Topology optimization via machine learning and deep learning: a review

Shin, Seungyeon; Shin, Dongju; Kang, Namwoo

doi:10.1093/jcde/qwad072

Cited by 17 publications

(5 citation statements)

References 131 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…[18][19][20][21] Various integration approaches have emerged, all with the common objective of speeding up the optimization process. 22,23 These approaches include acceleration of iteration, 24,25 generative design, 26,27 postprocessing, 28 and metamodeling. 29,30 In particular, metamodeling strategies focus on streamlining the optimization process by replacing partially or completely the resource-intensive FEA with a neural network-based (NN) model, thereby yielding a substantial reduction in the overall computational cost.…”

Section: Motivationmentioning

confidence: 99%

“…The constraint equations ( 17)-( 23) have been incorporated into the original objective function (15). This integration is used to enforce specific criteria, including definition of the frequency interval of interest ( 17)- (19), ensuring the coupling between the two resonant peaks (20), imposing an upper limit on the total reference mass of the MLAM structure (21), setting an upper bound on the MLAM core's thickness (22), and establishing both upper and lower bounds for each parameter 𝜑 i (23) in the set of design variables, 𝚿. For each parameter, denoted as 𝜑 i , the bounds are defined by 𝜑 i,U and 𝜑 i,L for the upper and lower bounds, respectively.…”

Section: Optimization Problem Formulationmentioning

confidence: 99%

See 1 more Smart Citation

Machine learning in solid mechanics: Application to acoustic metamaterial design

Yago,

Sal‐Anglada,

Roca

et al. 2024

Numerical Meth Engineering

View full text Add to dashboard Cite

Machine learning (ML) and Deep learning (DL) are increasingly pivotal in the design of advanced metamaterials, seamlessly integrated with material or topology optimization. Their intrinsic capability to predict and interconnect material properties across vast design spaces, often computationally prohibitive for conventional methods, has led to groundbreaking possibilities. This paper introduces an innovative machine learning approach for the optimization of acoustic metamaterials, focusing on Multiresonant Layered Acoustic Metamaterial (MLAM), designed for targeted noise attenuation at low frequencies (below 1000 Hz). This method leverages ML to create a continuous model of the Representative Volume Element (RVE) effective properties essential for evaluating sound transmission loss (STL), and subsequently used to optimize the overall topology configuration for maximum sound attenuation using a Genetic Algorithm (GA). The significance of this methodology lies in its ability to deliver rapid results without compromising accuracy, significantly reducing the computational overhead of complete topology optimization by several orders of magnitude. To demonstrate the versatility and scalability of this approach, it is extended to a more intricate RVE model, characterized by a higher number of parameters, and is optimized using the same strategy. In addition, to underscore the potential of ML techniques in synergy with traditional topology optimization, a comparative analysis is conducted, comparing the outcomes of the proposed method with those obtained through direct numerical simulation (DNS) of the corresponding full 3D MLAM model. This comparative analysis highlights the transformative potential of this combination, particularly when addressing complex topological challenges with significant computational demands, ushering in a new era of metamaterial and component design.

show abstract

Section: Motivationmentioning

confidence: 99%

Section: Optimization Problem Formulationmentioning

confidence: 99%

Machine learning in solid mechanics: Application to acoustic metamaterial design

Yago,

Sal‐Anglada,

Roca

et al. 2024

Numerical Meth Engineering

View full text Add to dashboard Cite

show abstract

“…However, despite the advancements, several studies still face various challenges. These encompass problematic aspects of encountered solutions such as try to improve low pixel resolution in artificially generated designs, or try to improve the rate of success of Machine Learning techniques to obtain successful solutions, also improving the overall ability of existing methodologies for the obtention of valid solution in three-dimensional spaces, also tackling the expenses related to data collection for the conformation of design databases for generative design, and finally, managing effectively the substantial computational load associated with both Finite Element Analysis (FEA) and the machine learning approach itself [162].…”

Section: Machine Learning Approaches For Solving the To Problemmentioning

confidence: 99%

Iterative design of satellite structures and topology optimisation methods

Hurtado-Pérez,

Pablo-Sotelo,

Ramírez-López

et al. 2024

J. Phys.: Conf. Ser.

View full text Add to dashboard Cite

Satellite launches are critical amid increasing demands for advanced services. Rising costs urge innovation in engineering. Enhancing satellite structural integrity during launches is crucial due to intense vibrations. Topology optimisation and additive manufacturing converge as promising solutions. Topology optimisation uses mathematical techniques to iteratively improve structures by reducing mass while enhancing attributes like load capacity. This field, active for about three decades, employs continuum and discrete algorithms to optimise structures. Recent trends show optimised structures through topology optimisation and additive manufacturing promise cost-effective and high-performance solutions. This study aims to extensively review topology optimisation methods, specifically focusing on satellite structures, to shape future developments within the years to come.

show abstract

“…Despite the progress in this field, there remain challenges in many studies. These include addressing the low resolution of generated designs, enhancing the performance of machine learning methods, adapting the methodologies to different design domains (three-dimensional spaces), grappling with the costs associated with gathering data and managing the high computational burden of both FEA and the machine learning approach [336].…”

Section: Machine Learning Applied To Topology Optimisationmentioning

confidence: 99%

On Topology Optimisation Methods and Additive Manufacture for Satellite Structures: A Review

Hurtado-Pérez,

Pablo-Sotelo,

Ramírez-López

et al. 2023

Aerospace

View full text Add to dashboard Cite

Launching satellites into the Earth’s orbit is a critical area of research, and very demanding satellite services increase exponentially as modern society takes shape. At the same time, the costs of developing and launching satellite missions with shorter development times increase the requirements of novel approaches in the several engineering areas required to build, test, launch, and operate satellites in the Earth’s orbit, as well as in orbits around other celestial bodies. One area with the potential to save launching costs is that of the structural integrity of satellites, particularly in the launching phase where the largest vibrations due to the rocket motion and subsequent stresses could impact the survival ability of the satellite. To address this problem, two important areas of engineering join together to provide novel, complete, and competitive solutions: topology optimisation methods and additive manufacturing. On one side, topology optimisation methods are mathematical methods that allow iteratively optimising structures (usually by decreasing mass) while improving some structural properties depending on the application (load capacity, for instance), through the maximisation or minimisation of a uni- or multi-objective function and multiple types of algorithms. This area has been widely active in general for the last 30 years and has two main core types of algorithms: continuum methods that modify continuous parameters such as density, and discrete methods that work by adding and deleting material elements in a meshing context. On the other side, additive manufacturing techniques are more recent manufacturing processes aimed at revolutionising manufacturing and supply chains. The main exponents of additive manufacturing are Selective Laser Melting (SLM) (3D printing) as well as Electron Beam Melting (EBM). Recent trends show that topology-optimised structures built with novel materials through additive manufacturing processes may provide cheaper state-of-the-art structures that are fully optimised to better perform in the outer-space environment, particularly as part of the structure subsystem of novel satellite systems. This work aims to present an extended review of the main methods of structural topology optimisation as well as additive manufacture in the aerospace field, with a particular focus on satellite structures, which may set the arena for the development of future satellite structures in the next five to ten years.

show abstract

Topology optimization via machine learning and deep learning: a review

Cited by 17 publications

References 131 publications

Machine learning in solid mechanics: Application to acoustic metamaterial design

Machine learning in solid mechanics: Application to acoustic metamaterial design

Iterative design of satellite structures and topology optimisation methods

On Topology Optimisation Methods and Additive Manufacture for Satellite Structures: A Review

Contact Info

Product

Resources

About