تجزیه و تحلیل سیستم جهت دار واقعه

This article presents an attempt towards a probabilistic event oriented system analysis in engineering. Engineering systems are represented as either complete or incomplete systems of events and as compounds of various subsystems of events. The event oriented system analysis investigates important subsystems in engineering systems, such as operational modes and failure modes and their interrelations. The analysis is also applicable to engineering systems with various relations among the sets of events, such as mutually exclusive and inclusive sets. Further, the systems and subsystems are subjected to probability and uncertainty analysis. The system uncertainty analysis is based on entropy. General relations among the probability, uncertainty of the system and uncertainties of the subsystems are derived by using information theory. Specific mathematical aspects and available methods in the uncertainty modelling of systems and subsystems are summarised. Numerical examples confirm the relevance of the event oriented system analysis and indicate potential improvements in system design.

Each object can be viewed as a system in different ways. An object is often a part or subsystem of a more complex system. The object itself can consist of many components and possibly of more subsystems. Systems in engineering are often viewed as objects of many discrete interacting components with uncertain capabilities. Moreover, systems are subjected to uncertain external demands. Sometimes, the components are grouped into a number of subsystems, each of them pertaining to some specific characteristic function of the system. Subsequently, many systems can be assumed to depend only on the current states of their components. Complex engineering systems can be subjected to service modes and effects analysis in order to identify the events that can occur at the component, subsystem and system levels. The goals of such an analysis are to determine the effects of known operational and failure modes on the overall behaviour of the systems [3], [8] and [11]. In addition to operational modes and effects analysis, semiquantitative and quantitative methods can be applied to predict the probabilities of safe operation or the accidents [10]. Redundancies [5] and [6] and robustness [4] can also be considered. Service modes and effects analysis is an essential step towards understanding complex systems without which reliability and uncertainty analysis cannot be performed. The procedures presented in this article are applied in addition to the traditional system analysis for the solution of practical numerical examples in engineering. The aim is to demonstrate the usefulness of the event oriented system analysis as a tool for assessment of system performances and improvements of engineering system design with respect to system uncertainties.

This article suggests that the traditional probabilistic engineering system analysis based on physical and/or technical components of a system, may be extended by an event oriented system analysis. Such an analysis should take into account different random events in the system's lifetime service. The presented procedure can be consistently applied to problems of exclusive or inclusive events by adequate partitioning of the event space. The uncertainties in system's operation originate from the unpredictability of possible events. A practical uncertainty measure, in addition to other complex system performance measures, convey knowledge about the number of operational and failure modes and their probabilities. The relation of the uncertainties of the system and of subsystems to the overall system performance, as it is defined in event oriented system analysis, may be helpful in different fields of engineering in the refinement of system performance. Shannon's entropy can be used for uncertainty assessment of complete systems and Renyi's entropy for incomplete systems. The theorems about the mixture of distributions and dependent systems can be applied for bringing into the relation, the probabilities and uncertainties of the systems and those of the subsystems. The entropy, as the only rational measure of system uncertainty, does not depend on anything else other than possible events and in this sense is entirely objective. The assessment of the uncertainty of systems by representing them by systems of events and the application of the entropy as defined in the information theory has been well known in engineering. The reason that the system uncertainty analysis is not widely adopted in engineering practice could be the fact that the entropy of a system itself in general is not particularly helpful in the assessment of system performance. However, the uncertainties of important subsystems of events, such as the operational and failure modes, as well as their relations to the uncertainty and reliability of the entire system, can provide a better insight into the system performance. In many engineering problems, the difficulty is to consider all relevant circumstances. An event may be random with respect to some circumstances, and at the same time it may be completely determined with respect to some other circumstances. The randomness or determinedness of an event depends on whether the circumstance do or do not determine the occurrence or non-occurrence of the event. The choice of circumstances depends on the observer and there is certain freedom of choice within the limits of possibilities. Within each subsystem, other groups or subgroups of modes can be of interest to the designers and to the users, like modes of equal operational capacity or modes with equal failure rates etc. The event oriented analysis can also be applied at any level of subsystem partitioning. The article tackles the problem of distinction among complex system, including also possible redundancy and robustness, performing identical function, with the same level of reliability but with various probability distributions or with different number of operational and failure modes. The system uncertainty can be thought of as a design decision attribute, which takes into account the number of events and the dispersion of their probabilities, over all possible events and important subsystems of events, which is not included in design considerations about safety and economy. Such an approach based on event oriented system analysis, could provide an improved alternative to strengthening lifeline networks, updating or inverse analysis with observations made on system behaviour and in general better system designs. At present, the event oriented system analysis faces possible numerical problems in dealing with larger systems. For a complete event oriented system analysis, an enumeration of all the possible events is needed. Most of the quantitative methods are economical in the use of only the most influential events in order to reduce the computational efforts. The methods presented in the paper also allows the uncertainty assessments of incomplete systems, consisting of only observable or only of important events, being then numerically more efficient and perhaps more practically applicable. An enormous increase in numerical capacities of recent computer systems could further encourage the development of even oriented system analysis.