Les évents de sécurité protègent de l'explosion les réacteurs chimiques sièges d'un emballement thermique de réaction. Pour les systèmes non tempérés (c'est à dire produisant majoritairement des gaz incondensables), les méthodes de dimensionnement des évents issues des travaux du DIERS sont très surdimensionnantes. Une méthode basée sur le principe de similitude, développée dans le cadre de l'ONU pour la famille des peroxydes, fournit des aires d'évent plus réalistes mais elle est très contraignante. Le présent travail a permis la réalisation d'un nouvel outil de dimensionnement en similitude pour scénario d'incendie : la maquette à 0,1 litre. Il s'agit d'une extension du calorimètre adiabatique VSP2. Cette maquette permet, à l'échelle du laboratoire, la réalisation de blowdowns et la détermination directe du rapport A/V de l'évent nécessaire, mais également le suivi en temps réel de la masse réactionnelle évacuée. <br />Nous avons validé l'utilisation de cette maquette à 0,1 litre (1 x 10-3 m-1 < A/V < 3,5 x 10-3 m-1) en comparant avec des blowdowns analogues effectués à l'INERIS dans le réacteur ONU 10 litres. Ces blowdowns ont été réalisés avec une solution d'hydroperoxyde de cumène (30% en masse) dans 2,2,4-triméthyl-1,3-pentanediol diiso-butyrate. Ces essais ont montré que la maquette à 0,1 litre conduit à des évents légèrement plus grands (0 à 50 %) que le réacteur ONU 10 litres. Elle se situe donc du côté de la sécurité, tout en étant beaucoup moins surdimensionnante que la méthode DIERS, et utilisable à l'échelle du laboratoire. La principale limite est due à des fuites thermiques dont il faut vérifier pour chaque système étudié que l'influence est négligeable. <br />Du point de vue compréhension, nos expériences montrent que, même si la décomposition de notre système ressemble à celle d'un système non tempéré (2 pics de pression), elle génère des vapeurs (produits de la décomposition) qui ont une forte influence sur le 2ème pic : ces vapeurs provoquent un ralentissement de la réaction et l'atténuation des températures maximales atteintes. On constate même une corrélation Pmax = f(Tmax). Ce comportement pourrait concerner la plupart (toutes ?) des décompositions. <br />Les mesures de masse évacuée ont permis de distinguer trois types de comportements qui illustrent l'influence de la pression dans le réacteur sur le « level swell ». La confrontation avec un modèle dynamique purement « gassy » a montré que l'évacuation de masse réactionnelle peut se traduire par une évacuation purement diphasique ou par une alternance gaz/ diphasique, que pour les hautes pressions l'évacuation est purement gazeuse au turnaround et que l'évacuation diphasique lors de la dépressurisation du second pic doit être imputée en grande partie à la présence de vapeur (ébullition). <br />Enfin, nous avons identifié et quantifié la contribution des différentes hypothèses au caractère surdimensionnant de la méthode DIERS appliquée à notre système hybride non tempéré. Parmi les hypothèses surdimensionnantes identifiées, celle qui suppose que le « turnaround » est gouverné par une égalité de débit volumique est de loin celle qui est la cause principale de surdimensionnement.

This thesis provides a structural and materials engineering explanation for many of the running fractures that occurred in steel structures during the destructive Kobe and Northridge earthquakes in the mid 1990s. A method is developed that allows the ductile endurance of structural steel members subjected to cyclic plastic deformation during earthquakes to be assessed and for pre-necking running fractures to be avoided. The study commenced following the 2000 World Earthquake Conference in Auckland. The conference brought together the findings of the huge research effort, in America, Japan, Europe and New Zealand, that followed the Kobe and Northridge earthquakes. The running fractures that had occurred in steel structures represented an unpredicted failure mode that structural engineers have not known how to predict or suppress through the engineering design process. A clear fundamental understanding of the causes and how to prevent the fractures did not arise from the conference. In fact apparently conflicting results were reported. Full scale cyclic tests in New Zealand on structural assemblies had not resulted in running fractures, whereas tests in American and Japan had. Structural engineers designing earthquake resistant structures rely on constructional steel to be materially homogeneous and nominally tri-linear in behaviour. Steel is expected to behave elastically under regular in-service loading, have a reliable and flat yield stress-strain characteristic, and under overload then develop predictable levels of strain-hardening in conjunction with significant plastic elongation up to its ultimate tensile strength. Steel is expected to eventually fracture after further plastic elongation and necking. Ductile design strategies and methods utilise the plastic elongation characteristics of steel to protect structures in earthquake. Plastic deformation is considered to beneficially dissipate energy generated in the structure by a severe earthquake and also dampen the structure’s response. The occurrence of running fracture without significant cyclic plastic deformation and before section necking in steelwork, therefore undermines the basis of the ductile seismic design approach. The initial part of the thesis is devoted to bringing together the fundamental aspects of materials engineering related to fracture of constructional steel. This is intended to provide a bridge of knowledge for structural engineering practitioners and researchers not fully conversant with materials engineering aspects of fracture. Fracture behaviour in steel is a broad and complex topic that developed rapidly in the twentieth century driven by the demands of technological growth. The unexpected fracture of welded liberty ships at sea in World War 2; the need for reliable long term containment for the nuclear reactors in the 1950s and 1960s; and prevention of fatigue failures in aircraft frames since the 1950s all drove engineering research into steel fracture behaviour. There are many subtle variations in definitions in the published literature on fracture that can be confusing. Therefore an attempt has been made to clarify terminology. The term brittle fracture in particular is only used in this thesis as applying to running fracture when the general or far field tensile stresses are below the yield stress of the steel. The term pre-necking or running fracture is preferred to describe the condition more broadly which may occur prior to and also after general yielding, but before section necking. Running fracture is a manifestation of pre-necking fracture in which insufficient plastic flow is available in the assembly to absorb the energy released upon fracture. The experimental studies investigated the behaviour of constructional steel commonly used in New Zealand, at various levels of plastic strain. This started with Charpy V-Notch (CVN) testing which revealed that a significant transition temperature shift and curve shape change occurs with increasing plastic strain and the associated strain-hardening. This showed that the ability of steel to avoid pre-necking or running fracture reduces as the level of plastic strain-hardening increases. Temperature controlled Crack Tip Opening Displacement (CTOD) testing was then undertaken. The setting of testing temperatures for the CTOD tests were guided by review of the CVN test results, using published CVN to fracture toughness correlation methods. However running cleavage fractures developed in the CTOD specimens at higher than predicted temperatures of 10 oC and 20 oC. These are typical service temperatures for structures in New Zealand and so are very likely to occur at the time of an earthquake. The implication from this is that there are levels of strain-hardening and conditions of material notching constraint that can lead to pre-necking and running fracture in New Zealand fabricated steel structures, under severe earthquake loading. Care was taken in the CTOD testing to monitor and maximise the capture of data electronically using a specially developed Direct Current Potential Drop method. This allowed the test results to be analysed and considered in varying ways, leading to a consistent assessment of the CTOD, crack growth, and the specific work of fracture in each test piece. While CTOD test results have sometimes been published by structural and welding engineering researchers in the wake of Kobe and Northridge, the results were typically of little use for this study as the CTOD initiation point was generally not identified effectively. The effect of remote plastic flow in the specimens was also not adequately accounted for. The CTOD test results were often simply used to help correlate other factors observed by the researchers. Side-grooving of specimens was not reported as having been used in any of the published results reviewed. When conducting CTOD test with highly ductile constructional steels it is very difficult to get useful CTOD results if the specimens are not side-grooved, as significant necking and tunnelling will otherwise occur and limit the usefulness of the results. Work by Knott and also by McRobie and Smith was seminal in terms of identifying some critical aspects of plane strain development in CTOD tests, and the links to non-metallic particle density with respect to fracture toughness and CTOD at initiation. Some of their findings with regards to the effect of pre-strain on CTOD initiation were subsequently found to confirm the experimental findings in this study. No effective methodology for prediction of pre-necking or running fracture in a structural member or assembly when subjected to gross plastic cyclic deformation was found to exist in the literature. It was concluded however that the principles of specific work of fracture, and monotonic and cyclic fracture similitude were particularly relevant. These were therefore utilised in the development of the design method proposed in this thesis. The CTOD test results were reviewed, isolating the remote plastic flow component, to determine the critical specific work of fracture property Rc of the steels tested. A meeting with Professor Kuwamura at the University of Tokyo was providential, allowing discussion of his similitude principle, and observations in person of some of the fractured specimens developed during his full scale test series’. Running fractures with cleavage were evident in the specimens, with their tell-tale chevron markings. He had predicted running fracture problems in structures in Japan ahead of the Kobe earthquake and been largely ignored. His insights were subsequently seriously considered in Japan after the earthquake. He and his colleagues developed the principle of structural similitude that relates monotonic fracture displacement ductility to cyclic fracture displacement ductility for a particular assembly. This arose from their observation that running fractures developed from ductile crack formation at blunt notches in structures. The similitude principle has echoes of the Coffin-Manson approach to ductile crack initiated low cycle fracture. The principle of similitude has a log–log relationship as does the Manson-Coffin relationship. So where notch plasticity controls the initiation of fracture in a structural assembly it is conceptually reasonable to expect that the number of cycles to initiation of fracture from a notch will have a log–log relationship to the amplitude of the cyclic strain developed in the notch. Kuwamura found that steel assemblies with lower CVN energy had reduced cyclic fracture endurance than the same assemblies made with steel with higher CVN impact energy. However no method of predicting performance of any particular assembly could be developed from his observations. The benefit of his method primarily relates to the minimising of testing necessary to assess the fracture limited cyclic displacement ductility of a structural assembly. However it doesn’t provide a means for designing a structural assembly to achieve specific levels of ductile endurance other than clearly identifying the need to use steel with good CVN characteristics. The most significant development arising from this thesis is therefore the development of a design method to assess cyclic ductile endurance. The method utilises the specific work of fracture properties obtained from CTOD specimens of the steel in conjunction with a relatively simple fracture mechanics assessment and an elasto-plastic finite element analysis (FEA). The FEA model is used to determine the displacement ductility of the assembly at the calculated onset of pre-necking fracture. The elasto-plastic stress–strain properties of the steel in various pre-strain states required for the FEA may be derived from tensile testing. Kuwamura’s similitude principle is then used to predict cyclic plastic endurance at various constant displacement ductility amplitudes. The method is extended using Miner’s rule to allow for the effects of increasing variable amplitude cyclic plastic loading. In summary the thesis explains why pre-necking and running fractures occur in steel members subjected to cyclic plastic deformation during a severe earthquake. In addition a method for consistently assessing the ability of structural steel assemblies to achieve a specified level of ductile endurance during earthquakes is proposed. The method is verified against published results for a cyclic test of a simple steel member with a crack at mid-span. / Whole document restricted, but available by request, use the feedback form to request access.

Les différences hommes/femmes sont un présupposé que nous considérons souvent comme une évidence. Les similitudes, en revanche, sont présentées comme des particularités propres à certains individus. Elles ne sont souvent pas perçues comme une caractéristique humaine intrinsèque. Il existe cependant un rapport entre similitude et différence. Ce qui sépare fondamentalement les similitudes des différences c'est le point de vue de l'observateur : celui qui décide ce à quoi il va accorder de l'importance, non pas en fonction de critères scientifiques et objectifs, mais plutôt en fonction de croyances et de présupposés. Les scientifiques, eux-mêmes, ne sont pas à l'abri de ce biais, et ce, en dépit du fait que les différences sont scientifiquement observables et mesurables. L'importance qu'on leur accorde n'est pas plus réelle que les similitudes auxquelles on accorde, en revanche, beaucoup moins de valeur, d'attention et que l'on étudie beaucoup moins, mais qui seraient néanmoins mesurables. / Gender differences are a presupposition that we often take for granted. Similarities, however, are presented as the particular traits of individuals. Often they are not perceived as an inherent human characteristic. But there is a relationship between similarity and difference. What fundamentally separates the similarities from the differences is the viewpoint of the observer: The importance accorded to the one or the other is not based on scientific and objectives criteria, but rather the observer's beliefs and assumptions. Scientists themselves are not immune to this bias, even though the differences are scientifically observable and measurable. The significance of differences is no more real than that of similarities which, nonetheless, are granted much less value, attention, and are less often studied, although they are measurable.

This thesis provides a structural and materials engineering explanation for many of the running fractures that occurred in steel structures during the destructive Kobe and Northridge earthquakes in the mid 1990s. A method is developed that allows the ductile endurance of structural steel members subjected to cyclic plastic deformation during earthquakes to be assessed and for pre-necking running fractures to be avoided. The study commenced following the 2000 World Earthquake Conference in Auckland. The conference brought together the findings of the huge research effort, in America, Japan, Europe and New Zealand, that followed the Kobe and Northridge earthquakes. The running fractures that had occurred in steel structures represented an unpredicted failure mode that structural engineers have not known how to predict or suppress through the engineering design process. A clear fundamental understanding of the causes and how to prevent the fractures did not arise from the conference. In fact apparently conflicting results were reported. Full scale cyclic tests in New Zealand on structural assemblies had not resulted in running fractures, whereas tests in American and Japan had. Structural engineers designing earthquake resistant structures rely on constructional steel to be materially homogeneous and nominally tri-linear in behaviour. Steel is expected to behave elastically under regular in-service loading, have a reliable and flat yield stress-strain characteristic, and under overload then develop predictable levels of strain-hardening in conjunction with significant plastic elongation up to its ultimate tensile strength. Steel is expected to eventually fracture after further plastic elongation and necking. Ductile design strategies and methods utilise the plastic elongation characteristics of steel to protect structures in earthquake. Plastic deformation is considered to beneficially dissipate energy generated in the structure by a severe earthquake and also dampen the structure’s response. The occurrence of running fracture without significant cyclic plastic deformation and before section necking in steelwork, therefore undermines the basis of the ductile seismic design approach. The initial part of the thesis is devoted to bringing together the fundamental aspects of materials engineering related to fracture of constructional steel. This is intended to provide a bridge of knowledge for structural engineering practitioners and researchers not fully conversant with materials engineering aspects of fracture. Fracture behaviour in steel is a broad and complex topic that developed rapidly in the twentieth century driven by the demands of technological growth. The unexpected fracture of welded liberty ships at sea in World War 2; the need for reliable long term containment for the nuclear reactors in the 1950s and 1960s; and prevention of fatigue failures in aircraft frames since the 1950s all drove engineering research into steel fracture behaviour. There are many subtle variations in definitions in the published literature on fracture that can be confusing. Therefore an attempt has been made to clarify terminology. The term brittle fracture in particular is only used in this thesis as applying to running fracture when the general or far field tensile stresses are below the yield stress of the steel. The term pre-necking or running fracture is preferred to describe the condition more broadly which may occur prior to and also after general yielding, but before section necking. Running fracture is a manifestation of pre-necking fracture in which insufficient plastic flow is available in the assembly to absorb the energy released upon fracture. The experimental studies investigated the behaviour of constructional steel commonly used in New Zealand, at various levels of plastic strain. This started with Charpy V-Notch (CVN) testing which revealed that a significant transition temperature shift and curve shape change occurs with increasing plastic strain and the associated strain-hardening. This showed that the ability of steel to avoid pre-necking or running fracture reduces as the level of plastic strain-hardening increases. Temperature controlled Crack Tip Opening Displacement (CTOD) testing was then undertaken. The setting of testing temperatures for the CTOD tests were guided by review of the CVN test results, using published CVN to fracture toughness correlation methods. However running cleavage fractures developed in the CTOD specimens at higher than predicted temperatures of 10 oC and 20 oC. These are typical service temperatures for structures in New Zealand and so are very likely to occur at the time of an earthquake. The implication from this is that there are levels of strain-hardening and conditions of material notching constraint that can lead to pre-necking and running fracture in New Zealand fabricated steel structures, under severe earthquake loading. Care was taken in the CTOD testing to monitor and maximise the capture of data electronically using a specially developed Direct Current Potential Drop method. This allowed the test results to be analysed and considered in varying ways, leading to a consistent assessment of the CTOD, crack growth, and the specific work of fracture in each test piece. While CTOD test results have sometimes been published by structural and welding engineering researchers in the wake of Kobe and Northridge, the results were typically of little use for this study as the CTOD initiation point was generally not identified effectively. The effect of remote plastic flow in the specimens was also not adequately accounted for. The CTOD test results were often simply used to help correlate other factors observed by the researchers. Side-grooving of specimens was not reported as having been used in any of the published results reviewed. When conducting CTOD test with highly ductile constructional steels it is very difficult to get useful CTOD results if the specimens are not side-grooved, as significant necking and tunnelling will otherwise occur and limit the usefulness of the results. Work by Knott and also by McRobie and Smith was seminal in terms of identifying some critical aspects of plane strain development in CTOD tests, and the links to non-metallic particle density with respect to fracture toughness and CTOD at initiation. Some of their findings with regards to the effect of pre-strain on CTOD initiation were subsequently found to confirm the experimental findings in this study. No effective methodology for prediction of pre-necking or running fracture in a structural member or assembly when subjected to gross plastic cyclic deformation was found to exist in the literature. It was concluded however that the principles of specific work of fracture, and monotonic and cyclic fracture similitude were particularly relevant. These were therefore utilised in the development of the design method proposed in this thesis. The CTOD test results were reviewed, isolating the remote plastic flow component, to determine the critical specific work of fracture property Rc of the steels tested. A meeting with Professor Kuwamura at the University of Tokyo was providential, allowing discussion of his similitude principle, and observations in person of some of the fractured specimens developed during his full scale test series’. Running fractures with cleavage were evident in the specimens, with their tell-tale chevron markings. He had predicted running fracture problems in structures in Japan ahead of the Kobe earthquake and been largely ignored. His insights were subsequently seriously considered in Japan after the earthquake. He and his colleagues developed the principle of structural similitude that relates monotonic fracture displacement ductility to cyclic fracture displacement ductility for a particular assembly. This arose from their observation that running fractures developed from ductile crack formation at blunt notches in structures. The similitude principle has echoes of the Coffin-Manson approach to ductile crack initiated low cycle fracture. The principle of similitude has a log–log relationship as does the Manson-Coffin relationship. So where notch plasticity controls the initiation of fracture in a structural assembly it is conceptually reasonable to expect that the number of cycles to initiation of fracture from a notch will have a log–log relationship to the amplitude of the cyclic strain developed in the notch. Kuwamura found that steel assemblies with lower CVN energy had reduced cyclic fracture endurance than the same assemblies made with steel with higher CVN impact energy. However no method of predicting performance of any particular assembly could be developed from his observations. The benefit of his method primarily relates to the minimising of testing necessary to assess the fracture limited cyclic displacement ductility of a structural assembly. However it doesn’t provide a means for designing a structural assembly to achieve specific levels of ductile endurance other than clearly identifying the need to use steel with good CVN characteristics. The most significant development arising from this thesis is therefore the development of a design method to assess cyclic ductile endurance. The method utilises the specific work of fracture properties obtained from CTOD specimens of the steel in conjunction with a relatively simple fracture mechanics assessment and an elasto-plastic finite element analysis (FEA). The FEA model is used to determine the displacement ductility of the assembly at the calculated onset of pre-necking fracture. The elasto-plastic stress–strain properties of the steel in various pre-strain states required for the FEA may be derived from tensile testing. Kuwamura’s similitude principle is then used to predict cyclic plastic endurance at various constant displacement ductility amplitudes. The method is extended using Miner’s rule to allow for the effects of increasing variable amplitude cyclic plastic loading. In summary the thesis explains why pre-necking and running fractures occur in steel members subjected to cyclic plastic deformation during a severe earthquake. In addition a method for consistently assessing the ability of structural steel assemblies to achieve a specified level of ductile endurance during earthquakes is proposed. The method is verified against published results for a cyclic test of a simple steel member with a crack at mid-span. / Whole document restricted, but available by request, use the feedback form to request access.

This thesis provides a structural and materials engineering explanation for many of the running fractures that occurred in steel structures during the destructive Kobe and Northridge earthquakes in the mid 1990s. A method is developed that allows the ductile endurance of structural steel members subjected to cyclic plastic deformation during earthquakes to be assessed and for pre-necking running fractures to be avoided. The study commenced following the 2000 World Earthquake Conference in Auckland. The conference brought together the findings of the huge research effort, in America, Japan, Europe and New Zealand, that followed the Kobe and Northridge earthquakes. The running fractures that had occurred in steel structures represented an unpredicted failure mode that structural engineers have not known how to predict or suppress through the engineering design process. A clear fundamental understanding of the causes and how to prevent the fractures did not arise from the conference. In fact apparently conflicting results were reported. Full scale cyclic tests in New Zealand on structural assemblies had not resulted in running fractures, whereas tests in American and Japan had. Structural engineers designing earthquake resistant structures rely on constructional steel to be materially homogeneous and nominally tri-linear in behaviour. Steel is expected to behave elastically under regular in-service loading, have a reliable and flat yield stress-strain characteristic, and under overload then develop predictable levels of strain-hardening in conjunction with significant plastic elongation up to its ultimate tensile strength. Steel is expected to eventually fracture after further plastic elongation and necking. Ductile design strategies and methods utilise the plastic elongation characteristics of steel to protect structures in earthquake. Plastic deformation is considered to beneficially dissipate energy generated in the structure by a severe earthquake and also dampen the structure’s response. The occurrence of running fracture without significant cyclic plastic deformation and before section necking in steelwork, therefore undermines the basis of the ductile seismic design approach. The initial part of the thesis is devoted to bringing together the fundamental aspects of materials engineering related to fracture of constructional steel. This is intended to provide a bridge of knowledge for structural engineering practitioners and researchers not fully conversant with materials engineering aspects of fracture. Fracture behaviour in steel is a broad and complex topic that developed rapidly in the twentieth century driven by the demands of technological growth. The unexpected fracture of welded liberty ships at sea in World War 2; the need for reliable long term containment for the nuclear reactors in the 1950s and 1960s; and prevention of fatigue failures in aircraft frames since the 1950s all drove engineering research into steel fracture behaviour. There are many subtle variations in definitions in the published literature on fracture that can be confusing. Therefore an attempt has been made to clarify terminology. The term brittle fracture in particular is only used in this thesis as applying to running fracture when the general or far field tensile stresses are below the yield stress of the steel. The term pre-necking or running fracture is preferred to describe the condition more broadly which may occur prior to and also after general yielding, but before section necking. Running fracture is a manifestation of pre-necking fracture in which insufficient plastic flow is available in the assembly to absorb the energy released upon fracture. The experimental studies investigated the behaviour of constructional steel commonly used in New Zealand, at various levels of plastic strain. This started with Charpy V-Notch (CVN) testing which revealed that a significant transition temperature shift and curve shape change occurs with increasing plastic strain and the associated strain-hardening. This showed that the ability of steel to avoid pre-necking or running fracture reduces as the level of plastic strain-hardening increases. Temperature controlled Crack Tip Opening Displacement (CTOD) testing was then undertaken. The setting of testing temperatures for the CTOD tests were guided by review of the CVN test results, using published CVN to fracture toughness correlation methods. However running cleavage fractures developed in the CTOD specimens at higher than predicted temperatures of 10 oC and 20 oC. These are typical service temperatures for structures in New Zealand and so are very likely to occur at the time of an earthquake. The implication from this is that there are levels of strain-hardening and conditions of material notching constraint that can lead to pre-necking and running fracture in New Zealand fabricated steel structures, under severe earthquake loading. Care was taken in the CTOD testing to monitor and maximise the capture of data electronically using a specially developed Direct Current Potential Drop method. This allowed the test results to be analysed and considered in varying ways, leading to a consistent assessment of the CTOD, crack growth, and the specific work of fracture in each test piece. While CTOD test results have sometimes been published by structural and welding engineering researchers in the wake of Kobe and Northridge, the results were typically of little use for this study as the CTOD initiation point was generally not identified effectively. The effect of remote plastic flow in the specimens was also not adequately accounted for. The CTOD test results were often simply used to help correlate other factors observed by the researchers. Side-grooving of specimens was not reported as having been used in any of the published results reviewed. When conducting CTOD test with highly ductile constructional steels it is very difficult to get useful CTOD results if the specimens are not side-grooved, as significant necking and tunnelling will otherwise occur and limit the usefulness of the results. Work by Knott and also by McRobie and Smith was seminal in terms of identifying some critical aspects of plane strain development in CTOD tests, and the links to non-metallic particle density with respect to fracture toughness and CTOD at initiation. Some of their findings with regards to the effect of pre-strain on CTOD initiation were subsequently found to confirm the experimental findings in this study. No effective methodology for prediction of pre-necking or running fracture in a structural member or assembly when subjected to gross plastic cyclic deformation was found to exist in the literature. It was concluded however that the principles of specific work of fracture, and monotonic and cyclic fracture similitude were particularly relevant. These were therefore utilised in the development of the design method proposed in this thesis. The CTOD test results were reviewed, isolating the remote plastic flow component, to determine the critical specific work of fracture property Rc of the steels tested. A meeting with Professor Kuwamura at the University of Tokyo was providential, allowing discussion of his similitude principle, and observations in person of some of the fractured specimens developed during his full scale test series’. Running fractures with cleavage were evident in the specimens, with their tell-tale chevron markings. He had predicted running fracture problems in structures in Japan ahead of the Kobe earthquake and been largely ignored. His insights were subsequently seriously considered in Japan after the earthquake. He and his colleagues developed the principle of structural similitude that relates monotonic fracture displacement ductility to cyclic fracture displacement ductility for a particular assembly. This arose from their observation that running fractures developed from ductile crack formation at blunt notches in structures. The similitude principle has echoes of the Coffin-Manson approach to ductile crack initiated low cycle fracture. The principle of similitude has a log–log relationship as does the Manson-Coffin relationship. So where notch plasticity controls the initiation of fracture in a structural assembly it is conceptually reasonable to expect that the number of cycles to initiation of fracture from a notch will have a log–log relationship to the amplitude of the cyclic strain developed in the notch. Kuwamura found that steel assemblies with lower CVN energy had reduced cyclic fracture endurance than the same assemblies made with steel with higher CVN impact energy. However no method of predicting performance of any particular assembly could be developed from his observations. The benefit of his method primarily relates to the minimising of testing necessary to assess the fracture limited cyclic displacement ductility of a structural assembly. However it doesn’t provide a means for designing a structural assembly to achieve specific levels of ductile endurance other than clearly identifying the need to use steel with good CVN characteristics. The most significant development arising from this thesis is therefore the development of a design method to assess cyclic ductile endurance. The method utilises the specific work of fracture properties obtained from CTOD specimens of the steel in conjunction with a relatively simple fracture mechanics assessment and an elasto-plastic finite element analysis (FEA). The FEA model is used to determine the displacement ductility of the assembly at the calculated onset of pre-necking fracture. The elasto-plastic stress–strain properties of the steel in various pre-strain states required for the FEA may be derived from tensile testing. Kuwamura’s similitude principle is then used to predict cyclic plastic endurance at various constant displacement ductility amplitudes. The method is extended using Miner’s rule to allow for the effects of increasing variable amplitude cyclic plastic loading. In summary the thesis explains why pre-necking and running fractures occur in steel members subjected to cyclic plastic deformation during a severe earthquake. In addition a method for consistently assessing the ability of structural steel assemblies to achieve a specified level of ductile endurance during earthquakes is proposed. The method is verified against published results for a cyclic test of a simple steel member with a crack at mid-span. / Whole document restricted, but available by request, use the feedback form to request access.

This thesis provides a structural and materials engineering explanation for many of the running fractures that occurred in steel structures during the destructive Kobe and Northridge earthquakes in the mid 1990s. A method is developed that allows the ductile endurance of structural steel members subjected to cyclic plastic deformation during earthquakes to be assessed and for pre-necking running fractures to be avoided. The study commenced following the 2000 World Earthquake Conference in Auckland. The conference brought together the findings of the huge research effort, in America, Japan, Europe and New Zealand, that followed the Kobe and Northridge earthquakes. The running fractures that had occurred in steel structures represented an unpredicted failure mode that structural engineers have not known how to predict or suppress through the engineering design process. A clear fundamental understanding of the causes and how to prevent the fractures did not arise from the conference. In fact apparently conflicting results were reported. Full scale cyclic tests in New Zealand on structural assemblies had not resulted in running fractures, whereas tests in American and Japan had. Structural engineers designing earthquake resistant structures rely on constructional steel to be materially homogeneous and nominally tri-linear in behaviour. Steel is expected to behave elastically under regular in-service loading, have a reliable and flat yield stress-strain characteristic, and under overload then develop predictable levels of strain-hardening in conjunction with significant plastic elongation up to its ultimate tensile strength. Steel is expected to eventually fracture after further plastic elongation and necking. Ductile design strategies and methods utilise the plastic elongation characteristics of steel to protect structures in earthquake. Plastic deformation is considered to beneficially dissipate energy generated in the structure by a severe earthquake and also dampen the structure’s response. The occurrence of running fracture without significant cyclic plastic deformation and before section necking in steelwork, therefore undermines the basis of the ductile seismic design approach. The initial part of the thesis is devoted to bringing together the fundamental aspects of materials engineering related to fracture of constructional steel. This is intended to provide a bridge of knowledge for structural engineering practitioners and researchers not fully conversant with materials engineering aspects of fracture. Fracture behaviour in steel is a broad and complex topic that developed rapidly in the twentieth century driven by the demands of technological growth. The unexpected fracture of welded liberty ships at sea in World War 2; the need for reliable long term containment for the nuclear reactors in the 1950s and 1960s; and prevention of fatigue failures in aircraft frames since the 1950s all drove engineering research into steel fracture behaviour. There are many subtle variations in definitions in the published literature on fracture that can be confusing. Therefore an attempt has been made to clarify terminology. The term brittle fracture in particular is only used in this thesis as applying to running fracture when the general or far field tensile stresses are below the yield stress of the steel. The term pre-necking or running fracture is preferred to describe the condition more broadly which may occur prior to and also after general yielding, but before section necking. Running fracture is a manifestation of pre-necking fracture in which insufficient plastic flow is available in the assembly to absorb the energy released upon fracture. The experimental studies investigated the behaviour of constructional steel commonly used in New Zealand, at various levels of plastic strain. This started with Charpy V-Notch (CVN) testing which revealed that a significant transition temperature shift and curve shape change occurs with increasing plastic strain and the associated strain-hardening. This showed that the ability of steel to avoid pre-necking or running fracture reduces as the level of plastic strain-hardening increases. Temperature controlled Crack Tip Opening Displacement (CTOD) testing was then undertaken. The setting of testing temperatures for the CTOD tests were guided by review of the CVN test results, using published CVN to fracture toughness correlation methods. However running cleavage fractures developed in the CTOD specimens at higher than predicted temperatures of 10 oC and 20 oC. These are typical service temperatures for structures in New Zealand and so are very likely to occur at the time of an earthquake. The implication from this is that there are levels of strain-hardening and conditions of material notching constraint that can lead to pre-necking and running fracture in New Zealand fabricated steel structures, under severe earthquake loading. Care was taken in the CTOD testing to monitor and maximise the capture of data electronically using a specially developed Direct Current Potential Drop method. This allowed the test results to be analysed and considered in varying ways, leading to a consistent assessment of the CTOD, crack growth, and the specific work of fracture in each test piece. While CTOD test results have sometimes been published by structural and welding engineering researchers in the wake of Kobe and Northridge, the results were typically of little use for this study as the CTOD initiation point was generally not identified effectively. The effect of remote plastic flow in the specimens was also not adequately accounted for. The CTOD test results were often simply used to help correlate other factors observed by the researchers. Side-grooving of specimens was not reported as having been used in any of the published results reviewed. When conducting CTOD test with highly ductile constructional steels it is very difficult to get useful CTOD results if the specimens are not side-grooved, as significant necking and tunnelling will otherwise occur and limit the usefulness of the results. Work by Knott and also by McRobie and Smith was seminal in terms of identifying some critical aspects of plane strain development in CTOD tests, and the links to non-metallic particle density with respect to fracture toughness and CTOD at initiation. Some of their findings with regards to the effect of pre-strain on CTOD initiation were subsequently found to confirm the experimental findings in this study. No effective methodology for prediction of pre-necking or running fracture in a structural member or assembly when subjected to gross plastic cyclic deformation was found to exist in the literature. It was concluded however that the principles of specific work of fracture, and monotonic and cyclic fracture similitude were particularly relevant. These were therefore utilised in the development of the design method proposed in this thesis. The CTOD test results were reviewed, isolating the remote plastic flow component, to determine the critical specific work of fracture property Rc of the steels tested. A meeting with Professor Kuwamura at the University of Tokyo was providential, allowing discussion of his similitude principle, and observations in person of some of the fractured specimens developed during his full scale test series’. Running fractures with cleavage were evident in the specimens, with their tell-tale chevron markings. He had predicted running fracture problems in structures in Japan ahead of the Kobe earthquake and been largely ignored. His insights were subsequently seriously considered in Japan after the earthquake. He and his colleagues developed the principle of structural similitude that relates monotonic fracture displacement ductility to cyclic fracture displacement ductility for a particular assembly. This arose from their observation that running fractures developed from ductile crack formation at blunt notches in structures. The similitude principle has echoes of the Coffin-Manson approach to ductile crack initiated low cycle fracture. The principle of similitude has a log–log relationship as does the Manson-Coffin relationship. So where notch plasticity controls the initiation of fracture in a structural assembly it is conceptually reasonable to expect that the number of cycles to initiation of fracture from a notch will have a log–log relationship to the amplitude of the cyclic strain developed in the notch. Kuwamura found that steel assemblies with lower CVN energy had reduced cyclic fracture endurance than the same assemblies made with steel with higher CVN impact energy. However no method of predicting performance of any particular assembly could be developed from his observations. The benefit of his method primarily relates to the minimising of testing necessary to assess the fracture limited cyclic displacement ductility of a structural assembly. However it doesn’t provide a means for designing a structural assembly to achieve specific levels of ductile endurance other than clearly identifying the need to use steel with good CVN characteristics. The most significant development arising from this thesis is therefore the development of a design method to assess cyclic ductile endurance. The method utilises the specific work of fracture properties obtained from CTOD specimens of the steel in conjunction with a relatively simple fracture mechanics assessment and an elasto-plastic finite element analysis (FEA). The FEA model is used to determine the displacement ductility of the assembly at the calculated onset of pre-necking fracture. The elasto-plastic stress–strain properties of the steel in various pre-strain states required for the FEA may be derived from tensile testing. Kuwamura’s similitude principle is then used to predict cyclic plastic endurance at various constant displacement ductility amplitudes. The method is extended using Miner’s rule to allow for the effects of increasing variable amplitude cyclic plastic loading. In summary the thesis explains why pre-necking and running fractures occur in steel members subjected to cyclic plastic deformation during a severe earthquake. In addition a method for consistently assessing the ability of structural steel assemblies to achieve a specified level of ductile endurance during earthquakes is proposed. The method is verified against published results for a cyclic test of a simple steel member with a crack at mid-span. / Whole document restricted, but available by request, use the feedback form to request access.

This thesis provides a structural and materials engineering explanation for many of the running fractures that occurred in steel structures during the destructive Kobe and Northridge earthquakes in the mid 1990s. A method is developed that allows the ductile endurance of structural steel members subjected to cyclic plastic deformation during earthquakes to be assessed and for pre-necking running fractures to be avoided. The study commenced following the 2000 World Earthquake Conference in Auckland. The conference brought together the findings of the huge research effort, in America, Japan, Europe and New Zealand, that followed the Kobe and Northridge earthquakes. The running fractures that had occurred in steel structures represented an unpredicted failure mode that structural engineers have not known how to predict or suppress through the engineering design process. A clear fundamental understanding of the causes and how to prevent the fractures did not arise from the conference. In fact apparently conflicting results were reported. Full scale cyclic tests in New Zealand on structural assemblies had not resulted in running fractures, whereas tests in American and Japan had. Structural engineers designing earthquake resistant structures rely on constructional steel to be materially homogeneous and nominally tri-linear in behaviour. Steel is expected to behave elastically under regular in-service loading, have a reliable and flat yield stress-strain characteristic, and under overload then develop predictable levels of strain-hardening in conjunction with significant plastic elongation up to its ultimate tensile strength. Steel is expected to eventually fracture after further plastic elongation and necking. Ductile design strategies and methods utilise the plastic elongation characteristics of steel to protect structures in earthquake. Plastic deformation is considered to beneficially dissipate energy generated in the structure by a severe earthquake and also dampen the structure’s response. The occurrence of running fracture without significant cyclic plastic deformation and before section necking in steelwork, therefore undermines the basis of the ductile seismic design approach. The initial part of the thesis is devoted to bringing together the fundamental aspects of materials engineering related to fracture of constructional steel. This is intended to provide a bridge of knowledge for structural engineering practitioners and researchers not fully conversant with materials engineering aspects of fracture. Fracture behaviour in steel is a broad and complex topic that developed rapidly in the twentieth century driven by the demands of technological growth. The unexpected fracture of welded liberty ships at sea in World War 2; the need for reliable long term containment for the nuclear reactors in the 1950s and 1960s; and prevention of fatigue failures in aircraft frames since the 1950s all drove engineering research into steel fracture behaviour. There are many subtle variations in definitions in the published literature on fracture that can be confusing. Therefore an attempt has been made to clarify terminology. The term brittle fracture in particular is only used in this thesis as applying to running fracture when the general or far field tensile stresses are below the yield stress of the steel. The term pre-necking or running fracture is preferred to describe the condition more broadly which may occur prior to and also after general yielding, but before section necking. Running fracture is a manifestation of pre-necking fracture in which insufficient plastic flow is available in the assembly to absorb the energy released upon fracture. The experimental studies investigated the behaviour of constructional steel commonly used in New Zealand, at various levels of plastic strain. This started with Charpy V-Notch (CVN) testing which revealed that a significant transition temperature shift and curve shape change occurs with increasing plastic strain and the associated strain-hardening. This showed that the ability of steel to avoid pre-necking or running fracture reduces as the level of plastic strain-hardening increases. Temperature controlled Crack Tip Opening Displacement (CTOD) testing was then undertaken. The setting of testing temperatures for the CTOD tests were guided by review of the CVN test results, using published CVN to fracture toughness correlation methods. However running cleavage fractures developed in the CTOD specimens at higher than predicted temperatures of 10 oC and 20 oC. These are typical service temperatures for structures in New Zealand and so are very likely to occur at the time of an earthquake. The implication from this is that there are levels of strain-hardening and conditions of material notching constraint that can lead to pre-necking and running fracture in New Zealand fabricated steel structures, under severe earthquake loading. Care was taken in the CTOD testing to monitor and maximise the capture of data electronically using a specially developed Direct Current Potential Drop method. This allowed the test results to be analysed and considered in varying ways, leading to a consistent assessment of the CTOD, crack growth, and the specific work of fracture in each test piece. While CTOD test results have sometimes been published by structural and welding engineering researchers in the wake of Kobe and Northridge, the results were typically of little use for this study as the CTOD initiation point was generally not identified effectively. The effect of remote plastic flow in the specimens was also not adequately accounted for. The CTOD test results were often simply used to help correlate other factors observed by the researchers. Side-grooving of specimens was not reported as having been used in any of the published results reviewed. When conducting CTOD test with highly ductile constructional steels it is very difficult to get useful CTOD results if the specimens are not side-grooved, as significant necking and tunnelling will otherwise occur and limit the usefulness of the results. Work by Knott and also by McRobie and Smith was seminal in terms of identifying some critical aspects of plane strain development in CTOD tests, and the links to non-metallic particle density with respect to fracture toughness and CTOD at initiation. Some of their findings with regards to the effect of pre-strain on CTOD initiation were subsequently found to confirm the experimental findings in this study. No effective methodology for prediction of pre-necking or running fracture in a structural member or assembly when subjected to gross plastic cyclic deformation was found to exist in the literature. It was concluded however that the principles of specific work of fracture, and monotonic and cyclic fracture similitude were particularly relevant. These were therefore utilised in the development of the design method proposed in this thesis. The CTOD test results were reviewed, isolating the remote plastic flow component, to determine the critical specific work of fracture property Rc of the steels tested. A meeting with Professor Kuwamura at the University of Tokyo was providential, allowing discussion of his similitude principle, and observations in person of some of the fractured specimens developed during his full scale test series’. Running fractures with cleavage were evident in the specimens, with their tell-tale chevron markings. He had predicted running fracture problems in structures in Japan ahead of the Kobe earthquake and been largely ignored. His insights were subsequently seriously considered in Japan after the earthquake. He and his colleagues developed the principle of structural similitude that relates monotonic fracture displacement ductility to cyclic fracture displacement ductility for a particular assembly. This arose from their observation that running fractures developed from ductile crack formation at blunt notches in structures. The similitude principle has echoes of the Coffin-Manson approach to ductile crack initiated low cycle fracture. The principle of similitude has a log–log relationship as does the Manson-Coffin relationship. So where notch plasticity controls the initiation of fracture in a structural assembly it is conceptually reasonable to expect that the number of cycles to initiation of fracture from a notch will have a log–log relationship to the amplitude of the cyclic strain developed in the notch. Kuwamura found that steel assemblies with lower CVN energy had reduced cyclic fracture endurance than the same assemblies made with steel with higher CVN impact energy. However no method of predicting performance of any particular assembly could be developed from his observations. The benefit of his method primarily relates to the minimising of testing necessary to assess the fracture limited cyclic displacement ductility of a structural assembly. However it doesn’t provide a means for designing a structural assembly to achieve specific levels of ductile endurance other than clearly identifying the need to use steel with good CVN characteristics. The most significant development arising from this thesis is therefore the development of a design method to assess cyclic ductile endurance. The method utilises the specific work of fracture properties obtained from CTOD specimens of the steel in conjunction with a relatively simple fracture mechanics assessment and an elasto-plastic finite element analysis (FEA). The FEA model is used to determine the displacement ductility of the assembly at the calculated onset of pre-necking fracture. The elasto-plastic stress–strain properties of the steel in various pre-strain states required for the FEA may be derived from tensile testing. Kuwamura’s similitude principle is then used to predict cyclic plastic endurance at various constant displacement ductility amplitudes. The method is extended using Miner’s rule to allow for the effects of increasing variable amplitude cyclic plastic loading. In summary the thesis explains why pre-necking and running fractures occur in steel members subjected to cyclic plastic deformation during a severe earthquake. In addition a method for consistently assessing the ability of structural steel assemblies to achieve a specified level of ductile endurance during earthquakes is proposed. The method is verified against published results for a cyclic test of a simple steel member with a crack at mid-span. / Whole document restricted, but available by request, use the feedback form to request access.

[Truncated abstract] To date, many subsurface contaminant modelling studies have focused on increasing model complexity and measurement requirements to improve model accuracy and widen model application. However, due to the highly complex and heterogeneous nature of flow in the subsurface, the greater benefit in model development may lie in decreasing complexity by identifying key processes and parameters, simplifying the relationships that exist between them, and incorporating these relationships into simple models that recognise or quantify the inherent complexity and uncertainty. To address this need, this study aims to identify and isolate the key processes and parameters that control dense nonaqueous phase liquid (DNAPL) and aqueous phase migration through single, onedimensional fractures. This is a theoretical representation which allows the study of processes through conceptual and mathematical models. Fracture systems typically consist of multiple two-dimensional fractures in a three-dimensional network; however, these systems are computationally and conceptually demanding to investigate and were outside of the scope of this study. This work initially focuses on DNAPL migration in single, one-dimensional fractures. The similitude techniques of dimensional and inspectional analysis are performed to simplify the system and to develop breakthrough time scale factors. This approach relies heavily on the limitations of the equation used for the analysis and on the difficulty in representing variable aperture scenarios. The complexity of the conceptual model is then increased by embedding the fracture in a two-dimensional, porous matrix. ... These tools can be readily applied by the field investigator or computer modeller to make order-of-magnitude estimates of breakthrough times, reduce or target measurement requirements, and lessen the need to employ numerical multiphase flow models. To determine the implications of the results found in the one-dimensional studies to applications at the field scale, the complexity of the conceptual model was increased to a single, two-dimensional, planar fracture embedded in a three-dimensional porous matrix. The focus of this study was not DNAPL breakthrough times but the relative importance and interaction of different mass transport processes and parameters on plume migration and evolution. Observations clearly show that estimates of the size, location and concentration of the plume is highly dependent on the geologic media, the temporal and spatial location and resolution of measurements, and on the history, mass and location of the DNAPL source. In addition, the processes controlling mass transport (especially matrix diffusion and back diffusion) act in combination at the field scale in ways not always expected from an analysis of processes acting individually at smaller spatial and temporal scales. Serious concerns over the application of the common '1% Rule of Thumb' to predict DNAPL presence and the use of remediation efforts that rely largely on natural attenuation are raised. These findings have major implications for the field worker and computer modeller, and any characterisation, monitoring or remediation program development needs to be sensitive to these findings.

Les futures missions d'exploration spatiale nécessitent le stockage d'ergols cryogéniques sur de longues durées.Sous l'effet d'entrées thermiques résiduelles, l'ergol se vaporise et le réservoir s'auto-pressurise, pouvant entraîner la rupture du réservoir pour des missionssuffisamment longues. Cette thèse s'intéresse à un système de contrôle, appelé Thermodynamic Venting System (TVS), reposant sur l'injection d'un jet sous-refroidi dans le réservoir.L'injection entraîne la condensation de la vapeur, la déstratification du bain liquide et donc une baisse de pression dans le réservoir.L'étude expérimentale a permis de mettre en place une technique originale d'isolation active générant une condition de paroi à flux de chaleurnet nul. Une base de données expérimentales, d'auto-pressurisation et de contrôle TVS, a été constituée avec ce nouveau dispositif.Elle a notamment permis de valider un modèle thermodynamique homogène permettant de prédire l'évolution de température et depression dans le réservoir. Ce modèle a été étendu pour de façon à décrire le comportement de tous les éléments constitutifs du système TVS.Un outil de dimensionnement du système complet a ainsi été mis en place.En le couplant à une plateforme d'optimisation, un système TVS optimal a pu être établi pour une mission de démonstration.Enfin une étude numérique a permis de mettre en évidence, pour notre cas d'étude, les faiblesses des modèles de changement de phase disponibles dans les solveursCFD commerciaux. Une modélisation prédictive du changement de phase a été mise en place dans un code de calcul recherche puis validée sur uncas académique 1D. / Future operations in space exploration require the ability to store cryogens for long duration. Residual heat loads induce cryogenic propellant vaporization andtank self-pressurization (SP), eventually leading to storage failure for long enough mission duration.This thesis focuses on a control strategy, called Thermodynamic Venting System (TVS), based on a recirculating liquid subcooled injection. The injection results inan ullage condensation, a liquid bath destratification and thus a tank pressure reduction.Experimentally, an original active insulation technique has been set up, yielding a net zero heat flux wall boundary condition. A data base ofself-pressurisation and TVS control experiments has been gathered with this new aparatus.It was used to validate an homogeneous thermodynamic model providing a fast prediction of tank temperature and pressure during control.This model has been extended to discribe the TVS system behaviour including all its components. This full system design tool has been coupled with an optimisationplatform and an optimal TVS design has been established for a demonstration mission.Furthermore, a numerical study has evidenced the weakness of commercial CFD software to simulate phase change, for TVS configuration.A predictive phase change formulation has been set up in a home-made software and validated on a 1D academic case.

Les bâtiments résidentiels et industriels munis d’un réseau de ventilation constituent des installations complexes, susceptibles d’être le siège de transferts de masse et d’énergie, selon les situations de fonctionnement. Afin d’étudier ces transferts de masse, une méthodologie permettant d’établir des expérimentations à échelle réduite pour l’étude des écoulements anisothermes a été développée. Cette méthodologie a été validée numériquement, puis appliquée à une configuration de référence, représentative du principe de fonctionnement des réseaux de ventilation qui sont rencontrés dans le domaine nucléaire. Les influences du vent et des phénomènes thermiques sur les transferts de masse au sein de cette configuration ont été étudiées dans la soufflerie climatique Jules Verne du CSTB pour différentes situations de fonctionnement du réseau de ventilation (ventilation en fonctionnement normal, arrêt de la ventilation ou régime de sauvegarde) et des scénarios de dégagement de chaleur. Ces sources thermiques peuvent être issues d’un processus industriel ou d’un incendie. Elles ont été reproduites expérimentalement par une injection d’hélium. Les effets des sources thermiques couplées ou non au vent sur les pertes ponctuelles ou totales du confinement des locaux ont été mis en évidence et analysés. La robustesse du code à zones SYLVIA, utilisé notamment pour appuyer les évaluations de sûreté des installations nucléaires, a été analysée à partir des résultats expérimentaux. La prise en compte des phénomènes physiques observés expérimentalement a été validée. Les inversions des débits de fuite causées par les phénomènes thermiques ont été reproduites avec le code SYLVIA. Une comparaison entre les calculs où la source de chaleur a été simulée avec une injection d’hélium et avec une puissance thermique a permis d’observer l’impact de l’injection de masse causé par l’hélium sur les pressions, les débits et les températures. / Residential and industrial buildings equipped with a ventilation system are complex facilities, where heat and mass transfers could occur according to the operating conditions. In order to study these mass transfers, a methodology has been developed to reduced-scale experimentations for non isothermal flows study. This methodology has been numerically validated, and then applied to a standard configuration, representing of the ventilation systems operating principle which are encounter in the nuclear field. The wind and the thermal phenomena influences on the mass transfers inside this configuration have been studied in the Jules Verne climatic wind tunnel of the CSTB for various operating ventilation system situations (normal operating ventilation system, stopping ventilation or protection rate of productivity) and scenarios of heat supply. These thermal sources can be generated by an industrial process or a fire. They have been reproduced experimentally with an helium injection. The effects of the heat sources coupled or not with wind on loss of building containment were highlighted and analyzed. The reliability of the zonal code SYLVIA, used notably to support safety assessment in nuclear buildings, has been analyzed from these experimental results. The modelling of the physical phenomena experimentally observed has been validated. The leakage flowrates reversals have been retrieved with the SYLVIA code. A comparison between the calculations where the heat source has been simulated with an helium injection and with a thermal power permitted to observe the mass injection effect has been caused by the helium on the pressures, the flowrates and the temperatures.