ComparisonInfrastructure

Bridge structural health monitoring compared with periodic inspection

Bridge structural health monitoring earns its cost on a minority of structures: those with a known defect to watch, a failure mode that develops between inspection visits, or a consequence of failure serious enough to justify continuous data. For most bridges, periodic hands-on and drone inspection remains the better buy. Monitoring never replaces mandated inspection; it changes what engineers know between visits.

Reviewed 10 October 2026 7 min read

On this page

Hands-on, drone and continuous monitoring side by side
The mandated inspection baseline that monitoring sits on top of
SHM sensors and the engineering question each one answers
From raw signals to an alert an engineer will act on
Which bridges justify continuous monitoring
A hypothetical steel truss bridge with fatigue concerns
Feeding monitoring results into bridge management and capital plans
Questions and answers
Sources

Hands-on, drone and continuous monitoring side by side

Criterion	Hands-on visual inspection	Drone and robotic inspection	Continuous SHM
Finds best	Cracks, section loss, bearing and joint defects within arm's reach	Surface defects on hard-to-reach elements, mapped imagery for comparison	Changes in strain, movement, vibration or scour between visits
Misses	Anything that develops and recovers between visits	Hidden defects, and anything needing touch, sounding or cleaning	Defects away from the sensors, and most local damage
Evidence frequency	At each scheduled inspection	At each flight, often alongside inspections	Continuous or event-triggered
Main cost drivers	Access equipment, traffic control, qualified inspectors	Pilots, data processing, permissions to fly	Installation, power, communications, maintenance and analysis for years
Output	Condition ratings and a narrative report	Images, 3D models and defect maps	Time series, alerts and trend reports
Regulatory standing	The basis of mandated inspection regimes	Supports inspection when it meets the required procedures	Supplementary evidence; not a substitute for mandated inspection

The columns are complements, not rivals. The decision is which structures add the third column on top of the first two.

The mandated inspection baseline that monitoring sits on top of

In the US, the National Bridge Inspection Standards set the floor. Under the routine-interval rules, a bridge is inspected at least every 24 months, every 12 months if a deck, superstructure, substructure, culvert or scour rating is 3 or less, and up to every 48 months only when strict condition, load rating, fatigue-detail and scour criteria are met and FHWA has been notified². Nonredundant steel tension members get their own hands-on inspections, and underwater elements have a separate interval².

On UK motorways and trunk roads, CS 450 requires general inspections at intervals of 24 months and principal inspections every 72 months unless a longer interval is agreed, with risk-based principal intervals capped at twelve years and unavailable for structures such as those in poor condition or at medium or high flood risk³.

Neither regime lets a sensor network stand in for an inspector. Monitoring therefore has to justify itself on what it adds between visits: earlier warning of a change, evidence to support a load posting or repair decision, or confidence to keep a structure open while a replacement is designed.

SHM sensors and the engineering question each one answers

Sensor	What it measures	Typical question	Watch out for
Strain gauges	Local strain at a detail, vibrating-wire or foil	Is stress at this fatigue detail growing or cycling more?	Temperature effects and gauge drift over years
Accelerometers	Vibration, from which natural frequencies are derived	Has global stiffness changed after an event?	Frequencies shift with temperature and traffic mass
Tilt, displacement and GNSS	Rotation, joint movement, settlement	Is a pier rotating or a bearing locking up?	Seasonal thermal movement that looks like a trend
Fiber-optic sensing	Strain and temperature along a cable, point or distributed	Where along a girder is strain concentrating?	Installation quality and cost of interrogators
Acoustic emission	Energy released by active crack growth	Is a known crack actively growing?	Noise from traffic, rain and fretting connections
Scour sensors	Bed level at piers by sonar, buried float-out devices or pier tilt	Is the foundation being undermined during this flood?	Debris damage and loss of sensors in the event itself

Choose sensors from the failure mode backwards. A system that measures everything usually answers nothing in particular.

From raw signals to an alert an engineer will act on

Most bridge data should be processed at the edge. Gateways on the structure compute summaries, detect events such as an overweight vehicle or a sudden tilt, and send compact records rather than raw high-rate streams over cellular links that may also run on batteries or solar power. Our page on edge AI and IoT covers the device side of that design.

Before any threshold, remove the environment. Temperature changes strain, movement and natural frequency every day and every season; a regression or machine learning model of normal behavior, trained on a year or more of healthy data where possible, turns the raw signal into a residual. Alerts are set on residuals, with first-level thresholds derived from design checks and load ratings rather than from statistics alone.

Modal analysis tracks natural frequencies and mode shapes and is good at detecting global changes such as a seized bearing or a lost support. It is poor at finding small local damage, which is why fatigue cracks are watched with strain or acoustic emission at the detail itself. Anomaly detection models help sort genuine changes from sensor faults, but every alert tier still maps to a written action: review the data, send an inspector, restrict traffic or close.

Which bridges justify continuous monitoring

If
A steel bridge has fatigue-prone details or recorded cracks on nonredundant members.
Then
Monitor strain and, where cracks exist, acoustic emission at those details.
Crack growth can progress between hands-on inspections and its consequence is high.
If
A river crossing is rated scour critical or has unknown foundations.
Then
Install scour or pier tilt monitoring tied to the agency's flood response plan.
Scour develops during floods, exactly when inspectors cannot safely look.
If
A load-posted bridge must stay open until its replacement is built.
Then
Monitor the governing members and weigh-in-motion if overweight traffic is suspected.
Data supports keeping the posting, tightening it or closing early on evidence.
If
A bridge in good condition has redundant load paths and no known defects.
Then
Rely on scheduled inspection, possibly with drone imagery for comparison between cycles.
Monitoring would mostly confirm what inspection already shows, at a long-term cost.

A hypothetical steel truss bridge with fatigue concerns

Hypothetical example

Monitoring a known detail until the retrofit

For example, a state agency might own a riveted steel truss carrying a regional highway over a river, where the last inspection found a small crack at a floor beam connection. A retrofit is designed but cannot be built for two construction seasons.

The engineers fit strain gauges and acoustic emission sensors at the cracked connection and at two similar details, plus a temperature sensor. An edge gateway records stress ranges per heavy vehicle and flags bursts of acoustic activity. After a season of baseline data, residual-based thresholds are agreed: a trend triggers a special inspection; a sharp change triggers an immediate visit and a traffic restriction decision.

The bridge stays on its normal inspection schedule throughout. The monitoring answers one narrow question, whether the crack is growing, until the retrofit removes the need to ask it.

Feeding monitoring results into bridge management and capital plans

Monitoring pays off twice when its findings flow into the bridge management system. Measured stress ranges can refine a fatigue assessment, scour data can update a bridge's risk score, and a stable trend can justify deferring a replacement. Those updates belong in the risk-based capital planning process, where they change the likelihood of failure for that structure. ColdAI designs these sensor networks and the analytics around them as part of its infrastructure work¹.

Questions and answers

Can structural health monitoring extend NBIS inspection intervals?

Not on its own. Under the NBIS, extended intervals depend on condition ratings and specific criteria in 23 CFR 650.311², or on a more rigorous risk-based method that FHWA must approve. Monitoring data may inform an owner's understanding of risk, but owners should confirm with their FHWA division office before relying on it for any interval decision.

Can drones replace hands-on bridge inspection?

Drones reduce the need for access equipment and lane closures and produce imagery that is easy to compare between cycles. They cannot sound concrete, clean off corrosion or touch a fatigue detail, so they support rather than replace hands-on inspection, especially for nonredundant steel tension members and other elements where the rules expect close examination.

How long does a bridge monitoring system last?

Much less than the bridge. Batteries, gateways, communications modules and some sensors need replacing during a monitoring program, and cellular technologies change. Budget for maintenance, data analysis and eventual decommissioning from the start, and set an end date or review point tied to the question the monitoring was installed to answer.

Who should own structural health monitoring alerts?

A named structures engineer at the bridge owner, not the vendor. Vendors can run data quality checks and first-line review, but the decision to send an inspector, restrict traffic or close a bridge belongs to the owner. Agree the alert tiers, response times and escalation contacts in writing before the system goes live.

Is modal analysis enough to detect damage on its own?

Rarely. Natural frequencies respond strongly to temperature and traffic and only weakly to small local damage, so modal tracking is better at spotting global changes such as a lost bearing than at catching an early crack. Pair it with local sensors at the details that govern the structure's risk.

Sources

Infrastructure: structural health monitoring and sensor integration — ColdAI
23 CFR § 650.311 — Inspection interval (National Bridge Inspection Standards) — Legal Information Institute, Cornell Law School · checked 10 October 2026
CS 450 Inspection of highway structures (Design Manual for Roads and Bridges) — Standards for Highways · checked 10 October 2026

More in Infrastructure

Back to Infrastructure

Next step

Tell us which structure keeps your bridge engineers up at night

Send the latest inspection report and the defect or risk you are worried about. We will say whether monitoring would answer that question, which sensors it would need, or whether inspection alone is the better choice.

Discuss a monitoring question