Tesis doctoral de Cristina Ibáñez Llano
The research of this thesis lies in the framework of the improvement of the mathematical and algorithmic foundations of the analytical quantification techniques of probabilistic safety assessments (psa) by means of the binary decision diagram (bdd) technique, trying to overcome the difficulties arising from the application of this technique with real industrial models. It has also explored how to embed these techniques into the dynamic and integrated approaches of safety analysis required to address the safety problem in a consistent and comprehensive manner. probabilistic safety assessments (psa) is a well-established technique for integrating various reliability models and to numerically quantify the frequency of damage in nuclear facilities. Its use is now widespread in nuclear regulation and it complements traditional deterministic analysis, providing a comprehensive and structured approach to identifying undesired accident scenarios, computing their likelihood in terms of occurrence frequency, and assessing the consequences and mitigation strategies. In terms of modelling methodologies, psa relies on the use of boolean models, in particular, in the combination of fault trees and event trees (ft/et). With such models, it is possible to schematically represent the causes and consequences of damage of complex accident sequences. most of the computational tools available for quantifying large psa models are unable to produce analytically correct results, since they rely on several simplifications and approximations. These tools have implemented what is often called the classical mcs approach, namely the kinetic tree theory. This approach is based on the computation of the minimal cutsets (mcs for short) by means of boolean reduction of the logic equation of the ft/et models and the neglection of those parts that have a minor contribution to the evaluation of the model in order to avoid a memory explosion of the computations due to the level of detail of the boolean models. In addition, negative logic is avoided or treated in an approximate way. binary decision diagrams (bdd) are a well-known alternative to the minimal cutsets approach to assess boolean models. After more than two decades of application, the bdd methodology has been applied successfully to improve fault tree assessment and its introduction in the field has permitted renewing its algorithmic framework, leading to important improvements. The main advantage of these compact data structures is that they can address the problems derived from the approximations performed by current techniques, providing more efficient and accurate analyses. Conversely to the classical methodology based on the computation of the mcs, the bdd approach involves no approximation in the quantification of the model and is able to handle correctly negative logic (success branches) at low additional complexity cost. However bdds are also subject to combinatorial explosion as the final size of the bdd is very sensitive to the variable ordering needed to convert the model into it. however, and despite remarkable advances, the application of this technique to solve large models, and in particular, the fault trees and event trees coming from the psa studies of the nuclear industry, which includes several thousand of basic events and logic gates, remains to the date out of reach of a full automatic treatment. In most of the cases, it becomes impossible to fully convert the model into a bdd without considering truncation or simplification of the model due to the explosion in memory requirements. despite the computational advances experienced in recent decades, which have allowed having more powerful computers, and the improvements and refinements introduced in the algorithms and software tools themselves, the large size and complexity of current psa models places the computational efforts close to computer limits. In view of the limitations arising from both quantification approaches, further developments and new approaches are required to improve, adequate and consolidate the bdd technique as a feasible solution for psa assessment in the nuclear field. the thesis proposes a new hybrid methodology based on combining the information provided by the classical mcs approach using truncation limits with the bdd approach, aimed to conciliate both existing approaches and make the application of the bdd technique feasible for real industrial problems. The core of the methodology lies on applying several reduction procedures to the model based on syntactic transformations in order to explicitly reduce its complexity and therefore make the construction of the bdd feasible while controlling both the degree and the tendency of the reduction performed, allowing to obtain a better control over the impact of the simplification on the final quantification results and to properly account for the success branches, which is very important for the correct treatment of non-coherent models. this new hybrid methodology has been combined with an incremental approach to compute the bdd of the sequence that is compatible with the dynamic generation of event trees and therefore amenable and adaptable to dynamic developments and extensions of psa studies. the added value of this methodology is that it is possible to ensure a real confidence interval of the exact value by computing two bounding transformations that converge to it. Subsequently, explicit knowledge of the error bound is obtained. Moreover, it can be used to measure the acceptability of the results obtained with the traditional techniques. the methodology presented has been mathematically founded and implemented in a computer tool which has been developed as part of the research carried out for this thesis. The probdd tool is a software tool based on the bdd formalism dedicated to the analysis and probabilistic quantification of the fault trees and event trees arising in the psa studies. the proposed method has been applied to a case study extracted from a real psa model of a spanish bwr power plant. The results obtained confirm the applicability of the methodology to estimate the exact results correctly, therefore confirming the soundness of the ideas developed in this thesis and the ultimate confirmation of the feasibility and adequacy of the methodology developed. The interest and relevance of this task not only lies in the validation of the proposed methodology from an academic point of view, but also in the fact that it is the first application of the bdd technology to a real psa study in the spanish nuclear context.
Datos académicos de la tesis doctoral «Aplicación de los diagramas binarios de decisión al análisis integrado de seguridad«
- Título de la tesis: Aplicación de los diagramas binarios de decisión al análisis integrado de seguridad
- Autor: Cristina Ibáñez Llano
- Universidad: Pontificia comillas
- Fecha de lectura de la tesis: 17/09/2010
Dirección y tribunal
- Director de la tesis
- Francisco Nieto Fuentes
- Tribunal
- Presidente del tribunal: sebastian salvador Martorell alsina
- antonie Rauzy (vocal)
- Luis Gonzalez sanchez (vocal)
- José Mira mcwilliams (vocal)