How Safe is Safe? (Some thoughts focused on Fire Safety in Buildings)
Rodrigo Machado Tavares, BEng, CEng, PhD
Technical Director (Head - Fire Safety) at Arcadis
This article solely represents my personal views on this topic.
When I started my journey within Fire Safety, back in 2001, I was mainly involved in carrying out PRA (Probabilistic Risk Assessments) and using QRA (Quantitative Risk Assessments or Analysis) methods for assessing the risks of explosions in underground natural gas pipelines as results of (i) soil settlement and/or (ii) pipeline’s fractures. On that time, it became clearer how the (good) understanding of the concepts of risks, hazards, probability, likelihood etc. is paramount to Safety Engineering/Safety Science and related areas, such as Fire (Safety) Engineering. Within Engineering (e.g., Civil Engineering; Structural Engineering; Mechanical Engineering; Electrical Engineering; Fire Engineering), we are intuitively embedded with the idea of risks (and associated concepts as mentioned in the previous lines of this paragraph) via the safety factors, margins of safety we use when developing calculations for structures, for example. Within this context and focusing on the Fire (Safety) Engineering area, it appears there is some confusion about the concepts mentioned above.
For instance, within our (fire safety) industry, it appears to be not uncommon to consider Probability and Likelihood as synonyms and as a matter of fact, they are not the same. The potential implication(s) of assuming these two are the same is to carry out Fire Risk Assessments (FRA) and/or Fire Safety Audits which can either become unrealistic, providing a plan of actions which are too simplistic (potentially compromising the life safety by defining insufficient mitigation measures) or too conservative (unfeasible mitigation measures). Statistically speaking, Probability and Likelihood are not the same. Probability is the chance that a particular outcome (i.e., an event consequence or severity from a hazard) occurs based on the values of parameters considered in a model/function/equation. While likelihood is how accurate and realistic a sample (with historic data) is for supporting the quantitative data (i.e., values) of a parameter. Reference [1] clarifies this difference in more detail.
This minor confusion is perhaps a result from another misunderstanding of the simple equation: R = f x C, where R: Risk, f: frequency and C: Consequence (or Severity). Frequency is intrinsically associated with Probability, thus indeed, both, frequency and probability are part of the definition of risk.
In simpler terms, probability is a mathematical construct (based on the axiomatic theory of probability), and it is a quantitative variable (i.e., a number), from the (0; 1) interval. And frequency is a physical number of events (such as a fire occurrence in a HRRB) within the “real world” (i.e., as I tend to call it as the tangible and measurable dimension of the observable Universe) per an interval of time. We can then put frequency with probability together and use them to the real problem(s), such as defining and quantifying Risks. From a practical point of view, both terms (i.e., frequency and probability) are interchangeable due to no influence on the real statistic problems solutions. There are several reliable sources of information about these concepts in Statistics, but references [2-4] are very didactic and clear for a quick read.
Few days ago, I had a look again on this report by The HSE (The Health & Safety Executive [5] here from the UK: https://www.hse.gov.uk), which was published last year. This is a very informative and still very useful document, particularly regarding some basic understanding on core-concepts.
Based on this and the brief digression/reflection on Risks and associated concepts, when thinking about the (recent) Building Safety Case and its associated report (i.e., the Building Safety Case report), I do think we should be very prudent on how we state how safe the building is to be occupied, caveating every single aspect to be considered holistically. There is a need (which should always be by default anyway) to spell out very clearly all the design parameters, their interconnections, the state of every AFP (Active Fire Protection) systems as well as the Passive Fire Protection (PFP) measures, the status of the Fire Safety Management etc. The Safety Management System (SMS) is/will be paramount for this. (There are several other official references from the UK Government via the Ministry of Housing, Communities & Local Government (MHCLG) and from the HSE which I listed in the end of this article, which provide, among other important information, further and detailed instructions about the Building Safety Case, how the Building Safety Case Report should be written and what factors should be considered etc.)
Within this rationale, in this report produced by the HSE, there is a reasonable expectation for the Accountable Person(s) [6] to answer (and clearly justify the answer, providing factual explanations) the question: “Why is your building safe to occupy?”
Having said that, I would suggest all those in charge in producing the Building Safety Case Report to be submitted to the Building Safety Regulator to consider some further engineering assessment, potentially aligned with principles found in PRA, QRA and similar approaches depending on the scenario. Obviously, each is case will be a particular case and every report needs to be bespoke, nevertheless, I would say, without over-complicating neither recreating the wheel, we can potentially use well-known and established approaches such as the ALARP (As Low As Reasonably Practicable) method also proposed by The HSE [7,8]. The following diagram presents the ALARP triangle.
Figure 1: ALARP model (Source: adapted from the Health Safety and Executive: https://www.hse.gov.uk)
Based in Figure 1 above, it is relevant to note that in the intolerable region, the risk cannot be justified. In the ALARP region, the risk is tolerable only if its reduction is impracticable or if the cost of reduction is grossly disproportionate to the improvement gained. In the broadly acceptable region, the risk is tolerable without reduction, but it is necessary to maintain assurance that it remains at this level.
Within this context, the questions, “how safe is the building to be occupied?” or “why is your building safe to be occupied?” can be answered as well using the ALARP concept to explain the assumptions and associated fire safety/protection measures considered in the Building Safety Case Report. The report should include all assessments, reviews, engineering judgment etc. which were carried out. It is the duty of care to explain all these logically to the client/residents and the authorities for: a) managing expectations and b) being as most realistic as possible. (The same can be applied to the new PAS 9980:2022 - Assessing the external wall fire risk in multi-occupied residential buildings (https://www.bsigroup.com/en-GB/standards/pas-9980).
I would also add, or perhaps make it clearer that in my view, we can/should take some of the approaches we already have in the Oil & Gas, Energy, Aviation, Nuclear sectors, and introduce gradually in our Building/Construction industry.
Please feel free to contact me directly to exchange ideas etc.
Wishing a good rest of Father’s Day for all us fathers (and mothers who fulfill this role too).
Regards, Rodrigo
Additional Note: a) If we carry out a PRA for a High-Rise Residential Building (HRRB), the likelihood of a fire to happen could be quantified in terms of a time-dependent average frequency of this fire to happen. Furthermore, frequency represents the (random, aleatory) uncertainty on whether this fire will happen in a given defined time-frame, whereas probability can be defined to express the epistemic uncertainty of this frequency estimation, particularly where there is lack of historic data. This paper published in 1981 (“old, but gold”) by GARRICK, J. & KAPLAN, S. "On the Quantitative Definition of Risk", Risk Analysis Vol 1 No. 1 is very insightful and informative. This link below as well:
https://www.psychologicalscience.org/observer/bayes-for-beginners-probability-and-likelihood
b) The understanding on Bayes' Theorem can also bring some additional light on some points discussed in this article.
Figure 2: Formula of Bayes' Theorem
REFERENCES:
[2]. https://www.stats.org.uk/probability/frequency.html
[3]. https://www.sciencedirect.com/topics/mathematics/relative-frequency-distribution
[4]. https://astrostatistics.psu.edu/samsi06/tutorials/tut1loredol3.pdf
[5]. https://www.hse.gov.uk/building-safety/news/safety-case-principles.pdf
[7]. https://www.hse.gov.uk/managing/theory/alarpglance.htm
[8]. https://www.hse.gov.uk/managing/theory/index.htm
Other additional relevant references on Building Safety Case:
https://www.hse.gov.uk/building-safety/safety-cases-reports.htm
https://www.gov.uk/government/publications/building-safety-bill-factsheets/safety-case-factsheet
https://constructionmanagement.co.uk/hse-creates-new-toolkit-for-high-rise-building-safety-cases/
Worked over 40 years in the Global Chemical Industry
2 年Interesting perspective and well thought out. Do the large insurance companies or their industry groups have access to global data on high rise residential building fires (frequency, severity) which could be useful in benchmarking the “best practices” and “worst practices”? Their interest of course is mainly protection against major financial losses from fires among their insured’s real estate holdings. But if there is sufficient history and data from previous fire events, fire prevention/safety engineers might benefit from a compilation of their data.