Performance-Based Engineering and Digital Analysis

Introduction and context

Structural fire safety has traditionally relied on prescriptive provisions: minimum fire-resistance ratings, limiting temperatures, compartment sizes, and specified protection thicknesses. These provisions are valuable because they are simple to administer and reflect accumulated experience. They may, however, be conservative or unsuitable for unusual buildings, innovative materials, complex geometries, large atria, transport facilities, or structures in which several safety systems interact.

Performance-based engineering (PBE) provides an alternative method. Instead of demonstrating only that a prescribed detail has been followed, the engineer defines safety objectives, develops credible fire scenarios, predicts the resulting thermal and structural response, and demonstrates that acceptance criteria are satisfied. Digital analysis supports this process through zone models, computational fluid dynamics (CFD), heat-transfer models, and finite-element analysis (FEA).

The method is not simply “running software.” It is an engineering argument based on a defined hazard, justified assumptions, appropriate models, verified calculations, validated methods, sensitivity studies, and transparent documentation. A sophisticated model with an unrealistic fire or poorly defined boundary conditions can produce less reliable conclusions than a simple, well-justified model.

Learning objectives

By the end of this chapter, the learner should be able to:

1. Explain the principles and workflow of performance-based structural fire engineering.
2. Develop credible design fire scenarios using fuel, ventilation, detection, suppression, and occupancy information.
3. Distinguish the capabilities and limitations of zone, CFD, heat-transfer, and finite-element models.
4. Apply fundamental heat-transfer and structural-fire equations with consistent units.
5. Carry out verification, validation, sensitivity, and uncertainty checks.
6. Interpret model outputs without confusing numerical precision with engineering accuracy.
7. Prepare a traceable technical report suitable for peer review, approval, and future checking.

1. The performance-based design framework

A performance-based fire design normally follows this chain:

objectives → hazards → scenarios → thermal environment → material response → structural response → acceptance criteria → decision and documentation

1.1 Safety objectives and acceptance criteria

Objectives should be stated in measurable terms. Typical objectives include:

• preventing structural collapse during the required fire duration;
• maintaining compartmentation and preventing fire spread;
• protecting occupants and emergency responders;
• preserving essential facilities and business continuity;
• limiting damage to neighbouring property or the environment.

Acceptance criteria convert objectives into checks. Examples include:

• a member temperature remaining below a specified limiting temperature;
• calculated deflection or strain remaining within a specified limit;
• no loss of load-bearing capacity during the design scenario;
• sufficient time for evacuation or firefighting intervention;
• fire-resisting boundaries maintaining integrity and insulation.

Criteria must come from the governing building regulations, accepted design guidance, product approvals, or an agreed engineering basis. They must not be selected merely because a software package reports them conveniently.

1.2 Performance criteria and limit states

Structural fire design concerns several limit states:

• thermal limit state: temperatures cause unacceptable degradation of strength, stiffness, connections, or protection;
• mechanical limit state: gravity and fire-induced actions exceed resistance;
• stability limit state: buckling, lateral torsional buckling, progressive collapse, or connection failure occurs;
• integrity limit state: fire barriers, joints, or penetrations permit unacceptable fire or smoke spread;
• functionality limit state: an essential structure cannot perform its required post-fire function.

A structure may remain below a member-temperature limit yet fail through connection rupture, restraint-induced forces, instability, or loss of composite action. Therefore, temperature is an input to structural assessment, not a complete structural criterion.

2. Developing credible fire scenarios

2.1 The design fire as an engineering model

A design fire is a simplified representation of a possible fire, not a prediction of one inevitable event. It should reflect the building’s occupancy, fuel load, ventilation, geometry, suppression systems, fire detection, response procedures, and likely fire locations.

A credible scenario identifies:

• ignition source and first fuel item;
• fire location and potential alternative locations;
• fuel type, quantity, arrangement, and heat-release characteristics;
• ventilation openings and their evolution during the fire;
• sprinkler or suppression operation and possible failure;
• detection, alarm, evacuation, and firefighting actions;
• exposure of structural members and protection systems;
• consequences of fire spread between compartments.

Scenarios should include both likely and severe credible cases. A very unlikely event may be relevant if its consequences are catastrophic, but the probability and consequence basis should be stated explicitly.

2.2 Heat-release rate

The heat-release rate (HRR) is commonly represented as a time-dependent function. A growing fire is often approximated by a t-squared curve:

where Q̇(t) is the HRR in kW, t is time in seconds, and α is the fire-growth coefficient in kW/s². This approximation applies only over the period in which the assumed growth behaviour is reasonable. It does not automatically represent fuel exhaustion, ventilation limitation, sprinkler control, or decay.

The time to reach a target HRR is:

where Q̇_max is the selected peak HRR. The selected growth rate should be supported by experimental data, recognized guidance, or a conservative engineering assumption appropriate to the fuel arrangement.

The total energy released over a period is:

For a t-squared fire continuing to time t, this becomes:

The result has units of kW·s, equivalent to kJ. This calculation is meaningful only if the assumed HRR function remains valid throughout the integration interval.

2.3 Ventilation and compartment effects

A fire may be fuel-controlled while sufficient oxygen is available, or ventilation-controlled when oxygen supply limits burning. Openings, doors, façade failures, smoke vents, and leakage paths can change the thermal environment substantially. A scenario should therefore state whether openings are fixed, automatically activated, manually operated, or allowed to fail at specified conditions.

Compartment fires may exhibit a growth phase, fully developed phase, and decay phase. A standard nominal fire curve can be useful for code comparison, but it is not necessarily a physically realistic representation of a particular compartment. A natural-fire model may be more appropriate when fuel load, ventilation, and compartment geometry are known with reasonable confidence.

2.4 Uncertainty in scenarios

Important uncertainties include fuel quantity, HRR, fire location, sprinkler reliability, door status, glazing failure, fire duration, smoke movement, and material properties. Rather than hiding uncertainty in a single “best estimate,” the engineer should identify parameters that could change the decision and test them through sensitivity analysis or scenario variation.

3. Digital modelling hierarchy

No single model is appropriate for every question. The modelling hierarchy should match the decision being made.

A practical workflow often uses these models sequentially: a zone model screens scenarios; CFD refines the gas environment where spatial effects matter; heat-transfer analysis calculates member temperatures; and FEA assesses structural response.

3.1 Zone models

A zone model divides a compartment into regions, commonly a hot upper layer and a cooler lower layer. Conservation of mass and energy is applied to each zone, with entrainment, vent flows, wall losses, and interface movement represented by simplified correlations.

Zone models are effective for preliminary design, parametric studies, and relatively simple compartments. They are less suitable where smoke stratification is strongly three-dimensional, openings produce complex jets, local plume impingement governs member heating, or the building contains large interconnected spaces.

The engineer must check whether the compartment geometry, ventilation arrangement, and fire location satisfy the assumptions of the selected zone formulation.

3.2 Computational fluid dynamics

CFD solves conservation equations over a discretized domain. In fire applications, the equations represent conservation of mass, momentum, energy, and species, together with turbulence, combustion, radiation, and boundary models. A common low-speed approximation treats the flow as variable-density and may use an LES turbulence model.

CFD outputs can include gas temperature, velocity, pressure, oxygen concentration, smoke concentration, and radiative exposure. These are field quantities, not exact measurements. Mesh size, time step, combustion model, radiation treatment, ventilation assumptions, and HRR prescription can materially influence results.

CFD should be used when spatial information affects a decision, such as smoke movement in an atrium, plume interaction with a roof, local heating of an unprotected beam, or cross-compartment smoke spread. It should not be selected solely because it appears more advanced.

3.3 Heat-transfer analysis

Heat-transfer analysis calculates temperatures through a member and its protection. A simplified one-dimensional conduction equation is:

where ρ is density in kg/m³, cₚ is specific heat in J/(kg·K), T is temperature in K or °C for temperature differences, t is time in s, k is thermal conductivity in W/(m·K), x is distance in m, and q̇ᵥ is volumetric heat generation in W/m³. For ordinary passive protection without internal heat generation, q̇ᵥ = 0.

At a surface, convection and radiation may be represented by:

where q̇'' is net heat flux into the surface in W/m², h is the convective heat-transfer coefficient in W/(m²·K), T_g is gas temperature in °C, T_s is surface temperature in °C, ε is effective emissivity, and σ = 5.670374419 × 10⁻⁸ W/(m²·K⁴) is the Stefan–Boltzmann constant. Absolute temperature must be used in the radiation terms; using Celsius directly in the fourth-power expression is incorrect.

Heat-transfer properties can depend strongly on temperature. Steel conductivity, concrete moisture, specific heat, spalling, protection conductivity, and contact resistance should be represented according to suitable standards, test data, or documented manufacturer information. An apparently precise thermal result may be misleading if the material data are uncertain.

3.4 Finite-element structural analysis

FEA represents members or assemblies as interconnected elements. In a nonlinear transient structural analysis, the general equilibrium statement is:

where M is the mass matrix, C is the damping matrix, u is the displacement vector, f_int is the internal resisting-force vector, f_ext is the external-load vector, and T denotes temperature-dependent effects. Dots indicate time derivatives.

Temperature changes may cause thermal strain. A simple free thermal strain relation is:

where ε_th is thermal strain, α_T is the coefficient of thermal expansion in 1/K, T is current temperature, and T₀ is the reference temperature. This relation is a simplification when the coefficient varies with temperature; in that case, the strain should be integrated over the temperature range.

If expansion is restrained, thermal strain generates force and may redistribute load. The analysis must represent realistic restraints, connection stiffness, composite action, imperfections, fire-induced degradation, and possible contact or separation. Boundary conditions are often more influential than mesh refinement.

4. Verification, validation, sensitivity and uncertainty

4.1 Verification

Verification asks: Was the mathematical model solved correctly? It includes checking equations, coding, units, convergence, conservation, mesh independence, time-step sensitivity, and implementation against benchmark problems with known solutions.

Examples include:

• checking that a heat-conduction model reproduces a one-dimensional analytical solution;
• confirming that a CFD model conserves mass and energy within acceptable numerical error;
• verifying that an FEA model reproduces a simply supported beam response before adding fire effects;
• repeating a calculation with a refined mesh or smaller time step.

Verification does not prove that the model represents the real building.

4.2 Validation

Validation asks: Does the model represent the physical system adequately for its intended use? Evidence may include comparison with fire experiments, furnace tests, compartment tests, published benchmark cases, or observed incidents. Validation should address the relevant regime: a model validated for a small burner test is not automatically validated for a large atrium fire.

Validation measures should be reported transparently. A useful error measure for a prediction y_i and reference value y_i^ref is:

and, when a nonzero reference permits normalization:

These are descriptive measures, not universal pass/fail criteria. The acceptable difference depends on the engineering decision, measurement uncertainty, and model purpose.

4.3 Sensitivity analysis

Sensitivity analysis determines how outputs respond to changes in inputs. A simple finite-difference local sensitivity is:

where R is the response of interest, x is an input parameter, and ∆x is a sufficiently small perturbation. A dimensionless normalized sensitivity is:

This indicates the approximate percentage change in response for a one-percent change in the input, provided the response and input are nonzero and the local relationship is meaningful.

Useful sensitivity variables include peak HRR, fire-growth rate, ventilation area, emissivity, protection thickness, concrete moisture, steel yield strength, restraint stiffness, load ratio, mesh size, and sprinkler activation time.

4.4 Uncertainty analysis

Uncertainty may be aleatory, arising from natural variability, or epistemic, arising from limited knowledge or modelling assumptions. A basic propagation approximation for independent inputs is:

where Var denotes variance, R is the response, and xᵢ are uncertain inputs. This first-order approximation assumes locally linear response and, in the displayed form, negligible covariance. If inputs are correlated, covariance terms must be added.

For nonlinear fire and structural models, Monte Carlo analysis may be used. Inputs are sampled from justified distributions, the model is executed for each sample, and the resulting response distribution is examined. A distribution must not be invented merely to obtain a probability; it requires evidence or an explicitly stated expert assumption.

5. Worked engineering example: protected steel beam

Consider an illustrative protected steel beam exposed on three sides. A heat-transfer analysis predicts a maximum steel temperature of 540 °C after 60 minutes. The project acceptance criterion is a limiting steel temperature of 550 °C for the specified load ratio and structural system.

The thermal margin is:

The result passes the stated temperature criterion by 10 °C. However, this is not automatically a robust pass. Suppose sensitivity runs produce 565 °C when protection conductivity is increased within its plausible range, and 575 °C when a ventilation assumption produces a hotter gas exposure. The decision is then controlled by the scenario and material uncertainty, not by the nominal 540 °C result.

A suitable engineering conclusion would state that the base case satisfies the criterion, but the design requires either confirmation of the protection properties, a more conservative exposure, additional protection thickness, or a structural analysis demonstrating adequate resistance at the sensitivity temperatures. The example illustrates why reporting one number without uncertainty and sensitivity information is inadequate.

6. Integrated case study: illustrative atrium and composite floor

Background

This fictional illustrative case concerns a four-storey office building with a central atrium. A composite steel-concrete floor surrounds the atrium, and several primary beams are partially protected. The prescriptive fire-resistance detail is difficult to apply because smoke movement, beam exposure, and compartment geometry differ from the standard test configuration.

Problem

The design team must determine whether the proposed protection and structural system maintain stability for credible fires on the ground floor and at an upper-level retail interface. Concerns include plume impingement on beams, smoke spread through the atrium, sprinkler activation, and restraint of floor beams by surrounding construction.

Analysis approach

First, the team defines objectives: prevent disproportionate collapse during the design fire, preserve protected escape routes, and maintain separation between the retail and office zones.

Second, four scenarios are developed:

1. a controlled sprinklered fire with early detection;
2. a sprinkler activation delay with a larger HRR;
3. a sprinkler failure and limited ventilation case;
4. an upper-level fire near the atrium edge with local beam exposure.

A zone model screens the scenarios and identifies the upper-level smoke layer and tenability concerns. Because the atrium flow is strongly three-dimensional, CFD is then used for the two most demanding scenarios. Gas temperatures and radiative conditions near the beams are transferred to a heat-transfer model. The resulting temperature fields are applied to a nonlinear FEA model of the composite floor, beams, columns, connections, and realistic restraint conditions.

The structural model includes temperature-dependent steel strength and stiffness, concrete degradation, thermal expansion, composite interaction, and geometric imperfections. It also examines connection rotation and the possibility of load redistribution after local yielding.

Sensitivity and validation

The team varies peak HRR, sprinkler activation time, protection thickness, beam restraint stiffness, and mesh size. The results show that sprinkler delay and protection thickness have the greatest effect on peak beam temperature, while connection stiffness strongly affects predicted deflection and axial force. A refined CFD mesh changes the maximum gas temperature near the beam by 4%, whereas changing the assumed opening condition changes it by 18%.

The zone and CFD models are checked against recognized benchmark cases. The heat-transfer model is checked against a protected-steel furnace result. The FEA procedure is first tested against a room-temperature composite beam calculation and then against available elevated-temperature test data.

Outcome

The nominal design satisfies the structural acceptance criteria, but the sprinkler-delay sensitivity case produces excessive beam temperatures and connection rotation. The team improves the design by increasing local protection, revising the atrium-edge barrier detail, and specifying monitoring and inspection requirements for the suppression system. The final report clearly distinguishes code requirements, modelling assumptions, calculated results, and conservative judgements.

Learning points

• Scenario definition can control the result more than numerical refinement.
• CFD is justified by the atrium geometry, not by its sophistication alone.
• Structural fire performance depends on connections, restraint, and load redistribution as well as member temperature.
• A nominal pass is insufficient when plausible uncertainties produce failure.
• Digital analysis is most credible when independent checks and validation evidence accompany it.

7. Documentation and professional governance

A performance-based report should be reproducible by a competent reviewer. It should include:

• project description, drawings, occupancy, and fire strategy;
• objectives, acceptance criteria, and governing regulations;
• scenario selection and excluded scenarios with reasons;
• HRR functions, fuel assumptions, ventilation, suppression, and detection;
• software name, version, solver settings, mesh, time step, convergence criteria, and material models;
• thermal and structural boundary conditions;
• verification and validation evidence;
• sensitivity and uncertainty studies;
• plots, tabulated results, critical locations, and failure definitions;
• limitations, engineering judgement, peer review, and conclusions;
• data files, model versions, units, and change-control records.

The report should avoid unsupported claims such as “the model proves safety.” Models provide evidence within their assumptions. The responsible engineer must explain why the evidence is sufficient for the stated decision.

Ethical and regulatory considerations

Engineers must not manipulate scenarios, material properties, mesh settings, or acceptance criteria to obtain a preferred result. Suppression reliability, maintenance condition, emergency access, and occupancy changes should not be assumed without an accountable basis. Software licensing, version control, cybersecurity, data retention, and independent checking are part of professional governance.

Artificial intelligence may assist with input checking, document comparison, mesh-quality screening, or surrogate modelling. It should not replace engineering judgement or validation. AI-generated assumptions and automatically detected patterns require human review, traceability, and protection against hidden bias or fabricated sources.

8. Best-practice checklist

Before accepting a digital fire-engineering conclusion, ask:

1. Is the safety objective explicit and measurable?
2. Are the fire scenarios credible, sufficiently severe, and linked to the building use?
3. Does the selected model match the physical question?
4. Are all units, material properties, boundary conditions, and initial conditions documented?
5. Has the model been verified numerically and validated physically where possible?
6. Have mesh, time-step, solver, and convergence effects been checked?
7. Have key scenario and material uncertainties been varied?
8. Are structural connections, restraint, load redistribution, and failure modes included?
9. Are acceptance criteria authoritative and applied consistently?
10. Could an independent engineer reproduce and challenge the analysis?

Chapter Summary

Performance-based engineering establishes safety by linking objectives, credible fire scenarios, thermal exposure, material response, structural behaviour, and acceptance criteria. Zone models are efficient for compartment-scale screening; CFD resolves complex spatial flow and smoke behaviour; heat-transfer models calculate member temperatures; and finite-element models evaluate deformation, stability, load redistribution, and failure.

The validity of a result depends on more than software selection. Scenario credibility, boundary conditions, material data, verification, validation, sensitivity analysis, uncertainty treatment, and documentation are essential. A model should be no more complex than necessary, but no simpler than the physical question requires. Professional conclusions must state assumptions and limitations and must distinguish a nominal calculation from a robust demonstration of safety.

Key Terms

Performance-based engineering: Engineering design that demonstrates measurable safety performance rather than relying exclusively on prescriptive details.

Design fire: An engineering representation of a credible fire scenario used for analysis.

Heat-release rate: The rate at which a fire releases thermal energy, usually expressed in kW.

Zone model: A compartment fire model that represents regions such as hot and cool layers using averaged properties.

Computational fluid dynamics: Numerical solution of fluid-flow and heat-transfer equations over a spatial domain.

Heat-transfer model: A model that predicts temperature distribution within protection and structural materials.

Finite-element analysis: Numerical analysis that divides a structure into elements to calculate response.

Thermal strain: Deformation caused by temperature change in a material.

Thermal boundary condition: A prescribed or calculated condition governing heat flow at a surface.

Verification: Demonstration that equations and numerical implementation have been solved correctly.

Validation: Demonstration that a model represents physical behaviour adequately for its intended use.

Sensitivity analysis: Study of how output changes when an input parameter is varied.

Uncertainty analysis: Evaluation of the effect of incomplete knowledge or variability on predicted results.

Aleatory uncertainty: Uncertainty associated with natural variability.

Epistemic uncertainty: Uncertainty associated with limited knowledge or modelling assumptions.

Acceptance criterion: A defined condition used to judge whether performance is satisfactory.

Load redistribution: Transfer of forces to alternative structural members after stiffness or strength loss.

Mesh independence: Condition in which further mesh refinement has an acceptably small effect on results.

Design fire scenario: A specified combination of fire, ventilation, suppression, occupancy, and structural exposure conditions.

Technical traceability: Ability to connect results to inputs, methods, software versions, checks, and decisions.

Review Questions

Short Answer Questions

1. What is the main difference between prescriptive and performance-based fire design?
2. State two situations in which CFD may be more appropriate than a zone model.
3. What is the purpose of verification?
4. Why must absolute temperature be used in the radiation term of a heat-flux equation?
5. Name four uncertain parameters that can affect structural fire predictions.

Long Answer Questions

1. Explain the complete workflow from safety objectives to documented performance-based approval.
2. Compare zone, CFD, heat-transfer, and finite-element models, including their outputs and limitations.
3. Discuss how ventilation, suppression, and fuel arrangement influence a credible design fire.
4. Explain why structural fire assessment must consider restraint, connections, and load redistribution rather than member temperature alone.
5. Describe how verification, validation, sensitivity analysis, and uncertainty analysis complement one another.

Analytical / Application Questions

1. A fire is represented by Q̇(t) = 0.012t² kW. Calculate the time required to reach 1,200 kW and state one reason why the t-squared model may not remain valid after that time.
2. A protected beam has a calculated temperature of 535 °C and a limiting temperature of 550 °C. Explain whether this is sufficient evidence for approval if a plausible ventilation sensitivity case predicts 570 °C.
3. For an atrium with strong plume interaction, smoke stratification, and complex openings, propose a model hierarchy and identify the principal verification, validation, and sensitivity checks you would require.

Multiple Choice Questions with Answers

Question 1

Which statement best describes performance-based fire engineering?

A. It eliminates the need to define acceptance criteria.

B. It demonstrates that measurable safety objectives are achieved for credible scenarios.

C. It requires CFD for every fire-safety assessment.

D. It replaces engineering judgement with software output.

Correct Answer: B. It demonstrates that measurable safety objectives are achieved for credible scenarios.

Question 2

In the relation Q̇(t) = αt², what does α represent?

A. The compartment volume

B. The fire-growth coefficient

C. The steel yield strength

D. The radiation constant

Correct Answer: B. The fire-growth coefficient

Question 3

Which model is generally most suitable for rapid screening of a relatively simple compartment fire?

A. Zone model

B. Three-dimensional nonlinear FEA only

C. Detailed fracture-mechanics model

D. Post-fire forensic model

Correct Answer: A. Zone model

Question 4

What is the primary purpose of model verification?

A. To prove that the building will never burn

B. To demonstrate that the numerical model has been solved and implemented correctly

C. To select the cheapest construction material

D. To replace experimental evidence in all cases

Correct Answer: B. To demonstrate that the numerical model has been solved and implemented correctly

Question 5

Which temperature scale must be used in the fourth-power radiation terms?

A. Celsius only

B. Fahrenheit only

C. Absolute temperature in kelvin

D. Any scale gives the same result

Correct Answer: C. Absolute temperature in kelvin

Question 6

What does validation examine?

A. Whether the model represents relevant physical behaviour adequately

B. Whether the report has the largest possible number of figures

C. Whether the mesh contains exactly one element per metre

D. Whether the software has a graphical interface

Correct Answer: A. Whether the model represents relevant physical behaviour adequately

Question 7

A normalized local sensitivity approximately equal to 2 means that, locally, a one-percent input change produces approximately what response change?

A. 0.5%

B. 1%

C. 2%

D. 200%

Correct Answer: C. 2%

Question 8

Which item is a structural consequence of restrained thermal expansion?

A. Thermal force and possible load redistribution

B. Automatic reduction of fire duration

C. Elimination of connection forces

D. Guaranteed prevention of buckling

Correct Answer: A. Thermal force and possible load redistribution

Question 9

Why may a nominal temperature pass be insufficient?

A. Because temperature is never relevant to structural fire design

B. Because plausible uncertainties may produce temperatures or structural responses exceeding criteria

C. Because all calculations must fail conservatively

D. Because only prescriptive rules can be used

Correct Answer: B. Because plausible uncertainties may produce temperatures or structural responses exceeding criteria

Question 10

Which information most directly improves reproducibility of a digital fire analysis?

A. Only the final maximum temperature

B. Software version, units, assumptions, model settings, input data, and checking evidence

C. A statement that the software is widely used

D. A graphical rendering without numerical data

Correct Answer: B. Software version, units, assumptions, model settings, input data, and checking evidence