Validation of the "FRAME" -method.

"FRAME" appeals to risk professionals as it "feels good", i.e. the results of the calculations fit the intuitive evaluation, based on knowledge and experience. But as for all empirical methods, the question is often raised where does it stand scientifically? The distrust for empirical methods is understandable, yet such calculation rules have been established and used with confidence in many fields and for many years before there was any scientific support for them. Men did have some empirical rules to build ships before Archimedes discovered the physical law for buoyancy.

It is fairly clear that organising a research program to check the validity of all the parameters by fire tests is an impossible task, which goes beyond the need for accuracy. However, "FRAME" users are entitled to know on which scientific knowledge and proof the method has been beased. The proposed approach is to check if the method fits with available knowledge on risk assessment and to check the compatibility of the calculation results with real fire cases.

Assessing Hazards and risks

Risk assessment has become a common growing practice in all safety related disciplines, and various methods are used to assist the decision making processes, not only to know what to do, but also to have a reasonable insight in the cost/benefit balance of a proposed set of provisions.

Fire risk assessment has been practiced for many years with variable success. The development of performance based fire codes and fire safety engineering has increased the interest for risk assessment tools suitable for fire safety issues. Some methods exist already for more than 40 years, others were developed and have been abandoned again, some in-house methods are only accessible for a few, but others are easily available at limited or even at no cost. All currently available methods use some kind of software to make calculations and to issue reports.

The WG6 workgroup of the EC-funded FiRE-tECH research programme, Fire Risk Evaluation to European Cultural Heritage, issued in 2004 an overview report describing some 12 fire risk assessment methods and their main characteristics. The reports of the 8 working groups issued reports were for some time available at the Internet, but unfortunately the website is closed and the reports are no more readily accessible. An overview of the project and some documents are available on the FRAME - website under "Extra - FiRE-Tech project ".

Guidelines for Fire Risk Assessments.

In English, two guidelines exist to help the user of fire risk assessments: NFPA 551, Guide for the Evaluation of Fire Risk Assessments, and CFPA Europe Guideline n° 4 "Introduction to Qualitative Fire Risk assessment", issued in 2006.

The NFPA 551 guide is intended to provide assistance, primarily to authorities having jurisdiction (AHJs), in evaluating the appropriateness and execution of a fire risk assessment (FRA) for a given fire safety problem. While this guide primarily addresses regulatory officials, it also is intended for others who review FRAs, such as insurance company representatives and building owners. The CFPA Guideline n°4 is mainly oriented towards safety in a workplace both for staff and visitors.

To confuse the reader, both documents have different definitions for hazard, risk, exposure, etc. To make the confusion even larger in the United Kingdom, a different type of fire risk assessment is imposed since 2006 for non-domestic premises by the Fire Regulatory Reform Order and the Fire (Scotland) Act. The obligation basically requires a list of existing fire related hazards and a code compliance check. The websites of the numerous British companies which offer this kind of service are a nuisance for anybody who searches the internet for more elaborate fire risk assessment tools. Luckily, you might have found the "FRAME" website too.

The French INRS has published a number of documents on the subject which follow basically the same checklist + code compliance approach as the British.
In Germany, the VFDB (the German Fire Protection Association) has published a 243 pages "Leitfaden Ingenieurmethoden des Brandschutzes", in which chapter 10 "Risikomodelle und Sicherheitsbetrachtungen" is based on the FiRE-tECH WG6 report.

Acceptable risk arguments.

All methods for risk assessment give some way of quantification of hazards, and in most of them there are also guidelines to compare the risk with a benchmark which is considered to be an "acceptable level" of risk.

The "acceptable risk" idea has been debated for a long time. One of the first publications on the subject was William W. Lowrances' book "Of Acceptable risk, Science and the determination of safety." published in 1976. Lowrance already wrote that people accept risks, even with deadly consequences, if the combination of probability, exposure and severity is low enough. He also indicates a number of factors that influence the way risk are accepted or tolerated.

Sometimes people are bound to accept a risk because they do not have the means to protect themselves or because they consider these risks as inherent to life. In this way, people accept in some countries floods or earthquakes as part of their living conditions. Risks are less acceptable when the consequences are more severe, when more people are exposed to the risk at the same time, and when the duration of exposure is lengthy. A risk is considered less acceptable when the consequences are readily visible.

A risk is more easily accepted when the consequences are reversible, of short duration or repairable, when there is a visible benefit in taking the risk, and when one thinks to have control of the causes of the undesirable event. An unknown or hidden risk will be less acceptable. In this way, people have more fear of a fire during the night than during daytime, although the probability of starting a fire is much higher when they do all kind of things, then when they are at rest.

The basic acceptable level of risk in a situation of permanent exposure to a hazard is a risk level that is lower than the natural death level. That level is usually defined as the risk of one death with a probability of less than 1 per million of persons per year, or 1.10^-6 / person*year.

For accidents with a potential of multiple deaths, it appears that the acceptance of risk is reduced by the square of the number of possible victims: for 3 deaths, the acceptability is 9 x less, for 10 victims, it is 100 x less, and for a 100 deaths, it is 10.000 x less.

When the exposure to the hazard is not permanent, a higher level of risk will be accepted. NFPA 551 makes no clear distinction between continuous and discontinuous exposure, unlike the CFPA Guideline n° 4 which uses the following approach:

For life threatening hazards such as aids, smoking cigarettes and also fire, it appears that the acceptable risk level is not defined by the frequency of occurrence but by the level of exposure. As the outcome is always death, the recommendations are meant to reduce the exposure to the hazard, not to reduce the effect of it.

In fire protection, the common practice to impose higher requirements on high rise buildings compared to low or medium height buildings with the same occupancy, cannot be explained by a difference in fire probability, but is fits perfectly with the higher exposure in high rise buildings, where the time required to evacuate the occupants is much higher.
Richard Bukowski defined this approach in one of his papers as "hazard based" regulation but I dare say that "exposure based" is a more appropriate definition.

There a several approaches possible to make a (fire) risk assessment ( see also : "risk evaluation" ), but the most elaborate methods use mathematical expressions,based on a combination of values representing the probability of fire, the severity of the fire scenario and the level of exposure.

Probability based risk quantification methods.

Probability based methods start from the observation that unwanted vents like fire have a variety of outcomes. Most methods are not specifically developed for fire risks and need some kind of interpretation to be used for fire risk assessments. The existing methods can be classified as two-dimensional or three-dimensional evaluation of risks.

The "American school" of fire protection engineers uses mostly two-dimensional risk quantification schemes, based on the technique described in NFPA 551, Guide for the Evaluation of Fire Risk Assessments. Two-dimensional methods were developed by the nuclear, chemical and aviation industry where the hazard exposure is continuous or exists over long periods. They express the risk as a combination of a probability of occurrence or frequency and a magnitude of the consequent loss or severity.

These two-dimensional schemes tend to create some confusion as they do not fit the usual practice for life safety, where a combination of severity and exposure is more often used in the risk assessments.

The most commonly used approach is the ASET/ RSET evaluation: A worst case RESET, Required safe egress time, calculated with an egress model is compared with a worst case ASET, the Available safe egress time, obtained with a zone or field model. Once the RSET is higher than the ASET, the occupants are considered to be exposed to a life threatening and unacceptable risk. In reality the probability of such a simultaneous occurrence of both scenarios is very low, and in many situations there will be still a comforting safety margin.

NFPA 101 "Life safety code" chapter 5.5 defines 8 design fire scenarios for approvingperformance based designs. Except for scenario 1, none of these scenarios can be found back in the fire statistics among the most probable fire cases. This means that the probability component of the risk is disregarded and that the focus lays on the exposure component, when life safety is at stake.

Methods based on a combined evaluation of exposure, probability and severity offer a better approach for fire, which is a non-continuous hazard and basically a rare but unwanted event. This approach fits better to the intuitive way of making fire risk assessments and to the reasoning underneath code requirements.

Three-dimensional methods use Exposure, Probability and Severity for risk quantification. This type of risk evaluation is practised in those areas where the exposure to the hazard is not continuous, such as workplace risks, machinery defects, and fire.

One of the oldest and widespread methods for workplace risk assessment was developed back in 1976 by Kinney, Wiruth e.a. and is widely used for the analysis of workplace hazards. The approach is known as the Kinney or ESP method. A similar approach can be found in the standards for the safety of machinery, such as EN-ISO 14121-1:2007 and the older EN 1050 and EN 954-1 standards, and this three-dimensional evaluation is also included in "FRAME".

Three-dimensional risk expressions.

For the safety of machinery, the way to combine severity and probability of occurrence into an acceptable risk is clearly defined and explained in EN1050 and EN954-1. The severity of a risk is evaluated as "worst case consequence" without consideration for the effectiveness of the protection or for the duration of exposure. Such "worst case" becomes acceptable when the combination of exposure and low probability balance the severity of the case. A generally used mathematical expression developed by KINNEY e.a. For such an acceptable situation is the formula:

• Sev = measure for severity
• Poc = measure of probability of occurrence
• Exp = measure of exposure
• C(constant) = measure of acceptable risk level

FRAME approach of fire risks

The similarity between the "FRAME" basic formulas and the expression used in the KINEY method clearly shows that "FRAME" is not a "point-system" like some checklist based other methods ( e.g. NFPA 101A FSES method) but a probability - exposure - severity based method.

Introducing a beginning phase in a fire model is more significant, as it gives an indication of the time delay before the severe thermal action starts, and influences greatly the effectiveness of defensive actions such as fire operations and sprinkler actuation.
Most fire models are very elementary when dealing with the heat release of fires. Yet, this aspect of fire development could be a key issue, especially for human safety, as the developing phase of the fire defines the time available for the escaping from the fire area. Scientific literature refers to a simple t²-curve with a growth parameter value for slow, medium, fast and ultra-fast fire development.
There is almost no research into what parameters influence fire growth. In FRAME, three influence factors have been identified as contributing to the fire growth and hence to the fire severity: the volume/area ratio of the combustibles, the combustibility of the surfaces and the ignition characteristics of the surface materials. These have been identified by three parameters and combined in the fire spread factor i.

Generally speaking, localised fires are easier to handle: They do not impose a severe action on the building elements and can be approached for extinguishment. The transition from a localised fire to a fully developed fire is described in the scientific literature and expressed as a function of the fire heat release, the (square root) of the height between ceiling and floor, and the area of available ventilation openings. (E.g. Thomas’s flashover correlation, ventilation limit theory by Kawagoe).
In FRAME this relationship is found back in the ventilation factor v, which is calculated in a similar way with the log of the mobile fire load, the venting ratio k, and the (square root) of the height:

The probability of occurrence.

In FRAME, the combination of probability related factors is spread between P and D to fit practical design conditions, as some parameters are more linked to the buildings’ location, the others to the design of fire protection systems.

What is really evaluated is not the probability of "a fire", but the probability that the fire grows beyond control and reached the severity of the worst case. Kinney proposes a single value for the probability, but in fire oriented developed methods using the event tree approach, the final "worst case" probability is split up in several sub factors: one for the probability of ignition, one for the probability of early control, one for extinguishing by the fire brigade, sprinklers etc., and finally one for the probability that the an uncontrolled fire engulfs the compartment and destroys it. A similar combination of probability related factors is used in FRAME.

Probability of ignition.

A number of fire safety studies consider the probability of ignition to be more or less uniform within compartments with similar occupancies, supported by statistical values. A few surveys have established such values for offices, housing, industrial building: they are in a range around 10^-6 events per m² per year. The probability of ignition is therefore linked to the compartment floor area: the larger the compartment, the more likely a fire will occur. In prescriptive codes, compartment size limitations are apparently not linked to probability, but inspired by a concern for controllability of the fire by limiting the total quantity of combustibles (area x fire load).

This approach is probably too much of a simplification: In fact the size of a compartment does not only define the number of (evenly) distributed ignition sources, but has also an impact on the time necessary to discover the fire, the occurrence of secondary ignition sources and the time necessary for the fire brigade to reach the seat of the fire.
In FRAME the occurrence of ignition is used as part of the exposure evaluation: A building and its users are only exposed to the fire once ignition has occurred: The more ignition sources available in building or compartment, the higher the exposure is and hence the less a fire risk becomes acceptable. This is included in the acceptable risk A.

Building configuration is a risk aggravating element and is built in the area factor g, the level factor e, and the access factor z. The shape of the compartment, the presence of intermediate galleries and multiple levels and the location versus the access level are also included.

In the "natural fire concept" approach the increase in compartment size from 2500 m² to 10.000 m² causes a 15 % increase of fire severity value. For the same situation, the g-factor in FRAME doubles the value of P, which means a 100 % increase in fire severity value, reflecting not only the increased probability of ignition but also the decrease in controllability of the fire, resulting from the reduced capacity of occupants and fire brigade to gain early control in a large building or less accessible spot.

It should be noted that in FRAME the g-factor does not intervene in the risk assessment for the occupants. As any developing fire is considered as "worst case for people", the size of the compartment is not considered as relevant for severity and/or probability of the risk to persons. However, the size and shape of the compartment is considered in the calculation of A1, but this is a measure for the "exposure" and is dealt with separately.

Probability of controlling the fire.

1/D indicates the probability that a fire develops fully into a catastrophic situation: A high level of protection reduces considerably the probability of such event.
Statistical fire studies estimate the probability of early control of the fire by the occupants to be between 45 and 75 % of the cases, based on comparisons of the number of insurance claims and the number of fire brigade interventions in areas where both data were well documented.
The probability of effective control by fire brigades and sprinklers again is derived from statistical information on insurance claims: e.g. from the ratio between medium value and high value insurance claims, an average fire brigade effectiveness (= limiting the fire to the room of origin) of 90% is deduced. Sprinkler reliability is reasonably documented, so effective sprinkler control can be evaluated. The main causes of sprinkler system failure are also well known and the reliability of a particular sprinkler protection can be fairly well assessed. Anchor points are also the premium rebate percentages used by the insurance industry for active fire protection: higher rebates mean that the final cost of the fire is statistically lower and thus that corresponding fire protection systems are more reliable.

Reliability of protection elements in FRAME.

FRAME protection degree sub factors W (water supply), N (normal protection), S (special protection), U (escape) and Y (salvage) deal with a large number of variants of design features, active fire protection devices and systems, fire fighting organisation, etc., as well with reliability aspects. Early control by the occupants is e.g. considered as part of the normal protection. It can be easily checked that the values used in the evaluation of these factors reflect the relative contribution of these features to the overall probability of successful control of fire before it reaches a critical situation. A lack of water supply on the premises results in a value for W, which just means that the fire brigade has one chance in two to extinguish the fire with the water in their trucks. The combined result is a "probability" correction for the risk assessment formula.

In FRAME the probability of control is written as a division by "protection factors". The values of N and W are in fact always £< 1, so 1/W and 1/N give "failure rates"³ 1: When the quality of the water supply and of the normal protection are substandard, the probability that a fire can be controlled is reduced. The values of S, F (and U and Y) are always ³ 1: the higher these reliability factors, the lower the probability of failure.

Probability of building collapse.

The probability of a final "victory" of fire depends in the end on the fire resistance of the structural and separating elements compared to the estimated duration of the fire. In general, codes require a certain level of fire resistance for structural and non-structural elements, compared with the available fire load as basis combined with a safety factor to reduce the probability of ruin by fire. A typical fire in a non-industrial environment has an average ISO 834 duration between 30 and 45 minutes.
Code requirements basically start with 30 minutes fire resistance for small and low-rise buildings, with increasing levels of requirements for medium height; taller and high rise buildings. As the fire duration does not basically change with the height of the building, the higher fire resisting requirements are in fact safety factor applications to reduce the probability of collapse in case the fire breaks out of the original fire compartment into other levels of the building.

FRAME deals with this aspect in the resistance factor F and reckons with three assumptions.
The first is that the available stability in case of fire, is the joined result of the stability of structure the roof, floors, walls and internal separations. There is no scientific proof for counting these in a 50 %, 25%, 12.5 % and 12.5 % combination, yet there is no research done to support a better guess.
The second is that the value of F has to reflect the increased reliability of high fire resistance performance components, but also that the higher fire resistance may not be needed, certainly if the fire load is limited. This is dealt with in the first part of the F-formula, which gives a "bent" increase for F versus fire resistance. The absolute value of F also increases in the same way as the safety factors applied in building codes and at the same time follows broadly the same curve as the e-factor, so that the traditional link between building height and fire resistance requirements is also observed.
The third assumption is that building designers shall neither rely entirely on active fire protection systems or on passive fire resistance. This is accomplished by the second term of the F-formula, where the value of the special protection S is used to decrease the final value of F.

The exposure component in FRAME.

The EN954-1 approach of risk evaluation indicates that a higher level of protection reliability is required when the exposure of the subjects to the risk is frequent or prolonged. There is thus a requirement for some measurement of the exposure.
As fire is a rather rare event, the main consideration that defines life safety in fire situations will be the exposure time. For the activities, the duration of the fire is only one element: the consequences of a fire are not ended when the fire is out, the business interruption or reduction can continue for several months, the reconstruction time for a building can be very long, and unique objects can be destroyed for ever.
These considerations have resulted in three slightly different formulas in FRAME to calculate the exposure, one for the property, one for the people, and one for the activities.

Exposure for people

Usually people are considered to be safe, when they have left building on fire: the most evident measurement for the exposure is the evacuation time. But experience learns that the fire propagation in a building is not a uniform phenomenon and that rapid fire spread is the major reason for fire victims. This means that to evaluate correctly the exposure of people, evacuation time and fire propagation shall be jointly considered. In FRAME this results in the formula:

A1 = {1.6 – a} - ( t + r )

Exposure for property

A = {1.6 – a} - (t + c1 + c2)

Exposure for the activities.

A2 = {1.6 – a} - (c1 + c2 + d)

Risk distribution.

The general KINNEY formula: { Sev * Poc * Exp < C} gives the impression that there is a constant level of acceptable risk. This is not entirely true, as the values used to define the components Sev, Poc and Exp are defined on a non-linear scale.

The general public expects that the risk level would be more or less uniform for similar risk situations, but real accidents prove from time to time the contrary. Code requirements are normally NOT retro-active, which means that very often older buildings are less fire safe that new ones, and that a uniform legally accepted risk level does not exist.

A fire risk assessment method cannot link fire safety to a construction date; on the contrary it should be a tool to upgrade existing situations to present standards by supporting the equivalency concept.

But even in comparable risk situations, the expression for the risk level should be modulated to reflect individual differences.
This is the basis for fire insurance premium rates. It is basically the price for transferring risk to the insurance company. Basically, there is a standard rate per business category, reflecting the statistical observed fire damage/ property value ratio, based on the average fire severity for that business category.
The basic fire insurance rate is then modified for the presence of well know risk varying circumstances, such as heating systems, use of flammable liquids, welding operations, etc. In such situations, the possibility of ignition is increased and in order to obtain the same premium rate (relevant for the risk level) additional safety precautions are required.

It means that instead of a constant acceptable limit of risk, there should be a modulated limit, linked to identifiable ignition sources.

In FRAME, this condition is met by the { 1.6 - a } common part of the acceptable risk A - formulas. This part of the formula transforms the risk value in a way similar to insurance tariffs. The value for that part of A means actually that a higher number of ignition sources in the compartment will increase the overall fire risk. It also emphasises the need for fire prevention, where the most elementary rule is to reduce as much as possible the combination of fire load and ignition sources in one place.

Risk aversion

People do not like high severity risks even with low probabilities. This risk aversion is the basis for the whole insurance industry. By buying an insurance policy, the owner a a building transforms a high severity / low probability loss ( my house being destroyed by a fire) into a high probability / low severity loss : I have to spent every year a bit of my money on insurance premiums.
This aspect of risk aversion must be reflected in the results of any risk assessment method. Kinney solved this by defining the values of the contributing factors on a non-linear scale. In FRAME, risk aversion is incorporated in the formulas used for P, A and D.

Risk value expression.

The expression of the fire risk on a numerical scale is very much a convention, in the same way as using metric or American units for the characteristics of a building. In the KINNEY method risk values vary between 0.05 and theoretically 10.000 (a continuous threat of catastrophe). Why does FRAME (and its predecessor Gretener) use a scale that locates the value of the risk in a range around 1?

The most elementary reason is that Gretener originally wanted to develop a technical system for insurance premium rates, and these happen to be around 1 ‰ of the insured value. To obtain this it was more convenient to work with logarithmic based expressions. A lot of work has been done by trial and error to find suitable coefficients to transform measurable and identifiable data such as building dimensions, system characteristics, reliability data into working formulas.

The output of a risk assessment has to be communicated to the stakeholders in a way they can reach informed decisions. Ranking systems with a benchmark, like NFPA 101A FSES, are therefore much appreciated by the inexperienced observer.

Probability and severity combinations, like 5.10 ^-6 deaths / million hours are understandable by scientists but mean very little for decision makers, unless compared with some benchmark, e.g. national averages.

A loss potential calculation in thousands of Euros or dollars will not ring a bell for an insurance broker, but a correspondingly high fire insurance premium rate will do. The cost estimate becomes even more complicated when the costs are unevenly distributed among the interested parties. A particular complex situation is e.g. a large warehouse where the owner of the building, the owner(s) of the content and the user of the building are different companies, with competing interests. Decision makers are more interested to know what level of fire safety they can get for the budget foreseen to be spent on fire protection.

A loss potential calculation is therefore more useful. FRAME results can be transformed into a) a fire loss estimate as % of the compartment; b) an insurance premium rate; and c) a relative probability of victims, compared to a residential fire risk. Most managers are happy with this kind of information that is easily understandable and still gives them the freedom of decision making.

Conclusion.