A topology optimisation framework to design test specimens for one - shot identification or discovery of material models
Saeid Ghouli, Moritz Flaschel, Siddhant Kumar, Laura De Lorenzis · 2025
Read A topology optimisation framework to design test specimens for one - shot identification or discovery of material models by Saeid Ghouli, Moritz Flaschel, Siddhant Kumar, Laura De Lorenzis (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2501-12756.txt.
Introduction
Saeid Ghouli and co-authors design a method to shape test samples so one test can reveal a material's behavior, in work from 2025. Full-field imaging now records how a sample surface moves in fine detail. Rich movement data could replace many simple tests with a single smart one.
The key is the sample shape, which decides how varied the test data will be. The authors use a density method to grow the best shape by computer. Their target was a material that stretches unevenly by direction. The design aims to make the fitted model stable and exact. Better samples mean fewer tests, lower cost, and faster checks for new materials.
Catalog record
| CREATOR | Saeid Ghouli, Moritz Flaschel, Siddhant Kumar, Laura De Lorenzis |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 117 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2501-12756
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
A topology optimisation framework to design test specimens for one - shot identification or discovery of material models Saeid Ghouli, Moritz Flaschel, Siddhant Kumar, Laura De Lorenzis Abstract The increasing availability of full-field displacement data from imaging techniques in experimental mechanics is determining a gradual shift in the paradigm of material model calibration and discovery, from using several simple-geometry tests towards a few, or even one single test with complicated geometry. The feasibility of such a “one-shot" calibration or discovery heavily relies upon the richness of the measured displacement data, i.e., their ability to probe the space of the state variables and the stress space (whereby the stresses depend on the constitutive law being sought) to an extent sufficient for an accurate and robust calibration or discovery process. The richness of the displacement data is in turn directly governed by the specimen geometry. In this paper, we propose a density-based topology optimisation framework to optimally design the geometry of the target specimen for calibration of an anisotropic elastic material model. To this end, we perform automatic, high-resolution specimen design by maximising the robustness of the solution of the inverse problem, i.e., the identified material parameters, given noisy displacement measurements from digital image correlation. We discuss the choice of the cost function and the design of the topology optimisation framework, and we analyse a range of optimised topologies generated for the identification of isotropic and anisotropic elastic responses. 1 Introduction Characterising the mechanical behaviour of materials typically requires multiple tests. For instance, fully characterising the linear elastic behaviour of a planar orthotropic material (such as a composite lamina) typically demands at least three separate experiments ( Tsai, 1965 ; Acosta-Flores and Eraña-Díaz, 2024 ) ; these could be, for instance, a longitudinal tensile test (measuring longitudinal and transverse strains simultaneously), a transverse tensile test, and an in-plane shear test. Also, traditional mechanical testing involves measurements with contact sensors such as strain gauges or displacement transducers, providing local experimental data only. The advent of advanced full-field measurement techniques such as digital image correlation ( DIC ) and digital volume correlation ( DVC ) has revolutionised the field of material law characterisation by opening up the perspective of fully characterising material behaviour with one single test ( Guélon et al., 2009 ; Fu et al., 2020 ) . In this new context, the optimal design of the target specimen is of utmost importance. A review paper by Pierron and Grédiac (2021) denotes this paradigm as “material testing 2.0". During the last decades, different groups of researchers have dealt with the specimen design problem suited for parameter identification via full-field measurements. As follows, we briefly highlight the major concepts and methodologies, focusing only on automatic (i.e., optimisation-based) approaches. Many studies have dealt with shape optimisation techniques to optimise the geometry of a pre-selected topology parametrised with a few variables. Different cost functions have been used for different classes of constitutive laws. In the context of orthotropic materials, Grédiac and Pierron (1998) aimed at equal co-existence of different strain components in the test specimen. Pierron et al. (2007) targeted maintaining the balance between the sensitivities to noise for the identification of the different stiffness components. Gu and Pierron (2016) accounted for the full identification process, thus optimising the DIC metrological parameters in addition to geometrical variables. The cost function was defined as the sum of systematic and random errors when calibrating the orthotropic stiffness components. Such a cost definition depends on the reference (i.e., ground-truth) material parameters. In the context of elastoplasticity, Souto et al. (2015) put forward the definition of a heterogeneity indicator (combining the standard deviation, range, maximum and average of different strain states) which could assess the suitability of a test specimen for sheet metal testing. The heterogeneity indicator was then employed in Souto et al. (2016) to design a butterfly-shaped specimen. Bertin et al. (2016) minimised the uncertainty in the parameter identification process by maximising the minimum eigenvalue of the Hessian matrix encountered in the finite element model updating ( FEMU ) approach. Chapelier et al. (2022) followed a similar approach and proposed a spline-based shape optimisation with special constraints to prevent excessive mesh distortion and load increase in a two-step shape optimisation, involving remeshing in between. Zhang et al. (2022) and Conde et al. (2023) employed the heterogeneity indicator proposed by Souto et al. (2015) in their cost functions, and confirmed that the initial configuration of the geometry has a strong influence on the final optimised design. More recently, Tung and Li (2024) devised a cost function quantifying the spread of data points in the deviatoric plane, while Ihuaenyi et al. (2024) utilised the information entropy concept to quantify the heterogeneity of the stress data points in the space of stress-triaxiality versus Lode angle parameter. Another stream of research has employed topology optimisation techniques to automatically design optimal test specimens for constitutive parameter calibration facilitated by DIC measurements. Within the popular framework of density-based topology optimisation, the design variables are the virtual densities which designate the existence of void or material in each element of a finite element ( FE ) mesh ( Bendsøe, 1989 ) . Also in this stream, different cost formulations have been proposed. For orthotropic materials, Chamoin et al. (2020) followed a similar strategy to Bertin et al. (2016) to minimise the uncertainty in the parameter identification process. Owing to the computational cost associated with numerical gradient calculation, only coarse-resolution topologies including grey scales were generated. Almeida et al. (2020) employed the sum of fractions of the specimen surface undergoing certain principal stress states as a heterogeneity function, which was then used in a multi-objective optimisation setup together with the compliance to maintain sufficient stiffness in the designed specimen. The final designs had low resolutions, contained grey scales, and were difficult to realise and manufacture. The weights balancing the multi-objective optimisation problem were adjusted heuristically, while they were recognised to highly influence the output. Barroqueiro et al. (2020) adopted the concept of compliant mechanisms and proposed the ratio between the output and input displacements (of the mechanism with linear elastic response) as the cost function. Through a two-stage algorithm, the authors first produced a large number of heterogeneous designs, and then ranked them using a performance indicator (favouring equal co-existence of different deformation modes) while assuming an elastoplastic response. The best topology underwent a manual redesign to ensure smooth outer boundaries. Gonçalves et al. (2023) improved the original framework proposed in Barroqueiro et al. (2020) by considering an additional, outer optimisation loop (leading to a two-level optimisation framework) which sought an optimal design domain configuration by changing parameters such as aspect ratio and material volume fraction. The authors performed numerous preliminary studies to define suitable bounds for the configurational parameters such that the heterogeneously designed specimen does not fail prematurely (i.e., prior to plasticity). The design strategy remained indirect and two-step while requiring expertise in selecting the parameter bounds and (dis-)approving the generated designs. The above short review reveals some gaps in the current state of the art. Shape optimisation approaches, while leading to much faster computations, require significant expertise in material testing for the initial selection of a suitable topology. On the other hand, when