Domain-aware-AI Augmented Design of Bridge Structures

May 1, 2022
Share this project


The architecture, engineering, and construction (AEC) industry currently adopts Generative Design (GD) approaches to enable computer-assisted design decision-making in early project stages. Existing approaches largely neglect structural engineering aspects such as design verification and construction processes. This holds especially true for bridges (yet they are the essential backbone of civil infrastructure) as there are no mature computational design nor optimization approaches available due to complex non-linear interactions of structural components and a tremendous amount of geometric and material design variables. Given the current situation, this project develops an agnostic toolkit for design of bridge structures and addresses four main goals: (a) overcoming the mentioned GD deficits by deriving an agnostic method using domain- aware artificial intelligence (AI)-related approaches to tackle the combination of GD with structural bridge design and optimization; (b) developing and implementing a toolkit “AI-BridgeGEN” with a generic approach through the case studies of tied-arch-network-bridges and concrete network-arch-bridges, (c) allowing further derivation of engineering domain knowledge through numerical investigations via this software tool, and (d) fostering dissemination of the developed approach into engineering research and practice by providing open source software and the generated data as benchmark data set for the scientific machine learning community. By its conception, this proposal substantially changes the way of research and practice in bridge design and delivers impact in three areas: (i) enriching applications of state-of-the-art AI models and potentially paving the way of developing new AI model classes (combining geometric deep learning with Generative Adversarial Networks), (ii) influencing future design concepts of bridge structures by expected new mechanical, technical and structural insights into the interactions between design variables of bridges, (iii) using the developed methodology for a ground-breaking pilot study in structural engineering, influencing research directions for structural optimization of a wider range of civil infrastructure besides bridges. The outcomes enable a significantly more efficient, economic and sustainable (yet reliable) design of bridges.



SDSC Team:
Luis Salamanca
Konstantinos Tatsis
Alessandro Maissen
Fernando Perez-Cruz

PI | Partners:

ETH Zurich, Chair of Concrete Structures and Bridge Design:

  • Dr. Ing. Michael A. Kraus
  • Prof. Dr. Walter Kaufmann
  • Sophia Kuhn (Ph.D. student)
  • Vera Balmer

More info



For bridge design, engineers currently rely on performance-driven parametric  models that allow to generate, simulate and evaluate a small number of design instances, and gather their performance feedback. Due to the complexity of dealing with a large set of concurrent objectives, this process heavily depends on prior experience, leading to the investigation of only a narrow spectrum of possible solutions. We need to implement tools that assist the engineer during the early phases of the design, enabling to more easily evaluate bridge candidates, discover unexplored areas previously intangible, and ensure the economic viability and sustainability of the tackled problems.

Proposed Approach / Solution

We aim at implementing methodologies to perform “Inverse evaluation“, providing the users with tools for automatically generating a diverse set of solutions given some requested performance measures. This will be facilitated through the implementation of a general toolbox that will allow to tackle different use cases, as well as varied data representations. The backbone of the toolbox will be a ML model, based on generative models such as autoencoders, that will allow learning a forward and inverse model from the available design instances. It will implement visualization tools to enable a further exploration of the design space, and more thorough understanding of the design problem.


The current project, and the methods envisioned on it, may substantially change the way of research and practice in bridge design from a currently passive use of parametric computer-aided design software into using an active, computationally intelligent partner (“co-pilot”). In the bridge engineering domain, insights on the mechanical and technical aspect could be also learnt through the exploration of the interrelation of design objectives and parametric parameters. Beyond, similar approaches for generative design could be extended to other civil infrastructures and related fields, where the design diversity might be also restricted by existing biases and/or limitations.

Figure 1: New design paradigm proposed by the toolbox. First, during the offline phase, samples are generated with the parametric model, and the model is trained. Then, during the online phase, the trained ML model is harnessed to carry out forward and inverse design
Figure 2: User interface implemented for the project “Brücke über den Graben St. Gallen“




  • Balmer, V., Kuhn, S. V., Bischof, R., Salamanca, L., Kaufmann, W., Perez-Cruz, F., & Kraus, M. A. (2024). Design space exploration and explanation via conditional variational autoencoders in meta-model-based conceptual design of pedestrian bridges. Automation in Construction, 163, 105411.
  • Kraus, M. A., Kuhn, S. V., Hodel, A., Bischof, R., Maissen, A., Salamanca Mino, L., & Pérez‐Cruz, F. (2024). Parametrische Modellierung und generatives tiefes Lernen für den Brückenentwurf. Bautechnik, 101(3), 174-180.

Additional resources


  1. W. Kaufmann and B. Meier, “Conceptual bridge design beyond signature structures,” in IABSE Conference, Geneva 2015: Structural Engineering: Providing Solutions to Global Challenges - Report, 2015, pp. 510–517, doi: 10.2749/222137815818357520.
  2. S. Abrishami, J. S. Goulding, F. P. Rahimian, and A. Ganah, “Integration of BIM and generative design to exploit AEC conceptual design innovation,” J. Inf. Technol. Constr., vol. 19, pp. 350–359, 2014.
  3. G. P. Monizza, C. Bendetti, and D. T. Matt, “Parametric and Generative Design techniques in mass-production environments as effective enablers of Industry 4.0 approaches in the Building Industry,” Autom. Constr., vol. 92, pp. 270–285, 2018.
  4. D. Holzer, R. Hough, and M. Burry, “Parametric Design and Structural Optimisation for Early Design Exploration,” Int. J. Archit. Comput., vol. 5, no. 4, pp. 625–643, 2007, doi: 10.1260/147807707783600780.
  5. M. Turrin, P. Von Buelow, and R. Stouffs, “Design explorations of performance driven geometry in architectural design using parametric modeling and genetic algorithms,” Adv. Eng. Informatics, vol. 25, no. 4, pp. 656–675, 2011, doi: 10.1016/j.aei.2011.07.009.
  6. L. Yang, D. Zhang, and G. E. M. Karniadakis, “Physics-informed generative adversarial networks for stochastic differential equations,” SIAM J. Sci. Comput., vol. 42, no. 1, pp. A292–A317, 2020, doi: 10.1137/18M1225409.


More projects


In Progress
Interpretable and Robust Machine Learning for Mobility Analysis
No items found.

Feature Learning for Bayesian Inference

In Progress
No items found.


In Progress
Personalized epidural electrical stimulation of the lumbar spinal cord for clinically applicable therapy to restore mobility after paralyzing spinal cord injury
No items found.


In Progress
Lensless Actinic Metrology for EUV Photomasks
No items found.


Latest news

Smartair | An active learning algorithm for real-time acquisition and regression of flow field data
May 1, 2024

Smartair | An active learning algorithm for real-time acquisition and regression of flow field data

Smartair | An active learning algorithm for real-time acquisition and regression of flow field data

We’ve developed a smart solution for wind tunnel testing that learns as it works, providing accurate results faster. It provides an accurate mean flow field and turbulence field reconstruction while shortening the sampling time.
The Promise of AI in Pharmaceutical Manufacturing
April 22, 2024

The Promise of AI in Pharmaceutical Manufacturing

The Promise of AI in Pharmaceutical Manufacturing

Innovation in pharmaceutical manufacturing raises key questions: How will AI change our operations? What does this mean for the skills of our workforce? How will it reshape our collaborative efforts? And crucially, how can we fully leverage these changes?
Efficient and scalable graph generation through iterative local expansion
March 20, 2024

Efficient and scalable graph generation through iterative local expansion

Efficient and scalable graph generation through iterative local expansion

Have you ever considered the complexity of generating large-scale, intricate graphs akin to those that represent the vast relational structures of our world? Our research introduces a pioneering approach to graph generation that tackles the scalability and complexity of creating such expansive, real-world graphs.

Contact us

Let’s talk Data Science

Do you need our services or expertise?
Contact us for your next Data Science project!