Development of a national risk acceptance criteria for managing major industrial hazards

Abstract

This work attempts to address the issue of managing risk to the safety of the public posed by Major Accident Hazards (MAH) from the Chemical Process Industry (CPI) in Sri Lanka. The research essentially focuses on the establishment of a suitable risk acceptance criteria as well as an appropriate framework that can be used in determining the level of safety offered by a particular MAH installation in Sri Lanka. The “level of safety” of an installation is then compared against the risk acceptance criteria to determine its acceptability in the Sri Lankan context. The history of process safety management as is understood at present was investigated and the different risk regulation regimes currently in practice globally were identified. The role of risk assessment in each risk regulatory regime was investigated and the need for risk informed decision making was firmly established. The thesis then focuses on the prevalent categories of approaches in risk assessment. The different risk assessment approaches are investigated further. Out of those approaches, the consequence assessment and probabilistic risk assessment approaches or methods were chosen for the development of the risk assessment framework. The different risk metrics used to express the risk for each approach and the respective risk acceptance criteria were identified. Then appropriate risk acceptance criteria were developed for the two approaches. The establishment of a safety distance corresponding to 1% fatality of the public was adopted for the consequence based assessment method whereas a FN criteria line with an anchor point of (10, 10-4) and slope -1 was chosen for the probabilistic risk assessment method. The applicability of the different risk acceptance criteria in the Sri Lankan context is carried out for the case of propane storage tank. Data gaps and constraints are identified. Both methods adopt a conservative decision making approach. A significant constraint is the lack of a nationally verified and validated set of failure rate data for process equipment and ignition probability data; these are essential for establishing conditional probabilities when calculating accident frequencies. The usage of generic data for failure rates is not recommended due to the wide variability in different data sources. Further, allowing room for choosing an arbitrary set of failure rate data could create an opportunity for biasing the risk acceptance decision. In this work, a framework is presented for applying the risk acceptance criteria developed. An FN curve based on upper bound data for the probabilistic risk assessment method and modified consequence assessment method are developed. The probabilistic risk assessment method is modified to accommodate the variability in generic failure rate data. The decision of acceptability is made by defining an FN curve using upper bound values of the FN curve and comparing it with the criterion line. A safety distance proportionate with the overall level of risk based on a relative risk reduction factor (RRRF) is introduced.