Optimisation of Structures Excited by Wind under Consideration of Stochastic Actions and Various Limit States
Applied Mechanics, Statics and Dynamics
Final Report Abstract
Wind-induced vibrations represent critical design and service criteria of slender structures such as long-span bridges and tall towers. The project ”Optimization of wind-excited structures taking into account stochastic effects and various limit states” dealt with the development of a framework to perform aerodynamic shape optimization of structures by utilizing advanced analysis methods for different aeroelastic phenomena under defined limit states. In order to develop such a computational framework, several challenges were identified and addressed scientifically, namely: i) fundamental investigations of the underlying processes occurring during fluidstructure interaction; ii) development of reduced-order models of the aerodynamic forces and coefficients; and iii) appropriate formulation of the aerodynamic optimization problem. The fundamental investigations emphasized that adequate modelling of the aerodynamic forces is crucial when predicting the dynamic structural response. Studying the nonlinear interaction between gust- and motion-induced forces showed that simple semianalytical aerodynamic models could be inadequate when predicting the structural response for any limit state. The developed CFD method for simulation of sinusoidal gusts proved to be effective for extracting the unsteady information in the gust-induced forces from numerical simulations, which is necessary during reduced-order modelling. Extracting the unsteadiness of gust-induced forces had previously been impossible without carrying out expensive wind tunnel tests. The development of reduced-order models enabled predicting the aeroelastic response for a large number of structural design alternatives. This is otherwise computationally infeasible through CFD simulations or economically impractical through wind tunnel tests. Reduced-order modelling was accomplished on two levels: i) by using data-driven methods to model the time-dependent aerodynamic forces for the prediction of the aeroelastic response; ii) by using metamodels to approximate the aerodynamic response based on the design variables. On the first level, it was shown that data-driven models based on machine learning methods, such as Gaussian Processes and Artificial Neural Networks, can adequately capture nonlinear aerodynamic phenomena. Such nonlinear phenomena are typically intractable for the conventional semi-analytical aerodynamic models. On the second level, the aerodynamic coefficients were approximated using metamodels and taking the design variables as input (e.g. nose angle of a bridge deck). This substantially lowered the number of simulations necessary for all viable design alternatives. The metamodels were then used as an input for the semi-analytical and/or data-driven models during optimization. Careful formulation of the optimization problem proved imperative when considering multiple limit states and the stochastic input parameters. The outcome of the optimization studies showed that it is difficult to determine the governing aeroelastic phenomenon (e.g. flutter or buffeting) for a distinct limit state a priori. Therefore, multi-objective optimization in a Pareto formulation is necessary to achieve a high-performance aerodynamic design. Further to this end, the uncertainty in the design variables was also considered through Reliability-Based Design Optimization. This proved to be a significant advancement from a practical perspective since a probabilistic approach considering input uncertainty is the basis of all building codes. All three aspects are an integral part of the developed framework for aerodynamic shape optimization of civil structures. Conducting the fundamental investigations was necessary to model the underlying aeroelastic phenomena appropriately. Developing reduced-order models made the prediction of the aeroelastic response for many design alternatives feasible. Finally, an appropriate formulation of the optimization problem showed to be a prerequisite for an aerodynamically optimal design. All model components were suitably validated and tested in practical applications related to bridge design. In conclusion, the outcome of the projects offers both a profound insight into the physical processes occurring in fluid-structure interaction and a practical methodology for the aerodynamic design of long-span bridges and tall towers. Therefore, it is of interest to academics as well as practitioners.
Publications
- A CFD Study on the Influence of Free-stream Deterministic Gusts on the Critical Flutter Velocity of Streamlined Bridge Deck. IABSE Congress New York, New York City, USA. 2019
Kavrakov, I., Tesfaye, S., and Morgenthal, G.
(See online at https://doi.org/10.2749/newyork.2019.2671) - Determination of complex aerodynamic admittance of bridge decks under deterministic gusts using the Vortex Particle Method. Journal of Wind Engineering & Industrial Aerodynamics, 193, pp.103971, 2019
Kavrakov, I., Argentini, T., Omarini, S., Rocchi, D. and Morgenthal, G.
(See online at https://doi.org/10.1016/j.jweia.2019.103971) - Prediction of aerodynamic behavior of bridge decks using artificial neural networks. The Sixteenth International Conference on Civil, Structural & Environmental Engineering Computing & Fifth International Conference on Soft Computing & Optimization in Civil, Structural and Environmental Engineering, Riva del Garda, Italy, 2019
Abbas, T., Kavrakov, I., Morgenthal, G. and Lahmer, T.
- Reliability-based design optimization for shape design of bridge decks to mitigate the vortex-induced vibration (VIV) phenomena. Case Study: Trans-Tokyo Bay Bridge. The Sixteenth International Conference on Civil, Structural & Environmental Engineering Computing & Fifth International Conference on Soft Computing & Optimization in Civil, Structural and Environmental Engineering, Riva del Garda, Italy, 2019
Jaouadi, Z., Abbas, T., Morgenthal, G. and Lahmer, T.:
- Shape optimization of bridge decks considering vortex–induced vibration phenomena. 90th Annual Meeting GAMM 2019. Vienna, Austria, 2019
Jaouadi, Z., Abbas, T., Morgenthal, G. and Lahmer, T.
- Comparison Metrics for Timehistories: Application to Bridge Aerodynamics, Journal of Engineering Mechanics, 146(9), pp.04020093, 2020
Kavrakov,I. , Kareem, A. and Morgenthal, G.
(See online at https://doi.org/10.1061/(ASCE)EM.1943-7889.0001811) - Prediciton of aeroelastic response of bridge decks using artificial neural networks, Computers and Structures, 231, pp.106198, 2020
Abbas, T., Kavrakov, I., Morgenthal, I. and Lahmer, T.
(See online at https://doi.org/10.1016/j.compstruc.2020.106198) - Single and multi-objective shape optimization of streamlined bridge decks. Structural and Multidisciplinary Optimization, 61, pp.1495-1514, 2020
Jaouadi, Z., Abbas, T., Morgenthal, G. and Lahmer, T.
(See online at https://doi.org/10.1007/s00158-019-02431-3) - Data-driven aerodynamic analysis of structures using Gaussian Processes. Journal of Wind Engineering & Industrial Aerodynamics, 222, pp.104911, 2022
Kavrakov, I., McRobie, A. and Morgenthal, G.
(See online at https://doi.org/10.1016/j.jweia.2022.104911) - Numerical investigation of the nonlinear interaction between the sinusoidal motion-induced and gust-induced forces acting on bridge decks
Tesfaye, S., Kavrakov, I. and Morgenthal, G.
(See online at https://doi.org/10.1016/j.jfluidstructs.2022.103680)
