Safety performance of hazardous systems: Approaches to risk estimations for operations and design

Abstract

Safety of chemical and petroleum process installations have received increased legislative and academicattention in most countries, to enhance protection for people from adverse effect of activities involving materials with dangerous properties. This may represent vigilance of current society against the potential for major process accidents such as explosion, fire, and toxic release which may result in fatalities and injuries. Despite knowledge on what characterizes a major accident is enhanced through experiences with past accidents, severe process accidents continue to occur until recently. This PhD study focuses on the safety challenges associated with the risk contribution from changes in plant design and operations, and suggests advanced methodologies for safety and risk analysis. The models and methods developed aim to support both industry practitioners and risk analysts. Operators of systems handling hazardous materials may confront with difficulties when making safety-related decisions. The cost for risk control and protective measures can be high, and therefore the control measures should be selected based on plant-specific contexts. For this reason, the industry ought to aim for better understanding of risk, instead of focusing on compliance to standards and regulations. However, this is challenging, because event scenarios involving major hazards are rare. This work therefore suggests representative case studies where available information and data are prerequisite for both constructing models and running the model for obtaining results. The results of this PhD study show how safety can be enhanced by continuous monitoring of safety barrier performance, based on forward-looking risk indicators as well as retrospective risk indicators for Dynamic Risk Analysis. However, modeling works based on Bayesian Networks, Multi-Phase Markov model and Petri Nets can be relatively challenging for a company and their adoption in practical application may still be in question. Thoughtful discussion on uncertainties and sensitivity analyses represent further required works for this research. In addition, operational modes of technical systems and their interaction with human operators should also be addressed in a more integrated manner. Despite these limitations, the proposed methodologies may provide insights on how to select and apply prevalent techniques of risk analysis in real industry cases. Furthermore, illustrations of the models and approaches in example cases lays the foundations for advances in safety and risk analysis. More importantly, this PhD study is expected to encourage continual learning about risk and safety analysis in the relevant industry sectors.