دانلود کتاب Uncertainty in Quantitative Risk Analysis – Characterisation and Methods of Treatment

In Sweden, it is possible to discern a considerable increase in the use of quantitative risk analysis (QRA) as part of the foundation for decision making regarding safety-related issues in various areas, for instance land use planning, licensing procedures for hazardous activities, infrastructure projects, and as an integrated part of environmental impact assessments. The QRA methodology has proven to be of substantial use regarding the determination of major contributions to risk, and for the evaluation of different decision options, e.g. different design alternatives. However, due to a lack of consensus concerning which methods, models and inputs should be used in an analysis, and how the, sometimes considerable, uncertainties that will inevitably be introduced during the process should be handled, questions arise regarding the credibility and usability of the absolute results from QRA. Without a description of and discussion on the uncertainties involved in such an analysis, the practical use of the results in absolute terms will be severely limited. For instance, comparison of the results with established risk targets, or tolerability criteria, something that is becoming increasingly common, becomes a fairly arbitrary exercise. The need for standardisation in this area is evident.
In this dissertation, the fundamental characteristics of different types of uncertainty introduced in QRA, together with different methods of treatment, are presented. Somewhat simplified, comprehensive uncertainty analysis can be regarded as having three major objectives. Firstly, it is a question of making clear to the decision-maker that we do not know everything, but decisions must be based on what we do know. Secondly, the task is to define how uncertain we are. Is the uncertainty involved acceptable in meeting the decision-making situations we face, or is it necessary to try to reduce the uncertainty in order to be able to place enough trust in the information? Consequently, the third step is to try to reduce the uncertainty involved to an acceptable level.
At an elementary level, two major groups of uncertainty can be discerned, i.e. aleatory (or stochastic) and epistemic (or knowledge-based) uncertainty. The most important distinction between these two types of uncertainty, at a practical level, is that the knowledge-based uncertainty can be reduced by further study, should a reduction in the overall uncertainty in the results from an analysis prove necessary. The aleatory uncertainty, on the other hand, is by definition irreducible. Inherent in the QRA process is the need to use expert judgement to estimate the values of unknown parameters (knowledge-based uncertainty). A discussion is presented on various methods of eliciting information from experts in a structured manner, together with a presentation of known pitfalls of such exercises. Knowledge about such procedures, and about the problems associated with them, is a key issue in keeping knowledge-based uncertainty to a minimum.
The core of the dissertation, however, is a structured survey of methods of propagating and analysing parameter uncertainty. The basic features of a number of different approaches and methods of uncertainty treatment are presented, followed by a discussion of the arguments for and against the different approaches, and on different levels of treatment based on the problem under consideration. To further exemplify the different features of the methods surveyed, a case study is presented, in which a simplified facility for ammonia storage is analysed with respect to the risk it poses to its surroundings. Emphasis is placed on the kind of information required for use of the different methods, and on the kind of results they produce.
Summary
i
Uncertainty in Quantitative Risk Analysis – Characterisation and Methods of Treatment
It is concluded that methods are available for the explicit treatment of uncertainty in risk analysis with sufficient sophistication for most problems, although some types of uncertainty, mainly those related to completeness and general quality issues, are inherently problematic to quantify. Furthermore it is concluded, regarding future standardisation work in this area, that the probabilistic (Bayesian) framework offers the most comprehensive “tool box” for uncertainty analysis, and appears to be the most promising approach for dealing with the uncertainties in QRA. This is due to its strong theoretical foundations and the possibility of quantifying, and analysing, uncertainties originating from fundamentally different sources (e.g. aleatory and epistemic uncertainty) separately.
Recommendations for future research and standardisation efforts in the area are given, and the main conclusion is that generic guidelines across all sectors of industry are not deemed viable, due to the different conditions under which they operate. Instead, differences between various industrial sectors, for instance, the chemical process industry and the transportation industry, would have to be acknowledged in such work, presumably resulting in separate guidelines. Furthermore, possible ways of differentiating the level of uncertainty description and analysis required in an analysis, based on, for instance, the complexity of the problem and the nature of the hazard source, should be examined within each sector of industry. In this dissertation, a discussion is presented on various levels of treatment, which may serve as a basis for further debate. This kind of work on standardisation is an absolute necessity for the general use of risk tolerability criteria to be meaningful.