The tuning of optimization software is of key interest to researchers solving mixed integer programming (MIP) problems. The efficiency of the optimization software can be greatly impacted by the solver’s parameter settings and the structure of the MIP. A designed experiment approach is used to fit a statistical model that would suggest settings of the parameters that provided the largest reduction in the primal integral metric. Tuning exemplars of six and 59 factors (parameters) of optimization software, experimentation takes place on three classes of MIPs: survivable fixed telecommunication network design, a formulation of the support vector machine with the ramp loss and L1-norm regularization, and node packing for coding theory graphs. This research presents and demonstrates a framework for tuning a portfolio of MIP instances to not only obtain good parameter settings used for future instances of the same class of MIPs, but to also gain insights into which parameters and interactions of parameters are significant for that class of MIPs. The framework is used for benchmarking of solvers with tuned parameters on a portfolio of instances. A group screening method provides a way to reduce the number of factors in a design and reduces the time it takes to perform the tuning process. Portfolio benchmarking provides performance information of optimization solvers on a class with instances of a similar structure.

Ultra-high performance liquid chromatography (UHPLC) provides a considerable increase in throughput compared to HPLC and a reduced solvent consumption. The implementation of UHPLC in pharmaceutical analysis, e.g. quality control, has accelerated in recent years and there is currently a mix of HPLC and UHPLC instrumentation within pharmaceutical companies. There are, however, technical and regulatory challenges converting a HPLC method to UHPLC making it difficult to take full advantage of UHPLC in regulatory-focused applications like quality control in pharmaceutical production. Using chromatographic modelling and fundamental theory, this thesis investigated method conversion between HPLC and UHPLC. It reports on the influence of temperature gradients due to viscous heating, pressure effects and stationary phase properties on the separation performance. It also presents a regulatory concept for less regulatory interaction for minor changes to approved methods to support efficient life cycle management. The higher pressure in UHPLC gave a retention increase of up to 40% as compared to conventional HPLC while viscous heating, instead, reduced retention and the net result was very solute dependent. Selectivity shifts were observed even between solutes with similar structure when switching between HPLC and UHPLC and an experimental method to predict such selectivity shifts was therefore developed. The peak shape was negatively affected by the increase in pressure for some solutes since secondary interactions between the solute and the stationary phase increased with pressure. With the upcoming ICH Q12 guideline, it will be possible for the industry to convert existing methods to UHPLC in a more flexible way using the deeper understanding and the regulatory concept presented here as a case example. / Ultra-high performance liquid chromatography (UHPLC) provides a considerable increase in throughput compared to conventional HPLC and a reduced solvent consumption. The implementation of UHPLC in pharmaceutical analysis has accelerated in recent years and currently both instruments are used. There are, however, technical and regulatory challenges converting a HPLC method to UHPLC making it difficult to take full advantage of UHPLC in regulatory-focused applications like quality control in pharmaceutical production. In UHPLC, the column is packed with smaller particles than in HPLC resulting in higher pressure and viscous heating. Both the higher pressure and the higher temperature may cause changes in retention and selectivity making method conversion unpredictable. Using chromatographic modelling and fundamental theory, this thesis investigates method conversion between HPLC and UHPLC. It reports on the influence of temperature gradients due to viscous heating, pressure effects and stationary phase properties on the separation performance. It also presents a regulatory concept for less regulatory interaction for minor changes to approved quality control methods and how predicable method conversion is achieved by improved understanding.

In composite materials, transverse shear modulus is a critical moduli parameter for designing complex composite structures. For dependable mathematical modeling of mechanical behavior of composite materials, an accurate estimate of the moduli parameters is critically important as opposed to estimates of strength parameters where underestimation may lead to a non-optimal design but still would give one a safe one. Although there are mechanical and empirical models available to find transverse shear modulus, they are based on many assumptions. In this work, the model is based on a three-dimensional elastic finite element analysis with multiple cells. To find the shear modulus, appropriate boundary conditions are applied to a three-dimensional representative volume element (RVE). To improve the accuracy of the model, multiple cells of the RVE are used and the value of the transverse shear modulus is calculated by an extrapolation technique that represents a large number of cells. Comparing the available analytical and empirical models to the finite element model from this work shows that for polymeric matrix composites, the estimate of the transverse shear modulus by Halpin-Tsai model had high credibility for lower fiber volume fractions; the Mori-Tanaka model was most accurate for the mid-range fiber volume fractions; and the Elasticity Approach model was most accurate for high fiber volume fractions. Since real-life composites have voids, this study investigated the effect of void fraction on the transverse shear modulus through design of experiment (DOE) statistical analysis. Fiber volume fraction and fiber-to-matrix Young’s moduli ratio were the other influencing parameters used. The results indicate that the fiber volume fraction is the most dominating of the three variables, making up to 96% contribution to the transverse shear modulus. The void content and fiber-to-matrix Young’s moduli ratio have negligible effects. To find how voids themselves influence the shear modulus, the transverse shear modulus was normalized with the corresponding shear modulus with a perfect composite with no voids. As expected, the void content has the largest contribution to the normalized shear modulus of 80%. The fiber volume fraction contributed 12%, and the fiber-to-matrix Young’s moduli ratio contribution was again low. Based on the results of this work, the influences and sensitivities of void content have helped in the development of accurate models for transverse shear modulus, and let us confidently study the influence of fiber-to-matrix Young’s moduli ratio, fiber volume fraction and void content on its value.

Fiber pushout test models are developed for a fiber-matrix-composite with a nonhomogeneous interphase. Using design of experiments, the effects of geometry, loading and material parameters on critical parameters of the pushout test such as the load-displacement curve and maximum interfacial shear and normal stresses are studied. The sensitivity analysis shows that initial load displacement curve is dependent only on the indenter type and not on parameters such as fiber volume fraction, interphase type, thickness of interphase, and boundary conditions. In contrast, interfacial shear stresses are not sensitive to indenter type, while the interfacial radial stresses are mainly sensitive to fiber volume fraction and the boundary conditions.

abstract: Nonregular designs are a preferable alternative to regular resolution four designs because they avoid confounding two-factor interactions. As a result nonregular designs can estimate and identify a few active two-factor interactions. However, due to the sometimes complex alias structure of nonregular designs, standard screening strategies can fail to identify all active effects. In this research, two-level nonregular screening designs with orthogonal main effects will be discussed. By utilizing knowledge of the alias structure, a design based model selection process for analyzing nonregular designs is proposed. The Aliased Informed Model Selection (AIMS) strategy is a design specific approach that is compared to three generic model selection methods; stepwise regression, least absolute shrinkage and selection operator (LASSO), and the Dantzig selector. The AIMS approach substantially increases the power to detect active main effects and two-factor interactions versus the aforementioned generic methodologies. This research identifies design specific model spaces; sets of models with strong heredity, all estimable, and exhibit no model confounding. These spaces are then used in the AIMS method along with design specific aliasing rules for model selection decisions. Model spaces and alias rules are identified for three designs; 16-run no-confounding 6, 7, and 8-factor designs. The designs are demonstrated with several examples as well as simulations to show the AIMS superiority in model selection. A final piece of the research provides a method for augmenting no-confounding designs based on a model spaces and maximum average D-efficiency. Several augmented designs are provided for different situations. A final simulation with the augmented designs shows strong results for augmenting four additional runs if time and resources permit. / Dissertation/Thesis / Doctoral Dissertation Industrial Engineering 2020

Massachusetts passed legislation in the fall of 2012 to allow the construction of three casinos and a slot parlor in the state. The prevalence of problem gambling in the state and in areas where casinos will be constructed is of particular interest. The goal is to evaluate the change in prevalence after construction of the casinos, using a multi-mode address based sample survey. The objective of this thesis is to evaluate and describe ways of using statistical inference to estimates prevalence rates in finite populations. Four methods were considered in an attempt to evaluate the prevalence of problem gambling in the context of the gambling study. These methods were evaluated unconditionally and conditionally, controlling for gender, using mean square error (MSE) as a measure of accuracy. The simple mean, the post-stratified mean, the best linear unbiased predictor (BLUP), and the empirical best linear unbiased predictor (EBLUP) were considered in three examples. Conditional analyses of a population with N=1,000 and a crude problem gambling rate of 1.5, samples of n=200 led to the simple mean and the post-stratified mean to perform better in certain situations, as measured by their low MSE values. When there are less females than expected in a sample, the post-stratified mean produces a lower mean MSE over the 10,000 simulations. When there are more females than expected in a sample, the simple mean produces a lower mean MSE over the 10,000 simulations. Conditional analysis provided more appropriate results than unconditional analysis.

As the energy efficiency of dishwashers continuously improves, differentmanufacturers need to find new innovative solutions to further improve thedishwashers. To do so one can divide the areas of improvement into 4 parameters,water temperature, washing time, chemical treatment and mechanical processing.In this thesis the focus lies on the mechanical processing and the washing time. The conventional spray arm of today's dishwashers sets its rotational velocitythrough the water pressure. To step away from this ASKO Appliances AB hasdeveloped a prototype with a motorized lower spray arm. The purpose of this studyis therefore to investigate how the motorized spray arm performs in terms of energyusage, wash results and sound. This when altering the factors water pressure,rotational velocity and washing time. To get reliable results with the least amount of tests, the methodology of Design ofExperiments was used through the software MODDE 7. A 3-level full factorialdesign was chosen. The three factors were varied in three levels -1, 0, 1 and fiveextra centerpoints were added resulting in a test schedule of 32 tests. To measurethe wash result a soiling method and an evaluation method was developed. Theresults of the executed test schedule was analysed through MODDEs in-builtregression model Multiple Linear Regression. The analysed test results showed that the wash time was the dominant factor forboth energy usage and wash result and the second most dominant factor was waterpressure. The rotational velocity had little to no effect on the energy usage and asmall effect on the wash result at a low water pressure but increasing effect for ahigher water pressure. The measured sound values were constant over time andincreased with both increased water pressure and rotational velocity. The optimalrunning conditions regarding wash result was investigated using MODDE and thentested. The optimal rotational velocity was the setting 1, the highest level, for allthree time settings. The optimal water pressure was in the mid to high region, andincreased with increasing wash time. The results from the three optimum tests for the prototype were then compared totests with a conventional dishwasher with a pressure driven spray arm. Theconventional dishwasher was runned with its normal running conditions for thethree different time settings. The prototype achieved a similar wash result as theconventional machine for the time setting -1 but slightly better for the time setting0 and 1. The energy usage for the prototype was a doubling of the conventionalmachines energy usage. The sound measurements were excluded due to muffelingdifferences between the machines

In the presented Master’s thesis “Hardness differences compensation of components during profile rolling” are the effects of hardness oscillation of finished components considered. The effects are through the Design of Experiments method at various spline profiles and process options tested. The evaluation in Minitab software showed, that the reduction of roll diameter oscillation is through the process regulation based on rolling force difference possible.

Cette thèse traite de l'approximation de fonctions coûteuses à évaluer et présentant un comportement hétérogène selon les régions du domaine d'entrées. Dans beaucoup d’applications industrielles comme celles issues de la sûreté nucléaire et menées à l’IRSN, les codes de calcul de simulation physique ont des coûts numériques importants et leurs réponses sont alors approchées par un modèle. Les modèles par processus gaussien (PG) sont répandus pour construire des plans d'expériences au budget d'évaluations limité. Alors que les méthodes de planifications standard explorent le domaine en partant souvent d'une hypothèse de stationnarité et en utilisant des critères d'échantillonnages basés sur la variance, nous abordons sous deux aspects le problème d'adapter la planification d'expériences par PG aux fonctions à variations hétérogènes : d'une part sur la modélisation, en proposant une nouvelle famille de covariances (WaMI-GP) qui généralise deux noyaux existants (noyaux à indice multiple et noyaux à déformation tensorielle), et d'autre part sur la planification, en définissant et calculant de nouveaux critères d'échantillonnage à partir des dérivées du PG et dédiés à l'exploration de régions à grandes variations. Une analyse théorique de la nouvelle famille de PG ainsi que des expériences numériques montrent à la fois une certaine souplesse de modélisation et un nombre modéré de paramètres de modèle à estimer. Par ailleurs, un travail sur le couplage entre ondelettes et PG a conduit au développement d'un modèle non stationnaire, appelé ici wav-GP, qui utilise l'échelle locale pour approcher la dérivées de la déformation de manière itérative et non paramétrique. Les applications de wav-GP sur deux cas d'étude mécaniques fournissent des résultats prometteurs pour la prédiction de fonctions ayant un comportement hétérogène. D'autres part, nous formulons et calculons de nouveaux critères de variance qui reposent sur le champs de la norme du gradient du PG, et cela quelques soient les propriétés de (non) stationnarité du modèle. Les critères et les modèles sont comparés avec des méthodes de références sur des cas d'étude en ingénierie. Il en résulte que certains des critères par gradient proposés sont plus efficaces que les critères de variance plus conventionnels, mais qu'il est encore mieux de combiner le modèle WaMI-GP avec ces critères de variance. Ce modèle est souvent le plus compétitif dans des situations où les évaluations sont séquentielles et à nombre réduit. D'autres contributions s'inscrivant dans le thème de l'optimisation globale concernent en particulier le critère d'échantillonnage d'amélioration espérée (connu en anglais comme le critère d'expected improvement) et sa version multipoint pour l'évaluation en parallèle par paquet. Des formules analytiques et des approximations rapides sont établies pour une version généralisée du critère et pour son gradient. / This work deals with approximating expensive-to-evaluatefunctions exhibiting heterogeneous sensitivity to input perturbationsdepending on regions of the input space. Motivated by real test caseswith high computational costs coming mainly from IRSN nuclear safetystudies, we resort to surrogate models of the numerical simulatorsusing Gaussian processes (GP). GP models are popular for sequentialevaluation strategies in design of experiments under limited evaluationbudget. While it is common to make stationarity assumptions for theprocesses and use sampling criteria based on its variance forexploration, we tackle the problem of accommodating the GP-based designto the heterogeneous behaviour of the function from two angles: firstvia a novel class of covariances (WaMI-GP) that simultaneouslygeneralises existing kernels of Multiple Index and of tensorised warpedGP and second, by introducing derivative-based sampling criteriadedicated to the exploration of high variation regions. The novel GPclass is investigated both through mathematical analysis and numericalexperiments, and it is shown that it allows encoding muchexpressiveness while remaining with a moderate number of parameters tobe inferred. Moreover, exploiting methodological links between waveletsanalysis and non-stationary GP modelling, we propose a new non-stationary GP (Wav-GP) with non-parametric warping. The key point is aniterated estimation of the so-called local scale that approximates thederivative of the warping. Wav-GP is applied to two mechanical casestudies highlighting promising prediction performance. Independently ofnon-stationarity assumptions, we conduct derivations for new variance-based criteria relying on the norm of the GP gradient field. Criteriaand models are compared with state-of-the-art methods on engineeringtest cases. It is found on these applications that some of the proposedgradient-based criteria outperform usual variance-based criteria in thecase of a stationary GP model, but that it is even better to usevariance-based criteria with WaMI-GP, which dominates mostly for smalldesigns and in sequential set up. Other contributions in samplingcriteria address the problem of global optimisation, focusing on theexpected improvement criterion and its multipoint version for parallelbatch evaluations. Closed form formulas and fast approximations areestablished for a generalised version of the criterion and its gradient. Numerical experiments illustrate that the proposed approachesenable substantial computational savings.

This thesis has been conducted at Cargotec Sweden AB as a case study on the loader body of the HIAB XS 144 crane. The loader body is the innermost part in the cranes arm-system and its fatigue life is critical to the operational life of the whole crane. Welding is the main joining process in Cargotec’s cranes and are often a limiting factor when it comes to fatigue life. The weld joining the column to the loader body is carrying the whole crane moment. Previous testing has shown that this weld often limits the fatigue life of the loader body, it has thus been evaluated. Weld fatigue life is affected by a large amount of parameters. To pinpoint the parameters mainly affecting the weld fatigue life and to understand their influence, calculations have been organized using factorial design. The evaluation has been carried out using 3D finite element calculations utilizing sub-modelling to calculate local stresses in the weld notches. Different parameters have been evaluated based on their influence on the local notch stresses. To estimate stresses from the evaluated parameters, regression equations have been fitted. The effective notch method has been used to estimate weld fatigue life. The evaluation has shown that a butt-weld design with root-support, only being welded from the outside of the loader body, as used on some other crane models, could not provide a robust design for the XS 144 crane. The evaluation could also point out several critical parameters that need to be considered when using such design. Apart from the local weld geometry, plate thickness, plate angle, material offset and thickness in the casted column were mainly affecting the weld notch stresses.