600 Technik, Medizin, angewandte Wissenschaften
Refine
Document Type
- Article (11)
- Doctoral Thesis (11)
- Bachelor Thesis (1)
- Book (1)
- Master's Thesis (1)
Institute
- Institut für Strukturmechanik (ISM) (11)
- Bauhaus-Institut für zukunftsweisende Infrastruktursysteme (b.is) (3)
- Professur Stahl- und Hybridbau (2)
- Professur Werkstoffe des Bauens (2)
- Fakultät Kunst und Gestaltung (1)
- Materialforschungs- und -prüfanstalt an der Bauhaus-Universität (1)
- Professur Baubetrieb und Bauverfahren (1)
- Professur Informatik in der Architektur (1)
- Professur Konstruktives Entwerfen und Tragwerkslehre (1)
- Professur Modellierung und Simulation - Mechanik (1)
Keywords
- OA-Publikationsfonds2022 (8)
- Finite-Elemente-Methode (3)
- Maschinelles Lernen (3)
- Beton (2)
- Finite Element Method (2)
- Isogeometric Analysis (2)
- Isogeometrische Analyse (2)
- Lebenszyklus (2)
- Partielle Differentialgleichung (2)
- Phasenfeldmodell (2)
Year of publication
- 2022 (25) (remove)
The Finite Element Method (FEM) is widely used in engineering for solving Partial Differential Equations (PDEs) over complex geometries. To this end, it is required to provide the FEM software with a geometric model that is typically constructed in a Computer-Aided Design (CAD) software. However, FEM and CAD use different approaches for the mathematical description of the geometry. Thus, it is required to generate a mesh, which is suitable for FEM, based on the CAD model. Nonetheless, this procedure is not a trivial task and it can be time consuming. This issue becomes more significant for solving shape and topology optimization problems, which consist in evolving the geometry iteratively. Therefore, the computational cost associated to the mesh generation process is increased exponentially for this type of applications.
The main goal of this work is to investigate the integration of CAD and CAE in shape and topology optimization. To this end, numerical tools that close the gap between design and analysis are presented. The specific objectives of this work are listed below:
• Automatize the sensitivity analysis in an isogeometric framework for applications in shape optimization. Applications for linear elasticity are considered.
• A methodology is developed for providing a direct link between the CAD model and the analysis mesh. In consequence, the sensitivity analysis can be performed in terms of the design variables located in the design model.
• The last objective is to develop an isogeometric method for shape and topological optimization. This method should take advantage of using Non-Uniform Rational B-Splines (NURBS) with higher continuity as basis functions.
Isogeometric Analysis (IGA) is a framework designed to integrate the design and analysis in engineering problems. The fundamental idea of IGA is to use the same basis functions for modeling the geometry, usually NURBS, for the approximation of the solution fields. The advantage of integrating design and analysis is two-fold. First, the analysis stage is more accurate since the system of PDEs is not solved using an approximated geometry, but the exact CAD model. Moreover, providing a direct link between the design and analysis discretizations makes possible the implementation of efficient sensitivity analysis methods. Second, the computational time is significantly reduced because the mesh generation process can be avoided.
Sensitivity analysis is essential for solving optimization problems when gradient-based optimization algorithms are employed. Automatic differentiation can compute exact gradients, automatically by tracking the algebraic operations performed on the design variables. For the automation of the sensitivity analysis, an isogeometric framework is used. Here, the analysis mesh is obtained after carrying out successive refinements, while retaining the coarse geometry for the domain design. An automatic differentiation (AD) toolbox is used to perform the sensitivity analysis. The AD toolbox takes the code for computing the objective and constraint functions as input. Then, using a source code transformation approach, it outputs a code for computing the objective and constraint functions, and their sensitivities as well. The sensitivities obtained from the sensitivity propagation method are compared with analytical sensitivities, which are computed using a full isogeometric approach.
The computational efficiency of AD is comparable to that of analytical sensitivities. However, the memory requirements are larger for AD. Therefore, AD is preferable if the memory requirements are satisfied. Automatic sensitivity analysis demonstrates its practicality since it simplifies the work of engineers and designers.
Complex geometries with sharp edges and/or holes cannot easily be described with NURBS. One solution is the use of unstructured meshes. Simplex-elements (triangles and tetrahedra for two and three dimensions respectively) are particularly useful since they can automatically parameterize a wide variety of domains. In this regard, unstructured Bézier elements, commonly used in CAD, can be employed for the exact modelling of CAD boundary representations. In two dimensions, the domain enclosed by NURBS curves is parameterized with Bézier triangles. To describe exactly the boundary of a two-dimensional CAD model, the continuity of a NURBS boundary representation is reduced to C^0. Then, the control points are used to generate a triangulation such that the boundary of the domain is identical to the initial CAD boundary representation. Thus, a direct link between the design and analysis discretizations is provided and the sensitivities can be propagated to the design domain.
In three dimensions, the initial CAD boundary representation is given as a collection of NURBS surfaces that enclose a volume. Using a mesh generator (Gmsh), a tetrahedral mesh is obtained. The original surface is reconstructed by modifying the location of the control points of the tetrahedral mesh using Bézier tetrahedral elements and a point inversion algorithm. This method offers the possibility of computing the sensitivity analysis using the analysis mesh. Then, the sensitivities can be propagated into the design discretization. To reuse the mesh originally generated, a moving Bézier tetrahedral mesh approach was implemented.
A gradient-based optimization algorithm is employed together with a sensitivity propagation procedure for the shape optimization cases. The proposed shape optimization approaches are used to solve some standard benchmark problems in structural mechanics. The results obtained show that the proposed approach can compute accurate gradients and evolve the geometry towards optimal solutions. In three dimensions, the moving mesh approach results in faster convergence in terms of computational time and avoids remeshing at each optimization step.
For considering topological changes in a CAD-based framework, an isogeometric phase-field based shape and topology optimization is developed. In this case, the diffuse interface of a phase-field variable over a design domain implicitly describes the boundaries of the geometry. The design variables are the local values of the phase-field variable. The descent direction to minimize the objective function is found by using the sensitivities of the objective function with respect to the design variables. The evolution of the phase-field is determined by solving the time dependent Allen-Cahn equation.
Especially for topology optimization problems that require C^1 continuity, such as for flexoelectric structures, the isogeometric phase field method is of great advantage. NURBS can achieve the desired continuity more efficiently than the traditional employed functions. The robustness of the method is demonstrated when applied to different geometries, boundary conditions, and material configurations. The applications illustrate that compared to piezoelectricity, the electrical performance of flexoelectric microbeams is larger under bending. In contrast, the electrical power for a structure under compression becomes larger with piezoelectricity.
BIM-basierte Digitalisierung von Bestandsgebäuden aus Sicht des FM am Beispiel von Heizungsanlagen
(2022)
Das Ziel der Arbeit ist, für das Facility Management relevante Informationen für die mit Building Information Modeling basierende Erstellung von Bestandsgebäuden am Beispiel einer Hei- zungsanlage zu definieren. Darauf basierend sind die notwendigen Arbeitsschritte der Objek- taufnahme abgeleitet. Für die Definition der Arbeitsschritte wurden das grundlegende Vorge- hen bei einer Objektaufnahme sowie die gesetzlichen Gegebenheiten für den Betrieb einer Heizungsanlage dargelegt. Darüber hinaus sind in der vorliegenden Ausarbeitung die Vorteile und Herausforderungen hinsichtlich des Zusammenspiels von Building Information Modeling und Facility Management analysiert. Die definierten Arbeitsschritte sind anhand eines Beispiel- projektes angewendet worden. Im Rahmen des Beispielprojekts sind die entscheidenden Be- triebsdaten je Anlagenteil in Form von Informationsanforderungen nach DIN 17412 definiert. Das Gebäudemodell ist durch Parameter mit den für das Facility Management relevanten In- formationen ergänzt. Die Resultate des Beispielprojektes sind mit aussagekräftigen Schnitten, Plänen sowie 3-D-Visualisierungen dargestellt. Abschließend sind die Ergebnisse in Bezug auf das FM validiert. Aus den Arbeitsschritten und Ergebnissen ist eine Leitlinie erstellt worden für den Digitalisierungsprozess von Bestandsgebäuden für das Facility Management.
Quantification of cracks in concrete thin sections considering current methods of image analysis
(2022)
Image analysis is used in this work to quantify cracks in concrete thin sections via modern image processing. Thin sections were impregnated with a yellow epoxy resin, to increase the contrast between voids and other phases of the concrete. By the means of different steps of pre-processing, machine learning and python scripts, cracks can be quantified in an area of up to 40 cm2. As a result, the crack area, lengths and widths were estimated automatically within a single workflow. Crack patterns caused by freeze-thaw damages were investigated. To compare the inner degradation of the investigated thin sections, the crack density was used. Cracks in the thin sections were measured manually in two different ways for validation of the automatic determined results. On the one hand, the presented work shows that the width of cracks can be determined pixelwise, thus providing the plot of a width distribution. On the other hand, the automatically measured crack length differs in comparison to the manually measured ones.
Material failure can be tackled by so-called nonlocal models, which introduce an intrinsic length scale into the formulation and, in the case of material failure, restore the well-posedness of the underlying boundary value problem or initial boundary value problem. Among nonlocal models, peridynamics (PD) has attracted a lot of attention as it allows the natural transition from continuum to discontinue and thus allows modeling of discrete cracks without the need to describe and track the crack topology, which has been a major obstacle in traditional discrete crack approaches. This is achieved by replacing the divergence of the Cauchy stress tensor through an integral over so-called bond forces, which account for the interaction of particles. A quasi-continuum approach is then used to calibrate the material parameters of the bond forces, i.e., equating the PD energy with the energy of a continuum. One major issue for the application of PD to general complex problems is that they are limited to fairly simple material behavior and pure mechanical problems based on explicit time integration. PD has been extended to other applications but losing simultaneously its simplicity and ease in modeling material failure. Furthermore, conventional PD suffers from instability and hourglass modes that require stabilization. It also requires the use of constant horizon sizes, which drastically reduces its computational efficiency. The latter issue was resolved by the so-called dual-horizon peridynamics (DH-PD) formulation and the introduction of the duality of horizons.
Within the nonlocal operator method (NOM), the concept of nonlocality is further extended and can be considered a generalization of DH-PD. Combined with the energy functionals of various physical models, the nonlocal forms based on the dual-support concept can be derived. In addition, the variation of the energy functional allows implicit formulations of the nonlocal theory. While traditional integral equations are formulated in an integral domain, the dual-support approaches are based on dual integral domains. One prominent feature of NOM is its compatibility with variational and weighted residual methods. The NOM yields a direct numerical implementation based on the weighted residual method for many physical problems without the need for shape functions. Only the definition of the energy or boundary value problem is needed to drastically facilitate the implementation. The nonlocal operator plays an equivalent role to the derivatives of the shape functions in meshless methods and finite element methods (FEM). Based on the variational principle, the residual and the tangent stiffness matrix can be obtained with ease by a series of matrix multiplications. In addition, NOM can be used to derive many nonlocal models in strong form.
The principal contributions of this dissertation are the implementation and application of NOM, and also the development of approaches for dealing with fractures within the NOM, mostly for dynamic fractures. The primary coverage and results of the dissertation are as follows:
-The first/higher-order implicit NOM and explicit NOM, including a detailed description of the implementation, are presented. 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. Combining with the method of weighted 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. For the sake of conciseness, the implementation in this chapter is focused on linear elastic solids only, though the NOM can handle more complex nonlinear problems. An explicit nonlocal operator method for the dynamic analysis of elasticity solid problems is also presented. The explicit NOM avoids the calculation of the tangent stiffness matrix as in the implicit NOM model. The explicit scheme comprises the Verlet-velocity algorithm. The NOM can be very flexible and efficient for solving 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. Several numerical examples are presented to show the capabilities of this method.
-A nonlocal operator method for the dynamic analysis of (thin) Kirchhoff plates is proposed. The nonlocal Hessian operator is derived from a second-order Taylor series expansion. NOM is higher-order continuous, which is exploited for thin plate analysis that requires $C^1$ 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 time discretization. After confirming the accuracy of the nonlocal Hessian operator, several numerical examples are simulated by the nonlocal dynamic Kirchhoff plate formulation.
-A nonlocal fracture modeling is developed and applied to the simulation of quasi-static and dynamic fractures using the NOM. The phase field's nonlocal weak and associated strong forms are derived from a variational principle. The NOM requires only the definition of energy. We present both a nonlocal implicit phase field model and a nonlocal explicit phase field model for fracture; the first approach is better suited for quasi-static fracture problems, while the key application of the latter one is dynamic fracture. To demonstrate the performance of the underlying approach, several benchmark examples for quasi-static and dynamic fracture are solved.
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.
Tropical coral reefs, one of the world’s oldest ecosystems which support some of the highest levels of biodiversity on the planet, are currently facing an unprecedented ecological crisis during this massive human-activity-induced period of extinction. Hence, tropical reefs symbolically stand for the destructive effects of human activities on nature [4], [5]. Artificial reefs are excellent examples of how architectural design can be combined with ecosystem regeneration [6], [7], [8]. However, to work at the interface between the artificial and the complex and temporal nature of natural systems presents a challenge, i.a. in respect to the B-rep modelling legacy of computational modelling.
The presented doctorate investigates strategies on how to apply digital practice to realise what is an essential bulwark to retain reefs in impossibly challenging times. Beyond the main question of integrating computational modelling and high precision monitoring strategies in artificial coral reef design, this doctorate explores techniques, methods, and linking frameworks to support future research and practice in ecology led design contexts.
Considering the many existing approaches for artificial coral reefs design, one finds they often fall short in precisely understanding the relationships between architectural and ecological aspects (e.g. how a surface design and material composition can foster coral larvae settlement, or structural three-dimensionality enhance biodiversity) and lack an integrated underwater (UW) monitoring process. Such a process is necessary in order to gather knowledge about the ecosystem and make it available for design, and to learn whether artificial structures contribute to reef regeneration or rather harm the coral reef ecosystem.
For the research, empirical experimental methods were applied: Algorithmic coral reef design, high precision UW monitoring, computational modelling and simulation, and validated through parallel real-world physical experimentation – two Artificial Reef Prototypes (ARPs) in Gili Trawangan, Indonesia (2012–today). Multiple discrete methods and sub techniques were developed in seventeen computational experiments and applied in a way in which many are cross valid and integrated in an overall framework that is offered as a significant contribution to the field. Other main contributions include the Ecosystem-aware design approach, Key Performance Indicators (KPIs) for coral reef design, algorithmic design and fabrication of Biorock cathodes, new high precision UW monitoring strategies, long-term real-world constructed experiments, new digital analysis methods and two new front-end web-based tools for reef design and monitoring reefs. The methodological framework is a finding of the research that has many technical components that were tested and combined in this way for the very first time.
In summary, the thesis responds to the urgency and relevance in preserving marine species in tropical reefs during this massive extinction period by offering a differentiated approach towards artificial coral reefs – demonstrating the feasibility of digitally designing such ‘living architecture’ according to multiple context and performance parameters. It also provides an in-depth critical discussion of computational design and architecture in the context of ecosystem regeneration and Planetary Thinking. In that respect, the thesis functions as both theoretical and practical background for computational design, ecology and marine conservation – not only to foster the design of artificial coral reefs technically but also to provide essential criteria and techniques for conceiving them.
Keywords: Artificial coral reefs, computational modelling, high precision underwater monitoring, ecology in design.
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.
The reduction of the cement clinker content is an important prerequisite for the improvement of the CO2-footprint of concrete. Nevertheless, the durability of such concretes must be sufficient to guarantee a satisfactory service life of structures. Salt frost scaling resistance is a critical factor in this regard, as it is often diminished at increased clinker substitution rates. Furthermore, only insufficient long-term experience for such concretes exists. A high salt frost scaling resistance thus cannot be achieved by applying only descriptive criteria, such as the concrete composition. It is therefore to be expected, that in the long term a performance based service life prediction will replace the descriptive concept.
To achieve the important goal of clinker reduction for concretes also in cold and temperate climates it is important to understand the underlying mechanisms for salt frost scaling. However, conflicting damage theories dominate the current State of the Art. It was consequently derived as the goal of this thesis to evaluate existing damage theories and to examine them experimentally. It was found that only two theories have the potential to describe the salt frost attack satisfactorily – the glue spall theory and the cryogenic suction theory.
The glue spall theory attributes the surface scaling to the interaction of an external ice layer with the concrete surface. Only when moderate amounts of deicing salt are present in the test solution the resulting mechanical properties of the ice can cause scaling. However, the results in this thesis indicate that severe scaling also occurs at deicing salt levels, at which the ice is much too soft to damage concrete. Thus, the inability of the glue spall theory to account for all aspects of salt frost scaling was shown.
The cryogenic suction theory is based on the eutectic behavior of salt solutions, which consist of two phases – water ice and liquid brine – between the freezing point and the eutectic temperature. The liquid brine acts as an additional moisture reservoir, which facilitates the growth of ice lenses in the surface layer of the concrete. The experiments in this thesis confirmed, that the ice formation in hardened cement paste increases due to the suction of brine at sub-zero temperatures. The extent of additional ice formation was influenced mainly by the porosity and by the chloride binding capacity of the hardened cement paste.
Consequently, the cryogenic suction theory plausibly describes the actual generation of scaling, but it has to be expanded by some crucial aspects to represent the salt frost scaling attack completely. The most important aspect is the intensive saturation process, which is ascribed to the so-called micro ice lens pump. Therefore a combined damage theory was proposed, which considers multiple saturation processes. Important aspects of this combined theory were confirmed experimentally.
As a result, the combined damage theory constitutes a good basis to understand the salt frost scaling attack on concrete on a fundamental level. Furthermore, a new approach was identified, to account for the reduced salt frost scaling resistance of concretes with reduced clinker content.
In recent years, lightweight materials, such as polymer composite materials (PNCs) have been studied and developed due to their excellent physical and chemical properties. Structures composed of these composite materials are widely used in aerospace engineering structures, automotive components, and electrical devices. The excellent and outstanding mechanical, thermal, and electrical properties of Carbon nanotube (CNT) make it an ideal filler to strengthen polymer materials’ comparable properties. The heat transfer of composite materials has very promising engineering applications in many fields, especially in electronic devices and energy storage equipment. It is essential in high-energy density systems since electronic components need heat dissipation functionality. Or in other words, in electronic devices the generated heat should ideally be dissipated by light and small heat sinks.
Polymeric composites consist of fillers embedded in a polymer matrix, the first ones will significantly affect the overall (macroscopic) performance of the material. There are many common carbon-based fillers such as single-walled carbon nanotubes (SWCNT), multi-walled carbon nanotubes (MWCNT), carbon nanobuds (CNB), fullerene, and graphene. Additives inside the matrix have become a popular subject for researchers. Some extraordinary characters, such as high-performance load, lightweight design, excellent chemical resistance, easy processing, and heat transfer, make the design of polymeric nanotube composites (PNCs) flexible. Due to the reinforcing effects with different fillers on composite materials, it has a higher degree of freedom and can be designed for the structure according to specific applications’ needs. As already stated, our research focus will be on SWCNT enhanced PNCs. Since experiments are timeconsuming, sometimes expensive and cannot shed light into phenomena taking place for instance at the interfaces/interphases of composites, they are often complemented through theoretical and computational analysis.
While most studies are based on deterministic approaches, there is a comparatively lower number of stochastic methods accounting for uncertainties in the input parameters. In deterministic models, the output of the model is fully determined by the parameter values and the initial conditions. However, uncertainties in the input parameters such as aspect ratio, volume fraction, thermal properties of fiber and matrix need to be taken into account for reliable predictions. In this research, a stochastic multiscale method is provided to study the influence of numerous uncertain input parameters on the thermal conductivity of the composite. Therefore, a hierarchical multi-scale method based on computational homogenization is presented in to predict the macroscopic thermal conductivity based on the fine-scale structure. In order to study the inner mechanism, we use the finite element method and employ surrogate models to conduct a Global Sensitivity Analysis (GSA). The SA is performed in order to quantify the influence of the conductivity of the fiber, matrix, Kapitza resistance, volume fraction and aspect ratio on the macroscopic conductivity. Therefore, we compute first-order and total-effect sensitivity indices with different surrogate models.
As stochastic multiscale models are computational expensive, surrogate approaches are commonly exploited. With the emergence of high performance computing and artificial intelligence, machine learning has become a popular modeling tool for numerous applications. Machine learning (ML) is commonly used in regression and maps data through specific rules with algorithms to build input and output models. They are particularly useful for nonlinear input-output relationships when sufficient data is available. ML has also been used in the design of new materials and multiscale analysis. For instance, Artificial neural networks and integrated learning seem to be ideally for such a task. They can theoretically simulate any non-linear relationship through the connection of neurons. Mapping relationships are employed to carry out data-driven simulations of inputs and outputs in stochastic modeling.
This research aims to develop a stochastic multi-scale computational models of PNCs in heat transfer. Multi-scale stochastic modeling with uncertainty analysis and machine learning methods consist of the following components:
-Uncertainty Analysis. A surrogate based global sensitivity analysis is coupled with a hierarchical multi-scale method employing computational homogenization. The effect of the conductivity of the fibers and the matrix, the Kapitza resistance, volume fraction and aspect ratio on the ’macroscopic’ conductivity of the composite is systematically studied. All selected surrogate models yield consistently the conclusions that the most influential input parameters are the aspect ratio followed by the volume fraction. The Kapitza Resistance has no significant effect on the thermal conductivity of the PNCs. The most accurate surrogate model in terms of the R2 value is the moving least square (MLS).
-Hybrid Machine Learning Algorithms. A combination of artificial neural network (ANN) and particle swarm optimization (PSO) is applied to estimate the relationship between variable input and output parameters. The ANN is used for modeling the composite while PSO improves the prediction performance through an optimized global minimum search. The thermal conductivity of the fibers and the matrix, the kapitza resistance, volume fraction and aspect ratio are selected as input parameters. The output is the macroscopic (homogenized) thermal conductivity of the composite. The results show that the PSO significantly improves the predictive ability of this hybrid intelligent algorithm, which outperforms traditional neural networks.
-Stochastic Integrated Machine Learning. A stochastic integrated machine learning based multiscale approach for the prediction of the macroscopic thermal conductivity in PNCs is developed. Seven types of machine learning models are exploited in this research, namely Multivariate Adaptive Regression Splines (MARS), Support Vector Machine (SVM), Regression Tree (RT), Bagging Tree (Bag), Random Forest (RF), Gradient Boosting Machine (GBM) and Cubist. They are used as components of stochastic modeling to construct the relationship between the variable of the inputs’ uncertainty and the macroscopic thermal conductivity of PNCs. Particle Swarm Optimization (PSO) is used for hyper-parameter tuning to find the global optimal values leading to a significant reduction in the computational cost. The advantages and disadvantages of various methods are also analyzed in terms of computing time and model complexity to finally give a recommendation for the applicability of different models.
The detailed structural analysis of thin-walled circular pipe members often requires the use of a shell or solid-based finite element method. Although these methods provide a very good approximation of the deformations, they require a higher degree of discretization which causes high computational costs. On the other hand, the analysis of thin-walled circular pipe members based on classical beam theories is easy to implement and needs much less computation time, however, they are limited in their ability to approximate the deformations as they cannot consider the deformation of the cross-section.
This dissertation focuses on the study of the Generalized Beam Theory (GBT) which is both accurate and efficient in analyzing thin-walled members. This theory is based on the separation of variables in which the displacement field is expressed as a combination of predetermined deformation modes related to the cross-section, and unknown amplitude functions defined on the beam's longitudinal axis. Although the GBT was initially developed for long straight members, through the consideration of complementary deformation modes, which amend the null transverse and shear membrane strain assumptions of the classical GBT, problems involving short members, pipe bends, and geometrical nonlinearity can also be analyzed using GBT. In this dissertation, the GBT formulation for the analysis of these problems is developed and the application and capabilities of the method are illustrated using several numerical examples. Furthermore, the displacement and stress field results of these examples are verified using an equivalent refined shell-based finite element model.
The developed static and dynamic GBT formulations for curved thin-walled circular pipes are based on the linear kinematic description of the curved shell theory. In these formulations, the complex problem in pipe bends due to the strong coupling effect of the longitudinal bending, warping and the cross-sectional ovalization is handled precisely through the derivation of the coupling tensors between the considered GBT deformation modes. Similarly, the geometrically nonlinear GBT analysis is formulated for thin-walled circular pipes based on the nonlinear membrane kinematic equations. Here, the initial linear and quadratic stress and displacement tangent stiffness matrices are built using the third and fourth-order GBT deformation mode coupling tensors.
Longitudinally, the formulation of the coupled GBT element stiffness and mass matrices are presented using a beam-based finite element formulation. Furthermore, the formulated GBT elements are tested for shear and membrane locking problems and the limitations of the formulations regarding the membrane locking problem are discussed.