This thesis presents a discussion of design tools for analysing block stability around a tunnel. First, it was determined that joint length and field stress have a significant influence on estimating block stability. The results of calculations using methods based on kinematic limit equilibrium (KLE) were compared with the results of filtered DFN-DEM, which are closer to reality. The comparison shows that none of the KLE approaches– conventional, limited joint length, limited joint length with stress and probabilistic KLE – could provide results similar to DFN-DEM. This is due to KLE’s unrealistic assumptions in estimating either volume or clamping forces. A simple mechanism for estimating clamping forces such as continuum mechanics or the solution proposed by Crawford-Bray leads to an overestimation of clamping forces, and thus unsafe design. The results of such approaches were compared to those of DEM, and it was determined that these simple mechanisms ignore a key stage of relaxation of clamping forces due to joint existence. The amount of relaxation is a function of many parameters, such as stiffness of the joint and surrounding rock, the joint friction angle and the block half-apical angle. Based on a conceptual model, the key stage was considered in a new analytical solution for symmetric blocks, and the amount of joint relaxation was quantified. The results of the new analytical solution compared to those of DEM and the model uncertainty of the new solution were quantified. Further numerical investigations based on local and regional stress models were performed to study initial clamping forces. Numerical analyses reveal that local stresses, which are a product of regional stress and joint stiffness, govern block stability. Models with a block assembly show that the clamping forces in a block assembly are equal to the clamping forces in a regional stress model. Therefore, considering a single block in massive rock results in lower clamping forces and thus safer design compared to a block assembly in the same condition of in-situ stress and properties. Furthermore, a sensitivity analysis was conducted to determine which is  the most important parameter by assessing sensitivity factors and studying the applicability of the partial coefficient method for designing block stability. It was determined that the governing parameter is the dispersion of the half-apical angle. For a dip angle with a high dispersion, partial factors become very large and the design value for clamping forces is close to zero. This suggests that in cases with a high dispersion of the half-apical angle, the clamping forces could be ignored in a stability analysis, unlike in cases with a lower dispersion. The costs of gathering more information about the joint dip angle could be compared to the costs of overdesign. The use of partial factors is uncertain, at least without dividing the problem into sub-classes. The application of partial factors is possible in some circumstances but not always, and a FORM analysis is preferable. / QC 20111201

The partial safety factor method is the main safety concept applied across structural design standards. This method is also presented in EN-1990 as the basis of structural design in Europe. In the review of this code for the new generation of Eurocodes, analysis of the partial safety factor method seems necessary. The origin of the partial safety factor method is related to probabilistic methods and reliability analysis. Therefore, the latter is selected as tools for the evaluation of the partial safety factor method in the EN-1990 framework. Consequently this research begins with an explanation of the background of partial safety factor methods and reliability analysis. Different aspects of this safety concept are investigated through this study. The analysis strategy is based on the study of partial safety factor method according to the different part of EN-1990. The research is divided into two main parts, according to the basic components of limit state functions: load and resistance. Aspects related to loading are investigated first. The available load combinations and the recommended partial factors are investigated based on their reliability levels. The load combinations are compared with each other according to the sustainability of their design. An increased factor for the application of snow load is proposed to overcome safety problems related to snow load on structures. Consequently, a proposal for simplifying these load combinations is offered and verified according to reliability analysis. In the final step, regarding the load’s partial factors, a method of calibration is proposed, based on Monte Carlo reliability analysis. Afterwards, the aspects related to the resistance are analyzed. Resistances depend mostly on experimental data. Therefore, the relationship between the partial safety factor of resistance and test numbers is investigated. A probabilistic analysis based on Annex D of EN-1990 is then applied to calculate the model uncertainty partial factor and the resistance partial factor for a database from masonry shear walls. A comparison is made to show the influence of different way of partial safety factor utilization in a limit state function.:1 Introduction 2 Partial safety factor method and EN-1990 3 Reliability analysis 4 Load combinations and partial safety factors 5 Resistance partial safety factor 6 Summary and outlook

Neste trabalho é realizado o estudo dos coeficientes parciais de segurança utilizados em normas de projeto estrutural. O trabalho é fundamentado na teoria de confiabilidade estrutural, que permite uma representação explícita das incertezas envolvidas em resistências e ações, e resulta em uma estimativa quantitativa da segurança estrutural: o índice de confiabilidade. O trabalho aborda uma metodologia de calibração de norma que permite obter o conjunto de coeficientes parciais de segurança que minimiza as variações dos índices de confiabilidade das mais diversas estruturas projetadas segundo uma norma de projeto, em relação ao índice de confiabilidade alvo utilizado na calibração. Uma calibração inicial é feita utilizando um índice de confiabilidade alvo igual a 3,0, mas diferentes índices de confiabilidade também são considerados. A análise é feita para elementos estruturais de aço. A calibração é feita para dois formatos de norma distintos. No formato das normas americanas (ANSI/AISC), um conjunto de coeficientes parciais de segurança é obtido para cada expressão de combinação de ações. No formato da norma Brasileira e Européia, um único conjunto de coeficientes parciais (e de coeficientes de combinação) é obtido, para uma única expressão de dimensionamento. Os coeficientes parciais de segurança obtidos para os dois formatos são comparados com os coeficientes utilizados atualmente na norma brasileira ABNT NBR8800:2008. Os índices de confiabilidade resultantes também são comparados, para as diferentes combinações de ações, em função da razão de proporcionalidade entre as diferentes ações. Os resultados obtidos mostram que os coeficientes parciais de segurança utilizados na norma brasileira ABNT NBR8800:2008 levam a uma variação significativa dos índices de confiabilidade. No trabalho, é obtido um conjunto de coeficientes que reduz esta variação, proporcionando maior uniformidade dos índices de confiabilidade. Estes resultados sugerem que uma revisão dos coeficientes parciais utilizados na norma brasileira ABNT NBR8800:2008 pode ser recomendável. Esta recomendação, no entanto, depende de um aprofundamento da investigação iniciada neste trabalho, uma vez que os resultados obtidos não refletem todas as situações de projeto cobertas por esta norma. / This study addresses the partial safety factors used in structural design codes. The study is based on the theory of structural reliability, which allows an explicit consideration of the uncertainties in material strengths and load actions, and results in a quantitative measure of structural safety: the reliability index. A calibration methodology is considered, which allows one to find a set of partial safety factors that minimizes the variations on reliability indexes, for all structures designed within a code, with respect to a pre-selected target reliability. An initial calibration is performed for a target reliability index 3,0, but other target reliabilities are also considered. The analysis is limited to steel structural members. The calibration is made for two distinct code formats. In the ANSI/AISC code format, a set of partial factors is obtained for each load combination expression. In the Brazilian and European code formats, one single set of partial (and load combination) factors is obtained, for a single load combination expression. Partial safety factors obtained for both code formats are compared with partial factors used in National Brazilian code ABNT NBR8800:2008. The resulting reliability indexes are also compared, for the distinct load combination expressions, in terms of the proportionality ratios between the distinct actions. Results obtained in the study show that the partial safety factors used in ABNT NBR8800:2008 lead to significant variation on reliability indexes. Another set of partial safety factors is obtained in the study, which provides greater uniformity of reliability indexes. These results suggest that a revision of partial safety factors adopted in ABNT NBR8800:2008 might be recommended. This recommendation, however, is dependent on a deepening of the investigation started in this study, which does not reflect all design situations covered by the design code.

Neste trabalho é realizado o estudo dos coeficientes parciais de segurança utilizados em normas de projeto estrutural. O trabalho é fundamentado na teoria de confiabilidade estrutural, que permite uma representação explícita das incertezas envolvidas em resistências e ações, e resulta em uma estimativa quantitativa da segurança estrutural: o índice de confiabilidade. O trabalho aborda uma metodologia de calibração de norma que permite obter o conjunto de coeficientes parciais de segurança que minimiza as variações dos índices de confiabilidade das mais diversas estruturas projetadas segundo uma norma de projeto, em relação ao índice de confiabilidade alvo utilizado na calibração. Uma calibração inicial é feita utilizando um índice de confiabilidade alvo igual a 3,0, mas diferentes índices de confiabilidade também são considerados. A análise é feita para elementos estruturais de aço. A calibração é feita para dois formatos de norma distintos. No formato das normas americanas (ANSI/AISC), um conjunto de coeficientes parciais de segurança é obtido para cada expressão de combinação de ações. No formato da norma Brasileira e Européia, um único conjunto de coeficientes parciais (e de coeficientes de combinação) é obtido, para uma única expressão de dimensionamento. Os coeficientes parciais de segurança obtidos para os dois formatos são comparados com os coeficientes utilizados atualmente na norma brasileira ABNT NBR8800:2008. Os índices de confiabilidade resultantes também são comparados, para as diferentes combinações de ações, em função da razão de proporcionalidade entre as diferentes ações. Os resultados obtidos mostram que os coeficientes parciais de segurança utilizados na norma brasileira ABNT NBR8800:2008 levam a uma variação significativa dos índices de confiabilidade. No trabalho, é obtido um conjunto de coeficientes que reduz esta variação, proporcionando maior uniformidade dos índices de confiabilidade. Estes resultados sugerem que uma revisão dos coeficientes parciais utilizados na norma brasileira ABNT NBR8800:2008 pode ser recomendável. Esta recomendação, no entanto, depende de um aprofundamento da investigação iniciada neste trabalho, uma vez que os resultados obtidos não refletem todas as situações de projeto cobertas por esta norma. / This study addresses the partial safety factors used in structural design codes. The study is based on the theory of structural reliability, which allows an explicit consideration of the uncertainties in material strengths and load actions, and results in a quantitative measure of structural safety: the reliability index. A calibration methodology is considered, which allows one to find a set of partial safety factors that minimizes the variations on reliability indexes, for all structures designed within a code, with respect to a pre-selected target reliability. An initial calibration is performed for a target reliability index 3,0, but other target reliabilities are also considered. The analysis is limited to steel structural members. The calibration is made for two distinct code formats. In the ANSI/AISC code format, a set of partial factors is obtained for each load combination expression. In the Brazilian and European code formats, one single set of partial (and load combination) factors is obtained, for a single load combination expression. Partial safety factors obtained for both code formats are compared with partial factors used in National Brazilian code ABNT NBR8800:2008. The resulting reliability indexes are also compared, for the distinct load combination expressions, in terms of the proportionality ratios between the distinct actions. Results obtained in the study show that the partial safety factors used in ABNT NBR8800:2008 lead to significant variation on reliability indexes. Another set of partial safety factors is obtained in the study, which provides greater uniformity of reliability indexes. These results suggest that a revision of partial safety factors adopted in ABNT NBR8800:2008 might be recommended. This recommendation, however, is dependent on a deepening of the investigation started in this study, which does not reflect all design situations covered by the design code.

Probabilistic assessment and life cycle management of engineering components and systems in a nuclear power plant is intended to ensure safe and efficient operation of energy generation over its entire life. The CANDU reactor core consists of 380-480 pressure tubes, which are like miniature pressure vessels that contain natural uranium fuel. Pressure tubes operate under severe temperature and radiation conditions, which result in degradation with ageing. Presence of flaws in a pressure tube makes it vulnerable to delayed hydride cracking (DHC), which may lead to rupture or break-before-leak situation. Therefore, assessment of flaws in the pressure tubes is considered an integral part of a reactor core assessment program. The main objective of the thesis is to develop advanced probabilistic and mechanical stress field models for the assessment of flaws. The flaw assessment models used by the industries are based on deterministic upper/lower bound values for the variables and they ignore uncertainties associated with system parameters. In this thesis, explicit limit state equations are formulated and first order reliability method is employed for reliability computation, which is more efficient than simulation-based methods. A semi-probabilistic approach is adopted to develop an assessment model, which consists of a mechanics-based condition (or equation) involving partial factors that are calibrated to a specified reliability level. This approach is applied to develop models for DHC initiation and leak-before-break assessments. A novel feature of the proposed method is that it bridges the gap between a simple deterministic analysis and complex simulations, and it is amenable to practical applications. The nuclear power plant systems are not easily accessible for inspection and data collection due to exposure to high radiation. For this reason, small samples of pressure tubes are inspected at periodic intervals and small sample of data so collected are used as input to probabilistic analysis. The pressure tube flaw assessment is therefore confounded by large sampling uncertainties. Therefore, determination of adequate sample size is an important issue. In this thesis, a risk informed approach is proposed to define sample size requirement for flaw assessment. Notch-tip stress field is a key factor in any flaw assessment model. Traditionally, linear elastic fracture mechanics (LEFM) and its extension, serves the basis for determination of notch-tip stress field for elastic and elastic-perfectly-plastic material, respectively. However, the LEFM solution is based on small deformation theory and fixed crack geometry, which leads to singular stress and strain field at the crack-tip. The thesis presents new models for notch and crack induced stress fields based on the deformed geometry. In contrast with the classical solution based on small deformation theory, the proposed model uses the Cauchy's stress definition and boundary conditions which are coupled with the deformed geometry. This formulation also incorporates the rotation near the crack-tip, which leads to blunting and displacement of the crack-tip. The solution obtained based on the final deformed configuration yields a non-singular stress field at the crack-tip and a non-linear variation of stress concentration factor for both elastic and elastic-perfectly-plastic material. The proposed stress field formulation approach is applied to formulate an analytical model for estimating the threshold stress intensity factor (KIH) for DHC initiation. The analytical approach provides a relationship between KIH and temperature that is consistent with experimental results.

Probabilistic assessment and life cycle management of engineering components and systems in a nuclear power plant is intended to ensure safe and efficient operation of energy generation over its entire life. The CANDU reactor core consists of 380-480 pressure tubes, which are like miniature pressure vessels that contain natural uranium fuel. Pressure tubes operate under severe temperature and radiation conditions, which result in degradation with ageing. Presence of flaws in a pressure tube makes it vulnerable to delayed hydride cracking (DHC), which may lead to rupture or break-before-leak situation. Therefore, assessment of flaws in the pressure tubes is considered an integral part of a reactor core assessment program. The main objective of the thesis is to develop advanced probabilistic and mechanical stress field models for the assessment of flaws. The flaw assessment models used by the industries are based on deterministic upper/lower bound values for the variables and they ignore uncertainties associated with system parameters. In this thesis, explicit limit state equations are formulated and first order reliability method is employed for reliability computation, which is more efficient than simulation-based methods. A semi-probabilistic approach is adopted to develop an assessment model, which consists of a mechanics-based condition (or equation) involving partial factors that are calibrated to a specified reliability level. This approach is applied to develop models for DHC initiation and leak-before-break assessments. A novel feature of the proposed method is that it bridges the gap between a simple deterministic analysis and complex simulations, and it is amenable to practical applications. The nuclear power plant systems are not easily accessible for inspection and data collection due to exposure to high radiation. For this reason, small samples of pressure tubes are inspected at periodic intervals and small sample of data so collected are used as input to probabilistic analysis. The pressure tube flaw assessment is therefore confounded by large sampling uncertainties. Therefore, determination of adequate sample size is an important issue. In this thesis, a risk informed approach is proposed to define sample size requirement for flaw assessment. Notch-tip stress field is a key factor in any flaw assessment model. Traditionally, linear elastic fracture mechanics (LEFM) and its extension, serves the basis for determination of notch-tip stress field for elastic and elastic-perfectly-plastic material, respectively. However, the LEFM solution is based on small deformation theory and fixed crack geometry, which leads to singular stress and strain field at the crack-tip. The thesis presents new models for notch and crack induced stress fields based on the deformed geometry. In contrast with the classical solution based on small deformation theory, the proposed model uses the Cauchy's stress definition and boundary conditions which are coupled with the deformed geometry. This formulation also incorporates the rotation near the crack-tip, which leads to blunting and displacement of the crack-tip. The solution obtained based on the final deformed configuration yields a non-singular stress field at the crack-tip and a non-linear variation of stress concentration factor for both elastic and elastic-perfectly-plastic material. The proposed stress field formulation approach is applied to formulate an analytical model for estimating the threshold stress intensity factor (KIH) for DHC initiation. The analytical approach provides a relationship between KIH and temperature that is consistent with experimental results.