In-situ measurements from a pressure shaft have been carried out to validate the new signal processing procedure. The prototype measurements use dynamic pressure and geophone sensors placed at both ends of the pressure shaft of the Grimsel II pumped-storage plant, in the Canton of Bern, in Switzerland. The data are acquired continuously and accessed on-line via internet. Different approaches to estimate the wave speed and wave dissipation generated inside the pressure shaft during start-up and shut-down of pumps and turbines have been applied. The relatively small water-hammer pressure fluctuations combined with the homogeneous quality of the rock mass surrounding the pressure shaft made it difficult to apply the entire localization procedure. Nevertheless, monitoring charts have been established based on the statistical quality control of the two indicators namely the water-hammer wave speed and the wave dissipation coefficient. The wave speed was assessed from the Fourier transformation spectrums (F) while the dissipation coefficient was determined by computing the root mean square (RMS) of the signal followed by an exponential regression fitting. Three control limits representing the actual state of the shaft wall have been set on these charts from the acquired and processed pressure data. These limits and the overall behaviour of the pattern of future measured points will be used for on-line monitoring of the shaft. The control limits of the monitoring charts for the water-hammer wave speed should be revised after acquiring a longer series of in-situ measurements. The control limits of the exponential dissipation coefficient computed during the pump and turbine start-up modes can be used for on-line monitoring. During the pump and turbine shut-down modes, the dissipation coefficient has encountered a shift of about 55 %. Additional measurements are needed to understand its global pattern behaviour.
EPFL2011
