Turbulence characterisation in complex fjord topography using measurement and numerical modelling for bridge design

Abstract

On Norway’s west coast, the Norwegian Public Road Administration initiated a major infrastructure development project that includes the design and construction of bridges to replace ferries. The width of the fjords, flow conditions, and topography will all influence the type of bridges considered. In this context, since 2014, a measurement program has been running with the purpose of providing data for turbulence characterization in bridge design. Since the available theoretical background, for turbulence characterization in the context of bridge design, is based on flat and homogeneous terrain, the proposed bridge will almost certainly be subjected to environmental conditions that could lead to a catastrophic disaster if turbulence is not properly quantified. As a result, the first section of this work focused on the similarities and differences in flow characteristics in flat and complex terrain. The second section focused on researching common turbulence characteristics for the purpose of bridge design in multiple fjords. Finally, the study considered the use of numerical simulation for the estimation of flow characteristics relevant to bridge design across the span of the bridge where measurements are limited. The quantification of turbulence for bridge design revealed that the topographic feature can be divided into two categories: long-fetch winds (winds blowing over an extended stretch of water) and short-fetch winds (winds blowing over irregular and heterogeneous terrain upstream of the measurement location). Wind sector dependency was observed in the estimation of turbulence intensity, normalized vertical standard deviation, and friction velocity. Furthermore, a significant deviation from the assumption of Gaussian fluctuations, which is commonly found in bridge aerodynamic theory, is observed. The spectral characteristics of turbulence were investigated using one-point and two-point power spectral densities. Surface layer scaling may be possible under neutral conditions provided the velocity spectra are normalized by an adequate friction velocity, according to the findings. The spectral analysis also revealed the importance of sector dependence. The along wind and horizontal cross wind velocity spectra exhibit noticeable features such as large amplitude, spectral plateau, and double spectral peak in the low frequency section. In the vertical velocity spectrum, surface layer scaling may be applicable to both short-fetch and long-fetch winds, similar to examples of flat terrain. In addition, as compared to the flat terrain counterpart, the vertical spectra have a much higher spectral peak. Furthermore, the influence of terrain on flow was evident in the cross-spectrum study, with the co-spectrum and quad-spectrum having similar magnitudes, which is not the case with flat and homogeneous terrain. These common features found in the along wind, cross wind, and vertical wind spectra could have a big impact on the dynamic load estimates for the proposed bridges. Vertical coherence was analyzed to assess the Davenport similarity, which is generally assumed in flat and homogeneous terrain. The results revealed that at large vertical separations, an alternative model incorporating separation distance and height agrees with the co-coherence estimate. This suggests that the Davenport similarity may be limited for substantial vertical separation in complex topography. The average Davenport coefficient, on the other hand, was found to be consistent with the suggested values in the Norwegian bridge design handbook (N400). Since a thorough characterization of turbulence for bridge design necessitates the analysis of flow conditions over the bridge span, a nested numerical simulation was established using the macroscale model AROME developed by M´et´eo-France and the microscale model SIMRA developed by SINTEF. The study’s initial part concentrated on the validation of the results obtained by estimating the correlation between measurement and numerical simulation. The validation was based on mean turbulent characteristics because the turbulence closure is based on two equations. If the mast is not placed downstream of a steep mountain, the results showed reasonable agreement for wind speed and direction on the fjord’s shore. Sector dependence is also apparent when it comes to the correlation of angle of attack, as it was previously observed in the case of the study of integral and spectral characteristics. Furthermore, the angle of attack validation revealed a dependence on the mast location and anemometer height, highlighting the main issue associated with the numerical prediction of vertical velocity in sheltered locations. To adequately establish the source of disparities between observation and numerical simulation when it comes to estimating angle of attack and wind direction at severely sheltered locations, more research is needed.