Risk assessment based on maximum allowable damage

Abstract

Demonstrating the adequate fire safety performance of a building design is a requisite part of the approvals process in the construction industry. Risk assessments provide a powerful tool to quantify fire risk, which is often used as a proxy for fire safety performance. However, the nature of the risk assessment process means that it may also open the door for unacceptable losses to be present in the form of latent risks. This research analyses this problem and puts forward key elements needed for better fire risk assessments and an inherently safer built environment.In a performance based environment, fire safety engineering can take advantage of frameworks and methodologies to explicitly demonstrate performance through qualitative or quantitative risk assessments. The latter are preferred as they evoke a sense of objectivity and scientific rigor and due to their extensive use in high-risk settings, e.g. chemical and nuclear safety. Quantitative risk assessments can be deterministic or probabilistic. The former focuses on specific consequences, while the latter quantifies the likelihood of a range of undesired consequences.Probabilistic risk quantification can result in significant consequences appearing to be acceptable conditional on the fact that their likelihood is low enough. However, three problems arise from this: i) public responses to fires which results in large loss of life completely invalidate the premise of acceptable risk based on negligible likelihood; ii) the subjective basis and lack of structured approaches to identify scenarios that can actually challenge the performance; and iii) the reliance of probabilistic risk assessments on scarce statistical data is a topic of great concern in other fields, with some practitioners suggesting that they be reserved for specialized applications. As probabilistic risk assessments are often promoted as the default choice for demonstrating fire safety performance, the danger of providing a false sense of safety is significant. This concern has motivated this research, which has the overarching objective of proposing an alternative methodology for fire risk assessment that promotes an inherently safer built environment.This research is divided in two parts. The first part details the limitations associated to risk assessments, to probabilistic risk assessments and to the existing guidance for fire risk assessments. These limitations include data availability, the subjective basis for scenario identification and an unknown degree of trustworthiness of the results. The second part proposes a consequence-driven fire risk assessment methodology, applies it to three distinct case studies and evaluates both its advantages and limitations.The proposed methodology focuses on the potential range of fire consequences. The upper end of this consequences range should be below a stakeholder-predefined maximum allowable damage for the performance to be adequate. This concept is the basis of the methodology, correspondingly named Maximum Allowable Damage (MAD). The consequence range is estimated for a ‘bare-bones’ design, which excludes those active safety measures that cannot be ensured to work over the life of a building (e.g. sprinklers). This estimation allows identifying possible modifications to its constitutive elements as to achieve adequate performance. This approach is aligned with the inherently safer design philosophy which designs out hazards and introduces safety measures only on an as needed basis.MAD uses a high-level representation to describe how the fire damage materializes, i.e. damage model. This representation provides an initial bounding of the scenarios. The variables, their values, and underlying assumptions within the damage model are logged, allowing to judge their trustworthiness. Trustworthiness in MAD is a function of the strength of knowledge and output sensitivity of each variable and assumption. To characterize the upper end of the damage potential, a deductive reasoning approach is implemented, where the central question is which conditions are necessary (and plausible) to exceed the maximum allowable damage. Hence, scenarios are iterated and both the damage model and the information registry are updated as needed.MAD allowed obtaining insight on the fire safety performance of three distinct case studies. This insight was formulated as actionable recommendations that enabled an adequate performance. The trustworthiness of each assessment was a key component in achieving and defining these recommendations. Both the damage model and the information registry serve as supports for the trustworthiness of the assessment and as a basis for third party reviews and future reassessments. As observed in all case studies, the acceptance criteria largely defines the degree of conservatism required in the assessment, addressing one of the main criticisms to consequence-driven analyses.While MAD does not produce a metric to be used as a universal yardstick for fire risk, it is a means to trustworthy insight into fire risk. This insight can be used to ensure that a building is inherently safe before other methods are applied to optimize the fire safety design.