How well do calculations methods prescribed in today's design codes and standards represent conditions in natural fires? Can the temperature and behavior of a steel member in fire be predicted from these calculations? A literature review of structural fire codes, full scale fire tests, published fire test data, the function and selection of design fires, mechanical and thermal behaviors of structural steel, and numerical calculation methods for the temperature of steel members was conducted as a foundation to analyze whether a not a structural fire engineer can answer these questions. Through comparisons of published data from four natural fires tests performed at the Cardington test facility in the United Kingdom to numerical calculations based upon prescribed methods from Eurocode 3 and the Swedish Design Manual, time-temperature curves were developed to demonstrate the variation in temperature of the recorded data in the natural fire tests at Cardington to the equivalent members being analyzed with numerical calculation methods. When available, fire compartment characteristics were replicated during numerical calculations to ensure the highest correlation between the recorded and calculated results. An Excel tool was created to rapidly calculate and produce the resulting time-temperature curves as well as yield strength, modulus of elasticity, and load carrying capacity using a variety of input parameters including design fire data and steel member selection. The goal of the Cardington fires study was to provide comparisons of published natural fire data to results of numerical calculation methods from the codes. Additional comparisons were developed using a US Office design to show the effects of changing compartment and design parameters on the steel temperature, yield strength, elastic modulus and load carrying capacity. Differences found in temperature of steel members between the published Cardington data and numerical calculations proved the difficulty of predicting the behavior of a structural steel beam throughout an entire length of a fire or even until failure. Discussion of results addressed the selection of design fires, input parameters, structural layouts of office buildings, heating and cooling phases of steel members, and failure criteria.

Les machines électriques synchrones à aimants permanents sont susceptibles de rencontrer un disfonctionnement suite à un échauffement non maîtrisé. L’objectif de cette étude est de développer un modèle thermique générique et prédictif pouvant simuler diverses situations d’intérêts: régime permanent, régime transitoire, mode dégradé, entrefer immergé, haute vitesse. Pour cela, la méthode nodale est utilisée pour développer le modèle thermique générique. En parallèle, un banc d’essai et un prototype sont conçus pour valider le modèle. L’étude de sensibilité des résultats du modèle à certains paramètres montrent que certains coefficients de convection, certaines conductances de contact et la conductivité thermique radiale du bobinage ont une influence considérable sur les résultats du modèle. Cependant ces paramètres sont mal connus, car ils sont issus des formules empiriques ou des abaques. Grâce au prototype et au modèle développé, ces paramètres sont identifiés. Trois méthodes d’identification sont testées pour aboutir à une stratégie d’identification: les algorithmes génétiques, la méthode de Gauss-Newton et la méthode de Levenberg-Marquardt. Plusieurs essais sont effectués sur le prototype instrumenté. La mesure des températures à des lieux précis du prototype permet d’identifier les paramètres mal connus et de valider le modèle. / Permanent magnet synchronous machines are likely to break down due to poorly controlled heating. The goal of this study was to develop a generic and predictive thermal model to calculate the temperature of machines during the design phase simulating temperatures at various states. These states include: steady state, transient state, fault mode, axial circulating of a cooling fluid in the air-gap and high speed. The lumped parameter method was used to develop this generic thermal model. Meanwhile, a test bench and a prototype instrumented with thermocouples were manufactured to validate the model at the same time. Sensitivity studies of the results of the model to some parameters demonstrated that some convective coefficients, contact conductances and the thermal conductivity of the winding in the radial direction influenced the model. However, these parameters are poorly known, because empirical formulas or abacus are used to calculate them. Using, the prototype and the developed model, these parameters were identified. Three methods of identification were tested in order to find a strategy for the identification: the genetic algorithms method, the Gauss-Newton method and the Levenberg-Marquardt method. Many tests were done on the prototype. The measure of the temperatures on the specific place allows to identify these parameters and to validate the model.

A soft quadruped robot is being developed at the Department of Machine Design and Department of Production Engineering at KTH. The legs of the robot consist of four continuum actuators that can achieve complex movements. In order to efficiently develop gaits for the robot, reinforcement learning will be used. The learning process will use data from simulation instead of directly from the real robot to save time and resources. However, it is significantly more computationally expensive to simulate soft robotics than rigid, because the physical laws of flexible materials are inherently complex. Because of this, soft robot simulations tend to be slower which limits their usability for reinforcement learning. This thesis explores simulation modelling options in Matlab Simscape for the soft quadruped robot, that can be used in reinforcement learning. Four simulation models of the soft actuator were implemented in order to be tested and compared. Two actuation methods and two build options were chosen based on the literature study and related works, and were then permuted for the different combinations. The tested combinations are: lumped-parameter method actuated by internal force, flexible beam actuated by internal force, lumped-parameter method actuated by cable/pulley network and flexible beam actuated by cable/pulley network. The four actuators were built and tested separately. Computational time and simulation-to-reality gap were used for evaluating the modeling methods. The results show that the best option when modelling the soft actuator for reinforcement learning in Matlab Simscape is to use the lumped-parameter method in combination with a cable and pulley network. High accuracy level can still be achieved despite not keeping the true number of attachment points between the cable and actuator. The number of pulleys in the model is linearly correlated to the time cost required to simulate the model. / En mjuk fyrbent robot är under utveckling vid institutionen för maskinkonstruktion och institutionen för industriell produktion på KTH. Robotens ben består av fyra kontinuerligt deformerbara ställdon som kan åstadkomma komplexa rörelser. För att effektivt utveckla gångstilar till roboten kommer förstärkt inlärning att användas. Inlärningsprocessen kommer att använda data från simulering istället för från den fysiska roboten för att spara tid och resurser. Det är dock betydligt dyrare beräkningsmässigt att simulera mjuk robotik än styv, eftersom flexibla material är mer komplexa. På grund av detta tenderar simuleringar av mjuka robotar att vara långsammare, vilket begränsar deras användbarhet för förstärkt inlärning. Detta examensarbete utforskar därför alternativ för modellering och simulering av den mjuka fyrbenta roboten i Matlab Simscape, med målet att den ska kunna användas med förstärkt inlärning. Fyra olika simuleringsmodeller av det mjuka ställdonet implementerades för att testas och jämföras. Två aktiveringsmetoder och två konstruktionsalternativ valdes baserat på litteraturstudien och relaterade arbeten, och permuterades sedan till möjliga versioner. De testade versionerna är således: klumpparametermetod som aktiveras av intern kraft, flexibel balk som aktiveras av intern kraft, klumpparametermetod som aktiveras av kabelnätverk och flexibel balk som aktiveras av kabelnätverk. De fyra ställdonen byggdes och testades separat. Beräkningstid och grad av verklighetstrogenhet, användes för att jämföra resultaten av dessa tester. Resultaten visar att det bästa alternativet vid modellering av det mjuka ställdonet för förstärkt inlärning i Matlab Simscape är att använda klumpparametermetoden i kombination med ett kabelnätverk. Hög noggrannhetsnivå kan uppnås trots att man inte bibehåller det verkliga antalet fästpunkter mellan kabeln och ställdonet. Antalet fästpunkter för kabeln i modellen är linjärt korrelerat till den tidskostnad som krävs för att simulera modellen.