A machine learning based surrogate constitutive model is presented for the elasto-damage response of transversely isotropic materials, such as fiber reinforced composites and wood. In the context of continuum damage mechanics, the nonlinear constitutive response of such materials is formulated by means of damage variables that correspond to various damage mechanisms. These damage variables, in combination with the initial (undamaged) elasticity tensor, are used to evaluate the nonlinear relation between stress and strain tensors. Instead of analytically expressing the relation between damage variables and strain state, an artificial neural network is employed. The training data is generated by performing single-element analyses for plane stress states using the Finite Element Program ABAQUS/Standard 2024 (Dassault Systemes Simulia Corp., Johnston, Rhode Island, USA). Radial strain loading paths are prescribed and the built-in Hashin elasto-damage material model for fiber reinforced composites is employed. The simulation output is the evolution of the stress tensor and the damage variables as a function of strain tensor input. The neural network is trained by minimizing the error between the damage variables predicted by the network and the Finite Element Method computations. The trained network is then implemented as constitutive material law into ABAQUS/Explicit via the VUMAT interface. Structural simulations are carried out to predict the nonlinear response of composite components.
Sep 11, 2025
Lattice structures consist of repeated similar shaped unit cells and are commonly found in modern engineering applications such as crash structures, acoustic components, or energy-efficient thermal applications. However, the design, analysis and optimization of such structures is still subject of current research. In this presentation we want to show how neural networks can be used to implicitly define their geometry and how immersed boundary methods can be used to analyse their physical behaviour as a part of an optimization process. Unlike other lattice structure optimization methods, we neither assume a large separation of scale nor periodicity. Instead, in each optimization step we perform a full-scale structural analysis. In theory, this enables the use of conventional topology optimization methods such as e.g. the SIMP or the Level-Set method. However, the complex geometry of lattice structures results in an infeasible high number of design parameters. One approach to reduce the number of design parameters is to rely on shape optimization and define a parametrized unit cell using explicit geometrical representations, e.g. [1]. However, only a comparatively small set of geometries can be represented and finding a parametrization for complex geometries is not straight forward, especially if topological changes are desired. In our approach, we employ the DeepSDF [2] method, where a continuous and low-dimensional latent space is introduced to encode the geometric information. Since each single unit cell is characterized by a different latent vector, a spatially graded lattice structure can be created by continuously varying the latent vector over the structure. The neural network is then used to map the geometry from the latent space back to its signed distance representation. This geometrical definition makes this approach especially suited for immersed boundary methods.
Jul 21, 2025
Lattice structures consist of repeated, similarly shaped unit cells and are commonly found in modern engineering applications such as crash structures, acoustic components, or energy-efficient thermal applications. However, the design, analysis and optimization of such structures is still subject of current research. In this presentation we want to show how neural networks can be used to represent the geometry of the individual cells and how this enables an efficient optimization of spatially graded lattice structures.
Feb 20, 2025
The aerodynamic efficiency in different flight conditions can be increased by morphing the wings. Morphing wings require high deformability and high strain recoverability, both of which can be achieved by using optimized structures made of additive manufactured superelastic lattice materials. To obtain an optimized structure of such a wing for certain flight conditions, a mechanical structural optimization of the corresponding trailing edge is presented.
Sep 18, 2024
An example talk using Hugo Blox Builder's Markdown slides feature.
Jun 3, 2024