Deep Neural Network and Monte Carlo Tree Search applied to Fluid-Structure Topology Optimization

Gaymann, Audrey; Montomoli, Francesco

doi:10.1038/s41598-019-51111-1

Cited by 17 publications

(13 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…As pointed out by Sigmund et al [195], gradient-free approaches seldomly make sense for topology optimisation, due to the high dimensional problems for increasing design resolutions. This is perfectly illustrated in the work of Gaymann and Montomoli [71], where the design resolution is absurdly coarse and useless in practise. However, gradient-free approaches can be useful when gradient information is not available, like when using a commercial solver as a black-box, or when dealing with discontinuous functions with non-well-defined gradients.…”

Section: Optimisation Methodsmentioning

confidence: 95%

“…Most of these use first-order methods, with notable exceptions being the work of Evgrafov [34,41] using higher-order schemes. Only three papers use gradient-free optimisation approaches, consisting of genetic algorithms [47,186] and neural networks [71]. As pointed out by Sigmund et al [195], gradient-free approaches seldomly make sense for topology optimisation, due to the high dimensional problems for increasing design resolutions.…”

Section: Optimisation Methodsmentioning

confidence: 99%

“…Gaymann et al [70] applied a density-based method to the design of fluidic diode valves in both two-and three-dimensions at medium-to-high Reynolds numbers. Gaymann and Montomoli [71] applied deep neural networks and Monte Carlo Tree search to an absurdly coarse design grid.…”

Section: Steady Laminar Flowmentioning

confidence: 99%

See 2 more Smart Citations

A Review of Topology Optimisation for Fluid-Based Problems

Alexandersen

Andreasen

2020

Fluids

186

View full text Add to dashboard Cite

This review paper provides an overview of the literature for topology optimisation of fluid-based problems, starting with the seminal works on the subject and ending with a snapshot of the state of the art of this rapidly developing field. “Fluid-based problems” are defined as problems where at least one governing equation for fluid flow is solved and the fluid–solid interface is optimised. In addition to fluid flow, any number of additional physics can be solved, such as species transport, heat transfer and mechanics. The review covers 186 papers from 2003 up to and including January 2020, which are sorted into five main groups: pure fluid flow; species transport; conjugate heat transfer; fluid–structure interaction; microstructure and porous media. Each paper is very briefly introduced in chronological order of publication. A quantititive analysis is presented with statistics covering the development of the field and presenting the distribution over subgroups. Recommendations for focus areas of future research are made based on the extensive literature review, the quantitative analysis, as well as the authors’ personal experience and opinions. Since the vast majority of papers treat steady-state laminar pure fluid flow, with no recent major advancements, it is recommended that future research focuses on more complex problems, e.g., transient and turbulent flow.

show abstract

Section: Optimisation Methodsmentioning

confidence: 95%

Section: Optimisation Methodsmentioning

confidence: 99%

See 1 more Smart Citation

A Review of Topology Optimisation for Fluid-Based Problems

Alexandersen

Andreasen

2020

Fluids

186

View full text Add to dashboard Cite

show abstract

“…Since its inception, MCTS was indeed successfully implemented in many board games [22][23][24][25], protein folding problems [26], chemical design applications [27], planning and logistics [28,29], and is currently intended as one of the best approaches in Artificial General Intelligence for games (AGI) [30]. Interestingly, as of 2021, only a small number of engineering-related applications of MCTS exist [31,32].…”

Section: Introduction and State Of The Artmentioning

confidence: 99%

Monte Carlo Tree Search as an intelligent search tool in structural design problems

Rossi

Winands

Butenweg

2021

Engineering with Computers

View full text Add to dashboard Cite

Monte Carlo Tree Search (MCTS) is a search technique that in the last decade emerged as a major breakthrough for Artificial Intelligence applications regarding board- and video-games. In 2016, AlphaGo, an MCTS-based software agent, outperformed the human world champion of the board game Go. This game was for long considered almost infeasible for machines, due to its immense search space and the need for a long-term strategy. Since this historical success, MCTS is considered as an effective new approach for many other scientific and technical problems. Interestingly, civil structural engineering, as a discipline, offers many tasks whose solution may benefit from intelligent search and in particular from adopting MCTS as a search tool. In this work, we show how MCTS can be adapted to search for suitable solutions of a structural engineering design problem. The problem consists of choosing the load-bearing elements in a reference reinforced concrete structure, so to achieve a set of specific dynamic characteristics. In the paper, we report the results obtained by applying both a plain and a hybrid version of single-agent MCTS. The hybrid approach consists of an integration of both MCTS and classic Genetic Algorithm (GA), the latter also serving as a term of comparison for the results. The study’s outcomes may open new perspectives for the adoption of MCTS as a design tool for civil engineers.

show abstract

“…Our algorithm reduces the computational cost by at least two orders of magnitude compared with directly applying heuristic methods. In addition to benchmarks from gradient-based solvers, we compare our algorithm with an offline version (where all training data are randomly generated), Generalized Simulated Annealing (GSA), BO, CMA-ES and a recent algorithm based on reinforcement learning 27 .…”

mentioning

confidence: 99%

Self-Directed Online Machine Learning for Topology Optimization

Deng

Wang

Qin

et al. 2021

Preprint

View full text Add to dashboard Cite

Topology optimization by optimally distributing materials in a given domain requires gradient-free optimizers to solve highly complicated problems. However, with hundreds of design variables or more involved, solving such problems would require millions of Finite Element Method (FEM) calculations whose computational cost is huge and impractical. Here we report a Self-directed Online Learning Optimization (SOLO) which integrates Deep Neural Network (DNN) with FEM calculations. A DNN learns and substitutes the objective as a function of design variables. A small number of training data is generated dynamically based on the DNN's prediction of the global optimum. The DNN adapts to the new training data and gives better prediction in the region of interest until convergence. Our algorithm was tested by compliance minimization problems and fluid-structure optimization problems. It reduced the computational time by 2 ~ 5 orders of magnitude compared with directly using heuristic methods, and outperformed all state-of-the-art algorithms tested in our experiments. This approach enables solving large multi-dimensional optimization problems.

show abstract

Deep Neural Network and Monte Carlo Tree Search applied to Fluid-Structure Topology Optimization

Cited by 17 publications

References 22 publications

A Review of Topology Optimisation for Fluid-Based Problems

A Review of Topology Optimisation for Fluid-Based Problems

Monte Carlo Tree Search as an intelligent search tool in structural design problems

Self-Directed Online Machine Learning for Topology Optimization

Contact Info

Product

Resources

About