-
Constitutive Manifold Neural Networks
Authors:
Wouter J. Schuttert,
Mohammed Iqbal Abdul Rasheed,
Bojana Rosić
Abstract:
Important material properties like thermal conductivity are often represented as symmetric positive definite (SPD) tensors, which exhibit variability due to inherent material heterogeneity and manufacturing uncertainties. These tensors reside on a curved Riemannian manifold, and accurately modeling their stochastic nature requires preserving both their symmetric positive definite properties and sp…
▽ More
Important material properties like thermal conductivity are often represented as symmetric positive definite (SPD) tensors, which exhibit variability due to inherent material heterogeneity and manufacturing uncertainties. These tensors reside on a curved Riemannian manifold, and accurately modeling their stochastic nature requires preserving both their symmetric positive definite properties and spatial symmetries. To achieve this, uncertainties are parametrized into scaling (magnitude) and rotation (orientation) components, modeled as independent random variables on a manifold structure derived from the maximum entropy principle. The propagation of such stochastic tensors through physics-based simulations necessitates computationally efficient surrogate models. However, traditional multi-layer perceptron (MLP) architectures are not well-suited for SPD tensors, as directly inputting their components fails to preserve their geometric properties, often leading to suboptimal results. To address this, we introduce Constitutive Manifold Neural Networks (CMNN). This approach introduces a preprocessing layer by mapping the SPD tensor from the curved manifold to the local tangent, a flat vector space, creating an information preserving map for input to the hidden layers of the neural networks. A case study on a steady-state heat conduction problem with stochastic anisotropic conductivity demonstrates that geometry-preserving preprocessing, such as logarithmic maps for scaling data, significantly improves learning performance over conventional MLPs. These findings underscore the importance of manifold-aware techniques when working with tensor-valued data in engineering applications.
△ Less
Submitted 16 June, 2025;
originally announced June 2025.
-
ChemZIP: Accelerated Modeling of Complex Aerothermochemical Interactions in Novel Turbomachines for Sustainable High-Temperature Chemical Processes
Authors:
Dylan Rubini,
Budimir Rosic
Abstract:
This paper introduces a new platform to accelerate the modeling of complex aerothermochemical interactions in new turbomachines, turbo-reactors, to decarbonise chemical processes. While previous work has aerothermally demonstrated the potential to decarbonize the heat input to the reaction, optimizing the reaction efficiency has been a challenge. This is because measuring reaction performance with…
▽ More
This paper introduces a new platform to accelerate the modeling of complex aerothermochemical interactions in new turbomachines, turbo-reactors, to decarbonise chemical processes. While previous work has aerothermally demonstrated the potential to decarbonize the heat input to the reaction, optimizing the reaction efficiency has been a challenge. This is because measuring reaction performance with aerochemical simulations is computationally prohibitive due to the uniquely complex aerodynamics and chemistry within turbomachines. To address this, we introduce a new multifidelity machine-learning-assisted methodology, called ChemZIP, to mitigate this bottleneck. Although data-driven methodologies exist for combustion, modeling reactive flows along the bladed path of a turbomachine poses new challenges. This has led to a novel training data generation process, which allows rich dynamic responses of the chemical system to be embedded into the training dataset at a fraction of the cost of reacting flow simulations. The resulting high-dimensional composition vector is compressed into a low-dimensional basis using an autoencoder-like neural network, inspired by but more universal than traditional flamelet-generated manifolds. Verification against 10,000 unseen one-dimensional test conditions shows an R2 score exceeding 95% across all quantities of interest. Following this, ChemZIP is coupled into a fully-fledged viscous computational fluid dynamics solver. For a set of process-relevant three-dimensional configurations entirely different from the training data, the predictive accuracy of the thermochemical state remains within 10% of an industry-standard solver while convergence is achieved 50 times faster, even for a small mechanism. Therefore, numerical computations are sufficiently fast that aerothermochemical optimization is now feasible for the first time in the design cycle
△ Less
Submitted 12 February, 2025;
originally announced February 2025.
-
Stochastic Modelling of Elasticity Tensor Fields
Authors:
Sharana Kumar Shivanand,
Bojana Rosić,
Hermann G. Matthies
Abstract:
We present a novel framework for the probabilistic modelling of random fourth order material tensor fields, with a focus on tensors that are physically symmetric and positive definite (SPD), of which the elasticity tensor is a prime example. Given the critical role that spatial symmetries and invariances play in determining material behaviour, it is essential to incorporate these aspects into the…
▽ More
We present a novel framework for the probabilistic modelling of random fourth order material tensor fields, with a focus on tensors that are physically symmetric and positive definite (SPD), of which the elasticity tensor is a prime example. Given the critical role that spatial symmetries and invariances play in determining material behaviour, it is essential to incorporate these aspects into the probabilistic description and modelling of material properties. In particular, we focus on spatial point symmetries or invariances under rotations, a classical subject in elasticity. Following this, we formulate a stochastic modelling framework using a Lie algebra representation via a memoryless transformation that respects the requirements of positive definiteness and invariance. With this, it is shown how to generate a random ensemble of elasticity tensors that allows an independent control of strength, eigenstrain, and orientation. The procedure also accommodates the requirement to prescribe specific spatial symmetries and invariances for each member of the whole ensemble, while ensuring that the mean or expected value of the ensemble conforms to a potentially 'higher' class of spatial invariance. Furthermore, it is important to highlight that the set of SPD tensors forms a differentiable manifold, which geometrically corresponds to an open cone within the ambient space of symmetric tensors. Thus, we explore the mathematical structure of the underlying sample space of such tensors, and introduce a new distance measure or metric, called the 'elasticity metric', between the tensors. Finally, we model and visualize a one-dimensional spatial field of orthotropic Kelvin matrices using interpolation based on the elasticity metric.
△ Less
Submitted 2 April, 2025; v1 submitted 25 September, 2024;
originally announced September 2024.
-
Stochastic Modelling of Symmetric Positive Definite Material Tensors
Authors:
Sharana Kumar Shivanand,
Bojana Rosić,
Hermann G. Matthies
Abstract:
Spatial symmetries and invariances play an important role in the behaviour of materials and should be respected in the description and modelling of material properties. The focus here is the class of physically symmetric and positive definite tensors, as they appear often in the description of materials, and one wants to be able to prescribe certain classes of spatial symmetries and invariances fo…
▽ More
Spatial symmetries and invariances play an important role in the behaviour of materials and should be respected in the description and modelling of material properties. The focus here is the class of physically symmetric and positive definite tensors, as they appear often in the description of materials, and one wants to be able to prescribe certain classes of spatial symmetries and invariances for each member of the whole ensemble, while at the same time demanding that the mean or expected value of the ensemble be subject to a possibly 'higher' spatial invariance class. We formulate a modelling framework which not only respects these two requirements$-$positive definiteness and invariance$-$but also allows a fine control over orientation on one hand, and strength/size on the other. As the set of positive definite tensors is not a linear space, but rather an open convex cone in the linear space of physically symmetric tensors, we consider it advantageous to widen the notion of mean to the so-called Fréchet mean on a metric space, which is based on distance measures or metrics between positive definite tensors other than the usual Euclidean one. It is shown how the random ensemble can be modelled and generated, independently in its scaling and orientational or directional aspects, with a Lie algebra representation via a memoryless transformation. The parameters which describe the elements in this Lie algebra are then to be considered as random fields on the domain of interest. As an example, a 2D and a 3D model of steady-state heat conduction in a human proximal femur, a bone with high material anisotropy, is modelled with a random thermal conductivity tensor, and the numerical results show the distinct impact of incorporating into the constitutive model different material uncertainties$-$scaling, orientation, and prescribed material symmetry$-$on the desired quantities of interest.
△ Less
Submitted 9 February, 2024; v1 submitted 16 September, 2021;
originally announced September 2021.
-
Scale-invariant multilevel Monte Carlo method and application to linear elasticity
Authors:
Sharana Kumar Shivanand,
Bojana Rosić
Abstract:
We propose a novel scale-invariant version of the mean and variance multi-level Monte Carlo estimate. The computation cost across grid levels is optimised using a normalized error based on t-statistics. By doing so, the algorithm achieves convergence independent of the physical scale at which the estimate is computed. The effectiveness of this algorithm is demonstrated through testing on a linear…
▽ More
We propose a novel scale-invariant version of the mean and variance multi-level Monte Carlo estimate. The computation cost across grid levels is optimised using a normalized error based on t-statistics. By doing so, the algorithm achieves convergence independent of the physical scale at which the estimate is computed. The effectiveness of this algorithm is demonstrated through testing on a linear elastic example, where material uncertainty incorporating both heterogeneity and anisotropy is considered in the constitutive law.
△ Less
Submitted 27 July, 2023; v1 submitted 17 June, 2021;
originally announced June 2021.
-
Long short-term relevance learning
Authors:
Bram van de Weg,
Lars Greve,
Bojana Rosic
Abstract:
To incorporate prior knowledge as well as measurement uncertainties in the traditional long short term memory (LSTM) neural networks, an efficient sparse Bayesian training algorithm is introduced to the network architecture. The proposed scheme automatically determines relevant neural connections and adapts accordingly, in contrast to the classical LSTM solution. Due to its flexibility, the new LS…
▽ More
To incorporate prior knowledge as well as measurement uncertainties in the traditional long short term memory (LSTM) neural networks, an efficient sparse Bayesian training algorithm is introduced to the network architecture. The proposed scheme automatically determines relevant neural connections and adapts accordingly, in contrast to the classical LSTM solution. Due to its flexibility, the new LSTM scheme is less prone to overfitting, and hence can approximate time dependent solutions by use of a smaller data set. On a structural nonlinear finite element application we show that the self-regulating framework does not require prior knowledge of a suitable network architecture and size, while ensuring satisfying accuracy at reasonable computational cost.
△ Less
Submitted 21 June, 2021;
originally announced June 2021.
-
Effect of Load Path on Parameter Identification for Plasticity Models using Bayesian Methods
Authors:
Ehsan Adeli,
Bojana Rosić,
Hermann G. Matthies,
Sven Reinstädler
Abstract:
To evaluate the cyclic behavior under different loading conditions using the kinematic and isotropic hardening theory of steel, a Chaboche viscoplastic material model is employed. The parameters of a constitutive model are usually identified by minimization of the distance between model response and experimental data. However, measurement errors and differences in the specimens lead to deviations…
▽ More
To evaluate the cyclic behavior under different loading conditions using the kinematic and isotropic hardening theory of steel, a Chaboche viscoplastic material model is employed. The parameters of a constitutive model are usually identified by minimization of the distance between model response and experimental data. However, measurement errors and differences in the specimens lead to deviations in the determined parameters. In this article, the Choboche model is used and a stochastic simulation technique is applied to generate artificial data which exhibit the same stochastic behavior as experimental data. Then the model parameters are identified by applying an estimation using Bayes's theorem. The Gauss-Markov-Kalman filter using functional approximation is introduced and employed to estimate the model parameters in the Bayesian setting. Identified parameters are compared with the true parameters in the simulation, and the efficiency of the identification method is discussed. In the end, the effect of the load path on parameter identification is investigated.
△ Less
Submitted 17 June, 2019;
originally announced June 2019.
-
Parameter Identification in a Probabilistic Setting
Authors:
Bojana V. Rosić,
Anna Kučerová,
Jan Sýkora,
Oliver Pajonk,
Alexander Litvinenko,
Hermann G. Matthies
Abstract:
Parameter identification problems are formulated in a probabilistic language, where the randomness reflects the uncertainty about the knowledge of the true values. This setting allows conceptually easily to incorporate new information, e.g. through a measurement, by connecting it to Bayes's theorem. The unknown quantity is modelled as a (may be high-dimensional) random variable. Such a description…
▽ More
Parameter identification problems are formulated in a probabilistic language, where the randomness reflects the uncertainty about the knowledge of the true values. This setting allows conceptually easily to incorporate new information, e.g. through a measurement, by connecting it to Bayes's theorem. The unknown quantity is modelled as a (may be high-dimensional) random variable. Such a description has two constituents, the measurable function and the measure. One group of methods is identified as updating the measure, the other group changes the measurable function. We connect both groups with the relatively recent methods of functional approximation of stochastic problems, and introduce especially in combination with the second group of methods a new procedure which does not need any sampling, hence works completely deterministically. It also seems to be the fastest and more reliable when compared with other methods. We show by example that it also works for highly nonlinear non-smooth problems with non-Gaussian measures.
△ Less
Submitted 19 January, 2012;
originally announced January 2012.
-
Acceleration of Uncertainty Updating in the Description of Transport Processes in Heterogeneous Materials
Authors:
A. Kucerova,
J. Sykora,
B. Rosic,
H. G. Matthies
Abstract:
The prediction of thermo-mechanical behaviour of heterogeneous materials such as heat and moisture transport is strongly influenced by the uncertainty in parameters. Such materials occur e.g. in historic buildings, and the durability assessment of these therefore needs a reliable and probabilistic simulation of transport processes, which is related to the suitable identification of material parame…
▽ More
The prediction of thermo-mechanical behaviour of heterogeneous materials such as heat and moisture transport is strongly influenced by the uncertainty in parameters. Such materials occur e.g. in historic buildings, and the durability assessment of these therefore needs a reliable and probabilistic simulation of transport processes, which is related to the suitable identification of material parameters. In order to include expert knowledge as well as experimental results, one can employ an updating procedure such as Bayesian inference. The classical probabilistic setting of the identification process in Bayes's form requires the solution of a stochastic forward problem via computationally expensive sampling techniques, which makes the method almost impractical. In this paper novel stochastic computational techniques such as the stochastic Galerkin method are applied in order to accelerate the updating procedure. The idea is to replace the computationally expensive forward simulation via the conventional finite element (FE) method by the evaluation of a polynomial chaos expansion (PCE). Such an approximation of the FE model for the forward simulation perfectly suits the Bayesian updating. The presented uncertainty updating techniques are applied to the numerical model of coupled heat and moisture transport in heterogeneous materials with spatially varying coefficients defined by random fields.
△ Less
Submitted 10 October, 2011;
originally announced October 2011.
