Thesis CUANTIFICACIÓN DE LA INCERTEZA EN SISTEMAS ESTRUCTURALES ESTÁTICOS Y DINÁMICOS MEDIANTE ESTRATEGIAS EFICIENTES EN EL CONTEXTO DE LÓGICA DIFUSA
Loading...
Date
2016
Journal Title
Journal ISSN
Volume Title
Program
INGENIERIA CIVIL
Campus
Universidad Técnica Federico Santa María UTFSM. Casa Central Valparaíso
Abstract
Las incertezas presentes en sistemas estructurales y en las solicitaciones a las que se ven sometidasgeneran un problema de alta complejidad a la hora de calcular las respuestas de interés útilesen el diseño, las cuales se ven inevitablemente afectadas por esta incertidumbre. El problema seha abordado de distintas formas, ya sea mediante probabilidades clásicas y técnicas Bayesianas osu cuanti cación mediante análisis de intervalos. Para esta última, los valores que puede tomar unparámetro de entrada se representan dentro de un intervalo, lo que finalmente se propaga tambiénen un intervalo para la respuesta de interés. El análisis estructural difuso se entiende como unasecuencia de análisis de intervalos, asignando a las variables inciertas funciones de pertenencia. Deesta forma, la respuesta estructural se caracteriza como un intervalo para cada nivel de pertenencia.Encontrar la función de pertenencia de las respuestas es una tarea que implica un alto costonumérico, para cuya solución se han propuesto métodos exactos y aproximados, entre los cuales eldesarrollo mediante expansiones ha llamado la atención debido a su bajo costo computacional, yaque se limita a un único análisis estructural (calcular la inversa de la matriz de rigidez o resolverel problema de valores propios para problemas estáticos y dinámicos, respectivamente). Esta tesispropone una mejora a los métodos de expansión existentes y su aplicación para análisis estructuraldifuso. Para el caso estático, se propone incluir el uso de una variable intermedia en una aproximación que permite modelar el efecto de la incertidumbre respecto de parámetros estructurales ycargas. Para el caso dinámico, la variable intermedia se aplica sobre una aproximación construidaen base a la función de respuesta en frecuencias, que incluye información acerca de la sensibilidad delas frecuencias naturales y modos de vibrar respecto de las variables difusas. La técnica propuestaserá aplicada a problemas de análisis estructural lineal, en diversos casos modelados mediante elmétodo de elementos nitos. Los resultados obtenidos del análisis estructural difuso son comparadoscon los métodos de expansión existentes y la solución exacta obtenida mediante técnicas deoptimización.
The uncertainties associated with structural systems and its loading conditions generate a highcomplexity problem when design responses are needed, which are unavoidable affected by this uncertainty.This problem has been tackled in different ways, either through classical probabilitiesand Bayesian techniques or its quanti cation via interval analysis. For the latter, the values thatan input parameter can assume are represented within an interval, which is propagated as an intervalfor the response of interest as well. Fuzzy structural analysis consists in a sequence of intervalanalysis, assigning membership functions to each uncertain parameter. In this way, the structuralresponse is characterized as an interval for each membership level. Calculating the membershipfunction of the responses is a task that implies a high numerical cost, for whose solution exact andapproximate methods have been proposed, where expansion procedures have drawn attention dueto its low computational cost, implying a single structural analysis (stiness matrix inversion oreigenproblem solution, for static or dinamic case, respectively). This thesis proposes an improvementfor the existing expansion methods and its application in fuzzy structural analysis. For thestatic loading case, intervening variables are included in an approximation that allows to modelthe uncertainty effect due to structural parameters and loads. For the dynamic loading case, theintervening variables are applied on an approximation constructed based on the frecuency responsefunction, including information from the natural frequencies and shape-modes derivatives with respectto fuzzy variables. The proposed technique its applied on linear structural analysis problemsfor different cases modeled through the nite element method. The obtained results for fuzzy structuralanalysis are compared with existing expansion methods and with the exact solution obtainedthrough optimization techniques.
The uncertainties associated with structural systems and its loading conditions generate a highcomplexity problem when design responses are needed, which are unavoidable affected by this uncertainty.This problem has been tackled in different ways, either through classical probabilitiesand Bayesian techniques or its quanti cation via interval analysis. For the latter, the values thatan input parameter can assume are represented within an interval, which is propagated as an intervalfor the response of interest as well. Fuzzy structural analysis consists in a sequence of intervalanalysis, assigning membership functions to each uncertain parameter. In this way, the structuralresponse is characterized as an interval for each membership level. Calculating the membershipfunction of the responses is a task that implies a high numerical cost, for whose solution exact andapproximate methods have been proposed, where expansion procedures have drawn attention dueto its low computational cost, implying a single structural analysis (stiness matrix inversion oreigenproblem solution, for static or dinamic case, respectively). This thesis proposes an improvementfor the existing expansion methods and its application in fuzzy structural analysis. For thestatic loading case, intervening variables are included in an approximation that allows to modelthe uncertainty effect due to structural parameters and loads. For the dynamic loading case, theintervening variables are applied on an approximation constructed based on the frecuency responsefunction, including information from the natural frequencies and shape-modes derivatives with respectto fuzzy variables. The proposed technique its applied on linear structural analysis problemsfor different cases modeled through the nite element method. The obtained results for fuzzy structuralanalysis are compared with existing expansion methods and with the exact solution obtainedthrough optimization techniques.
Description
Catalogado desde la version PDF de la tesis.
Keywords
ANALISIS ESTRUCTURAL DIFUSO, CARGAS DIFUSAS, DERIVADAS DE VALORES Y VECTORES PROPIOS, FUNCION DE RESPUESTA EN FRECUENCIAS, FUNCIONES DE PERTENENCIA, METODO DE ELEMENTOS FINITOS, PARAMETROS ESTRUCTURALES DIFUSOS, SENSIBILIDAD ESTRUCTURAL, SERIES DE TAYLOR, VARIABLES INTERMEDIAS