Overview of Standards for Fire Risk Assessment

Fire Science and Technorogy Vol.25 No.2(2006) 55-62 55 Overview of Standards for Fire Risk Assessment 1. INTRODUCTION John R. Hall, Jr. National Fire Protection Association In the past decade, the world has seen the publication of a number of national, regional, and international standards and guides for fire risk assessment. In order to make sense of these diverse documents, it is useful to review their similarities and differences systematically. Some of these documents were discussed in detail in other presentations at this symposium. For the sake of efficiency, the general framework for describing fire risk assessment standards presented here is illustrated primarily by standards-making bodies or working groups headquartered in the U.S. Specifically, these documents are used as examples: ISO TS 16732 SFPE Guide to Fire Risk Assessment NFPA 551 ASTM E1776 2. ELEMENTS OF CHOICE FOR FIRE RISK ASSESSMENT STANDARDS Here are some specifications, characteristics or elements of choice that distinguish different fire risk assessment standards: Purpose of the standard Who has authority to make decisions? Measure of loss or consequence Threshold of acceptability Analysis methods used Treatment of scenarios Data sources 3. PURPOSE OF THE STANDARD Support traditional engineering analysis. An option in ISO TS 16732 is to use simple fire risk analysis methods to identify a short list of diverse scenarios that collectively represent a large share of total risk. These scenarios are then assessed through a traditional engineering analysis, focusing solely on the magnitude of consequence (loss) and the effectiveness of alternative designs to reduce that loss. This use of fire risk analysis is not a fire risk assessment, because

56 JOHN R. HALL, JR the evaluation is not done in a risk format. This is a popular approach because many people engineers and sponsors are more comfortable if a design can be shown to perform safely under any circumstances, and although this is never possible, they are willing to treat performance under a range of scenarios as an acceptable approximation. Help officials review proposals. NFPA 551 is designed not to guide the performance of fire risk assessment but rather to assist responsible officials (called "authorities having jurisdiction" in the NFPA system) to confirm or refute the code equivalency of a design proposal whose justification is stated in terms of a supporting fire risk assessment. Again, this is not guidance for those performing a fire risk assessment; it is guidance for those reviewing a fire risk assessment, as part of a traditional design review and approval. Compare alternative products. ASTM E 1776 is a guide for people writing guides for fire risk assessment of alternative products within a product class. This is not a standard for fire risk assessment; it is a standard for fire risk assessment standards. It is not ready for direct use in assessment but it is available to support the development of guides that can be used directly. Also, the guides and standards to be developed according to ASTM E 1776 are not documents for the assessment of buildings, the subject of assessment for all other fire risk assessment standards, but rather the assessment of products, such as furniture or room lining materials. Compare alternative designs. ISO TS 16732 and the SFPE Guide to Fire Risk Assessment are guides to fire risk assessment of a proposed design for a building or similar sized constructed object, such as a bridge or a transportation vehicle. Such standards are intended to either replace conventional, so-called prescriptive codes or complement them by providing a standardized protocol for demonstrating equivalency to the conventional code. This is the principal type of fire risk assessment standard requested by users. 4. WHO HAS AUTHORITY? Most of the elements of choice in a fire risk assessment involve some latitude for choice on particular projects. Any such latitude places power in the hands of the person or group with decision-making authority. The kinds of decisions where latitude often exists include the level of risk deemed to be acceptable, the choice of scenarios and other assumptions for analysis, and the models, calculation methods, and data sources considered appropriate and acceptable. There are three principal alternatives for vesting final authority over decisions: Sponsor. In this alternative, final authority rests, explicitly or implicitly, with the person or group who will pay for the work to be done. Note that if the engineers conducting the fire risk assessment are given broad latitude to

Overview of Standards for Fire Risk Assessment 57 apply their professional judgment in the execution of the assessment, this will implicitly vest authority for decision-making in the hands of the engineers clients, who will tend to be the project sponsors. Designated authorities. In this alternative, all discretionary choices in the fire risk assessment must be explicitly and individually approved by the authorities with jurisdiction over the type of decisions being made. If the decision is whether to approve construction of a building, then building officials will be the responsible authorities. If the decision is whether to allow a business to operate, then fire code officials or permit-issuing authorities will be the responsible authorities. If the decision is whether to issue insurance for a property, then the insurance company will be the responsible authority. Stakeholder consensus. In this alternative, approval of choices is made by a group, assembled early in the project s lifetime, where all the interested parties are represented. This approach can vary considerably based on the type of group decision-making process employed. ISO TS 16732 does not address the question of authority. The SFPE guide allows for any possible answer but clearly favors the stakeholder consensus approach. NFPA 551 probably devotes more time and thought to the question of authority than any other guide or standard for fire risk assessment. NFPA 551 assumes that authority rests with the designated authorities, but that authority may be constrained by applicable standards (such as NFPA 551 or any of the other fire risk assessment or performance-based standards). NFPA 551 also recognizes that knowledge can be power. Large differences in knowledge and expertise between sponsor and authority can result in a shift of power and influence from the official authority to the more knowledgeable sponsor and his agents, the engineers. 5. MEASURE OF LOSS The most common measure of loss cited in these documents is fatal injury, although the onset of life-threatening conditions may be easier to measure through engineering means and is often used as a substitute measure for calculation purposes. Injuries to firefighters are sometimes treated differently, and non-fatal injuries may be cited as a secondary measure of loss. Other common measures of loss are economic in form. Direct damage to property is the most common, although fire size and extent of fire effects may be easier to measure through engineering means and is often used as a substitute measure for calculation purposes. Business interruption, loss of continuity of operations, and other measures of indirect loss are often of concern and may be included as measures of loss. Less often, measures of the impact of fire or of firefighting on the environment or on cultural heritage may be measures of loss.

58 JOHN R. HALL, JR 6. THRESHOLD OF ACCEPTABILITY The threshold of acceptability, which is probably the most important choice in terms of design approval, is a level drawn on some summary measure of risk. The scale used for that summary measure is in turn based on a mathematical function applied to the probability and consequence functions over the universe of scenarios. This is how the threshold operates mathematically. In practice, the choice of thresholds is rarely so fundamentally grounded but is typically based on some typical statistical summary measures. None of the cited fire risk assessment guides or standards includes guidance on appropriate thresholds of acceptability, only on the process to be used in selecting them. Expected value. If you weight the consequence for a scenario by the probability for that scenario and sum the products, you have an expected value summary measure. If low-probability, high-consequence extreme scenarios are unrepresented or under-represented in the scenario mix, then the calculated expected value will under-estimate the true expected value of risk. If the threshold of risk is based on general arguments and analogies, it may not be set low enough to compensate for the under-estimation. Probability of large loss. Insurance companies may be more concerned with the small probability of a very large loss, sufficient to bankrupt the company, than with the exact size of the more frequent and more routine losses. Cost-risk ratio. If all consequences are converted to monetary terms, it is possible to compared the expected-value risk to the cost associated with achieving that level of risk. In its pure form, the cost-risk ratio can be used to argue that anything other than a very high level of risk (a very low level of safety) is too expensive. When used in combination with a constraint that any acceptable design must at least provide safety equivalent to prescriptive codes, the cost-risk ratio can be a useful tool for deciding on lowest-cost options and affordable safety beyond the minimum requirement. Equivalence to reference. Instead of setting a fixed threshold of acceptable risk for all projects, one can specify that a fire risk assessment be performed on a representative standard design for a fully code-compliant building, then use the calculated risk from that reference design as the threshold for an alternative design. Allowances for uncertainty. All calculations involve some uncertainty. A key question is who receives the benefit of doubt from that uncertainty. Should we assume that a design is safe unless we are certain it is unsafe or that a design is unsafe unless we are certain it is safe? None of the existing fire risk assessment standards and guides clearly address the interpretation of uncertainty analysis, and it is generally acknowledged that even the calculation of uncertainty is limited to non-existent in the field.

Overview of Standards for Fire Risk Assessment 59 7. ANALYSIS METHODS Qualitative methods include index numbers and risk matrices. The latter is featured prominently in NFPA 551 and is identified as an option in ISO TS 16732 and the SFPE Guide to Fire Risk Assessment. There are no qualitative methods described or referenced in ASTM E 1776. Note that qualitative methods are not the same as quantitative methods using expert judgment as a source for data. Qualitative methods are difficult to validate and are rarely validated but need to be validated, because it is rare that their construction is sufficiently well grounded as to earn them a presumption of validity. Outputs of qualitative methods are not easy to compare to costs. On the advantage side, they are usually quick and easy to use and they may be easy to calibrate against a reference code-compliant building. An important disadvantage is that they may not produce the same or similar results for different users. Quantitative methods are primarily event trees and fault trees. Event trees are well suited to produce expected value outcome measures, and fault trees are well suited to produce large loss probability outcome measures or to provide ignition probabilities for the front end of an event tree. Traditional engineering models can be used to calculate or estimate consequences for scenarios. When done thoroughly, quantitative methods require considerable time, money, and data. When simplified to avoid these costs, these methods often lose low-probability or other atypical scenarios, but the loss of these scenarios raises serious questions about the accuracy and validity of the calculations. Except for NFPA 551, all published fire risk assessment standards and guides emphasize quantitative methods. 8. TREATMENT OF SCENARIOS Scenarios should be treated as a representative and sufficient sample from the universe of possible fires. It should be possible to assign every fire that could occur to one of the scenarios selected for analysis in such a way that the underlying probability and consequence distribution is fully and accurately represented in the more manageably sized set of scenarios for analysis. More commonly, scenarios are built up to represent the list of identified major hazards and fire challenges associated with the property. The use of a list of major hazards has value in that it reduces the likelihood that important but rare fires will be overlooked. However, the list approach can divert attention from the common fire challenges that all properties share (and that typically account for most of any property s fire experience) while also deemphasizing the role of probability. The SFPE guide has the most extensive treatment of the process of identifying hazards, developing hazards into scenarios, grouping scenarios into clusters, and establishing the sufficiency of the resulting scenario structure to represent the universe

60 JOHN R. HALL, JR of possible fires. ISO TS 16732 is less explicit and less detailed in its treatment of hazards, and ASTM E 1776 does not mention hazards as an element in scenario identification at all. 9. DATA SOURCES Some national design guides include or reference some national data sets that can be used. These may be approved sets of well documented experiments or national fire incident data bases, but more often are statistics derived from analysis of experiments and incident data bases. Even more often, the design guides limit their treatment of data to discussions of types of data sources, issues in the selection of data sources, and issues in the analysis and interpretation of numbers from certain data sources. Typically, the least attention is given to standardized data sets of field observations other than fire incident reports. This would include the kind of data required for empirical estimation of reliability and the results of exit drills or other exercises designed to obtain realistic data on human behavior leading to fire or reacting to fire. It is important not to base ignition probability or reliability parameter estimates exclusively on experiments or laboratory data that captures only equipment performance but not human error. Human behavior is consistently the primary contributor to probabilities of importance in fire risk assessment. Subjective judgment (often called expert judgment or engineering judgment) is often needed to complete the generation of data. It is important that the process of using judgment be done explicitly and that it pay attention to potential biases and errors. The selection of experts should reflect the diversity of interested parties and not just the sources of greatest technical expertise. Expertise in physical phenomena need not imply expertise in the relative frequency of different types of fire events or in related human behavior. Expertise in the mechanical or electrical reliability of equipment need not imply expertise in the field reliability of such equipment in ordinary use. The SFPE guide is the only document cited that devotes substantial space and guidance to the process of developing data from judgments. 10. NATIONAL STANDARDS VS. ISO STANDARDS An ideal standard captures the best available science on matters of fact and the best consensus of values of interested parties wherever preferences need to be standardized. One would not expect science to vary from nation to nation, but relevant fire experience, hence probabilities of types of scenarios, may well vary, as may human behavior, field reliability, and levels of acceptable risk. ISO expects national standards to provide more detail but to be compatible with the ISO standards on issues that both standards address. Accordingly, the ISO documents

Overview of Standards for Fire Risk Assessment 61 avoid specifying preferred data sources, recommended levels of acceptable risk, or recommended data sources. It is possible for a national standard to be "captured" by a single interest group or a single technical discipline. Some national standards may empower designers to bypass effective oversight by code officials. This is a recipe for unsafe buildings. Some national standards may substantially raise the standard of acceptable performance by chasing minimum feasible risk. This is a recipe for unaffordable buildings or no new buildings. The dangers of "capture" and creation of biased or unworkable national standards can best be avoided by aligning and harmonizing national standards with ISO standards and with each other. Each such harmonization creates pressure to broaden the group of interested parties with some influence on the decisions, and that in turn improves the chances that the preferences and values incorporated in those decisions are a broad consensus of values. The dangers of overriding legitimate national differences and stifling technical innovation can best be avoided by creating and maintaining national and regional standards as value-added extensions of ISO standards, that is, as documents that use the ISO standards as a base. 11. CONCLUSIONS Each paper in this workshop can be viewed as an example of some approach international standard, national standard, industry standard, individual approach to the general problem of applying fire risk concepts to important decisions about buildings and their safety. By comparing the approaches carefully and critically, each person can gain insight into legitimate vs. illegitimate differences. Legitimate differences reflect objectively different circumstances or preferences among the affected parties. Illegitimate differences reflect unsubstantiated departures from best science or best data or they reflect the capture of decision-making by one interested party from among the many who are affected and deserve to be fully involved in the decision. If we can distinguish legitimate from illegitimate differences, we will be closer to the ideal of standards that standardize where they should and only where they should. All the existing standards are still relatively new, and so they are still relatively flexible and open to change. If any standard ought to change, in light of the choices made in other standards, then there will never be a better opportunity to propose such changes and move them to implementation. Now is the time to create a global structure of harmonized standards with the best science and the broadest value consensus possible.