Project Details
Projekt
iLattice: Mechanistic machine learning for interactive design of flexible lattice structures
Applicant
Professor Dr. Oliver Weeger
Subject Area
Mechanics
Engineering Design, Machine Elements, Product Development
Engineering Design, Machine Elements, Product Development
Term
since 2025
Project identifier
Deutsche Forschungsgemeinschaft (DFG) - Project number 564439980
Additively manufactured, flexible, compliant lattice structures with nonlinear, functional dynamic mechanical behavior have significant potential for application in tailored actuation, energy absorption and dissipation, vibration mitigation, ventilation, or weight reduction in consumer products, soft robotics, or biomedicine. However, current research and commercial design tools are limited to elastostatic simulation and optimization capabilities of lattice structures, neglecting to consider multiple design objectives or uncertainties in geometry, material, or loading. Multiscale modeling techniques are often not applicable to additively manufactured structures without scale separation and periodicity, which poses a significant challenge to computational design. Consequently, the numerical effort associated with the highly nonlinear and dynamic simulation of flexible lattice structures, whether modeled as 3D solids or beams, is substantial. To address this challenge, physics-aware, mechanistic machine learning presents as a viable alternative. It can provide computationally efficient, accurate, and reliable structural dynamic surrogate models. Furthermore, to help engineers obtain optimized lattice designs for diverse applications, robust and multi-objective optimization capabilities are essential. Thus, the objective of the iLattice project is to develop a computational assistant for the interactive design of flexible lattice structures with highly nonlinear dynamic behavior. This objective will be achieved through the creation of accurate, efficient, and robust nonlinear surrogate models for lattice unit cells and structures. These surrogate models will be derived using mechanistic machine learning to ensure thermodynamically consistent dynamics and will be parameterized in terms of design variables such as unit cell geometries, topologies, strut diameters, or material properties. Then, they will be integrated with nonlinear and Bayesian multi-objective optimization methods, which will be formulated for diverse objectives and constraints. Furthermore, uncertainties will be considered by stochastic, Bayesian robust optimization approaches. These techniques will be integrated into a user-friendly computer-aided design tool. The resulting interactive design assistant will help engineers achieve Pareto-optimal, robust design suggestions and perform rapid design space exploration guided by design criteria, experience, and intuition. Overall, this project will expedite the development of reliable, mechanistic machine learning approaches for structural dynamics and their integration with machine learning-based multi-objective and robust optimization methods. While this framework will be specifically tailored for the creation of an interactive design assistant for flexible lattice structures, it holds the potential for applications far beyond this field in engineering design.
DFG Programme
Priority Programmes
