Vibrating wire sensors: why they are still the standard

In short

• A vibrating wire sensor converts strain, pressure, force, or tilt into a change in wire tension, and tension into the wire's natural frequency, which is read by a coil.
• Frequency readout is resistant to cable resistance, line length and amplitude interference. Manufacturers specify cable runs of more than 2 km. This is the source of 20+ year stability.
• Strain is calculated from the general relation ε ≈ K · (f₁² − f₀²) plus a temperature correction; the coefficients come from the manufacturer's calibration certificate.
• The wire is a quasi-static sensor. For vibration, impacts and traffic events, you need MEMS or piezoelectric sensors.
• The biggest enemy of vibrating wire sensors is not the technology, but the lack of a zero reading, the lack of temperature compensation and forgotten verification checks, in other words, "calibration debt".

Vibrating wire technology has roots in the 1920s and 1930s: Davidenkoff's work in Russia and André Coyne's work in France in 1928, Coyne's first installations on the Bromme dam in 1930, the 1931 patent for a vibrating wire strain sensor, and commercialization by Maihak in Germany and Telemac in France (Simmonds 2013). Since then, the principle has not changed, only the materials, housings and readers have. In this post I break it down: how it works, why it is resistant to what breaks other sensors, how to calculate temperature compensation, where it drifts, how to defend the choice when someone asks "why not a cheaper one", and when it is better to give it up. This is for the engineer who is specifying a system for years ahead or inheriting an asset with sensors installed long ago. The broader context, what quantities are measured on structures and by what means, is in the series pillar: structural monitoring (SHM): a complete guide.

Operating principle in three paragraphs

A vibrating wire sensor is a transducer in which the measured physical quantity changes the tension of a thin steel wire stretched between two anchors, and that tension is then read indirectly through the wire's natural vibration frequency. The wire sits in a sealed tube, with an electromagnetic coil and core beside it. The anchors are coupled to what you are measuring: concrete (cast-in strain gauge), a girder flange (surface-mounted strain gauge), a diaphragm (piezometer, pressure sensor), a ring loaded by a ground anchor (force sensor), or a pendulum (tiltmeter).

The measurement has two phases. First, the reader sends a pulse or a short sequence of pulses with swept frequency to the coil. That is the "pluck" that sets the wire vibrating. Then the same coil becomes a microphone: the wire vibrating in the magnetic field induces a sinusoidal voltage in it at a frequency equal to the wire's natural frequency. The reader measures the period of that signal, counts zero crossings over a fixed time, and returns the frequency in hertz. The whole cycle takes from a fraction of a second to a few seconds, depending on the reader. The differences between Hz, period, digits and the final result are explained in a separate guide on vibrating wire sensor readout calibration.

The physics is straightforward. The natural frequency of a tensioned wire increases with the square root of tension (f ∝ √σ, for constant length and density). Tension in the wire is proportional to its strain (Hooke's law), so strain is proportional to the square of frequency. When the concrete around a strain gauge is compressed, the anchors move closer together, the wire slackens, and the frequency drops. When water pressure acts on a piezometer diaphragm, the diaphragm deflects and unloads the wire, again reducing the frequency. The manufacturer calibrates the sensor so that the difference between the squares of the frequencies gives the quantity you are interested in directly.

What is measured with a wire

The thermistor in the last row is not an accessory. It is a condition for a meaningful reading, which I will return to in the temperature compensation section.

Why the frequency signal does not age

This is the core advantage of vibrating wire sensors. The reader does not ask the wire "how hard do you vibrate", but "how fast do you vibrate". The signal amplitude may drop tenfold because of a long cable, a corroded connector or moisture in a junction box, but the period of the signal stays the same. The information sits in the time between zero crossings, not in volts.

Long cables

In an electrical resistance strain gauge, the information is a change in resistance of fractions of an ohm. Every metre of cable, every temperature change in it, and every connector adds its own resistance to the signal. This can be compensated for, using three-wire and four-wire circuits, but it is a fight you wage for the entire service life of the sensor. In a vibrating wire sensor, cable resistance attenuates the amplitude, but does not change the frequency. Geokon states in its short form catalogue that frequency can be transmitted over cable longer than 2 000 m without noticeable signal degradation caused by changes in cable resistance, water ingress, temperature fluctuations, contact resistance or leakage to ground (Geokon, short form catalogue). You can therefore splice the cable in the field, extend it, or replace it after damage, and the sensor will read the same as before the change. For a strain gauge cast into a girder, that is the difference between "we have a measurement" and "we have a relic".

One caveat: cable resistance immunity applies to frequency, not to the thermistor. Temperature readout is resistive, so for very long lines, manufacturers' instructions provide a correction for conductor resistance (Geokon gives an order of 48.5 Ω per kilometre for its cable). Record the cable length in the measurement point metadata. It will be useful when someone, ten years later, has to calculate compensation.

Electromagnetic interference

Interference induced in the cable, from the power network, traction systems or converters, superimposes stray voltage on the wire signal. For amplitude measurement, that is a direct error. For frequency measurement, it is noise that a good reader filters out: the wire signal has a known, narrow frequency within the sensor operating band, and the period count averages over many cycles. A vibrating wire sensor operating on a railway asset with electric traction or beside a switchgear room does not require the exotic shielding that an electrical resistance strain gauge would need in the same place.

Where this resistance ends

To be fair, the resistance is not infinite. If the interference has a frequency component close to the wire frequency, for example a 50 Hz mains harmonic falling into the sensor band, a low-cost reader may latch onto noise instead of the wire. Reader manufacturers solve this with swept excitation, band-pass filters and signal quality control. Some readers return, in addition to frequency, an amplitude metric. Record it. A drop in amplitude at constant frequency is a sign of a cable problem before the reading stops arriving. For you, that means one thing: you will see a cable fault before you see a gap in the data. When the reading disappears or jumps, go through a structured vibrating wire sensor diagnosis before changing the band or coefficient.

Temperature compensation: formula and illustrative example

The wire is steel, so it expands with temperature. If the sensor heats up while the structure stays still, the wire lengthens, loses tension and frequency drops. The reader then shows compression that is not really there. The measured element itself also changes dimensions with temperature. That is why every vibrating wire calibration certificate includes at least two coefficients: the gauge factor and the thermal coefficient.

General form of the relationship

Strain is calculated using the general form:

ε = K · (f₁² − f₀²) + C_T · (T₁ − T₀)

where:

• f₀, T₀ - frequency and temperature in the zero (initial) reading,
• f₁, T₁ - frequency and temperature in the current reading,
• K - calibration coefficient (gauge factor) in µε/Hz², specific to the sensor unit or type,
• C_T - temperature correction coefficient in µε/°C.

Some manufacturers provide the result in a unit of convenience instead of frequency squared ("digit", usually f²/1000), and instead of a single C_T they give the wire thermal expansion coefficient and the construction material thermal expansion coefficient separately. The meaning is the same. What matters is distinguishing which strain you are asking about:

• Total strain (how much the element really changed length) - the correction covers only wire expansion.
• Load-induced strain (how much of that came from the load, not the weather) - the correction is the difference between the wire expansion and the material in which the sensor sits.

For alarm thresholds on a structure, you usually need the second one. For comparison with a calculation model, sometimes the first one. The coefficients must not be guessed; they are taken from the sensor certificate and from the material data for the structure.

Illustrative example (assumed numbers, not from a deployment)

Let us assume, for illustration, a vibrating wire strain gauge on a steel girder, zero reading f₀ = 2 500 Hz at T₀ = 10 °C, current reading f₁ = 2 600 Hz at T₁ = 18 °C. Let us assume K = 0.0004 µε/Hz² and C_T = 2 µε/°C (as the difference between the wire and steel girder thermal expansion; assumed values).

Frequency term: 2 600² - 2 500² = 6 760 000 - 6 250 000 = 510 000 Hz², times 0.0004 gives 204 µε of tension.

Temperature term: (18 - 10) · 2 = 16 µε.

Compensated result: 204 + 16 = 220 µε.

With these numbers, the error from ignoring temperature is 16 µε for an 8 °C difference. On a bridge that goes through several degrees each day and several tens across the seasons, that error eats a significant part of the margin between warning and alarm thresholds, and it does so cyclically. Without compensation, you get false alarms in hot weather or miss a trend hidden by winter. That is why a strain plot without a temperature axis beside it is only half a plot. We write about reading such plots on bridges in the guide bridge monitoring: what to measure and how to interpret it.

Practical note: the correction assumes that the thermistor measures the wire temperature, and the wire has the temperature of the element. On sunlit steel, the sensor heats up faster than the girder mass. A sun shield costs less than a year spent explaining 2:00 p.m. peaks. When the result must separate sensor, material and load effects, you need the full temperature compensation procedure for vibrating wire sensors, not one coefficient in a spreadsheet.

Drift, calibration and "calibration debt"

Stability in vibrating wire sensors does not mean nothing changes inside them. Vibrating wire sensor drift is a slow change in reading at constant measured quantity, caused by phenomena in the sensor itself rather than in the structure. The sources are well known: wire relaxation (steel under constant tension slowly creeps), micro-slip at the anchors, fatigue after a very large number of cycles, corrosion and, in piezometers, filter clogging and diaphragm ageing. Good manufacturers pre-age and pre-load the wire before installation so that most relaxation happens in the factory. That is why it is reasonable to talk about 20+ year stability: not because there is no drift, but because it is small and predictable.

What you cannot do is remove a strain gauge from concrete and send it to a laboratory. A cast-in sensor is calibrated once, in the factory. After installation, you have three tools left:

1. Zero reading. Without it, f₀ and T₀ in the formula are guesses, and the entire measurement shows changes relative to a moment you do not know. Take the zero reading after the installation has stabilised, after the adhesive has cured or the frame has been preloaded, record it, and treat it as a document.
2. Verification through physics. A sensor that responds to temperature as expected, to test loading as expected, and returns at night to the state from the previous night is working. A sensor whose "zero" drifts in one direction regardless of temperature and load is drifting, or the structure is creeping. Compare it with adjacent channels to decide.
3. Redundancy. Two sensors in the same section buy confidence cheaply. When one drifts and the other does not, you know which one.

Accessible sensors, such as those for force, displacement and tilt, can and should be checked periodically, removed and verified, or compared against a reference instrument. Manufacturers do not publish one universal interval in years; they recommend verification "at the appropriate time" so that drift can be detected and corrected before it distorts the assessment (Simmonds 2013). You therefore set the interval according to the manufacturer's recommendations for the specific sensor and the monitoring project requirements, and you put it into the schedule so it does not depend on one person's memory.

Calibration debt is the sum of sensors in a system that have exceeded the planned verification interval or do not have a documented zero reading, and therefore reduce the credibility of the alarms from the whole asset. The term is borrowed from "technical debt": every overdue sensor adds interest in the form of uncertainty. The debt grows quietly, because the sensor still sends numbers, and the numbers still look reasonable.

It surfaces at the worst possible moment, when you have to defend an alarm, or the absence of one, to someone who asks: "When was that sensor checked last?" Eurocode 7, in the clause on the observational method (PN-EN 1997-1, clause 2.7), requires replacement or supplementation of instrumentation that does not provide reliable data, and you cannot meet that requirement if nobody knows which channels are reliable. That is why calibration debt should be a report, not tribal knowledge held by one engineer. For you, a list of channels "to be checked" with dates is a document that protects you faster than the best memory.

How to justify choosing wire sensors for 20 years: arguments and counterarguments

The decision on measurement technology for an asset that is to be monitored for two decades is rarely made among engineers alone. It is made in a meeting where someone asks why not choose something cheaper, and someone else asks whether it will survive a logger replacement. Below are the arguments that work, and the counterarguments you will hear and should be ready to answer before they are raised.

Argument 1: the cost of changing technology later is higher than the price difference of the sensor. You cannot replace a strain gauge cast into concrete. If, after five years, it turns out that the sensor drifts or its electronics are no longer produced, you lose not the sensor, but the history of the measurement point, and history is the reason the monitoring exists. A wire sensor in concrete is passive: steel, coil, thermistor. There is no integrated circuit in it that will be discontinued.

Argument 2: a wire sensor does not tie you to one electronics supplier. A vibrating wire sensor returns frequency and thermistor resistance. Any vibrating wire reader can read them, regardless of manufacturer, and the coefficients are in the certificate. You can replace the logger or the platform and keep the point history continuous, provided you have preserved the zero reading and the certificate. This is the argument that works with procurement: you are not buying a closed ecosystem.

Argument 3: the stability is documented, not promised. The technology has almost a hundred years of field practice on dams, tunnels and bridges, literature on drift and field verification exists (Dunnicliff 1988), and Eurocode 7 for assets that adversely affect their surroundings or carry high risk allows monitoring for more than 10 years after construction or for the full service life. If the project horizon is years, the technology must also have a horizon of years.

Counterargument 1: "MEMS are cheaper and measure more." True, for dynamics and where the sensor can be replaced, MEMS are the right choice. The answer is: this is not either-or. Use wire sensors where the measurement point must outlive the structure, and MEMS where time and bandwidth matter. On one asset, in one system.

Counterargument 2: "A dedicated reader is needed, so there is an extra cost." True. A vibrating wire sensor will not connect to any generic analog input; a vibrating wire interface in the logger is needed. The answer is: you pay this cost once, and the logger is replaceable, unlike the sensor in concrete.

Counterargument 3: "Without a zero reading and compensation the data are useless." Also true, and this is the most serious counterargument because it concerns organisation, not physics. The answer is: this requirement must be written into the monitoring project as a procedure: who takes the zero reading, where it is stored, where the certificates are, who tracks the verification schedule. A technology that survives 20 years requires documentation that also survives 20 years.

What this means for you: in the meeting, you are not defending "the best sensor", but the lowest life-cycle cost of maintaining a continuous measurement history throughout the asset life. That is an argument that also makes sense to a non-technical decision maker.

Vibrating wire vs MEMS vs electrical resistance vs fibre optic

The comparison below concerns classes of solutions, not specific products. The ratings are descriptive, because real values depend on the manufacturer and the application.

The conclusion from the table is that this is not a ranking. It is a division of roles. Wire sensors win where multi-year stability of the measurement point in a harsh environment matters. MEMS win where time matters. Fibre optics win where "where exactly" matters, not "at which point". Mature systems combine these classes; we write separately about combining automatic point sensors with periodic fibre optic measurements: hybrid systems: point sensors and DFOS.

Where not to use vibrating wire sensors

The biggest misunderstanding about vibrating wire sensors is the expectation that, because they are "the best", they are suitable for everything. They are not.

Dynamics. A wire reading is a cycle of excitation, decay and counting. Even a fast reader gives a few readings per second, and a typical logger polls a channel every few minutes or less often. A truck crossing a bridge, an impact, a pile driver, vibration from a compactor, the wire will not see it, and if it does, it will see it as a random sample that lies through aliasing. For vibration events, use MEMS accelerometers or piezoelectric sensors, with time history, spectrum and third-octave analysis. What to do with that and how it relates to PN-B-02170 is described in the post building vibration monitoring during construction.

Modal identification. The natural frequencies of a structure and their change over time require a continuous acceleration signal. The wire gives point strain, useful for static assessment, useless for deriving vibration modes.

Very fast thermal and load transients. Proof loading with rapid travel, a hydraulic pulse in a pipeline, sudden cooling, when the phenomenon lasts less than the interval between readings, the wire gives an average or a random instant.

Places without access to sensors that require verification. A vibrating wire force or displacement sensor built in without the possibility of removal is a sensor that will have calibration debt in a few years and no chance of repayment. If access cannot be ensured, design redundancy from the start.

Assets with a short measurement life. A short campaign under proof load does not need 20-year stability. An electrical resistance strain gauge will be cheaper and faster to install, and its weaknesses will not have time to appear.

The common denominator: a wire sensor is for the question "how does the structure change over months and years". For the question "what happened at 14:07:32", other tools are used. Most often on the same asset, in the same system.

What it looks like in Inclify

Vibrating wire sensors are one of more than 40 sensor types selected and installed by the Inclify team, alongside chain inclinometers (in-place), MEMS, piezometers and temperature sensors. The platform accepts data from any measurement channel: loggers and hubs, including those already installed on the asset, send HTTP/JSON frames with a timestamp, channel ID and raw values, for a wire sensor the frequency and thermistor reading. Loggers in Inclify systems normally send readings every 15 minutes by default, or more often, but the platform accepts any step and calculates the actual cadence itself. How to connect what you already have is described in the text how to connect existing loggers to the online platform.

The conversion to µε, including the temperature compensation from the formula above, is defined as project equations: you enter K, C_T and the zero reading, and the platform calculates the result from each incoming frame. The zero reading is a separate object in the platform, a reference that can be derived from the device's first readings or entered manually, and every equation change remains in the audit log with values before and after, so in ten years it is clear who changed which coefficient and when. The temperature compensation report checks whether the channel no longer correlates with temperature after compensation, and the calibration debt report compares each channel's deviation from the reference and drift over the last 7 and 30 days, alongside a 7/30-day risk assessment and data SLA (completeness, gaps, freshness). A multi-axis chart overlays strain against temperature so that daily "breathing" can be separated from a lasting trend in a single view. Raw frames from devices, what the logger sent before anyone converted them to µε, can be viewed and downloaded by the administrator from the communication log, stored for the configured period, by default 7 days. Details of the sensor-logger-platform layer are on the page /platform/iot.

Sources and further reading

• A.J. Simmonds, "Long term monitoring using vibrating wire sensors", SHMII-6, Hong Kong 2013 - history of the technology, drift, verification of removable sensors - ishmii.org (PDF)
• Geokon, short form vibrating wire sensor catalogue (cable range, resistance to line resistance) - geokon.com (PDF); 4000 strain gauge and 4500 piezometer manuals (calibration formulas, thermal coefficients, cable resistance) - manual 4000, manual 4500
• Tunnels & Tunnelling, "How does it work? - Vibrating wire instruments" - history and cable range - tunnelsandtunnelling.com
• Campbell Scientific documentation for vibrating wire sensor interfaces (excitation, frequency measurement, signal quality) - campbellsci.com
• J. Dunnicliff, "Geotechnical Instrumentation for Monitoring Field Performance", Wiley 1988 (paperback ed. 1993) - chapters on vibrating wire transducers, drift and field verification
• PN-EN 1997-1 Eurocode 7 - clause 2.7 observational method (requirement to replace instrumentation that does not provide reliable data) and chapter 4 (monitoring during and after construction) - sklep.pkn.pl; in the second generation of Eurocode 7 (EN 1997-1:2024) the clause numbering is different
• "Hybrid systems - an introduction using bridges as an example" (industry publication, 2025) - vibrating wire sensors as part of the automated system alongside DFOS measurements - PDF
• On this blog: AI in structural monitoring: what really works - on how anomaly detection helps catch a drifting vibrating wire channel before a person does

What next

If you are specifying a system with vibrating wire sensors, or you have an asset where the wires have been installed for years and the data sit in the logger, we will show, using data from a similar asset, what the temperature-compensated equations, the calibration debt report and the multi-axis chart look like. The lowest entry point is connecting the existing logger to the platform on one asset: it takes a few days and does not require touching the sensors. It is worth doing this before the season when temperature interferes most with the readings, before the winter trend disappears under compensation that nobody has calculated. We reply within 24 hours: let's talk about your asset.