Thursday January 7th 2021,12.00 UK time
Dr John Owen, School of Engineering, University of Nottingham, United Kingdom, The Response of Bridges to Wind – Some Lessons from Monitoring Large Bridges
- Prof Steve Cai, Louisiana State University, Time domain simulation of turbulence effects on the aerodynamic flutter of long span bridges.
- Prof Claudio Mannini, University of Florence, Nonlinear modelling of self-excited forces for a long-span bridge under turbulent wind
- Prof Ole Andre Øiseth, Norwegian University of Science and Technology. Lessons learned from long-term wind and acceleration monitoring of the Hardanger Bridge.
- Prof Luca Caracoglia, North Eastern University, Boston, Relevance of Uncertainty Quantification to Study Wind Load Variability and its Effects on Long-Span Bridge Aeroelasticity
The Registration link can be found here.
The Response of Bridges to Wind – Some Lessons from Monitoring Large Bridges. This presentation will consider the lessons that can be learnt from monitoring the wind induced response of long span bridges by reviewing monitoring exercises on three different bridges. Developments in instrumentation and data analysis will briefly be reviewed before looking at specific wind response phenomena. The presentation will consider observations of vortex induced vibration and buffeting response and consider how these can be used to improve design predictions and modelling methods. With regard to buffeting, the influence of uncertainty in wind field parameters on the response of the bridge will be examined and the consequences for design identified. An example of non-stationary response due to thunderstorm activity will be presented, which demonstrates that non-stationary wind features can lead to significant structural response. The presentation will conclude by looking forward to how best to exploit the data from monitoring systems installed as “standard” on new bridges.
Time-Domain Simulations of Turbulence Effects on the Aerodynamic Flutter of Long-Span Bridges. Though turbulence effects on bridge flutter have been studied in the last few decades, its true effects remain a debate due to the limitation of previous wind tunnel facilities. In order to investigate and explain the effect of wind turbulence on the flutter instability, a time-domain simulation is carried out, which avoids the complicated random parametric excitation analysis used in previous studies. The simulations show the turbulence can change the vibration patterns and weaken the spatial vibration correlation to some extent. Due to this stabilizing effect, the critical flutter velocity is increased by 5% to 10% over that under smooth flow.
Nonlinear modelling of self-excited forces for a long-span bridge under turbulent wind. The prediction of the dynamic response of a long-span bridge under turbulent wind is a complicated task due to the important role played by nonlinearities and fluid memory. It is well known that a key issue is the nonlinear dependence of self-excited forces on the unsteady angle of attack induced by large-scale turbulence. The advantages of a relatively simple time-variant model over the classical linear time-invariant approaches are discussed and quantified based on a specific wind tunnel test campaign. In particular, the experiments for two different bridge sections revealed the viability for practical engineering applications of the basic assumption of slowly-varying angle of attack.
Lessons learned from long-term wind and acceleration monitoring of the Hardanger Bridge. The Hardanger Bridge is spanning the Hardanger fjord which is located at the west coast of Norway. A comprehensive monitoring campaign started in 2013 shortly after the bridge was opened to traffic and it is still ongoing. The monitoring data has revealed that uncertain turbulence parameters have a significant impact on the observed dynamic response and that the current design practice underestimates the dynamic response of the bridge severely. The monitoring data also underlines that nonstationary events can be severe and govern the bride design. This presentation outlines the lessons learned and gives an introduction to the methods developed for improved response predictions.
Relevance of Uncertainty Quantification to Study Wind Load Variability and its Effects on Long-Span Bridge Aeroelasticity. This short presentation will examine past and recent research activities in the field of bridge aeroelasticity under the influence of uncertain, experimentally measured loads and modeling simplifications. Description will include: (1) probability-based, stochastic algorithms for evaluating buffeting response influenced by various error sources, (2) Monte-Carlo sampling used to analyze bridge performance over time through life-cycle cost estimation, and (3) flutter reliability contaminated by random Scanlan (flutter) derivatives. In this context, the presentation will briefly introduce how Artificial Intelligence may be employed to investigate flutter occurrence without requiring solution of the multimode equations. Application examples will be derived from models of either existing or simulated long-span bridges.
Dr John S Owen is an Associate Professor in the Department of Civil Engineering at the University of Nottingham, where he has been for 27 years, serving two terms as Head of Department. He is a past chair of the UK Wind Engineering Society and was co-chair of the 6th European and African Conference on Wind Engineering. John’s principal research interests are in the dynamic response of structures to the wind and the use of dynamic data in structural health monitoring. He has been involved in monitoring the response of long span bridges to wind for many years, initially leading the monitoring programme on the Kessock Bridge (Scotland) and more recently working on the Forth Road Bridge (Scotland) and Phu My Bridge (Vietnam). John has also worked closely with colleagues in computational wind engineering to simulate the aero-elastic behaviour of bridge sections and has led a number of wind tunnel studies on section and full aero-elastic bridge models. He was responsible for the design and commissioning of the atmospheric boundary layer wind tunnel at Nottingham. Most recently, John has been working on the resilience of infrastructure in Typhoons developing a risk based methodology for networks in Vietnam.
Steve Cai is the coordinator of Structures Group at Louisiana State University and the holder of the Edwin B. and Norma S. McNeil Distinguished Professorship since 2010. He had his BS, MS and PhD from Zhejiang University, Tsinghua University, and Univ. of Maryland, respectively. His research interests include bridge performance evaluation, hazard mitigation of costal infrastructures (and vehicles) under wave/wind actions, and long-span bridge aerodynamics. He has served on many editorial boards and technical committees. Other major professional services include served as Secretary and Treasurer of American Association for Wind Engineering for more than 10 years.
Claudio Mannini is an assistant professor of the Department of Civil and Environmental Engineering, University of Florence, Italy. He got his Ph.D. in 2006. He received the ANIV Award in 2008, the IAWE Junior Award in 2011, and the EASD Junior Research Prize in 2014. His main research interests are bluff-body aerodynamics and wind-induced vibrations, addressed from theoretical, computational and experimental points of view. Since his Ph.D. thesis, he has always been enthralled by the aerodynamics of long-span bridges.
Ole Øiseth is a full professor in structural dynamics at the Department of Structural Engineering at the Norwegian University of Science and Technology. He is the head of the structural mechanics research group and the educational coordinator of the study program in civil and environmental engineering. The dynamics of structures subjected to environmental loading is his main research field. His research interests are stochastic dynamics, wind engineering, marine engineering, structural reliability and structural health monitoring. He has supervised 20 PhD candidates and 80 master students within these research fields since 2012.
Luca Caracoglia is currently an Associate Professor in the Department of Civil and Environmental Engineering of Northeastern University, Boston, Massachusetts, USA. His research interests are in structural dynamics, random vibration, wind engineering, fluid-structure interaction of civil engineering structures, linear and nonlinear cable network dynamics, wind-based energy harvesting systems and wind energy. Luca Caracoglia received the NSF CAREER Award in 2009. He was elected Fellow of the American Society of Civil Engineers in 2020.
Questions and answers
From Stoyan Stoyanoff to Prof. Cai: Procedure 3 is very problematic – cannot introduce U+u(t) leaving out the static forces and moments – it defies the quasi static expansion model double counting for those terms.
Prof Cai Thanks for the comment. In Procedure 3, namely the Random Parametric Excitation analysis (RPE) that was originally proposed by Lin and Bucher, the U+(t) term was introduced into the self-excited force. The static force and moment were handled separately, similar to the way typically used in other flutter analysis. People believe it is problematic for a few reasons as I discussed in my presentation and are listed below:
- Only along wind turbulence is included
- very complicated RPE theory is involved
- Wind is assumed as a white noise process
- flutter is based on statistical moment instability
- not convenient for FEM formulation and applications
- not sure if this approach can reflect the structure-fluid interaction mechanism
With that said, RPE analysis predicted the turbulence could be harmful for flutter. While this conclusion contradicted with most people’s opinion, it was partially verified based on some lab observations. For these reasons: we proposed the Procedure 4 that can remove some of those constraints and the RPE problem is directly numerically solved in the time-domain. One of the attractive features is that flutter derivatives are measured in the laminar flow, which can drastically reduce the lab work. Otherwise, aerodynamic parameters need to be measured for different turbulence conditions. Surely, validation of the procedure is needed.
From Aswathy M S to Prof Cai. In procedure 4, are the equations still going to be Stochastic Differential eqns, since white noise assumption is not there?
Prof Cai: Yes, the nature of the equation does not change, but equation can be solved numerically in the time-domain. Please see my Reply to Prof. Stoyanoff
Question from Mingshui Li to Prof. Cai: Thanks for your nice presentstion. You just showed the critical wind speed of flutter instability increases with turbulent intensity increasing. Have you verified your results via wind tunnel testing? The other issue is how the turbulent length scale effect the critical wind speed?
Prof Cai: No, we have not validated the procedure. This is an important issue. Please see my Reply to Prof. Stoyanoff. As noted in my presentation, in this preliminary results, the aerodynamic parameters were not updated based on the equivalent angle of wind attack in each time step. The along wind turbulence typically de-correlates the force and response, resulting in a beneficial factor for flutter as shown in the preliminary results. However, the vertical turbulence equivalently increases the wind attack angle, resulting in a harmful factor for flutter. The overall effect, with many different combinations of along wind and vertical turbulence, needs a systematic study to draw a general conclusion. The integral length in lab could be one to two orders less than that based on scaling principle due to facility limitation. There are nine integral lengths. It deserves a systematic study to know which one is most important for flutter performance (and buffeting response), though my best guess is Lu_x(along wind) and Lw_z(vertical).
From Yufen Zhou to Prof. Cai: How would the lack of integral length scale in the model scale compared with full scale affect the accuracy of the buffeting and flutter response? Would the wind tunnel response predict the buffeting and flutter response in a conservative way in terms of integral length scale similarity? If there is an influence, which integral length scale is dominated in this influence, e.g., the longitudinal integral length scale of horizontal turbulence Lux, or from vertical turbulence Lwx, or the lateral Luy, Lwy?
Prof Cai: These are good questions I do not have answers and deserve systematic studies. Please see my Reply to Prof. Li
From Guangzhong Gao to Claudio Mannini: Thank you for the presentation. Have you tried to apply nonlinear quasi-steady theory to consider the effect of time-varying angle of attack?
Claudio Mannini: Thank you for this interesting question, which allows me to explain an important feature that I had no time to point out during the presentation. We did apply quasi-steady theory, even correcting it based on some flutter derivative measurements to account for the effect of angular velocity. Nevertheless, this simplified nonlinear approach always provided inaccurate results. This was expected by looking at flutter derivatives. Indeed, the reduced velocity associated with the high-frequency motion was always between about 6 and 13 (a range of interest for buffeting calculations), and the aeroelastic coefficients for both case studies (the single-box deck of the Hardanger Bridge and a twin-box deck) show significant unsteady features up to values of the reduced velocity higher than the considered ones.
From Teng Wu to Claudio Mannini: Nice presentation! 2D indicial function (rational function) has been introduced to consider nonstationary aerodynamics, and now I see you successfully used 2D rational function in nonlinear aerodynamics. Does this mean this concept may provide a promising approach to build a unified bridge aerodynamics model for simultaneously consider both nonlinear and nonstationary effects?
Claudio Mannini: Thank you for this very relevant question. The theory of 2D indicial function is developed based on the assumption of a slow time-varying parameter, either the flow velocity as in some applications to transient wind effects or the angle of attack as in nonlinear buffeting. Nevertheless, I believe that an important result we obtained is that such an assumption does not seem as restrictive as one could expect. Indeed, from a practical engineering standpoint, it was possible to reasonably interpret as “slow” even variations of the angle of attack at a frequency not much smaller than the frequency of the motion. Though it needs to be confirmed for other test cases and different scenarios, this result corroborates the possibility to use a simple model based on 2D indicial functions also for nonstationary aerodynamics.
Question to Ole Øiseth: Very interesting presentation. Was the non-stationary response you showed for Hardanger during the Storm Tor associated with a rapid change in vertical velocity also?
Dr. Ole Oiseth: Thanks for the question and the positive feedback. A change in vertical velocity is also present, but it is not as pronounced as for the horizontal component. Only preliminary results are shown in the presentation. We will hopefully be able to study this phenomenon more in-depth soon.
From Khawaja Ali to Dr. Ole Oiseth: To what extent the selection of moving averaging time matters to visualize the non-stationary wind effects in the wind speed data? On what basis we can select moving averaging time for the analysis?
Dr. Ole Oiseth: Thanks for the question. We selected a three-second averaging period because it is commonly used to characterize wind gusts. The increase in velocity lasts for many seconds, so I think we could have used a more extended averaging period and still be able to study the nonstationary events in details. I don’t have a firm recommendation for the averaging period for a nonstationary event.
From Yufen Zhou to Dr. Øiseth: About the actual damping on the bridge, you showed that the damping decreased as the frequency increased. Could you explain where the damping level is obtained, on the bridge deck or hanger? In addition, could you explain the physical meaning of this phenomenon, i.e., damping ratio decreases significantly in high frequency modes. We usually think that the lowest fundamental frequency dominate the bridge response. Could you indicate how much the low damping level in the high frequency range affect the total response of the bridge? How conservative or unconservative would it be if we estimate the bridge response using the lowest frequency in each dominating direction?
Dr. Ole Oiseth: Thank you for your detailed questions. The damping was estimated by experimental modal analysis of selected hangers. We used a modal hammer and accelerometers. I find it hard to understand damping mechanisms fully. It tends to be a general understanding that damping decreases as the frequency increase, but I have observed that this is not the case for some structures. We only studied hanger vibration in this particular work. I agree that it is the lowest vibration modes that contribute most to the entire bridge’s buffeting response. Still, one should carefully consider higher vibration modes when considering the load effects. Possible low structural damping is perhaps not critical in buffeting response analysis since the aerodynamic damping is significant. The vortex-induced vibrations are discussed in this paper: https://journals.sagepub.com/doi/10.1177/1475921717721873