Assessing the capabilities of advanced risk quantification methods for engineering systems management
PhD Student: Miroslava Tegeltija
Project Duration: 2014-2018
When risk management considerations are integrated into the engineering systems design, both overall system performance and quality of the developed solutions improve. A central part of integrating risk management in engineering systems design is to ensure that the design process benefits from employing risk and uncertainty methods with different levels of sophistication. That is, namely through the application of risk analyses that model risk and uncertainty in different ways. Traditionally, especially in engineering fields, risk analyses have largely been expressed in a quantitative, probabilistic form. However, such quantitative information, either as customized input to decision making or as general-purpose statistics, is itself becoming increasingly problematic and afflicted by severe uncertainty. Both the precision in estimates and the quality of background knowledge, on which probabilities are based, have been challenged in practice and academia.
This PhD thesis investigates advanced risk and uncertainty quantification methods in the context of engineering systems to better address, reflect, and utilize available information and background knowledge in design. The investigation was guided by the four research questions focusing on: 1) challenges in current design risk management quantification, 2) advanced risk and uncertainty methods, introduced under the non-probabilistic framework: the first group of methods is based on imprecise probabilities, the second represents a group of semi-quantitative approaches and the third group of methods is based on exploratory modeling, 3) prototypical applications of the non-probabilistic methods in different engineering systems design contexts, and 4) the transfer and integration of these methods and their results into overall risk management and associated processes.
The results are presented in corresponding chapters from which four core findings are extracted: 1) currently widely used risk and uncertainty methods do not appropriately describe all uncertainty - especially uncertainty due to lack of knowledge, so called epistemic uncertainty – that remains a challenge, 2) advanced methods have been developed in other fields (i.e. outside of engineering design) to deal with similar issues and have provided valuable results in those fields, but have not yet been applied or tested in engineering design contexts, 3) for the engineering design situations and scenarios tested in this thesis, the non-probabilistic methods provided more credible representation of uncertainty, and 4) finding and employing a satisfactory quantification method from the available options is context dependent, and a broader process view needs to be considered when tailoring risk management to specific design situations.
This study contributes in four ways to the extension of our knowledge base on risk management in engineering systems design. First, the study synthesizes the challenges in current risk management from the literature and through empirical work regarding: modeling, quality of background knowledge and use and integration of results (research question 1/contribution 1). Second, this study systematically collects and categorizes advanced methods from the literature in other domains, conceptually develops them for the design context and provides a unique platform for their application through the non-probabilistic framework (research question 2/contribution 2). Third, it transfers these methods into usable tools through examples in case study applications in the oil and gas industry, followed by their comparison with several traditional probability approaches in representative situations (research question 3/contribution 3). Fourth, this study facilitates and enables a more adequate choice of a quantification method depending on the design context in question by developing a risk management tailoring approach (research question 4/contribution 4). The overall conclusion is that non-probabilistic methods have a high potential in engineering systems design, but their integration to the overall risk management and associated processes must be carefully and knowingly planned and carried out, to harness this potential and to achieve an actual design impact in practice.