Structural Health Monitoring: A Complete Guide for Infrastructure Owners
Infrastructure is ageing faster than it is being replaced. In the UK alone, over 70,000 highway bridges have a mean age exceeding 50 years. Worldwide, trillions of pounds of built assets are approaching or exceeding their original design life. Structural health monitoring (SHM) offers a way to manage this challenge — extending asset life, prioritising maintenance spending, and preventing catastrophic failures. This guide explains how SHM works, what technologies are available, and how to turn sensor data into confident engineering decisions.
What Is Structural Health Monitoring?
Structural health monitoring is the process of continuously or periodically measuring the physical behaviour of a structure — its movements, deformations, vibrations, temperature, and loading — and using that data to assess its condition and predict its future performance. Unlike a one-off inspection, SHM provides an ongoing record of how a structure is actually behaving under real-world conditions.
At its core, SHM answers a simple but critical question: is this structure performing as expected, and if not, what is changing and how fast?
The concept is not new — engineers have been placing instruments on dams and bridges for decades. What has changed dramatically in the past ten years is the availability of low-cost sensors, wireless connectivity, cloud-based data platforms, and machine learning algorithms that can process vast quantities of monitoring data automatically. SHM has moved from a niche research activity to a practical, scalable tool for infrastructure management.
Why SHM Is Becoming Essential
Several converging trends are driving the adoption of structural health monitoring across the infrastructure sector:
- Ageing infrastructure: Much of the world's critical infrastructure — bridges, tunnels, dams, retaining walls — was built in the mid-20th century and is now approaching or exceeding its intended service life. Wholesale replacement is financially and logistically impossible, so extending safe service life through monitoring is the pragmatic alternative.
- Increasing loads: Traffic volumes and vehicle weights have increased well beyond what many structures were designed for. Climate change is imposing new loading patterns — more intense rainfall, higher flood levels, thermal extremes — that were not in the original design codes.
- Shrinking budgets: Infrastructure owners must justify every pound of maintenance spending. SHM provides objective evidence to prioritise interventions: repair what needs repairing, and leave alone what is performing adequately. This is vastly more efficient than time-based maintenance schedules.
- Regulatory and safety pressure: High-profile failures — the Morandi Bridge collapse in Genoa (2018), the Champlain Towers South residential building collapse in Florida (2021) — have focused public and regulatory attention on the need for better monitoring of critical infrastructure.
- Construction-phase monitoring: Modern construction methods often require real-time monitoring during excavation, tunnelling, or adjacent construction to protect existing structures. SHM systems installed during construction frequently remain in place for long-term asset management.
The SHM Workflow: From Sensors to Decisions
A well-designed SHM system is not just a collection of sensors — it is a complete workflow that transforms physical measurements into actionable engineering decisions. The workflow has five stages, and weaknesses in any one stage undermine the entire system.
1. Sensor Selection and Design
The first step is defining what you need to measure and why. This sounds obvious, but it is where many monitoring programmes go wrong. A common mistake is installing sensors because they are available or fashionable, rather than because they answer a specific engineering question.
Good sensor selection starts with the failure modes you are trying to detect or track. For a bridge, this might be bearing settlement, deck cracking, cable force loss, or scour at the foundations. For a tunnel, it might be lining convergence, segment joint opening, or water ingress. Each failure mode suggests specific measurands (displacement, strain, tilt, temperature, water pressure) and specific sensor locations.
2. Installation and Commissioning
Even the best sensors produce poor data if they are badly installed. Installation quality is critically important and often underestimated. Key considerations include surface preparation, bonding methods, cable routing and protection, weatherproofing, and electromagnetic interference shielding. The installation must survive the full service life of the monitoring programme, which may be 20 years or more.
Commissioning involves verifying that every sensor is reading correctly, that data is flowing to the acquisition system, and that baseline readings are recorded before the structure is loaded or disturbed. Skipping proper commissioning is one of the most common and costly mistakes in SHM projects.
3. Data Acquisition
Data acquisition systems collect readings from the sensors at defined intervals and transmit them to a central location for storage and processing. This may be as simple as a datalogger in a weatherproof enclosure, or as complex as a multi-node wireless mesh network feeding a cloud platform in real time.
Key decisions at this stage include measurement frequency (once per hour? once per second? 1,000 times per second?), on-site processing vs. raw data transmission, power supply (mains, solar, battery), and communications (wired Ethernet, cellular 4G/5G, satellite, LoRaWAN).
4. Data Processing and Visualisation
Raw sensor data must be processed before it is useful. This includes applying calibration factors, compensating for temperature effects, filtering noise, detecting sensor faults, and converting raw measurements into engineering units. The processed data is then visualised in dashboards, time-series plots, and spatial maps that engineers can interpret quickly.
Automated processing pipelines are essential for any monitoring programme with more than a handful of sensors. Manual data processing does not scale, introduces human error, and creates unacceptable delays in identifying problems.
5. Interpretation and Decision-Making
The ultimate purpose of SHM is to inform decisions: is the structure safe to continue operating? Does it need maintenance? Should loads be restricted? Can it be given a longer service life? These decisions require engineering judgement informed by monitoring data, not just raw numbers on a screen.
This stage often involves comparing measured behaviour to design predictions, tracking trends over time, and assessing whether observed changes are within acceptable limits. It is the most valuable part of the SHM workflow — and the part most often neglected.
Types of Sensors Used in SHM
The sensor landscape for structural health monitoring is broad and evolving. Each sensor type has specific strengths and limitations. Choosing the right combination for a given application requires understanding these trade-offs.
Vibrating Wire Sensors
Vibrating wire (VW) sensors have been the backbone of geotechnical and structural monitoring for over 50 years. They work by measuring the resonant frequency of a tensioned steel wire, which changes when the wire is stretched or compressed. Because the output is a frequency rather than a voltage, VW sensors are inherently stable over long periods and resistant to electrical noise, moisture, and cable length effects.
Available configurations include strain gauges, piezometers (pore water pressure), load cells, earth pressure cells, crack meters, and settlement gauges. VW sensors are reliable, well-understood, and relatively inexpensive. Their main limitation is that they are point sensors — each one measures at a single location, so dense coverage requires many individual instruments and extensive cabling.
Electrical Resistance Strain Gauges
Foil strain gauges are the most widely used strain measurement technology in structural engineering. They are inexpensive, well-characterised, and available in a huge range of configurations. However, they are susceptible to drift over time, temperature effects, moisture ingress, and electrical noise over long cable runs. For short-term testing they are excellent; for long-term monitoring (years to decades), they require careful installation and periodic verification.
Accelerometers
Accelerometers measure dynamic response — vibrations, natural frequencies, and mode shapes. They are essential for monitoring structures subject to dynamic loading (bridges under traffic, buildings in seismic zones, towers in wind) and for operational modal analysis, which can detect changes in structural stiffness long before visible damage appears.
Modern MEMS accelerometers are small, low-cost, and low-power, making them ideal for wireless SHM nodes. Higher-performance piezoelectric accelerometers are used when greater sensitivity or bandwidth is required.
Fibre Optic Sensors
Fibre optic sensing is the fastest-growing segment of the SHM sensor market. Two main categories are relevant:
- Fibre Bragg Gratings (FBGs): Discrete sensors inscribed into an optical fibre. Each FBG reflects a specific wavelength of light that shifts with strain or temperature. Multiple FBGs can be multiplexed on a single fibre, with typical installations using 10–50 sensors per channel. FBGs offer excellent long-term stability, immunity to electromagnetic interference (EMI), and the ability to embed sensors in concrete, composites, or soil.
- Distributed fibre optic sensing (DFOS): Turns the entire length of an optical fibre into a continuous sensor, measuring strain, temperature, or vibration at every point along cables that can extend for kilometres. Technologies include Brillouin, Rayleigh, and Raman scattering-based systems. DFOS is particularly powerful for linear infrastructure (pipelines, tunnels, embankments) where thousands of point sensors would be impractical.
Fibre optic sensors are increasingly the technology of choice for new-build monitoring and for critical infrastructure where long-term reliability and EMI immunity are paramount.
GNSS (Global Navigation Satellite Systems)
GNSS receivers (GPS, GLONASS, Galileo, BeiDou) can measure the absolute position of a point on a structure to millimetre-level accuracy using real-time kinematic (RTK) processing or precise point positioning (PPP). GNSS monitoring is used for tracking the movement of bridges, dams, landslides, tall buildings, and retaining walls.
The main advantages are that GNSS measures absolute position (not relative displacement), works in all weather, and requires no line of sight between measurement points. The limitations are cost (survey-grade receivers are expensive), power consumption, and the need for a clear sky view — ruling out underground or indoor applications.
Tiltmeters (Inclinometers)
Tiltmeters measure angular rotation, typically with a resolution of a few arc-seconds. They are widely used for monitoring the tilt of retaining walls, bridge piers, building foundations, and dam crests. Both electrolytic and MEMS-based tiltmeters are common. They are simple to install, reliable over long periods, and provide a direct measure of rotational movement that is often the earliest indicator of foundation problems.
Crack Meters and Joint Meters
These sensors measure the opening, closing, or sliding of cracks and joints in concrete or masonry structures. They are conceptually simple — a displacement transducer spanning a crack — but critically important for monitoring heritage structures, tunnel linings, and concrete elements where crack growth is a key deterioration mechanism. Both vibrating wire and potentiometric types are available.
Wired vs. Wireless and IoT Systems
One of the most significant decisions in SHM system design is the choice between wired and wireless data acquisition. Both approaches have strong advocates, and the right choice depends on the specific application.
Wired Systems
Traditional wired systems connect sensors to a central datalogger via multi-conductor cables. The datalogger powers the sensors, takes readings, stores data locally, and transmits it to a remote server.
- Advantages: Proven reliability over decades. No battery replacement needed for sensors. High data throughput. Synchronised measurements across all channels. Well-understood by the monitoring industry.
- Disadvantages: Cable installation is expensive and disruptive, especially on existing structures. Cables are vulnerable to damage (construction activities, rodents, vandalism). Adding sensors after initial installation is difficult and costly. Cable runs exceeding 100–200 m can introduce signal degradation for some sensor types.
Wireless and IoT Systems
Wireless SHM nodes contain a sensor (or sensor interface), microcontroller, radio transceiver, and battery or energy harvester. They communicate via protocols such as LoRaWAN, NB-IoT, Zigbee, Wi-Fi, or proprietary radio links.
- Advantages: Dramatically reduced installation cost and time. Easy to add, move, or replace sensors. Ideal for retrofit monitoring of existing structures. Can reach locations where cabling is impractical. Scalable from a few nodes to hundreds.
- Disadvantages: Battery life is a persistent challenge (typically 1–5 years depending on measurement frequency and radio protocol). Data throughput is limited, especially for low-power wide-area networks like LoRaWAN. Time synchronisation between nodes is less precise than wired systems. Radio propagation can be unreliable in steel and concrete structures. The technology is less mature, and long-term reliability data is still accumulating.
The Practical Trade-Off
In practice, many modern SHM systems are hybrid: wired connections for sensors requiring high sampling rates or very long service life, and wireless nodes for supplementary measurements, hard-to-reach locations, or temporary monitoring during construction. The key is to match the communication architecture to the requirements rather than adopting a one-size-fits-all approach.
Key Applications
SHM is applied across virtually every category of infrastructure, but the monitoring strategy varies significantly depending on the structure type, its failure modes, and the consequences of failure.
Bridges
Bridges are the most monitored infrastructure type worldwide. Typical measurands include deck deflection under live load, bearing displacement, cable forces (for cable-stayed and suspension bridges), pier tilt, scour depth at foundations, and crack widths in concrete elements. Dynamic monitoring (accelerometers) is used for modal analysis to detect stiffness degradation. Fibre optic sensing is increasingly used for continuous strain profiling of bridge decks and girders.
Tunnels
Tunnel monitoring focuses on lining convergence (diameter changes), segment joint opening, water ingress, and ground settlement above the tunnel. During construction, monitoring is critical for verifying that excavation-induced ground movements remain within acceptable limits to protect surface structures. Distributed fibre optic sensing is particularly well-suited to tunnels, where a single cable along the tunnel crown can provide a continuous strain profile over kilometres.
Buildings
Building monitoring is typically triggered by adjacent construction (deep excavations, piling, tunnelling), suspected foundation settlement, or seismic risk. Common measurements include floor-level settlement, wall tilt, crack width, and dynamic response. For tall buildings, GNSS and accelerometer systems track wind-induced sway and long-term settlement.
Dams
Dams have the longest history of SHM, with many large dams instrumented since their original construction. Monitoring typically includes pore water pressure within and beneath the dam, seepage flow rates, crest settlement and displacement, internal temperature (for concrete dams during curing and service), and uplift pressure. Given the catastrophic consequences of dam failure, monitoring programmes are mandated by regulation in most countries.
Retaining Walls
Retaining wall monitoring focuses on wall tilt, lateral displacement, earth pressures, anchor loads (for tied-back walls), and pore water pressures behind the wall. Inclinometer casings installed within or behind the wall provide depth-resolved lateral displacement profiles. Increasing numbers of retaining wall projects now use fibre optic sensing for distributed strain measurement along the wall height.
Heritage Structures
Monitoring heritage buildings, bridges, and monuments presents unique challenges: the structure cannot be drilled, bolted, or modified in ways that compromise its heritage value. Low-impact sensor installation (surface bonding, clamp mounting, wireless nodes) is essential. Crack monitoring is almost always required, often supplemented by tilt and vibration measurements. Long monitoring durations (decades) are common, placing a premium on sensor stability and system maintainability.
Data Management Challenges
Modern SHM systems generate enormous volumes of data. A single distributed fibre optic sensing system measuring once per minute along 5 km of fibre at 1 m spatial resolution produces over 7 million data points per day. A 50-sensor vibrating wire system logging every 15 minutes produces nearly 4,800 readings per day, but over a 20-year monitoring programme that still totals over 35 million readings.
Managing this data presents several challenges:
- Storage and scalability: Cloud-based time-series databases (such as InfluxDB, TimescaleDB, or AWS Timestream) have largely replaced local file-based storage for modern systems. However, data storage costs can accumulate significantly over multi-year programmes, and data migration between platforms is non-trivial.
- Data quality: Sensor faults, communication dropouts, power interruptions, and environmental interference all introduce gaps and anomalies. Automated quality control routines are essential to flag suspect data before it enters the analysis pipeline.
- Real-time vs. periodic: Not all monitoring needs to be real-time. Construction-phase monitoring of adjacent structures may require readings every few minutes with immediate alerts. Long-term asset condition monitoring may only need daily or weekly readings with monthly reporting. Designing the system for the appropriate data rate avoids unnecessary cost and complexity.
- Data ownership and access: Who owns the monitoring data — the asset owner, the monitoring contractor, or the platform provider? Can the data be exported in open formats? These questions are often overlooked during procurement and become painful issues later.
- Long-term preservation: Infrastructure monitoring data has value over decades. Ensuring that data remains accessible and interpretable as technology platforms evolve requires careful attention to metadata, documentation, and format standards.
From Data to Decisions
The gap between collecting data and making decisions is where many SHM programmes underperform. Sensors are installed, data is collected, but no one is systematically analysing the data or translating it into maintenance actions. Bridging this gap requires a structured approach.
Threshold-Based Alerts
The simplest decision-support mechanism: define upper and lower limits for each measurand, and trigger an alert when a threshold is breached. For example, an alert when a crack width exceeds 0.3 mm, or when a tilt rate exceeds 1 mm/m per week. Thresholds are typically set in three tiers — green (normal), amber (requires review), red (requires immediate action) — based on design limits, code requirements, or engineering judgement.
Threshold-based alerts are straightforward to implement and easy to understand. However, they are reactive (the problem must already exist to trigger an alert), and setting appropriate thresholds requires experience. Too tight, and the system generates constant false alarms; too loose, and real problems are missed.
Trend Analysis
More valuable than absolute thresholds is the ability to detect trends: is a measurement increasing, decreasing, or stable over time? Is the rate of change accelerating? Trend analysis can identify developing problems weeks or months before a threshold is breached, providing time for planned intervention rather than emergency response.
Effective trend analysis requires removing environmental effects (temperature cycles, seasonal groundwater changes, tidal loading) that can mask or mimic structural changes. Statistical techniques such as regression analysis, moving averages, and seasonal decomposition are standard tools. The key is separating the signal (structural change) from the noise (environmental variation).
Predictive Models
The most sophisticated approach combines monitoring data with structural models to predict future behaviour. Finite element models calibrated against monitoring data (so-called "digital twins") can forecast how a structure will respond to future loading, deterioration, or repair scenarios. This enables proactive, evidence-based decision-making rather than reactive maintenance.
Building and maintaining these models requires both monitoring expertise and structural engineering knowledge — a combination that is relatively rare and is a key area where specialist consulting adds value.
The Role of AI and Machine Learning in Modern SHM
Artificial intelligence and machine learning are increasingly applied to SHM data, and the results are promising — though the reality is more nuanced than the marketing suggests.
Where AI/ML adds genuine value:
- Anomaly detection: Unsupervised learning algorithms (autoencoders, isolation forests, clustering) can identify unusual sensor readings without requiring pre-defined thresholds. This is particularly useful in the early stages of monitoring when baseline behaviour is still being established.
- Pattern recognition: ML models can learn the relationship between environmental variables (temperature, humidity, wind, traffic) and structural response, then flag deviations from the expected pattern. This is far more sensitive than simple thresholds.
- Damage classification: When trained on sufficient data, ML models can distinguish between different types of structural change (cracking, settlement, bearing failure) based on the pattern of sensor responses.
- Predictive maintenance: Time-series forecasting models can predict when a deterioration process will reach a critical level, enabling maintenance to be scheduled at the optimal time — not too early (wasting resources) and not too late (risking failure).
- Data compression and summarisation: Neural networks can reduce massive DFOS datasets to compact representations that capture the essential structural information, making long-term data storage and review tractable.
Where caution is needed:
- Training data scarcity: Most structures have never failed, so training data for damage states is extremely limited. ML models trained only on healthy data may not recognise damage when it occurs.
- Interpretability: Black-box models that flag anomalies without explaining why are of limited use to engineers who must justify decisions to asset owners and regulators. Explainable AI approaches are essential in safety-critical applications.
- Overfitting to environmental effects: Without careful feature engineering, ML models can learn temperature cycles and traffic patterns rather than structural behaviour, producing impressive-looking but ultimately meaningless results.
- No substitute for engineering judgement: AI is a tool that augments human expertise, not a replacement for it. The final decision on whether a structure is safe must rest with a qualified engineer who understands both the data and the structural mechanics.
The most effective use of AI in SHM is as part of a hybrid workflow: automated algorithms handle the data processing, quality control, and pattern recognition at scale, while experienced engineers focus their attention on the cases that the algorithms flag as unusual. This combination is far more effective than either pure manual review or fully automated decision-making.
Why Independent Consulting Matters
The SHM market is populated by sensor manufacturers, system integrators, and platform providers — each with products to sell. This creates a natural bias: a company that sells vibrating wire sensors will recommend vibrating wire solutions; a company that sells wireless IoT platforms will recommend wireless solutions. Neither is necessarily wrong, but neither is looking at your problem from a purely technical perspective.
An independent consultant provides vendor-neutral advice based solely on what is best for your project. This includes:
- Technology selection: Recommending the right sensor types, communication architecture, and data platform for your specific application — not the ones that generate the highest margin for a supplier.
- Specification and procurement support: Writing clear, performance-based specifications that allow competitive tendering and prevent lock-in to proprietary systems.
- Installation quality assurance: Reviewing installation procedures and witnessing critical installation activities to ensure the system will perform as intended.
- Data interpretation: Providing the engineering analysis that turns raw data into maintenance decisions — the most valuable and most often missing part of the SHM workflow.
- Peer review: Reviewing monitoring proposals, designs, or reports prepared by others, providing an independent check on technical adequacy and value for money.
At GeoMonix, I combine hands-on research experience in fibre optic sensing and geotechnical monitoring with practical consulting capability. My PhD research at the University of Birmingham focused on advanced monitoring techniques for infrastructure, and I bring that depth of technical understanding to every project.
Summary: Getting SHM Right
Structural health monitoring is not just about installing sensors. It is a complete system that must be designed with the end goal in mind: better decisions about the safety, maintenance, and service life of infrastructure assets. The key principles are:
- Start with the engineering question — what do you need to know, and what will you do with the answer?
- Choose sensors for the application — not the other way around. There is no universally "best" sensor; there is only the right sensor for your specific problem.
- Invest in data processing — raw data is not information. Automated processing pipelines, quality control, and visualisation are not optional extras.
- Plan for the long term — monitoring systems must survive years to decades. Consider maintainability, data continuity, and future expansion from the outset.
- Close the loop — ensure that monitoring data feeds into actual maintenance decisions. The most expensive monitoring system in the world is worthless if nobody acts on the data.
Need Help With Structural Health Monitoring?
Whether you are planning a new monitoring programme, reviewing an existing system that is underperforming, or trying to extract more value from data you already have, I can help. From sensor selection and specification through to data analysis and engineering interpretation, GeoMonix provides independent, vendor-neutral consulting tailored to your project.