A strategy with reduced models dedicated to parametrized nonlinear strongly coupled thermo-poroelasticity problems

In this work, we attempt to overcome the “curse of dimensionality” inherent to neutron diffusion kinetics problems by employing a novel reduced-order modeling technique known as proper generalized decomposition (PGD). The novelty of this work is that it represents the first attempt at applying PGD reduced-order modeling to time-dependent multigroup neutron diffusion kinetics. The performance of PGD reduced-order models (ROMs) will be quantified by comparing PGD ROMs to reference high-fidelity solutions using Rattlesnake for the TWIGL problem, a standard reactor kinetics benchmark. We show that for problems that exhibit sufficient spatial regularity, our proposed PGD algorithm computes accurate ROMs in less time than the reference high-fidelity calculation. By considering a variation of the TWIGL benchmark that maintains an analogous delayed supercritical behavior but has a smooth spatial solution, we compute PGD ROMs with a maximum relative difference in total power of less than 2.2% using 103 fewer degrees of freedom and a speedup of nearly 13× when compared to reference solutions. However, when introducing the stronger spatial irregularities of the reference benchmark, the accuracy and timing of the proposed PGD algorithm diminishes. We show that by using continuous finite elements, PGD ROMs are subject to undesirable numerical oscillations. In this paper, we motivate the use of PGD in neutron diffusion kinetics, discuss the adopted mathematical framework, and using our results, discuss the challenges and unique aspects of our implementation.

The Power Flow model is extensively used to predict the behavior of electric grids and results in solving a nonlinear algebraic system of equations. Modeling the grid is essential for design optimization and control. Both applications require a fast response for multiple queries to a parametric family of power flow problems. Different solvers have been introduced especially designed for the algebraic nonlinear power flow equations, providing efficient solutions for single problems, even when the number of degrees of freedom is considerably large. However, there is no existing methodology providing an explicit solution of the Parametric Power Flow problem (viz. a computational vademecum, explicit in terms of the parameters). This work aims precisely at designing algorithms producing computational vademecums for the Parametric Power Flow problem. Once these solutions are available, solving for different values of the parameters is an extremely fast (real-time) post-process and therefore both the optimal design and the control problem can readily be addressed. In a first phase, a new family of iteratives solvers for the non-parametric version of the problem is devised. The method is based on a hybrid formulation of the problem combined with an alternated search directions scheme. These methods are designed such that it can be generalized to deal with the parametric version of the problem following a Proper Generalized Decomposition (PGD) strategy. The solver for the parametric problem is conceived by performing the operations involving the unknowns in a PGD fashion. The algorithm follows the basic steps of the algebraic solver, but some operations are carried out in a PGD framework, that is requiring a nested iterative algorithm. The PGD solver is accompanied with an error assessment technique that allows monitoring the convergence of the iterative procedures and deciding the number of terms required to meet the accuracy prescriptions. Different examples of realistic grids and standard benchmark tests are used to demonstrate the performance of the proposed methodologies. El modelo de flujo de potencias se usa para predecir el comportamiento de redes eléctricas y desemboca en la resolución de un sistema de ecuaciones algebraicas no lineales. Modelar una red es esencial para optimizar su diseño y control. Ambas aplicaciones requieren una respuesta rápida a las múltiples peticiones de una familia paramétrica de problemas de flujo de potencias. Diversos métodos de resolución se diseñaron especialmente para resolver la versión algebraica de las ecuaciones de flujo de potencias. Sin embargo, no existe ninguna metodología que proporcione una solución explícita al problema paramétrico de flujo de potencias (esto quiere decir, un vademecum computacional explícito en términos de los parámetros). Esta tesis tiene como objetivo diseñar algoritmos que produzcan vademecums para el problema paramétrico de flujo de potencias. Una vez que las soluciones están disponibles, resolver problemas para diferentes valores de los parámetros es un posproceso extremadamente rápido (en tiempo real) y por lo tanto los problemas de diseño óptimo y control se pueden resolver inmediatamente. En la primera fase, una nueva familia de métodos de resolución iterativos para la versión algebraica del problema se construye. El método se basa en una formulación híbrida del problema combinado con un esquema de direcciones alternadas. Estos métodos se han diseñado para generalizarlos de forma que puedan resolver la versión paramétrica del problema siguiendo una estrategia llamada Descomposición Propia Generalizada (PGD). El método de resolución para el problema paramétrico calcula las incógnitas paramétricas usando la técnica PGD. El algoritmo sigue los mismo pasos que el algoritmo algebraico, pero algunas operaciones se llevan a cabo en el ambiente PGD, esto requiere algoritmos iterativos anidados. El método de resolución PGD se acompaña con una evaluación del error cometido permitiendo monitorizar la convergencia de los procesos iterativos y decidir el número de términos que requiere la solución para alcanzar la precisión preescrita. Diferentes ejemplos de redes reales y tests estándar se usan para demostrar el funcionamiento de las metodologías propuestas.

Characterization of the proper generalized decomposition method for fixed-source diffusion problems

The potential of reduced order models to improve computational efficiency without any loss of behavioral fidelity, is attracting many researchers. Indeed, the reduced order models appear to be efficient for linear models. However, the challenge isn't won yet facing the nonlinear models. The Proper Generalized Decomposition (PGD) is one of the popular reduced order models techniques. In fact, it reduces the computation time by separating the space dimensions and therefore reducing the dimensionality of the problem. Moreover, the PGD treats nonlinearity by a linearization step, using iterations for example. However, the aim of using reduced order models is the computation time reduction. Using iterative linearization techniques, computation time reduction becomes irrelevant and therefore new techniques should be proposed. In this work we propose a new linearization method by combining the PGD and the POD (Proper Orthogonal Decomposition). The treated problem rises from thermoset materials' curing where a coupling between the nonlinear heat equation and the nonlinear curing kinetics exists.

High fidelity structural problems that involve nonlinear material behaviour, when subjected to cyclic loading, usually demand infeasible computational resources; this demonstrates the need for efficient model order reduction (MOR) techniques in order to shrink these demands to fit into the available means. The solution of cyclic damage problems in a model order reduction framework is investigated in this chapter. A semi-incremental framework based on a large time increment (LATIN) approach is proposed to tackle cyclic damage computations under variable amplitude and frequency loadings. The involved MOR approach provides a low-rank approximation in terms of proper generalised decomposition (PGD) of the solution. The generated PGD basis can be interpreted as a set of linear subspaces altered on the fly to the current problem settings. The adaptation of PGD to new settings is based on a greedy algorithm that may lead to a large-sized reduced order basis (ROB). Thus, different orthonormalisation and compression techniques were evaluated to ensure the optimality of the generated ROB in [1] and will be overviewed here. The proposed implementation and a displacement formulated finite element (FE) incremental framework are compared to illustrate their differences in terms of memory footprint and computational time. Numerical examples with variable loadings are discussed, and a typical implementation is provided as open-source code, available at https://gitlab.com/shadialameddin/romfem.

Proper Generalized Decomposition with time adaptive space separation for transient wave propagation problems in separable domains

The proper generalized decomposition (PGD) is a model order reduction method which allows to reduce the computational time of a numerical problem by seeking for a separated representation of the solution. The PGD has been already applied to study an electrical machine but at standstill without accounting the motion of the rotor. In this paper, we propose a method to account for the rotation in the PGD approach in order to build an efficient metamodel of an electrical machine. Then, the machine metamodel will be coupled to its electrical and mechanical environment in order to obtain accurate results with an acceptable computational time on a full simulation.

In the domain of numerical computation, proper generalized decomposition (PGD), which consist in approximating the solution by a truncated sum of separable functions, is applied more and more in mechanics, and has shown its efficiency in terms of computation time and memory requirements. We propose to evaluate the PGD method to solve three-dimensional (3-D) quasi-static field problems coupling with an external electric circuit. The numerical model, obtained from the PGD formulation, is used to study the 3-D examples. The results are compared with those obtained when solving the full original problem. It is shown in this paper that the computation time rate versus the number of time steps is very small compared with the one in a classical time-stepping method, and can be very efficient to solve problems when small time steps are required.

Among the model order reduction techniques, the proper generalized decomposition (PGD) has shown its efficiency to solve a large number of engineering problems. In this paper, the PGD approach is applied to solve a multi-physics problem based on a magnetoelectric device. A reduced model is developed to study the device in its environment based on an offline/online approach. In the offline step, two specific simulations are performed in order to build a PGD reduced model. Then, we obtain a model very well fitted to study in the online stage the influence of parameters like the frequency or the load. The reduced model of the device is coupled with an electric load ( $R$ – $L$ ) to illustrate the possibility offered by the PGD.

Space-energy separated representations for multigroup neutron diffusion using proper generalized decompositions

Thermal transient problems, essential for modeling applications like welding and additive metal manufacturing, are characterized by a dynamic evolution of temperature. Accurately simulating these phenomena is often computationally expensive, thus limiting their applications, for example for model parameter estimation or online process control. Model order reduction, a solution to preserve the accuracy while reducing the computation time, is explored. This article addresses challenges in developing reduced order models using the proper generalized decomposition (PGD) for transient thermal problems with a specific treatment of the moving heat source within the reduced model. Factors affecting accuracy, convergence, and computational cost, such as discretization methods (finite element and finite difference), a dimensionless formulation, the size of the heat source, and the inclusion of material parameters as additional PGD variables are examined across progressively complex examples. The results demonstrate the influence of these factors on the PGD model's performance and emphasize the importance of their consideration when implementing such models. For thermal example, it is demonstrated that a PGD model with a finite difference discretization in time, a dimensionless representation, a mapping for a moving heat source, and a spatial domain non‐separation yields the best approximation to the full order model.

Despite the existence of computationally efficient tools, the effort for parametric investigations is currently high in industry. In this paper, within the context of Li-Ion batteries, an efficient meta-modelling approach based on the Proper Generalized Decomposition (PGD) is considered. From a suitable design of experiments, a parametric model is trained and then exploited to predict, in real time, the system response to a specific parameter combination. In particular, two different methods are considered, the sparse PGD (sPGD) and the anchored-ANOVA based one (ANOVA-PGD). As a use case for the method the dynamic indentation test of a commercial lithium-ion pouch cell with a cylindrical impactor is selected. The cell model considers a homogenised macroscopic structure suitably calibrated for explicit finite element simulations. Four parameters concerning the impactor are varied, both non-geometric (mass and initial velocity) and geometric (diameter and orientation). The study focuses on multi-dimensional outputs, such as curves and contour plots. Inspired by earlier studies, the sPGD is used to predict the force-displacement curves. As a further development, the impactor kinetic energy curve and the displacement contours are both predicted using its recently developed variant ANOVA-PGD. Moreover, a novel curve alignment technique based on the Gappy Proper Orthogonal Decomposition (Gappy-POD) is suggested here. The meta-model is compared to the results of an FE simulation and the resulting deviations are then discussed.

FE2 multiscale in linear elasticity based on parametrized microscale models using proper generalized decomposition

Numerical approach based on the collection of the most significant modes to solve cyclic transient thermal problems involving different time scales

The effective modeling of fresh cementitious materials during setting is an active and challenging research topic. The material behavior as well as the parameters and properties are not fully understood yet. Therefore, the flow during casting is the subject of ongoing research. Previous works attempted to model cement paste as a solid subjected to yielding, or as a fluid modeled by a Brinkman law or a power law, both in 2D or 3D. Of existing models, 3D simulation of power-law fluids seems to carry the best predictive abilities, considering the shear-thinning behavior of the cement paste. In this work, we model the vane test of cement paste using a power-law pseudoplastic fluid. We also add all material and model parameters as extra coordinates of the problem, an appealing approach when identifying material properties, at the expense of increasing the problem’s dimensionality and computational time, often referred to as the combinatory explosion. However, using model order reduction techniques, we have the ability to circumvent the curse of dimensionality by solving a full-scale multidimensional problem as a sequence of lower dimensionality problems. In what follows, we use the Proper Generalized Decomposition (PGD) to solve a multidimensional (6D) nonlinear power-law fluid problem and identify the cement paste properties.