Conclusions and Outlook
Chapter 6 Summary
The complexity of today’s industrial production networks constitutes a new challenge for industrial risk and safety management. A single company, complete industrial production systems as well as the society as a whole are exposed to various different types of risk every day. In order to handle potential risks in industry and their respective impact on mankind and the environment, an integrated approach to risk management and industrial environmental policy is needed, since complex decision situations need to be resolved in a wide variety of circumstances.
Understanding the risks emanating from technical failures in industrial operating proce-dures and the potentially arising emergency situations is very important for a sustainable development. Special attention must be paid when the individual production processes are coupled and when their respective time and length scales within a production network are strongly disparate. The modelling and handling of risks arising from such processes necessitates integrated, interdisciplinary and multiscale approaches.
The complex decision situations which need to be resolved in the context of industrial risk management require the consideration of various conflicting criteria. In particular, aspects of health and safety, the technical feasibility and the environmental impact need to be considered besides the purely economic factors. An explicit examination of the trade-offs between these conflicting objectives plays an important role in providing a sound understanding of a decision situation. Furthermore, different scientific expert groups are usually involved with heterogeneous technical background knowledge in different disci-plines. Know-how from economic, ecological, engineering and natural sciences must be brought together, taking account of political as well as socio-psychological factors.
Providing the basis for the evaluation of conflicting criteria and for bringing together existing knowledge from different disciplines, Multi-Criteria Decision Analysis (MCDA) is helpful to resolve the complexity of the occurring decision situations. Furthermore, seeking to facilitate the communication of decisions and contributing to form an audit trail, MCDA provides valuable support in explaining how decisions are taken, which is important in the light of the rising demand from the mass media and the public for information and justification from authorities.
Within the field of MCDA, the approaches Multi-Attribute Value Theory (MAVT) and Multi-Attribute Utility Theory (MAUT) have been described in detail because of their understandable nature and their suitability to support decision making in relation to industrial risk management. Contributing to transparency and traceability of decision making, MAVT provides a basis for participatory processes and group decisions in the context of the case study.
In order to address the various types of uncertainty, which may arise in a decision making process in industrial risk management, a framework for uncertainty handling has been proposed. On the basis of a structured uncertainty classification, methods based on Monte Carlo simulation can be used for a consistent modelling, propagation and visualisation of the different types of uncertainty. Special focus has been put on approaches that allow to explicitly illustrate the spread, i.e. the ranges in which the Multi-Attribute Decision Making (MADM) results can vary in consequence of the uncertainties.
The added value of the developed methods particularly lies in the proposed framework for multi-dimensional sensitivity analysis which allows to explore the robustness of the MADM results with respect to simultaneous variations of the subjective preference param-eters. Additionally, they contribute to facilitate the preference elicitation and consensus building in decision making groups. The elaborated approaches provide valuable insights into the robustness of decisions and also allow to investigate trade-offs between the dif-ferent conflicting criteria. Thus, these different analyses lead to a deeper understanding of decision problems. Especially, a comprehensive approach for value function sensitivity analysis (i.e. the analysis of the intra-criteria preferential uncertainties), including an in-vestigation of the impact of varying the boundaries of the value functions’ domains, has not previously been mentioned in literature. Furthermore, the introduced approaches to perform backwards calculations and the corresponding graphical visualisations are par-ticularly valuable allowing to consistently and transparently link the uncertainties in the results to the uncertainties in the MADM input data and parameters. Finally, the in-tegration of the proposed multi-dimensional sensitivity analyses into the framework of
Summary 155
expected utility theory allows, for the first time, to investigate the sensitivity of MAUT results with respect to simultaneous variations of different decision parameters.
The developed methods have been implemented in a software prototype in MATLAB in order to demonstrate their capabilities in the context of a case study from industrial risk management. Within the large field of industrial risk management, nuclear emergency and remediation management is an important and challenging area because, firstly, the resulting consequences of a potential emergency are severe and far-reaching and, secondly, electricity supply is a very relevant part of critical infrastructure. Since a large part of electricity is generated by nuclear energy and an area-wide, secure electricity supply is es-sential for the functioning of modern industrial production networks as well as the society in general, the creation of awareness of the possibility of technical failure and an improved preparedness to deal with the risks and to cope with emergencies are indispensable.
Multi-attribute decision analysis constitutes an important contribution to transparently resolving complex decision situations in the context of the case study. The modelling of the different types of uncertainty occurring in such decision situations has been dealt with in detail. The developed Monte Carlo framework allows a consistent modelling of the uncertainties of the empirical input data and their propagation through the model chain of the Real-time Online Decision Support System for Nuclear Emergency Manage-ment (RODOS). Additionally, the different types of uncertainty can be simultaneously considered and visualised. In particular, the proposed sensitivity analyses provide valu-able support in analysing the robustness of decisions in emergency management and also allow to explore trade-offs between conflicting objectives such as, for instance, radiological effectiveness and resources.
Finally, potential for future research has been pointed out. Especially, a methodological extension of the decision making framework in order to reflect the sequential character of decision problems in practice as well as an improvement of the models for economic consequence assessment have been emphasised.
From applying the developed methods to the case study, it can be concluded that the total set of decision alternatives could be significantly reduced by analysing the impact of the data uncertainties. Subsequently, the multi-dimensional sensitivity analyses allow to explore which parameter combinations result in which of the remaining alternatives as the most preferred. The knowledge acquired by conducting such uncertainty analyses within the scope of the case study can be transferred to strategic countermeasure planning problems, and can consequently contribute to an improved risk awareness and emergency preparedness.
In this thesis, a framework has been developed for an integrated handling of risks ema-nating from emergency situations arising from technical failures in industrial production.
Providing an integrated picture of decision situations by allowing the simultaneous consid-eration of technical, economic, environmental as well as socio-psychological and political aspects, the application of multi-criteria decision analysis has been demonstrated in the scope of a case study from emergency and remediation management in the nuclear power generation sector. Additionally, being suitable to integrate knowledge and experiences from various scientific approaches, MCDA leads to an enriched insight into the mod-elling of real world decision problems. A comprehensive uncertainty analysis, allowing to simultaneously consider and visualise the various types of uncertainty that can arise in any decision process, has not been mentioned in literature so far. Making use of the new multi-dimensional sensitivity analysis techniques and of the combination of Monte Carlo simulation and principal component analysis to support the visualisation, provides a deeper understanding of the effects of the data and parameter uncertainties on the overall results in the process of evaluating decision alternatives.