31.80 Angewandte Mathematik
Refine
Document Type
- Conference Proceeding (358)
- Article (261)
- Master's Thesis (3)
- Doctoral Thesis (2)
- Bachelor Thesis (1)
Institute
- Professur Informatik im Bauwesen (281)
- Institut für Strukturmechanik (ISM) (202)
- In Zusammenarbeit mit der Bauhaus-Universität Weimar (82)
- Professur Stochastik und Optimierung (42)
- Graduiertenkolleg 1462 (32)
- Professur Angewandte Mathematik (17)
- Institut für Konstruktiven Ingenieurbau (IKI) (4)
- Professur Baubetrieb und Bauverfahren (2)
- Professur Computer Vision in Engineering (2)
- Professur Modellierung und Simulation - Mechanik (2)
Keywords
- Angewandte Mathematik (328)
- Strukturmechanik (183)
- Computerunterstütztes Verfahren (153)
- Angewandte Informatik (146)
- Architektur <Informatik> (75)
- Computer Science Models in Engineering; Multiscale and Multiphysical Models; Scientific Computing (74)
- Modellierung (44)
- Stochastik (40)
- Building Information Modeling (35)
- CAD (35)
The study presents a Machine Learning (ML)-based framework designed to forecast the stress-strain relationship of arc-direct energy deposited mild steel. Based on microstructural characteristics previously extracted using microscopy and X-ray diffraction, approximately 1000 new parameter sets are generated by applying the Latin Hypercube Sampling Method (LHSM). For each parameter set, a Representative Volume Element (RVE) is synthetically created via Voronoi Tessellation. Input raw data for ML-based algorithms comprises these parameter sets or RVE-images, while output raw data includes their corresponding stress-strain relationships calculated after a Finite Element (FE) procedure. Input data undergoes preprocessing involving standardization, feature selection, and image resizing. Similarly, the stress-strain curves, initially unsuitable for training traditional ML algorithms, are preprocessed using cubic splines and occasionally Principal Component Analysis (PCA). The later part of the study focuses on employing multiple ML algorithms, utilizing two main models. The first model predicts stress-strain curves based on microstructural parameters, while the second model does so solely from RVE images. The most accurate prediction yields a Root Mean Squared Error of around 5 MPa, approximately 1% of the yield stress. This outcome suggests that ML models offer precise and efficient methods for characterizing dual-phase steels, establishing a framework for accurate results in material analysis.
Identification of modal parameters of a space frame structure is a complex assignment due to a large number of degrees of freedom, close natural frequencies, and different vibrating mechanisms. Research has been carried out on the modal identification of rather simple truss structures. So far, less attention has been given to complex three-dimensional truss structures. This work develops a vibration-based methodology for determining modal information of three-dimensional space truss structures. The method uses a relatively complex space truss structure for its verification. Numerical modelling of the system gives modal information about the expected vibration behaviour. The identification process involves closely spaced modes that are characterised by local and global vibration mechanisms. To distinguish between local and global vibrations of the system, modal strain energies are used as an indicator. The experimental validation, which incorporated a modal analysis employing the stochastic subspace identification method, has confirmed that considering relatively high model orders is required to identify specific mode shapes. Especially in the case of the determination of local deformation modes of space truss members, higher model orders have to be taken into account than in the modal identification of most other types of structures.
As an optimization that starts from a randomly selected structure generally does not guarantee reasonable optimality, the use of a systemic approach, named the ground structure, is widely accepted in steel-made truss and frame structural design. However, in the case of reinforced concrete (RC) structural optimization, because of the orthogonal orientation of structural members, randomly chosen or architect-sketched framing is used. Such a one-time fixed layout trend, in addition to its lack of a systemic approach, does not necessarily guarantee optimality. In this study, an approach for generating a candidate ground structure to be used for cost or weight minimization of 3D RC building structures with included slabs is developed. A multiobjective function at the floor optimization stage and a single objective function at the frame optimization stage are considered. A particle swarm optimization (PSO) method is employed for selecting the optimal ground structure. This method enables generating a simple, yet potential, real-world representation of topologically preoptimized ground structure while both structural and main architectural requirements are considered. This is supported by a case study for different floor domain sizes.
We present a physics-informed deep learning model for the transient heat transfer analysis of three-dimensional functionally graded materials (FGMs) employing a Runge–Kutta discrete time scheme. Firstly, the governing equation, associated boundary conditions and the initial condition for transient heat transfer analysis of FGMs with exponential material variations are presented. Then, the deep collocation method with the Runge–Kutta integration scheme for transient analysis is introduced. The prior physics that helps to generalize the physics-informed deep learning model is introduced by constraining the temperature variable with discrete time schemes and initial/boundary conditions. Further the fitted activation functions suitable for dynamic analysis are presented. Finally, we validate our approach through several numerical examples on FGMs with irregular shapes and a variety of boundary conditions. From numerical experiments, the predicted results with PIDL demonstrate well agreement with analytical solutions and other numerical methods in predicting of both temperature and flux distributions and can be adaptive to transient analysis of FGMs with different shapes, which can be the promising surrogate model in transient dynamic analysis.
Nonlocal theories concern the interaction of objects, which are separated in space. Classical examples are Coulomb’s law or Newton’s law of universal gravitation. They had signficiant impact in physics and engineering. One classical application in mechanics is the failure of quasi-brittle materials. While local models lead to an ill-posed boundary value problem and associated mesh dependent results, nonlocal models guarantee the well-posedness and are furthermore relatively easy to implement into commercial computational software.
In this work, we present a deep collocation method (DCM) for three-dimensional potential problems in non-homogeneous media. This approach utilizes a physics-informed neural network with material transfer learning reducing the solution of the non-homogeneous partial differential equations to an optimization problem. We tested different configurations of the physics-informed neural network including smooth activation functions, sampling methods for collocation points generation and combined optimizers. A material transfer learning technique is utilized for non-homogeneous media with different material gradations and parameters, which enhance the generality and robustness of the proposed method. In order to identify the most influential parameters of the network configuration, we carried out a global sensitivity analysis. Finally, we provide a convergence proof of our DCM. The approach is validated through several benchmark problems, also testing different material variations.
In machine learning, if the training data is independently and identically distributed as the test data then a trained model can make an accurate predictions for new samples of data. Conventional machine learning has a strong dependence on massive amounts of training data which are domain specific to understand their latent patterns. In contrast, Domain adaptation and Transfer learning methods are sub-fields within machine learning that are concerned with solving the inescapable problem of insufficient training data by relaxing the domain dependence hypothesis. In this contribution, this issue has been addressed and by making a novel combination of both the methods we develop a computationally efficient and practical algorithm to solve boundary value problems based on nonlinear partial differential equations. We adopt a meshfree analysis framework to integrate the prevailing geometric modelling techniques based on NURBS and present an enhanced deep collocation approach that also plays an important role in the accuracy of solutions. We start with a brief introduction on how these methods expand upon this framework. We observe an excellent agreement between these methods and have shown that how fine-tuning a pre-trained network to a specialized domain may lead to an outstanding performance compare to the existing ones. As proof of concept, we illustrate the performance of our proposed model on several benchmark problems.
In this paper, we present an open-source code for the first-order and higher-order nonlocal operator method (NOM) including a detailed description of the implementation. The NOM is based on so-called support, dual-support, nonlocal operators, and an operate energy functional ensuring stability. The nonlocal operator is a generalization of the conventional differential operators. Combined with the method of weighed residuals and variational principles, NOM establishes the residual and tangent stiffness matrix of operate energy functional through some simple matrix without the need of shape functions as in other classical computational methods such as FEM. NOM only requires the definition of the energy drastically simplifying its implementation. The implementation in this paper is focused on linear elastic solids for sake of conciseness through the NOM can handle more complex nonlinear problems. The NOM can be very flexible and efficient to solve partial differential equations (PDEs), it’s also quite easy for readers to use the NOM and extend it to solve other complicated physical phenomena described by one or a set of PDEs. Finally, we present some classical benchmark problems including the classical cantilever beam and plate-with-a-hole problem, and we also make an extension of this method to solve complicated problems including phase-field fracture modeling and gradient elasticity material.
In this study, we propose a nonlocal operator method (NOM) for the dynamic analysis of (thin) Kirchhoff plates. The nonlocal Hessian operator is derived based on a second-order Taylor series expansion. The NOM does not require any shape functions and associated derivatives as ’classical’ approaches such as FEM, drastically facilitating the implementation. Furthermore, NOM is higher order continuous, which is exploited for thin plate analysis that requires C1 continuity. The nonlocal dynamic governing formulation and operator energy functional for Kirchhoff plates are derived from a variational principle. The Verlet-velocity algorithm is used for the time discretization. After confirming the accuracy of the nonlocal Hessian operator, several numerical examples are simulated by the nonlocal dynamic Kirchhoff plate formulation.
We present a stochastic deep collocation method (DCM) based on neural architecture search (NAS) and transfer learning for heterogeneous porous media. We first carry out a sensitivity analysis to determine the key hyper-parameters of the network to reduce the search space and subsequently employ hyper-parameter optimization to finally obtain the parameter values. The presented NAS based DCM also saves the weights and biases of the most favorable architectures, which is then used in the fine-tuning process. We also employ transfer learning techniques to drastically reduce the computational cost. The presented DCM is then applied to the stochastic analysis of heterogeneous porous material. Therefore, a three dimensional stochastic flow model is built providing a benchmark to the simulation of groundwater flow in highly heterogeneous aquifers. The performance of the presented NAS based DCM is verified in different dimensions using the method of manufactured solutions. We show that it significantly outperforms finite difference methods in both accuracy and computational cost.
This paper presents numerical analysis of the discrete fundamental solution of the discrete Laplace operator on a rectangular lattice. Additionally, to provide estimates in interior and exterior domains, two different regularisations of the discrete fundamental solution are considered. Estimates for the absolute difference and lp-estimates are constructed for both regularisations. Thus, this work extends the classical results in the discrete potential theory to the case of a rectangular lattice and serves as a basis for future convergence analysis of the method of discrete potentials on rectangular lattices.
Realistic uncertainty description incorporating aleatoric and epistemic uncertainties can be described within the framework of polymorphic uncertainty, which is computationally demanding. Utilizing a domain decomposition approach for random field based uncertainty models the proposed level-based sampling method can reduce these computational costs significantly and shows good agreement with a standard sampling technique. While 2-level configurations tend to get unstable with decreasing sampling density 3-level setups show encouraging results for the investigated reliability analysis of a structural unit square.
When it comes to monitoring of huge structures, main issues are limited time, high costs and how to deal with the big amount of data. In order to reduce and manage them, respectively, methods from the field of optimal design of experiments are useful and supportive. Having optimal experimental designs at hand before conducting any measurements is leading to a highly informative measurement concept, where the sensor positions are optimized according to minimal errors in the structures’ models. For the reduction of computational time a combined approach using Fisher Information Matrix and mean-squared error in a two-step procedure is proposed under the consideration of different error types. The error descriptions contain random/aleatoric and systematic/epistemic portions. Applying this combined approach on a finite element model using artificial acceleration time measurement data with artificially added errors leads to the optimized sensor positions. These findings are compared to results from laboratory experiments on the modeled structure, which is a tower-like structure represented by a hollow pipe as the cantilever beam. Conclusively, the combined approach is leading to a sound experimental design that leads to a good estimate of the structure’s behavior and model parameters without the need of preliminary measurements for model updating.
Das Hauptziel der vorliegenden Arbeit war es, eine stetige Kopplung zwischen der ananlytischen und numerischen Lösung von Randwertaufgaben mit Singularitäten zu realisieren. Durch die inter-polationsbasierte gekoppelte Methode kann eine globale C0 Stetigkeit erzielt werden. Für diesen Zweck wird ein spezielle finite Element (Kopplungselement) verwendet, das die Stetigkeit der Lösung sowohl mit dem analytischen Element als auch mit den normalen CST Elementen gewährleistet.
Die interpolationsbasierte gekoppelte Methode ist zwar für beliebige Knotenanzahl auf dem Interface ΓAD anwendbar, aber es konnte durch die Untersuchung von der Interpolationsmatrix und numerische Simulationen festgestellt werden, dass sie schlecht konditioniert ist. Um das Problem mit den numerischen Instabilitäten zu bewältigen, wurde eine approximationsbasierte Kopplungsmethode entwickelt und untersucht. Die Stabilität dieser Methode wurde anschließend anhand der Untersuchung von der Gramschen Matrix des verwendeten Basissystems auf zwei Intervallen [−π,π] und [−2π,2π] beurteilt. Die Gramsche Matrix auf dem Intervall [−2π,2π] hat einen günstigeren Konditionszahl in der Abhängigkeit von der Anzahl der Kopplungsknoten auf dem Interface aufgewiesen. Um die dazu gehörigen numerischen Instabilitäten ausschließen zu können wird das Basissystem mit Hilfe vom Gram-Schmidtschen Orthogonalisierungsverfahren auf beiden Intervallen orthogonalisiert. Das orthogonale Basissystem lässt sich auf dem Intervall [−2π,2π] mit expliziten Formeln schreiben. Die Methode des konsistentes Sampling, die häufig in der Nachrichtentechnik verwendet wird, wurde zur Realisierung von der approximationsbasierten Kopplung herangezogen. Eine Beschränkung dieser Methode ist es, dass die Anzahl der Sampling-Basisfunktionen muss gleich der Anzahl der Wiederherstellungsbasisfunktionen sein. Das hat dazu geführt, dass das eingeführt Basissys-tem (mit 2 n Basisfunktionen) nur mit n Basisfunktion verwendet werden kann.
Zur Lösung diese Problems wurde ein alternatives Basissystems (Variante 2) vorgestellt. Für die Verwendung dieses Basissystems ist aber eine Transformationsmatrix M nötig und bei der Orthogonalisierung des Basissystems auf dem Intervall [−π,π] kann die Herleitung von dieser Matrix kompliziert und aufwendig sein. Die Formfunktionen wurden anschließend für die beiden Varianten hergeleitet und grafisch (für n = 5) dargestellt und wurde gezeigt, dass diese Funktionen die Anforderungen an den Formfunktionen erfüllen und können somit für die FE- Approximation verwendet werden.
Anhand numerischer Simulationen, die mit der Variante 1 (mit Orthogonalisierung auf dem Intervall [−2π,2π]) durchgeführt wurden, wurden die grundlegenden Fragen (Beispielsweise: Stetigkeit der Verformungen auf dem Interface ΓAD, Spannungen auf dem analytischen Gebiet) über-
prüft.
Rapid advancements of modern technologies put high demands on mathematical modelling of engineering systems. Typically, systems are no longer “simple” objects, but rather coupled systems involving multiphysics phenomena, the modelling of which involves coupling of models that describe different phenomena. After constructing a mathematical model, it is essential to analyse the correctness of the coupled models and to detect modelling errors compromising the final modelling result. Broadly, there are two classes of modelling errors: (a) errors related to abstract modelling, eg, conceptual errors concerning the coherence of a model as a whole and (b) errors related to concrete modelling or instance modelling, eg, questions of approximation quality and implementation. Instance modelling errors, on the one hand, are relatively well understood. Abstract modelling errors, on the other, are not appropriately addressed by modern modelling methodologies. The aim of this paper is to initiate a discussion on abstract approaches and their usability for mathematical modelling of engineering systems with the goal of making it possible to catch conceptual modelling errors early and automatically by computer assistant tools. To that end, we argue that it is necessary to identify and employ suitable mathematical abstractions to capture an accurate conceptual description of the process of modelling engineering systems.
Scalarization methods are a category of multiobjective optimization (MOO) methods. These methods allow the usage of conventional single objective optimization algorithms, as scalarization methods reformulate the MOO problem into a single objective optimization problem. The scalarization methods analysed within this thesis are the Weighted Sum (WS), the Epsilon-Constraint (EC), and the MinMax (MM) method. After explaining the approach of each method, the WS, EC and MM are applied, a-posteriori, to three different examples: to the Kursawe function; to the ten bar truss, a common benchmark problem in structural optimization; and to the metamodel of an aero engine exit module.
The aim is to evaluate and compare the performance of each scalarization method that is examined within this thesis. The evaluation is conducted using performance metrics, such as the hypervolume and the generational distance, as well as using visual comparison.
The application to the three examples gives insight into the advantages and disadvantages of each method, and provides further understanding of an adequate application of the methods concerning high dimensional optimization problems.