Split Concrete Model (SCM) is a unified approach towards modeling shear behavior in concrete. SCM is essentially a rational model which is evaluated and modified using a large experimental database.The shear strength of the concrete beam is modeled as the sum of the contribution of concrete, transverse reinforcement, longitudinal reinforcement and bond between concrete and longitudinal reinforcement. Concrete does not contribute to the shear strength after the formation of the crack. In SCM, this is shown to be accurately modeled by only considering the second branch of the critical crack while computing the contribution of concrete towards shear strength of the beam. Formation of the second branch of the critical crack and immediate subsequent failure of the beam has been compared to the split-cylinder test, which forms the conceptual basis of SCM.SCM computes the concrete contribution using the split tensile strength and the area under compression of the concrete beam. For cases where a split-cylinder test is not performed, a mathematical model is proposed to compute the split tensile strength using the compressive strength of concrete available from experimental results. This model is proposed using advanced statistical methods, including weighted residuals and Box-Cox transformation and is validated using various statistical procedures. The transverse reinforcement contributes to the shear strength of the concrete beam only after the formation of the crack. In SCM, this is shown to be accurately modeled by only considering the first branch of the critical crack while computing the contribution of the transverse reinforcement towards shear strength of the beam, instead of the conventional approach of considering the entire length of the crack. The contribution of the longitudinal steel and bond between concrete and longitudinal steel and concrete is accurately modeled unlike the conventional approaches which do not consider this contribution.Evaluation using the database shows that SCM can predict accurate results for all ranges of strength, depth, reinforcement ratio, and shear span to depth ratio of the beam. This shows that all the influencing parameters for concrete shear strength have been correctly modeled in SCM. SCM gives more accurate results as compared to current codified approaches as verified with design examples. Finally, specific recommendations have been made indicating how the shear design requirements in the current ACI code can be modified.

Rate-dependent effects introduced during rapid and/or dynamic events have typically been oversimplified to compensate for deficiencies in present analyses. As load test results are generally considered as the basis of performance from which foundations can be designed, it is imperative that the analyzed load test data be as accurate as possible. In an attempt to progress the state of load test data regression, this dissertation addresses two common assumptions made during the regression process: (1) the statnamic damping coefficient is constant throughout the entire load test and (2) the concrete stress-strain relationship is linear-elastic. Also presented is a case study where the inherent features of a rapid load test proved useful in identifying the occurrence and proximity of a structural failure within a drilled shaft.

Sociologie des techniques / Du point de vue d’une sociologie de la connaissance, la recherche contemporaine en intelligence artificielle tire son originalité de l’intérêt porté à la modélisation de la connaissance ordinaire, ce savoir engagé dans la vie de tous les jours (ex. aller l’épicerie, prendre son petit déjeuner). Les entreprises Facebook ou Google, le fameux GAFAM ainsi que les services de musique et de vidéos en ligne capturent et modélisent les faits et gestes du quidam afin de remettre les résultats de leurs calculs en jeu dans cette même vie courante. La recherche en IA porte également sur la mise en forme de savoirs savants et professionnels tout comme durant les années 1960 et 1970, mais cette thèse ne s’y intéressera pas. Dans le cadre d’une sociologie des sciences et des techniques, je me questionne sur ce que nous, en tant que civilisation occidentale, faisons du développement technologique, du monde que nous construisons à l’aide des diverses techniques. Ma préoccupation ne porte pas sur les choix des objets privilégiés par la recherche des laboratoires publics et privés, mais sur les usages, les débouchés selon la question très générale : en quoi telle technique modifie-t-elle ou non notre façon de vivre ? Or, cette question relève d’un exercice de prospective dans la mesure où bien souvent nous ne possédons pas assez de recul sur ces techniques. Elle demeure malgré tout une préoccupation d’arrière-plan de mes questions de recherche. En effet, ces modèles d’apprentissage machine, trouveraient-ils leur place au sein d’une civilisation qui n’accorderait pas autant d’importance à l’efficacité, à la productivité, à la rentabilité, à la science ou encore au travail ? Aussi, viennent-ils entériner l’ordre établi ou bien offrent-ils de nouvelles possibilités d’existence ? Comprendre minimalement l’artefact d’un point de vue technique et saisir du point de vue de la sociologie la façon dont les chercheurs pensent leurs objets nous éclaire sur les catégories de pensées principales qui orientent ces usages et le cas échéant sur les éventuels effets sociétaux. En l’occurrence, l’idée de modéliser de nombreuses activités de la vie quotidienne repose sur une représentation à priori de celle-ci de la part de chercheurs localisés socialement par leur profession et plus généralement leurs relations sociales diverses. Quelle est cette représentation et comment est-elle opérationnalisée dans les modèles ? Posée autrement, de quels rapports au monde témoignent ingénieurs et informaticiens par l’intermédiaire de leurs connaissances professionnelles, savantes et ordinaires ? Ainsi, dans cette thèse, mon travail se ramène à informer de la dimension sociale propre à la technique étudiée. À partir d’entretiens auprès de chercheurs en IA, la question de recherche est la suivante : quels sont les savoirs et raisonnements chez les chercheurs au fondement de l’élaboration des algorithmes relevant de l’intelligence artificielle contemporaine et qui construisent une représentation opératoire particulière de la vie sociale ? Exprimée en terme sociologique, cette question devient : en quoi les modèles d'apprentissage sont-ils un nouveau « modèle concret de connaissance » pour les usagers et informaticiens selon le concept développé par le sociologue Gilles Houles ? Les modèles dits d’apprentissage sous-tendent une conception relationnelle de la constitution de la connaissance humaine et d’un rapport à la réalité médié par l’action comme moyen d’actualisation de cette connaissance. Résumé simplement, le concept sociologique de « modèle concret de connaissance » objective les deux modalités d’existence de la vie humaine que nous retrouvons empiriquement sous deux concepts informatiques : symbolique (leur modèle mathématique) et l’action ou « l’agent informatique » ou « celui qui agit », que ce concept soit mobilisé ou non par les chercheurs. En somme, ces modèles en relation les uns avec les autres et matérialisés dans les objets dits connectés ou « smart » (ex. téléphones, électroménagers) forment un schéma opératoire organisateur de la vie sociale. Ce côté opératoire repose sur la position de « tiers médiateur » ou de « mémoire sociale technicisée » dans les relations humains-humains via machines. Je m'appuierai sur le concept de mémoire sociale et de morphologie sociale développée par le sociologue Maurice Halbwachs. Autrement dit, ce réseau d’objets connectés et d’êtres humains se ramène à l’instauration d’un cadre cognitif collectif particulier, issu des représentations sociales de groupes professionnels précis, mais mises en jeu dans la vie courante de tous, soit une institutionnalisation en cours. En effet, la diffusion et l’adoption des modèles découlent d’un processus de reconnaissance publique de savoirs sous-jacents et déjà institutionnalisés, pour l'essentiel, les mathématiques et l’ingénierie, le savoir implicite des « sciences de la gestion » et un savoir dit « scientifique » par les chercheurs. Plus précisément, le processus en cours consiste en l’institution d’êtres humains et de machines apprenantes en liens permanents via les objets connectés (« Internet of Things »). Au final, elle consiste en une mise en réseau des « régularités sociales » obtenues par classifications et régressions effectuées par les détenteurs des données. Je parlerai d'une « morphologie sociotechnique » ou d’une « configuration sociotechnique ». À la figure du robot polyvalent anthropomorphique censé condenser toute la mise en pratique du savoir sur l’IA, se substitue celle de relations sociales informatisées comme lieux du maintien des liens entre êtres humains par un concentré de savoirs et idées hétérogènes tels qu'un être humain ayant besoin d'aide ou la nécessité de la créativité par exemple. À l’usage, s’établit un type de lien social entre êtres humains via les machines pris entre la réduction propre à tout modèle de l’objet sur lequel il porte, la flexibilité qu’offre la possibilité d’ajustement (le côté « learning ») et le sens donné à l’action par l’utilisateur lambda. L’idéologie comme « mode de connaissance » du réseau sociotechnique est partie prenante de cette institutionnalisation en cours. Elle offre un cadre cognitif qui remet en cause la diversité produite par la division habituelle du travail sémantique au sein des ensembles sociaux en fournissant un modèle de légitimité, soit le discours du « partage ». L’accent mis par cette thèse sur les « sciences de la gestion » et la notion de réseau l’inscrit dans le prolongement des études des trente dernières années sur cette « société informationnelle » de Manuel Castells, une « nouvelle représentation générale des sociétés » en réseau en suivant Luc Boltanski ou encore « l’esprit gestionnaire » qui s’empara des fonctionnaires d’État selon le sociologue Albert Ogien. / From the point of view of a sociology of knowledge, contemporary research in Artificial Intelligence (AI) draws its originality by its interest in the modeling of ordinary knowledge, that knowledge engaged in everyday life. The companies Facebook or Google, the so-called GAFAM, as well as online music and video services, capture and model the facts and gestures of the average person in order to put the results of their calculations back into play in its very everyday life. The research in AI also deals with academic and professional knowledge as it did in the 1960s and 1970s, but this thesis will not focus on it. Within the framework of a sociology of science and technology, I question what We, as the Western civilization, do with technological development, about the world we build with the help of various techniques. My concern is not with the choice of research objects privileged by public and private laboratories, but with the uses, the outlets according to the very general question: In what way does such a technique modify or not our way of living? Now, this question is an exercise in foresight insofar as we often do not have enough hindsight on these techniques. Nevertheless, it remains a background concern of my research. Indeed, would these machine learning models find their place in a civilization that would not give as much importance to efficiency, productivity, profitability, science or work? Also, do they endorse the established social organization, or do they offer new possibilities of existence? Understanding the artifact minimally from a technical perspective and grasping from a sociological point of view the way in which researchers think about their objects sheds light on the main categories of thought that guide these uses and, if applicable, on the possible societal effects. In this case, the idea of modeling many activities of daily life is based on an a priori representation of it by researchers who are socially situated by their profession. What is this representation and how is it operationalized in the models? Put differently, what relationships to the world do engineers and computer scientists have through their professional, academic and ordinary knowledge? Thus, in this thesis, my work comes down to informing the social dimension specific to the studied technique. Based on interviews with AI researchers, the question is as follows: What is the knowledge and reasoning of the researchers at the core of the algorithms of contemporary artificial intelligence and which build a specific operational representation of social life? Expressed in sociological terms, this question becomes: In what way are learning models a new "concrete model of knowledge" for users et researchers according to the concept developed by the sociologist Gilles Houles? The so-called learning models underlie a relational constitution of human knowledge and of a relation to reality mediated by actions as a means of actualization of this knowledge. Summarized simply, the sociological concept of "concrete model of knowledge" objectifies the two modalities of existence of the human life that we find empirically under two computing concepts: symbolic (their mathematical model) and the “action” as "the computing agent" or "the one who acts", whether this concept is used or not by the researchers. In sum, these models in relation to each other and materialized in the notorious connected or "smart" objects (e.g. telephones, household appliances) turn social life into a sociotechnical network. Its operational side relies on the position of "third-party intermediary" or "technical social memory" in human-human relations via machines. I will draw upon the concept of "social memory" and "social morphology" developed by the sociologist Maurice Halbwachs. In other words, this network of connected objects and human beings comes down to the establishment of a particular collective cognitive framework, stemming from the social representations of specific professional groups, but put into play in the everyday life of all, that is to say an institutionalization in progress. Indeed, the diffusion and adoption of the models stem from a process of public recognition of underlying and already institutionalized knowledge, essentially mathematics and engineering, the implicit knowledge of the "management sciences" and a knowledge called "scientific" by researchers. More precisely, the process underway consists in the institution of human beings and learning machines in permanent links via connected objects ("Internet of Things"). In the end, it consists in the networking of "social regularities" obtained by classifications and regressions carried out by the data's owners. I will speak of a "sociotechnical morphology" or a "sociotechnical configuration". The figure of the anthropomorphic multipurpose robot, which is supposed to condense all the practical application of knowledge on AI, is replaced by that of computerized social relations as places where links between human beings are maintained by a concentration of heterogeneous knowledge and ideas, such as a human being in need of help or the need for creativity, for example. In use, a type of social link between human beings via the machines is established, caught between the reduction proper to any modelling of the object on which it concerns, the flexibility offered by the possibility of adjustment (the "learning" side) and the meaning given to the action by the lambda user. Ideology as a "mode of knowledge" of the socio-technical network is part of this ongoing institutionalization. It offers a cognitive framework that challenges the diversity produced by the usual division of semantic labor within social groups by providing a model of legitimacy, namely the discourse of "sharing". The emphasis placed by this thesis on the "management sciences" and the notion of network places it in the continuity of the studies of the last thirty years on this "informational society" of Manuel Castells, a "new general representation of societies" into a network according to Luc Boltanski, or the "managerial spirit" which took hold of the State civil servants according to the sociologist Albert Ogien

I detta examensarbete har en utredning gjorts med syftet att undersöka lämpligheten att genomföra en djupstabilisering med multicementpelare (MC-pelare) i ett område med högt ställda sättningskrav. Detta genom att numeriskt simulera hur djupstabiliserad lera kan bete sig vid belastning med avseende på långtidsdeformationer. Vidare utvärderas hur väl olika materialmodeller kan simulera sättningar med hänsyn till kryp samt huruvida krypsättningar bör beaktas vid denna typ av dimensionering. Simuleringarna är genomförda enligt de geotekniska förutsättningarna och den planerade byggnationen på Galoppfältet i Täby Park. Som en grund för beräkningarna genomfördes en litteraturstudie där bland annat valet att simulera pelarna med Mohr-Coulomb (MoC), Soft Soil Creep (SSC) och Concrete Model (CM) i Plaxis 2D motiveras. Pelarna simulerades med en axissymmetrisk 2D-modell och i samtliga beräkningar simulerades leran med kryp medan pelarna simulerades både med och utan krypsättningar. En kalibrering mot tidigare forskning på området samt sättningsberäkningar av lera på Galoppfältet genomfördes för att säkerställa den valda modellens lämplighet. Därefter simulerades pelare med olika geometriska förutsättningar och varierande uppfyllnadshöjd. En sensitivitetsanalys gjordes även för att åskådliggöra vilka av de osäkra parametrarna som ger stor påverkan på resultatet. Simuleringarna visade att det antagligen är lämpligt att djupförstärka leran med multicementpelare för att uppnå låga sättningar över tid. För en pelardiameter på 800 mm och ett centrumavstånd på 1200 mm uppnåddes följande sättningar med de olika uppfyllnadshöjderna samt de olikamaterial modellerna. Av de testade pelarna ansågs denna geometri mest lämplig för de givna förutsättningarna på Galoppfältet. Detta baseras på den totala sättningen och att vissa pelare gick till brott i simuleringarna för övriga geometrier med större centrumavstånd och mindre pelardiameter. Beräkningarna där pelarna simulerades med SSC- och CM med kryp visade på stor variation sinsemellan. Resultaten med SSC anses felaktigt låga på grund av bristfälliga materialdata och visas därför inte i sammanfattningen. Resultaten med CM utan kryp och MoC utan kryp anses mest trovärdiga om än dock potentiellt för låga eftersom krypsättningar i pelarna inte är inkluderade. Vidare anses CM överlag vara en lämplig modell för att simulera sättningar med kryp i djupförstärkt lera förutsatt att ett tillräckligt underlag för val av materialparametrar finns att tillgå. SSC kan potentiellt vara lämplig enligt resultat från sensitivitetsanalysen förutsatt att materialparametrar kan undersökas för de lokala förutsättningarna. Dock förordas CM för fortsatta studier med krypsättningar i djupförstärkt lera. MoC anses som en god modell om kryp försummas. Med hänsyn till detta anser författarna att kryptester på multicementpelare bör övervägas för att skapa bättre förutsättningar för dimensionering av djupförstärkt lera med krypsättningar. Något som kan ge underlag för dimensionering med högre precision och potentiellt ekonomisk vinning. / This master thesis investigates the suitability of using multicement, a mixture of cement and cement kiln dust (CKD), for deep cement mixing (DCM) as soil improvement in an area with high requirements regarding the maximum allowed settlements. This was done by numerical modelling of long-term settlement behaviour for a DCM improved soil. Furthermore, the usability of different material models when considering creep behaviour are studied. The report also investigates whether creep behaviour should be considered or not in this type of simulation. The simulations are done according to the geotechnical conditions and the planned design for an old gallop track in Täby Park, Sweden. A literature review is done as a basis for the simulations performed in Plaxis 2D with an axis symmetric 2D-model. The use of Mohr-Coulomb (MoC), Soft SoilCreep (SSC) and Concrete Model (CM) for the columns are motivated. Creep settlements are included in the clay for all simulations while the columns are simulated both with and without creep settlements. To ensure the suitability of the model, a calibration is done by replication of a research study as well as replication of settlement calculations of the unimproved clay on Galoppfältet. Thereafter different geometric options for the columns and varying fill depths are simulated. Lastly a sensitivity analysis is conducted to display how much the uncertain material properties affect the result. Using multicement columns (MC columns) for DCM seems to be a good method for stabilizing the soil in Galoppfältet to achieve small settlements over time. For columns with a diameter of 800 mm and a center distance of 1200 mm the following settlements were calculated with the different material models for different fill heights. This geometry was considered most suitable for the conditions at Galoppfältet.  The two material models which account for creep in the columns show significantly different levels of total settlement. Due to lack of material data the results from simulations with SSC is deemed incorrect and thus not shown in the abstract. Thereby the results from MoC and CM without creep are deemed most credible even though they may be slightly low since they do not consider creep settlements in the columns. Furthermore, the Concrete Model is considered a good model to simulate settlements including creep in improved clay given that the material properties are carefully selected based on testing of the material. Soft Soil Creep may potentially be suitable for columns according to the calibration given that the different material parameters are investigated for the local conditions. However, CMare favoured. Mohr-Coulomb is deemed as a good model if creep settlements are neglected. With regards to this, the authors believe that creep tests of multicement columns should be considered to create improved preconditions for dimensioning of DCM columns with creep. This to improve the accuracy of design and potentially create cost reductions.

The aim of the diploma thesis was to design the precast reinforced concrete prefabricated structures of the football tribune. It was modernized sports center in Brno. According to the investor was determined the shape of the football tribune. It was compiled the design documentation. For the calculation of inner strengths was used the software Dlubal RFEM, which operates on the Finite Element Method - FEM. All elements have been designed for ultimate and serviceability limit state. My thesis contains textual part, inner strengths, static calculations, drawing documentation of composition of structure, drawing documentation of reinforcement and design documentation of the object.

This thesis is focusing on general issues of floor concrete slabs, technology, implementation, subsoil models, calculation of internal forces and design approaches. Based on these data the thesis monitors the behaviour of floor slabs according to the various parameters. It also includes three versions of design of the real floor slab, made of steel fibre reinforced concrete, reinforced concrete and prestressed concrete. Economic comparison is also included. In addition it developes a design of the chalice base and reinforced concrete column of the supporting hall system, instead of the original usual base with steel column.