14.2 Failure Analysis & Analytical Tools

14.2 Failure Analysis & Analytical Tools

Introduction: Beyond Origin and Cause

NFPA 921 Chapter 21, "Failure Analysis and Analytical Tools," addresses the structured analytical methods investigators use once initial data collection is complete, particularly in complex investigations involving product failures, system malfunctions, or disputed timelines. Where earlier chapters focus on collecting physical evidence at the scene, Chapter 21 focuses on organizing, sequencing, and testing that evidence against physical and engineering principles. The chapter covers three principal analytical tool sets: timeline construction, systems analysis, and fire modeling/fire testing. Each tool supports the scientific method's hypothesis-testing step (Chapter 4) by giving the investigator a structured, defensible way to confirm or eliminate candidate hypotheses—an increasingly important consideration given Daubert and Frye admissibility standards for expert testimony.

Timeline Analysis

A timeline is a chronological sequence of events, correlated from multiple independent sources, that establishes when key incidents occurred relative to one another. Constructing an accurate, well-sourced timeline is often the single most powerful analytical tool available to an investigator because it can confirm or eliminate hypotheses about fire growth, occupant behavior, and system performance without requiring destructive testing.

Sources commonly integrated into a fire investigation timeline include:

• 911 call logs and dispatch records, which establish when the incident was first reported and often capture caller observations about visible fire or smoke conditions.
• Fire alarm and building automation system logs, which time-stamp detector activations, sprinkler flow-switch activations, and HVAC state changes—often to the second.
• Fire department response records (dispatch, arrival, initial size-up reports, tactical logs), which document conditions observed by first-arriving crews.
• Video evidence from security cameras, doorbell cameras, traffic cameras, and bystander cell phones, which can be time-synchronized (accounting for clock drift) to show smoke or flame progression.
• Witness statements, weighted carefully since witness time estimates are frequently imprecise and subject to memory distortion, but valuable when corroborated by other sources.
• Electronic data sources, including vehicle event data recorders, smart-home device logs, utility smart-meter data (showing electrical load changes), and cell phone records.

The investigator plots these sourced events on a single chronology and looks for consistency or contradiction: does the sprinkler activation time correlate with the fire growth curve implied by witness observations? Does the 911 call time align with when smoke would plausibly have become visible from outside the structure given the hypothesized origin and fuel package? Timeline construction is also central to civil litigation, where the sequence of notice, response, and system activation frequently determines liability.

Systems Analysis

Systems analysis examines the interaction and sequencing of the physical and mechanical systems involved in an incident—structural, electrical, mechanical, and fire protection systems—to understand how a failure or malfunction may have propagated. Rather than looking at a single component in isolation, systems analysis maps how components interact, which failures are causal versus which are merely coincident, and where in a sequence a critical failure occurred.

A common systems-analysis tool is the fault tree, a top-down logic diagram that starts with the undesired event (for example, "ignition of flammable vapor") and branches downward through the necessary and sufficient conditions (fuel present, oxidizer present, ignition source present, and the absence of any barrier that should have prevented ignition) using AND/OR logic gates. Fault trees are particularly useful in product-failure and industrial-incident investigations because they force the investigator to explicitly identify every contributing condition rather than jumping directly to a single suspected cause. Event-sequence diagramming, a related technique, lays out the discrete events (a component overheats, a thermal fuse fails to open, insulation degrades, arcing begins) in sequence and tests each link against the physical evidence recovered from the scene.

Fire Modeling

Fire modeling uses mathematical and computational representations of fire dynamics to estimate conditions—temperature, smoke layer height, toxic gas concentration, heat flux—at specific locations and times during a fire's development. NFPA 921 identifies two broad categories of models relevant to investigators, and it is important for the investigator to understand their purpose and limitations even without personally running the software:

• Zone models (such as CFAST, the Consolidated Model of Fire and Smoke Transport) divide a compartment into a small number of relatively uniform zones—typically an upper hot-gas layer and a lower cooler layer—and calculate how conditions in each zone change over time. Zone models are computationally fast and well-suited to multi-room, multi-compartment scenarios, but they rely on simplifying assumptions that can reduce accuracy in complex geometries or where the two-layer assumption breaks down.
• Field models / computational fluid dynamics (CFD) models (such as FDS, Fire Dynamics Simulator) divide the modeled space into a fine three-dimensional grid and solve the underlying fluid dynamics and combustion equations at each grid cell, producing much more spatially detailed output at substantially greater computational cost and required user expertise.

Investigators use fire models to test whether a hypothesized origin, fuel package, and growth rate are physically consistent with observed conditions (detector activation times, witness observations of smoke or flame at a given time, structural damage patterns). A model's output is only as reliable as its input assumptions—fuel load, ventilation conditions, compartment geometry, and material properties must all be reasonably well established from scene evidence before the model is run, and every model requires sensitivity analysis to understand how uncertainty in the inputs affects the outputs. ASTM E1355, the Standard Guide for Evaluating the Predictive Capability of Deterministic Fire Models, provides the accepted framework for validating a model's applicability to a specific investigation before its results are relied upon.

Fire Testing

Fire testing—ranging from small bench-scale material tests to full-scale reconstructions—provides empirical rather than computational verification of a hypothesis. Full-scale reconstruction (rebuilding a room or replicating a fuel package under controlled conditions) can validate or refute a specific ignition scenario, but it is resource-intensive, must be carefully documented on video and with instrumentation, and must faithfully replicate the material properties, geometry, and ventilation conditions of the original scene to be probative. As with fire modeling, the investigator must document every deviation between the test setup and the actual scene conditions, since opposing experts will scrutinize those deviations in litigation.

Limitations and the Investigator's Analytical Discipline

NFPA 921 is explicit that fire modeling and testing are tools to support the scientific method, not substitutes for it, and that an investigator without adequate training should not attempt to run or interpret model output independently—this typically requires either specialized training or retaining a qualified fire protection engineer. Regardless of which analytical tool is used, the investigator must document: the specific hypothesis being tested, the inputs and assumptions used, the tool's known limitations, and how the output was weighed against the totality of the physical evidence. A model or test result that is inconsistent with well-documented physical scene evidence should prompt re-examination of the model's inputs, not automatic rejection of the physical evidence—and vice versa. This disciplined, tool-supported hypothesis testing is precisely what distinguishes a scientifically defensible conclusion from a purely opinion-based one under Daubert and Frye admissibility review.