Este trabalho apresenta os resultados obtidos para carga crítica de flambagem de placas de material composto laminado com duas bordas engastadas e duas bordas simplesmente apoiadas sujeitas a diversas combinações de carregamento. A carga crítica ée obtida através do método de Rayleigh-Ritz, utilizando uma série polinomial e uma senoidal. Foi desenvolvido um programa computacional para que os resultados numéricos fossem obtidos. Os resultados numéricos são analisados e discutidos, e são comparados com resultados apresentados pelo ESDU (Engineering Science Data Unit) e com aproximações obtidas por elementos finitos.

A instabilidade de placas laminadas se tornou um problema de grande importância no projeto de estruturas leves e seguras. Desta forma, o estudo de métodos e o desenvolvimento de programas computacionais para a obtenção de cargas críticas de flambagem neste tipo de estrutura são muito úteis. Este trabalho apresenta o estudo da aplicação da série trigonométrica de duplo seno na obtenção de cargas críticas de laminados retangulares simétricos, com os bordos simplesmente apoiados, baseado nas hipóteses de Kirchhoff, utilizando o método de Rayleigh-Ritz. Um programa computacional no Matlab 6.5 foi desenvolvido e validado por comparação com resultados encontrados na literatura. Em seguida, comparou-se os resultados do programa com os resultados obtidos pelo método dos elementos finitos utilizando-se o Nastran. Verificou-se que a série de duplo seno não ée adequada para a solução de laminados altamente anisotrópicos.

Tese de Mestrado Integrado. Engenharia Mecânica. Faculdade de Engenharia. Universidade do Porto. 2013

No description available.

La thèse a pour objectif le développement de méthodes performantes pour la résolution de problèmes non linéaires ne mécanique des solides. Il est coutume d'utiliser une méthode de type Newton qui conduit à la résolution d'une séquence de systèmes linéaires. De plus, la prise en compte des relations linéaires imposées à l'aide de multiplicateurs de Lagrange confère aux matrices une structure de point-selle. Dans un cadre plus général, nous proposons, étudions et illustrons deux classes d'enrichissement de préconditionneurs (limited memory preconditioners) pour la résolution de séquences de systèmes linéaires par une méthode de Krylov. La première est un extension au cas symétrique indéfini d'une méthode existante, développée initialement dans le cadre symétrique défini positif. La seconde est plus générale dans le sens où elle s'applique aus systèmes non symétriques. Ces deux familles peuvent être interprétées comme des variantes par blocs de formules de mise à jour utilisées dans différentes méthodes d'optimisation. Ces techniques ont été développées dans le logiciel de mécanique des solides Code_Aster (dans un environnement parallèle distribué via la bibliothèque PETSc) et sont illustrées sur plusieurs études industrielles. Les gains obtenus en terme de coût de calcul sont significatifs (jusqu'à 50%), pour un surcoût mémoire négligeable. / The thesis aims at developing efficient numerical methods to solve nonlinear problems arising un solid mechanics. In this field, Newton methods are currently used, requiring the solution of a sequence of linear systems. Furthermore, the imposed linear relations are dualized with the Lagrange multipliers, leading to matrices with a saddle point structure. In a more general framework, we propose two classes of preconditioners (named limited memory preconditioners) to solve sequences of linear systems with a Krylov subspace method. The first class is based on an extension of a method initially developed for symmetric positive definite matrices to the symmetric indefinite case. Both families can be interpreted as block variants of updating formulas used in numerical optimization. They have been implemented into the Code_Aster solid mechanics software (in a parallel distributed environement using the PETSc library). These new preconditioning strategies are illustrated on several industrial applications. We obtain significant gains in computational cost (up to 50%) at a marginal overcost in memory.

The use of energy methods and variational principles is widespread in many fields of engineering of which structural mechanics and curve and surface design are two prominent examples. In principle many different types of function can be used as possible trial solutions to a given variational problem but where piecewise polynomial behaviour and user controlled cross segment continuity is either required or desirable, B-splines serve as a natural choice. Although there are many examples of the use of B-splines in such situations there is no common thread running through existing formulations that generalises from the one dimensional case through to two and three dimensions. We develop a unified approach to the representation of the minimisation equations for B-spline based functionals in tensor product form and apply these results to solving specific problems in geometric smoothing and finite element analysis using the Rayleigh-Ritz method. We focus on the development of algorithms for the exact computation of the minimisation matrices generated by finding stationary values of functionals involving integrals of squares and products of derivatives, and then use these to seek new variational based solutions to problems in the above fields. By using tensor notation we are able to generalise the methods and the algorithms from curves through to surfaces and volumes. The algorithms developed can be applied to other fields where a variational form of the problem exists and where such tensor product B-spline functions can be specified as potential solutions.

A transmission line format is presented which takes the form of a Multilayer Ridged Substrate Integrated Waveguide, for which signal energy is transmitted within standard PCB substrates, within a wave-guiding structure formed from conducting tracks in the horizontal plane and arrays of through-plated vias in the vertical plane. The Substrate Integrated Waveguide (SIW) is a recent development into which research is so far concentrated on single-layer rectangular variants which, like traditional rectangular waveguide, are amenable to analytic computation of the cutoff eigenvalues. Recent publications have offered empirically-derived relationships with which a Substrate Integrated Waveguide can be analysed by equivalence of the horizontal dimensions with a conventional waveguide, allowing such structures to be designed with minimal effort. We propose a ridged form of this structure, in which multiple PCB layers are stacked to obtain the desired height and the published equivalent width is used to obtain the horizontal dimensions. The proposed structure combines the increased bandwidth of ridged waveguide with SIW’s greatly reduced cost of manufacture and integration, relative to conventional waveguide, and improved power handling capacity and loss susceptibility relative to microstrip. Ridged variants have not yet been studied in the literature, however, in part because the eigenspectrum can not be obtained analytically. We thus present a semi-analytical software model with which to synthesise and analyse the cutoff spectrum in ridged Substrate Integrated Waveguide, verified by comparison with analytical solutions, where they exist, simulation in finite-element software and a physical prototype. Agreement with simulated and measured results is within 1 % in certain subsets of the parameter space and 11 % generally, and individual results are returned in times of the order of seconds. We use the model to analyse the relationship between geometry and frequency response, constructing an approximating function for the early modes which is significantly faster, such that think it can be used for first-pass optimisation. A range of optimal parameters are presented which maximise bandwidth within anticipated planar geometric constraints, and typical design scenarios are explored.

Please read the abstract in the section 00front of this document / Dissertation (MEng)--University of Pretoria, 2010. / Mechanical and Aeronautical Engineering / unrestricted

In this study, Ritz method has been employed to analyze the following problems: free vibrations of plates with curvilinear stiffeners, the lowest 100 frequencies of thick isotropic plates, free vibrations of thick quadrilateral laminates and free vibrations and static deformations of rectangular laminates, and sandwich structures. Admissible functions in the Ritz method are chosen as a product of the classical Jacobi orthogonal polynomials and weight functions that exactly satisfy the prescribed essential boundary conditions while maintaining orthogonality of the admissible functions. For free vibrations of plates with curvilinear stiffeners, made possible by additive manufacturing, both plate and stiffeners are modeled using a first-order shear deformation theory. For the thick isotropic plates and laminates, a third-order shear and normal deformation theory is used. The accuracy and computational efficiency of formulations are shown through a range of numerical examples involving different boundary conditions and plate thicknesses. The above formulations assume the whole plate as an equivalent single layer. When the material properties of individual layers are close to each other or thickness of the plate is small compared to other dimensions, the equivalent single layer plate (ESL) theories provide accurate solutions for vibrations and static deformations of multilayered structures. If, however, sufficiently large differences in material properties of individual layers such as those in sandwich structure that consists of stiff outer face sheets (e.g., carbon fiber-reinforced epoxy composite) and soft core (e.g., foam) exist, multilayered structures may exhibit complex kinematic behaviors. Hence, in such case, C_z⁰ conditions, namely, piecewise continuity of displacements and the interlaminar continuity of transverse stresses must be taken into account. Here, Ritz formulations are extended for ESL and layerwise (LW) Nth-order shear and normal deformation theories to model sandwich structures with various face-to-core stiffness ratios. In the LW theory, the C⁰ continuity of displacements is satisfied. However, the continuity of transverse stresses is not satisfied in both ESL and LW theories leading to inaccurate transverse stresses. This shortcoming is remedied by using a one-step well-known stress recovery scheme (SRS). Furthermore, analytical solutions of three-dimensional linear elasticity theory for vibrations and static deformations of simply supported sandwich plates are developed and used to investigate the limitations and applicability of ESL and LW plate theories for various face-to-core stiffness ratios. In addition to natural frequency results obtained from ESL and LW theories, the solutions of the corresponding 3-dimensional linearly elastic problems obtained with the commercial finite element method (FEM) software, ABAQUS, are provided. It is found that LW and ESL (even though its higher-order) theories can produce accurate natural frequency results compared to FEM with a considerably lesser number of degrees of freedom. / Doctor of Philosophy / In everyday life, plate-like structures find applications such as boards displaying advertisements, signs on shops and panels on automobiles. These structures are typically nailed, welded, or glued to supports at one or more edges. When subjected to disturbances such as wind gusts, plate-like structures vibrate. The frequency (number of cycles per second) of a structure in the absence of an applied external load is called its natural frequency that depends upon plate's geometric dimensions, its material and how it is supported at the edges. If the frequency of an applied disturbance matches one of the natural frequencies of the plate, then it will vibrate violently. To avoid such situations in structural designs, it is important to know the natural frequencies of a plate under different support conditions. One would also expect the plate to be able to support the designed structural load without breaking; hence knowledge of plate's deformations and stresses developed in it is equally important. These require mathematical models that adequately characterize their static and dynamic behavior. Most mathematical models are based on plate theories. Although plates are three-dimensional (3D) objects, their thickness is small as compared to the in-plane dimensions. Thus, they are analyzed as 2D objects using assumptions on the displacement fields and using quantities averaged over the plate thickness. These provide many plate theories, each with its own computational efficiency and fidelity (the degree to which it reproduces behavior of the 3-D object). Hence, a plate theory can be developed to provide accurately a quantity of interest. Some issues are more challenging for low-fidelity plate theories than others. For example, the greater the plate thickness, the higher the fidelity of plate theories required for obtaining accurate natural frequencies and deformations. Another challenging issue arises when a sandwich structure consists of strong face-sheets (e.g., made of carbon fiber-reinforced epoxy composite) and a soft core (e.g., made of foam) embedded between them. Sandwich structures exhibit more complex behavior than monolithic plates. Thus, many widely used plate theories may not provide accurate results for them. Here, we have used different plate theories to solve problems including those for sandwich structures. The governing equations of the plate theories are solved numerically (i.e., they are approximately satisfied) using the Ritz method named after Walter Ritz and weighted Jacobi polynomials. It is shown that these provide accurate solutions and the corresponding numerical algorithms are computationally more economical than the commonly used finite element method. To evaluate the accuracy of a plate theory, we have analytically solved (i.e., the governing equations are satisfied at every point in the problem domain) equations of the 3D theory of linear elasticity. The results presented in this research should help structural designers.

This thesis is about the development of a new numerical method for the analysis of composite shells. The present work is based on Reissner Mixed Variational Theorem (RMVT), the Sublaminate Generalized Unified Formulation (S-GUF), and the Ritz approximation. The present work investigates a more efficient way to compute transverse stresses (sigma_xz, sigma_yz, sigma_zz) based upon RMVT, allowing assigning their order of continuity a priori. This is a great advantage compared to a conventional displacement-based approach. In order to enable computing of both global and local responses (depending on the user’sneeds) the S-GUF framework was adopted. The Generalized Unified Formulation (GUF) enables the implementation of approximations with virtually unlimited algebraic order within a single code, and the order could also vary for different variables. In addition to the GUF, the concept of Sublaminate was utilized, allowing for sub-sectioning of the domain in the thickness direction into sublaminates, and it is then possible to apply different formulations in each of these sub-domains. The curvature of the shells is strictly defined by their radius-to- thickness ratio. The flexibility of S-GUF is helpful in the sense that curvature is only introduced and treated if needed by the particular case at hand. The governing equations obtained applying S-GUF to RMVT were solved in a weak formulation using the Ritz approximation. This choice was made to save computational time, which is one of the main benefits of the presented method. Validation of the code was made by comparing results from the present formulation with solutions available in the literature. Good to excellent agreement was found for several benchmark cases, supporting that the formulation is valid and provides reliable solutions.Finally, numerical and analytical considerations about the developed method were made: its numerical stability, how to tune its parameters, and which models result more correct from an analytical standpoint. / Denna avhandling handlar om utvecklingen av en ny numerisk metod för analys av kompositskal.Det aktuella arbetet är baserat på Reissner Mixed Variational Theorem (RMVT), Sublaminate Generalized Unified Formulation (S-GUF) och Ritz-approximationen.Arbetet går ut på att ta fram ett mer effektivt sätt att beräkna spänningar ut ur planet (sigma_xz, sigma_yz, aigma_zz) och utnyttjar RMVT, vilket möjliggör lokal hantering av kontinuitet i varierande ordning, definierad a priori. Detta innebär en stor fördel jämfört med en konventionell förskjutningsbaserad metod. För att möjliggöra beräkning av både global och lokal lösning, beroende påanvändarens behov, antogs S-GUF-ramverket. Den generaliserade enhetliga formuleringen (GUF) gördet möjligt att inom samma formulering implementera approximationer med i princip obegränsad algebraisk ordning, vilka dåocksåkan skilja mellan olika variabler. Förutom GUF används även konceptet Sublaminate som gör det möjligt att dela upp domänen i tjockleksriktningen i underregioner (sublaminate), och det är dåmöjligt att tillämpa olika formuleringar i var och en av dessa subdomäner. Krökningen hos ett skal definiers strikt av förhållandet mellan dess radie och tjocklek. Flexibiliteten hos S-GUF är fördelaktig dåkrökning endast hanteras för de specifika fall där det förekommer. De ekvationer som erhålls genom att applicera S-GUF på RMVT löses påsvag formmed användning av Ritz approximation. Detta val gjordes för att möjliggöra en snabbare beräkningstid, vilket är en av fördelarna med denna metod. Genom att jämföra de resultat med lösningar tillgängliga i litteraturen var det möjligt att validera resultaten och därmed även själva formuleringen. God till utmärkt överensstämmelse påvisades för ett antal olika standardfall vilket styrker att metoden fungerar och attdess lösningar är pålitliga. Slutligen ritades numeriska och analytiska överväganden om metoden här utvecklad, såsom dess numeriska stabilitet, hur man ställer in dess parametrar och vilka modeller somär mer korrekta ur en analytisk synvinkel.