-
General form of the Gauss-Seidel equation to linearly approximate the Moore-Penrose pseudoinverse in random non-square systems and high order tensors
Authors:
Luis Saucedo-Mora,
Luis Irastorza-Valera
Abstract:
The Gauss-Seidel method has been used for more than 100 years as the standard method for the solution of linear systems of equations under certain restrictions. This method, as well as Cramer and Jacobi, is widely used in education and engineering, but there is a theoretical gap when we want to solve less restricted systems, or even non-square or non-exact systems of equation. Here, the solution g…
▽ More
The Gauss-Seidel method has been used for more than 100 years as the standard method for the solution of linear systems of equations under certain restrictions. This method, as well as Cramer and Jacobi, is widely used in education and engineering, but there is a theoretical gap when we want to solve less restricted systems, or even non-square or non-exact systems of equation. Here, the solution goes through the use of numerical systems, such as the minimization theories or the Moore-Penrose pseudoinverse. In this paper we fill this gap with a global analytical iterative formulation that is capable to reach the solutions obtained with the Moore-Penrose pseudoinverse and the minimization methodologies, but that analytically lies to the solutions of Gauss-Seidel, Jacobi, or Cramer when the system is simplified.
△ Less
Submitted 28 March, 2025;
originally announced March 2025.
-
Mechanostat-type effective density correction for Carter-Hayes growth applied to topology optimization and its efficient interpolation for a target strain energy and volume fraction
Authors:
Luis Irastorza-Valera,
Ricardo Larraínzar-Garijo,
Javier Montoya-Adárraga,
Luis Saucedo-Mora
Abstract:
The need for optimized structures with good mechanical performance for the minimum weight is common in industry. Solid Isotropic Material with Penalization (SIMP) is a Topology Optimization (TO) method offering a trade-off between minimum compliance (i.e., maximum stiffness) and a fixed material amount for a given set of boundary conditions. Since TO is a non-convex problem, its gradient can be tu…
▽ More
The need for optimized structures with good mechanical performance for the minimum weight is common in industry. Solid Isotropic Material with Penalization (SIMP) is a Topology Optimization (TO) method offering a trade-off between minimum compliance (i.e., maximum stiffness) and a fixed material amount for a given set of boundary conditions. Since TO is a non-convex problem, its gradient can be tuned by filtering the topology's contour, creating sharper material profiles without necessarily compromising optimality. However, despite simplifying the layout, some filters fail to address manufacturability concerns such as capillarity (thin tweaks as struts) generated by uncertain loading, vibration or fatigue.
A tailored density-based filtering strategy is offered to tackle this issue. Additionally, volume fraction is left unconstrained so material can be strategically replenished through a logarithmic rule acting on the updated compliance. In doing so, an interpolation space with three degrees of freedom (volume, compliance, minimum thickness) is created, yielding diverse topologies for the same boundary conditions and design values along different stages of evolving topological families with distinct features.
The optimization process is further accelerated by introducing the volume-compliance iterative scheme as a physical loss function in a Double Distance Neural Network (D$^2$NN), obtaining similar results to 2,000 steps worth of vanilla iteration within 500 training epochs. This proposal offers a novel topology optimization design space based on minimum strut thickness - via filtering - and topological families defined by minimum volume fraction and compliance. The methodology is tested on several examples with diverse loading and boundary conditions, obtaining similarly satisfactory results, and then boosted via Machine Learning, acting as a fast and cheap surrogate.
△ Less
Submitted 27 March, 2025;
originally announced March 2025.
-
On the data-driven description of lattice materials mechanics
Authors:
Ismael Ben-Yelun,
Luis Irastorza-Valera,
Luis Saucedo-Mora,
Francisco Javier Montáns,
Francisco Chinesta
Abstract:
In the emerging field of mechanical metamaterials, using periodic lattice structures as a primary ingredient is relatively frequent. However, the choice of aperiodic lattices in these structures presents unique advantages regarding failure, e.g., buckling or fracture, because avoiding repeated patterns prevents global failures, with local failures occurring in turn that can beneficially delay stru…
▽ More
In the emerging field of mechanical metamaterials, using periodic lattice structures as a primary ingredient is relatively frequent. However, the choice of aperiodic lattices in these structures presents unique advantages regarding failure, e.g., buckling or fracture, because avoiding repeated patterns prevents global failures, with local failures occurring in turn that can beneficially delay structural collapse. Therefore, it is expedient to develop models for computing efficiently the effective mechanical properties in lattices from different general features while addressing the challenge of presenting topologies (or graphs) of different sizes. In this paper, we develop a deep learning model to predict energetically-equivalent mechanical properties of linear elastic lattices effectively. Considering the lattice as a graph and defining material and geometrical features on such, we show that Graph Neural Networks provide more accurate predictions than a dense, fully connected strategy, thanks to the geometrically induced bias through graph representation, closer to the underlying equilibrium laws from mechanics solved in the direct problem. Leveraging the efficient forward-evaluation of a vast number of lattices using this surrogate enables the inverse problem, i.e., to obtain a structure having prescribed specific behavior, which is ultimately suitable for multiscale structural optimization problems.
△ Less
Submitted 23 May, 2024; v1 submitted 30 October, 2023;
originally announced October 2023.
