A continuous adjoint formulation for exterior optimization of Dirichlet data on the boundary for potential flow applications has been developed in this thesis. This has been performed by utilizing boundary integral methods for both the primal problem (Laplace’s equation) and for the corresponding adjoint equation (Poisson’s equation) on the unit disc. A considerable portion is devoted to the estimation of the boundary flux for Poisson’s equation. The boundary flux is the (unique) Riesz-representer that determines the search direction in descent methods where the optimal control problem is to reconstruct the Dirichlet data on the boundary in order to minimize a quadratic functional. / I denna uppsats behandlas adjoint baserad optimering av Dirichlet data på randen för exteriöra problem relaterat till fluiddynamik. Detta har gjorts med randintegral-metoder för det primära problemet (Laplace’s ekvation) samt för den korresponderande adjoint- ekvationen (Poisson’s ekvation). Ett stort fokus är estimering av flödet vid randen för Poisson’s ekvation på enhetsdisken. För det primära Dirichlet problemet så bestämmer detta flöde sökriktningen i gradientbaserade optimeringmetoder där målet är att rekonstruera Dirichlet-datat på randen för att minimera en kvadratisk funktional.

Modeling surface adsorption requires systems of hundreds of atoms. To model such systems at an ab initio level successfully, we need to avoid traditional quantum chemical methods. In the present work we have shown that density functional theory is a powerful modeling tool for large chemical systems especially in combinations with pseudopotentials This is validated by an initial study of ethyl and heptyl xanthates and their sodium/potassium salts. In this study, all electron calculations using both Hartree-Fock and density functional theory methods are compared with experimental infrared results. To do this the influence of basis sets and modeling approaches on the geometrical structure and the vibrational modes are examined. This includes comparing the pseudopotential and full electron potential approaches. Results obtained from pseudopotential methods are in close agreement with both all electron calculations as well as experimental results, here used to study adsorption of heptyl xanthate ZnS surfaces. Vibrational frequencies of the adsorbed species is presented, together with calculations of the tilt angles. The investigation of the tilt angles resulted in 20.3° 20.6° and 25.2° for the 100, 110 and 111 surfaces respectively. Heptyl xanthate forms a bridging confirmation on both the 110 and 111 surfaces and a bidentate confirmation on the 100 surface. Assignments of vibrational modes of ethyl/heptyl xanthate molecule and its corresponding potassium/sodium salts are also reported. / <p>Godkänd; 2005; 20070108 (haneit)</p>

This thesis is focused on the Occupational Pension, an important part of the retiree’s total pension. It is paid by private insurance companies and determined by an annuity divisor. Regression modeling of the annuity divisor is done by using the monthly paid pension as a response and a set of 24 explanatory variables e.g. the expected remaining lifetime and advance interest rate. Two machine learning algorithms, artificial neural networks (ANN) and support vector machines for regression (SVR) are considered in detail. Specifically, different transfer functions for ANN are studied as well as the possibility to improve the SVR model by incorporating a non-linear Gaussian kernel. To compare our result with prior experience of the Swedish Pensions Agency in modeling and predicting the annuity divisor, we also consider the ordinary multiple linear regression (MLR) model. Although ANN, SVR and MLR are of different nature, they demonstrate similar performance accuracy. It turns out that for our data that MLR and SVR with a linear kernel achieve the highest prediction accuracy. When performing feature selection, all methods except SVR with a Gaussian kernel encompass the features corresponding to advance interest rate and expected remaining lifetime, which according to Swedish law {Swedish law: 5 kap. 12 § lagen (1998:674) om inkomstgrundad ålderspension} are main factors that determine the annuity divisor. The results of this study confirm the importance of the two main factors for accurate modeling of the annuity divisor in private insurance. We also conclude that, in addition to the methods used in previous research, methods such as MLR, ANN and SVR may be used to accurately model the annuity divisor. / Denna uppsats fokuserar på tjänstepensionen, en viktig del av en pensionärs totala pension. Den utbetalas av privata försäkringsbolag och beräknas med hjälp av ett så kallat delningstal. Regressionsmodellering av delningstalet görs genom att använda den månatliga utbetalda pensionen som svar och en uppsättning av 24 förklarande variabler såsom förväntad återstående livslängd och förskottsränta. Två maskininlärningsalgoritmer, artificiella neuronnät (ANN) och stödvektormaskiner för regression (SVR) betraktas i detalj. Specifikt så studeras olika överföringsfunktioner för ANN och möjligheten att förbättra SVR modellen genom att införa en ickelinjär Gaussisk kärna. För att jämföra våra resultat med tidigare erfarenhet från Pensionsmyndigheten vid modellering och förutsägande av delningstalet studerar vi även ordinär multipel linjär regression (MLR). Även om ANN, SVR och MLR är av olika natur påvisar dem liknande noggrannhet. Det visar sig för vår data att MLR och SVR med en linjär kärna uppnår den högsta noggrannheten på okänd data. Vid variabel urvalet omfattar samtliga metoder förutom SVR med en Gaussisk kärna variablerna motsvarande förväntad återstående livslängd och förskottsränta som enligt svensk lag är huvudfaktorer vid bestämning av delningstalet. Resultatet av denna studie bekräftar betydelsen av huvudfaktorerna för noggrann modellering av delningstalet inom privat försäkring. Vi drar även slutsatsen att utöver metoderna som använts i tidigare studier kan metoder såsom ANN, SVR och MLR användas med framgång för att noggrant modellera delningstalet.

The present thesis treats the numerical simulation of immiscible incompressible two-phase flows with moving contact lines. The conventional Navier–Stokes equations combined with a no-slip boundary condition leads to a non-integrable stress singularity at the contact line. The singularity in the model can be avoided by allowing the contact line to slip. Implementing slip conditions in an accurate way is not straight-forward and different regularization techniques exist where ad-hoc procedures are common. This thesis presents the first steps in developing the macroscopic part of an accurate multiscale model for a moving contact line problem in two space dimensions. It is assumed that a micro model has been used to determine a relation between the contact angle and the contact line velocity. An intermediate region is introduced where an analytical expression for the velocity field exists, assuming the solid wall is perfectly flat. This expression is used to implement boundary conditions for the moving contact line, at the macroscopic scale, along a fictitious boundary located a small distance away from the physical boundary. Model problems where the shape of the interface is constant throughout the simulation are introduced. For these problems, experiments show that the errors in the resulting contact line velocities converge with the grid size h at a rate of convergence p ≈ 2. Further, an analytical expression for the velocity field in the intermediate region for the case with a curved solid wall is derived. The derivation is based on perturbation analysis. / eSSENCE

In this thesis dual-pair summation-by-parts-operators were optimized to propagate seismic surface waves as accurately as possible. The resulting operators were tested on a problem consisting of pure surface waves, so-called Rayleigh waves, as well as a point-source receiver problem with a vertical point force. An error study as well as a convergence study was carried out, and the results were compared to errors and convergence rates obtained using previously derived summation-by-parts-operators. The results show that there is something to be gained from optimizing for seismic surface waves. For pure Rayleigh waves, we observe smaller errors than with existing summation-by parts operators. This improvement is most apparent for elastic materials with low shear wave speed. The optimization technique used did not ensure that a global minimum of the objective function was reached, which makes exploring other optimization techniques a possible extension of this work. Other examples of future work are to explore other objective functions as well as carry out more experiments on other kinds of waves and topographies.

No description available.

During the last two decades, advances in molecular technology have facilitated the sequencing and analysis of ancient DNA recovered from archaeological finds, contributing to novel insights into human evolutionary history. As more ancient genetic information has become available, the need for specialized methods of analysis has also increased. In this thesis, we investigate statistical and computational models for analysis of genetic data, with a particular focus on the context of ancient DNA. The main focus is on imputation, or the inference of missing genotypes based on observed sequence data. We present results from a systematic evaluation of a common imputation pipeline on empirical ancient samples, and show that imputed data can constitute a realistic option for population-genetic analyses. We also discuss preliminary results from a simulation study comparing two methods of phasing and imputation, which suggest that the parametric Li and Stephens framework may be more robust to extremely low levels of sparsity than the parsimonious Browning and Browning model. An evaluation of methods to handle missing data in the application of PCA for dimensionality reduction of genotype data is also presented. We illustrate that non-overlapping sequence data can lead to artifacts in projected scores, and evaluate different methods for handling unobserved genotypes. In genomics, as in other fields of research, increasing sizes of data sets are placing larger demands on efficient data management and compute infrastructures. The last part of this thesis addresses the use of cloud resources for facilitating such analysis. We present two different cloud-based solutions, and exemplify them on applications from genomics. / eSSENCE

This thesis focuses on developing an appropriate stochastic model for temperature dynamics as a means of pricing weather derivative contracts based on temperature. There are various methods for pricing weather derivatives ranging from simple one like historical burn analysis, which does not involve modeling the underlying weather variable to complex ones that require Monte Carlo simulations to achieve explicit weather derivatives contract prices, particularly the daily average temperature (DAT) dynamics models. Among various DAT models, appropriate regime switching models are considered relative better than single regime models due to its ability to capture most of the temperature dynamics features caused by urbanization, deforestation, clear skies and changes of measurement station. A new proposed model for DAT dynamics, is a two regime switching models with heteroskedastic mean-reverting process in the base regime and Brownian motion with nonzero drift in the shifted regime. Before using the model for pricing temperature derivative contracts, we compare the performance of the model with a benchmark model proposed by Elias et al. (2014), interms of the HDDs, CDDs and CAT indices. Using ve data sets from dierent measurement locations in Sweden, the results shows that, a two regime switching models with heteroskedastic mean-reverting process gives relatively better results than the model given by Elias et al. We develop mathematical expressions for pricing futures and option contracts on HDDs, CDDs and CAT indices. The local volatility nature of the model in the base regime captures very well the dynamics of the underlying process, thus leading to a better pricing processes for temperature derivatives contracts written on various index variables. We use the Monte Carlo simulation method for pricing weather derivatives call option contracts.

Investigating the properties, explaining, and predicting the behaviour of a physical system described by a system (matrix) pencil often require the understanding of how canonical structure information of the system pencil may change, e.g., how eigenvalues coalesce or split apart, due to perturbations in the matrix pencil elements. Often these system pencils have different block-partitioning and / or symmetries. We study changes of the congruence canonical form of a complex skew-symmetric matrix pencil under small perturbations. The problem of computing the congruence canonical form is known to be ill-posed: both the canonical form and the reduction transformation depend discontinuously on the entries of a pencil. Thus it is important to know the canonical forms of all such pencils that are close to the investigated pencil. One way to investigate this problem is to construct the stratification of orbits and bundles of the pencils. To be precise, for any problem dimension we construct the closure hierarchy graph for congruence orbits or bundles. Each node (vertex) of the graph represents an orbit (or a bundle) and each edge represents the cover/closure relation. Such a relation means that there is a path from one node to another node if and only if a skew-symmetric matrix pencil corresponding to the first node can be transformed by an arbitrarily small perturbation to a skew-symmetric matrix pencil corresponding to the second node. From the graph it is straightforward to identify more degenerate and more generic nearby canonical structures. A necessary (but not sufficient) condition for one orbit being in the closure of another is that the first orbit has larger codimension than the second one. Therefore we compute the codimensions of the congruence orbits (or bundles). It is done via the solutions of an associated homogeneous system of matrix equations. The complete stratification is done by proving the relation between equivalence and congruence for the skew-symmetric matrix pencils. This relation allows us to use the known result about the stratifications of general matrix pencils (under strict equivalence) in order to stratify skew-symmetric matrix pencils under congruence. Matlab functions to work with skew-symmetric matrix pencils and a number of other types of symmetries for matrices and matrix pencils are developed and included in the Matrix Canonical Structure (MCS) Toolbox.

The understanding of flow problems, and finding their solution, has been important for most of human history, from the design of aqueducts to boats and airplanes. The use of physical miniature models and wind tunnels were, and still are, useful tools for design, but with the development of computers, an increasingly large part of the design process is assisted by computational fluid dynamics (CFD). Many industrial CFD codes have their origins in the 1980s and 1990s, when the low order finite volume method (FVM) was prevalent. Discontinuous Galerkin methods (DGM) have, since the turn of the century, been seen as the successor of these methods, since it is potentially of arbitrarily high order. In its lowest order form DGM is equivalent to FVM. However, many existing codes are not compatible with standard DGM and would need a complete rewrite to obtain the advantages of the higher order. This thesis shows how to extend existing vertex-centered and edge-based FVM codes to higher order, using a special kind of DGM discretization, which is different from the standard cell-centered type. Two model problems are examined to show the necessary data structures that need to be constructed, the order of accuracy for the method, and the use of an hp-adaptation scheme to resolve a developing shock. Then the method is further developed to solve the steady Euler equations, within the existing industrial Edge code, using acceleration techniques such as local time stepping and multigrid. With the ever increasing need for more efficient and accurate solvers and algorithms in CFD, the modified DGM presented in this thesis could be used to help and accelerate the adoption of high order methods in industry.