Efficient reliability analysis of complex systems

点击翻译按钮获取中文摘要

In this project a theoretical and algorithmic basis is developed in order to analyse the time dependent reliability of complex systems of industrial size with high numerical efficiency and close to reality. These fundamental developments are particularly important in order to cope with the rapidly growing complexity of our industrial systems and to ensure their reliability. Those systems, such as supply and sewage systems, industrial and production facilities, control systems in buildings or on construction sites or in vehicles, power plants or structural systems, are of increasing criticality for the functionality of our society and daily life. The planned project is devoted to securing the reliability of such systems. The envisaged development addresses the key discrepancy of the current methods and technologies; for systems of industrial size they require a critical compromise between very strong simplifications of the model and an infeasible numerical effort associated with a detailed modelling. This discrepancy is rapidly growing with the complexity of systems and, thus, causes growing uncertainties in the modelling, in addition.The concept of the survival signature forms the basis for the development. This concept facilitates a generally applicable, powerful, flexible and transparent modelling of a system. Starting from this concept the systematic evaluation of the system states is replaced by a statistical estimation realized through efficient stochastic simulation. As a consequence the quality of the results becomes independent of the dimensionality of the problem, i.e. independent of the complexity of the system. The numerical efficiency of the analysis of complex systems is increased by several magnitudes. On this basis the time dependent reliability analysis of complex systems can be analysed in a detailed and realistic manner and simultaneously with high numerical efficiency. Moreover, the sophisticated simulation approaches are used for a targeted identification of critical failure scenarios, system components and system domains. The following hypothesis is derived from the statistical approach as a fundamental finding. When analysing complex systems, an increase of complexity does not necessarily require the analysis of an increased number of system states to ensure a consistent quality of the prognoses regarding system behaviour. The validity of this hypothesis is verified and demonstrated in the project.The uncertainties and the lack of information in the modelling, which result from growing complexity, are captured quantitatively with the aid of a concept of imprecise probabilities and are taken into account in the reliability analysis. In inverse direction the results are used to define minimal requirements for the model accuracy and the numerical effort in order to provide a sufficient quality of the prognosis to derive decisions regarding repair, maintenance etc. of the complex systems.