Vibration-Based Structural Health Monitoring of Highway Bridges Using Accelerometer Arrays

A bridge's natural vibration characteristics — the frequencies at which it naturally wants to sway or bounce, and the shapes it takes while doing so — are a direct function of its stiffness and mass distribution. Vibration-based structural health monitoring (SHM) puts arrays of accelerometers on a structure to continuously measure that vibration, under nothing more than ambient traffic and wind excitation, and tracks how those vibration characteristics change over time as a way to detect structural change between scheduled inspections.

Why Vibration Reveals Structural Change

A structure's natural frequencies are governed by the classic relationship between stiffness and mass: frequency increases with stiffness and decreases with mass, for a given mode shape. Structural damage — a crack, section loss from corrosion, a failed connection, a degraded bearing — almost always reduces local or global stiffness, which in turn produces a small but measurable downward shift in one or more natural frequencies, and can also subtly change the mode shapes themselves (the pattern of relative motion across the structure at that frequency).

The practical challenge is that frequency shifts from real damage are often small — a percent or less for early-stage damage — and environmental factors like temperature also shift natural frequencies (a bridge deck is measurably stiffer when cold than when hot, for concrete and asphalt both), so distinguishing damage-induced change from environmental noise is the central technical problem SHM systems have to solve.

Sensor Network Design

Accelerometer placement follows the target mode shapes: sensors are positioned at locations expected to show meaningful motion in the vibration modes of interest — commonly at quarter-span and mid-span locations for the primary vertical bending modes, with additional sensors near supports and at locations of particular structural concern (a specific girder line, a known deficient connection). Sampling rates are chosen well above the highest frequency of interest — typically 100 Hz or more for highway bridge applications where relevant modes are usually below 10–15 Hz — and data is either transmitted continuously to a central logging system or stored locally with periodic retrieval, depending on the project's power and connectivity constraints.

Most highway bridge SHM systems rely on ambient (operational) vibration from passing traffic and wind rather than artificial excitation, since ambient response is available continuously at no operational cost — a distinction from a formal ambient/forced vibration test typically done once at commissioning to establish the initial baseline.

Modal Analysis & Damage-Sensitive Features

Raw acceleration time-histories are processed through operational modal analysis techniques to extract natural frequencies, mode shapes, and damping ratios from ambient response data alone (without knowing the input excitation, unlike classical forced-vibration modal testing). Several damage-sensitive features are then tracked over time against the established baseline:

Natural frequency shift — the simplest and most widely used feature, but sensitive to both damage and environmental effects, requiring some form of environmental normalization to be reliable.

Mode shape curvature / modal strain energy change — more sensitive to the location of damage than frequency shift alone, since damage at a specific location distorts the mode shape locally even when the overall frequency shift is small.

Statistical pattern recognition on raw response data — machine-learning and statistical approaches that skip explicit modal extraction and instead flag when the overall vibration response pattern deviates from the established normal-condition baseline.

How It Works: Video Explainer

The clip below walks through the underlying concept in more depth than text alone conveys well — how sensor data becomes a usable structural indicator, and where the real engineering challenges (noise, environmental effects, sensor placement) actually show up in practice.

Implementation Challenges

Despite two decades of active research and increasing commercial deployment, vibration-based SHM implementation still faces well-recognized practical challenges: establishing a reliable baseline requires enough data across enough environmental conditions (temperature range, traffic loading variation) to separate normal variability from genuine change; sensor and data-acquisition reliability over years of outdoor deployment is itself a maintenance burden, with sensor drift or failure needing to be distinguished from actual structural signal; and false-positive management is critical for owner trust — an SHM system that generates frequent unwarranted alarms quickly gets ignored, so alarm thresholds and confirmation protocols (does a flagged change get confirmed by targeted visual inspection before triggering a response?) are as much a program-design question as a technical one.

1. Farrar, C.R. and Worden, K., Structural Health Monitoring: A Machine Learning Perspective, John Wiley & Sons.
2. ISO 18649, Mechanical Vibration — Evaluation of Measurement Results from Dynamic Tests and Investigations on Bridges, International Organization for Standardization.
3. FHWA, Bridge Structural Health Monitoring Reference Manual, Federal Highway Administration.