Summary: On a dam or a bridge pier, most of the movement a tiltmeter records is supposed to be there. Concrete dam deformation is elastic and driven by reservoir level and temperature, and instrumented arch dam records put the annual irreversible trend at less than 1 percent of the reversible amplitude, about 0.1 mm a year against a 20 mm annual swing. The work is therefore separation, not detection, and dam engineering has done it with the HST statistical model since the 1960s. On a pier the constraint is different: rotation is only one of three movement signatures, and the tilt caused by scour is a consequence of the erosion rather than a measurement of it. A tilt sensor is referenced to gravity, so it needs no external monument, but it gives rotation at a point and never replaces a deformation survey.
What a Tilt Reading Measures on a Dam or a Pier
A tilt sensor measures the rotation of its own mounting plane relative to gravity. Mounted on a dam crest or a bridge pier, it therefore reports the change in inclination at one point and nothing else. That single sentence rules out more of the loose reasoning about dam and bridge monitoring than it enables, so it is worth starting there.
The Federal Highway Administration’s Bridge Inspector’s Manual sets out the problems an inspector looks for on piers and bents: vertical movement, rotational movement, lateral movement, material defects, scour and undermining, areas subjected to high stresses, and fatigue prone details. Three of those entries are movement signatures, and a tilt sensor sees exactly one of them.
| Movement signature | How the inspector looks for it | Does a tilt sensor see it | What carries the information instead |
| Vertical movement | Joint opening or closing above the pier, changes in the parapet line, cracking or buckling in the pier, bearing seat elevations | No, unless the settlement is differential across the sensor base | A level survey along the deck line, and the deck joints themselves |
| Rotational movement (tipping) | Checking vertical alignment of the pier with a plumb bob or a level, clearance between beam ends, unusual cracking or spalling | Yes, this is the signature it measures | The tilt sensor is the automated version of the plumb bob check |
| Lateral movement (sliding) | General alignment, evidence of lateral displacement at the bearings, deck joint openings against the recorded temperature | No, a pure translation produces no rotation | Bearing inspection, joint measurement against temperature |
| Scour and undermining | Inspection at and below the footing, underwater inspection of the bed around the pier | Only as the consequence, once the pier has already rotated | Bed elevation measurement and the dedicated scour instrumentation listed in FHWA guidance |
The consequences follow directly. A pier that settles uniformly is a vertical movement, and it announces itself through deck joints that open or close and through a change in the parapet line, not through tilt. A pier that slides is a lateral movement, and it shows at the bearings and the joints. Only rotation, or differential settlement across the base of the sensor, produces an angle. The manual gives the practical version of this for multiple span bridges: deck joint openings should be consistent with the recorded temperature, so a joint reading that falls outside the temperature relationship is evidence of something other than thermal movement.
The classical field method for the rotational case is worth knowing because it is exactly what a tilt sensor automates. For rotational movement, the manual directs the inspector to check the vertical alignment of the pier using a plumb bob or a level. That is a periodic, manual, qualitative measurement of pier verticality. A wireless tiltmeter turns the same check into a continuous numerical record at the same location, which changes how often you know, not what you are looking at.
On a Dam, Most of the Movement Is Supposed to Happen
The United States Army Corps of Engineers publishes the technical guidance for deformation monitoring of its dams and locks in EM 1110-2-1009, Structural Deformation Surveying. The manual separates structures by material, and for concrete dams it states that deformation is highly dependent on reservoir water pressure and temperature variations with an overall elastic behavior. Elastic means reversible. The dam is designed to lean downstream when the reservoir fills and to return when it draws down.
The manual then describes what is not elastic. Permanent deformation can occur as the subsoil adapts to new loads, as concrete ages, or as foundation rock fatigue is experienced, and such deformation is not considered unsafe if it does not go beyond a predetermined critical value.
The relative size of the two components is the part that should change how a monitoring scheme is designed. Dam engineering literature gives the ratio for a large arch dam: at a height of 130 m the thermal displacement amplitude is about 20 mm, while the irreversible trend is about 0.1 mm per year. Over one year, the irreversible part of the displacement is less than 1 percent of the reversible part.
Place the two numbers side by side and the problem is clear. A threshold applied to the raw reading cannot work, because the reversible component alone is two orders of magnitude larger than the signal of interest. What has to be measured is not movement but the difference between two movements.
The standard solution comes from dam engineering rather than from sensing. Electricite de France developed the HST model, Hydrostatic, Season, Time, in the 1960s, and it remains the reference method. The measured displacement is modelled as the sum of a hydrostatic component that follows reservoir level, a seasonal component that follows temperature, and a time component. Fit the first two from the structure’s own history, subtract them, and the residual is the irreversible trend.
The same literature is direct about the simpler alternatives: plotting measured displacement against time, or against reservoir level, is described as difficult to analyse because of dispersion due to external reversible influences. A tilt record on its own cannot be separated, because the separation needs the load. Water level and temperature have to be logged on the same clock as the tilt. That is the first configuration decision on a dam, and it is made before any sensor is ordered.
Note the distinction this exposes between a structure whose load is instrumented and one whose load is not. Where the driving load is measured, as reservoir level is on a dam, the reversible response can be regressed out and the residual trend becomes a real number. Where the driving load is not measured, no amount of post-processing recovers the separation, and the only defensible criterion becomes a rate.
The Material Decides What the Number Means
The same manual reaches a different conclusion for embankment dams, and the difference matters more than any specification on a datasheet. For earth and rockfill dams the Corps describes deformation as largely characterized as more permanent, arising from consolidation under self weight and from the hydrostatic load of the reservoir, and it lists the surface expressions of a developing problem as a crest that cracks, or starts with a divot, or sometimes a heaved boil and a slough.
The identical reading therefore has opposite meanings on the two dam types. On a concrete dam the reversible part is expected behaviour and a new non-recoverable offset is the signal. On an embankment the permanent deformation is the normal mechanism, and the value itself tells you very little. What matters is the rate and whether it is accelerating.
| Structure type | How it deforms under load | What a new permanent offset means |
| Concrete dam | Elastic and reversible, driven by reservoir water pressure and temperature | The signal. The reversible part is expected, a non-recoverable offset is not |
| Earth or rockfill dam | Largely permanent, from consolidation under self weight and hydrostatic load | The normal mechanism. Rate and acceleration matter, the absolute value does not |
| Bridge pier | Rotation, vertical settlement and lateral sliding are separate signatures | Rotation only. Uniform settlement and pure sliding do not register as tilt |
This is also why dam monitoring cadence looks slow to anyone arriving from machine condition monitoring. The same manual notes that for concrete dams, highly accurate short term deflections or relative movements between monoliths due to varying temperature or hydraulic loading are more rarely required, because it is the long term trend that carries the safety information. Observations are read monthly rather than every minute, and that is a deliberate engineering choice rather than a limitation of the instruments.
On a Bridge, the Tilt Is a Consequence, Not the Measurement
Bridges invert the logic. On a bridge the concern is often a mechanism that is invisible between inspections, rather than a load response that can be reconstructed. The Federal Highway Administration requires every state to identify, monitor and improve highway bridges where scour is or could become critical, and its guidance is published as three linked documents: HEC-20 on stream stability, HEC-18 on evaluating scour, and HEC-23 on countermeasures. HEC-23 groups countermeasures into hydraulic, structural and monitoring families, and it places monitoring first among the recommended actions.
Scour is the erosion of the bed at piers and abutments, and it accelerates during high flow. It happens underwater, so between inspections the cause is largely unseen. A tilt sensor on the pier records the rotation that scour causes. That is a real and useful signal, but it is the consequence rather than the cause, which means that by the time the angle moves, the foundation has already moved.
The instrument is nonetheless part of the recognised toolbox. A Texas Department of Transportation study on real time monitoring of bridge scour reviews the fixed instrumentation available and includes the tiltmeter among them, alongside the float out device, the water stage sensor, the sonar sensor, the tethered buried switch, the accelerometer and the acoustic doppler velocimeter. Each instrument in that review comes with advantages and disadvantages, and the tiltmeter’s distinct role is that it measures movement of the structure rather than bed elevation.
Timing is where a configuration decision becomes urgent. A flood lasts hours. A tiltmeter reporting on a 24 hour heartbeat takes one sample a day, so a pier that rotates during a flood and then recovers, or fails, can appear in the record as two clean frames with the event between them. The datasheet separates the two reporting channels explicitly, and getting them the right way round is the most useful single decision on a scour critical bridge.
| Channel | Parameter | Published value | What it is for |
| Periodic path | Heartbeat interval | 60 to 131,071 s, default 86,400 s (24 h) | Uploads while the alarm is cancelled; this is what sets battery life |
| Event path | Alarm angle | Default 3 deg, settable across the measuring range | Wakes and reports when the tilt exceeds the configured angle |
| Event path | Alarm accuracy | plus or minus 0.01 deg in the -5 to +5 deg band, plus or minus 0.03 deg in the -15 to +15 deg band, plus or minus 0.1 deg in the -30 to +30 deg band | The angular error at the moment of triggering |
| Event path | Alarm delay | 0.3 to 25.5 s, default 2 s | How long the tilt must persist beyond the alarm angle before the alarm is raised |
Each of those items has a precise published definition, and the definitions are where the design intent sits. The heartbeat interval is defined as the interval at which the device periodically uploads while the alarm is cancelled, which is why it governs battery life. The alarm angle is settable across the measuring range, and the datasheet’s own worked example shows what the alarm accuracy means in practice: with an alarm angle of 3 deg and an alarm accuracy of 0.03 deg, an axis below 2.97 deg will not trigger an alarm, between 2.97 and 3.03 deg the trigger is uncertain, and above 3.03 deg the alarm is certain. The alarm delay is defined as the time the device must continue to tilt beyond the alarm angle before it raises the alarm.
Two conclusions follow. First, for an event driven risk such as scour the useful configuration is a short alarm delay together with a sensible alarm angle, not a faster heartbeat. A faster heartbeat spends battery every day of the year to gain nothing on the day that matters, because it still samples the flood at whatever the heartbeat interval happens to be. Second, a default alarm angle of 3 deg is an overturning alarm rather than a monitoring threshold. Nothing on a pier is behaving acceptably at 3 deg of rotation, so on a bridge the alarm angle is a fail-safe backstop while the trend work is done on the periodic record.
The Reference Problem That Tilt Does Not Solve
There is one limitation that no amount of sensor quality removes, and the Corps manual states it in a single line: assessment of permanent deformations requires absolute data. Absolute data in turn requires reference points outside the structure, established on stable ground in the foundation or the surrounding terrain and beyond the area that may be affected by the dam or the reservoir. The manual adds a step that is easy to forget: the reference network must itself be monitored, to confirm that the reference points have not moved.
The same manual records a requirement that reveals what this kind of surveying costs in practice. Deformation surveys are to be performed at night, to avoid troublesome optical distortions due to sunlight and heat radiation. A stable monument network, a survey crew, a dark site and a repeatable procedure are the price of an absolute displacement number, and the price is paid again every time the campaign is repeated.
A tilt sensor takes a different route to a reference. Its reference is gravity, which is not a monument that can settle, and it needs no line of sight, no ground station outside the structure’s influence and no night window. That is why it is viable on a long crossing where a survey network of comparable coverage would be impractical, and why it can be installed on an operating railway bridge without disturbing service.
What it cannot do is give position. A tilt reading is rotation at one point. It cannot tell you how far a pier has moved, only how far it has turned, and it cannot give absolute values for anything. Absolute deformation still requires a survey or a satellite network. Tilt monitoring is complementary to deformation surveying, and the honest version of the proposal says so.
The comparison with the reference instruments clarifies where a tiltmeter belongs. A study of a concrete arch dam notes that pendulums are the most sensitive deformation measurement device installed in a dam, with a potential accuracy of 0.5 mm, and then adds that in practice this accuracy is often not achieved, because of external interference from wind in the shaft, water condensation and resulting dripping, spider webs, and general bumping of the device by technicians. The same study states that the geodetic survey is the most reliable form of dam displacement monitoring. A tiltmeter competes with neither. It fills the interval between surveys with a continuous, gravity referenced record at a point, at an installation cost no survey campaign can match.
The Error Budget: Why One Temperature Channel Does Two Jobs
The published datasheet for the LoRaWAN wireless tilt sensor, model ZCT-IOTH-WL-2x-JLA8, gives a resolution of 0.001 deg, an accuracy of plus or minus 0.005 deg in the -15 to +15 degrees Celsius band and plus or minus 0.01 deg in the -30 to +30 degrees Celsius band, a zero point deviation of plus or minus 0.05 deg, and a zero point temperature drift of plus or minus 0.001 deg per degree Celsius typical, with a maximum of plus or minus 0.065 deg across the -40 to +85 degrees Celsius range. Read as a selection table they look like five numbers on one scale. Read as an error budget they behave in five different ways.
| Term | Published value | Test condition | How it behaves over a year |
| Resolution | 0.001 deg | 25 degrees Celsius | The floor on any trend, and not the limiting term here |
| Accuracy, tight band | plus or minus 0.005 deg | -15 to +15 degrees Celsius | Applies only while the structure stays inside the band |
| Accuracy, wide band | plus or minus 0.01 deg | -30 to +30 degrees Celsius | The figure to plan against in a continental climate |
| Zero point deviation | plus or minus 0.05 deg | 25 degrees Celsius | A fixed offset that removes itself against a baseline |
| Zero point temperature drift | plus or minus 0.001 deg per degree Celsius typical, plus or minus 0.065 deg max | -40 to +85 degrees Celsius | About 0.04 deg over an annual swing of 40 degrees Celsius, reversible and correlated with temperature |
The first consequence is that accuracy is a function of temperature rather than a fixed property of the instrument. A structure in a continental climate spends part of the year outside the tight band, so the figure to plan against is the wider one unless the band can be guaranteed for the whole campaign. Two piers in different climates can be running different accuracy specifications on the same purchase order.
The second consequence is the reason the temperature channel does double duty. At plus or minus 0.001 deg per degree Celsius, an annual swing of 40 degrees Celsius at the site moves the sensor zero by about 0.04 deg. That is small in absolute terms, but it is the same order as the smallest angle a dam or bridge engineer might want to trend. The important part is that this drift is reversible and correlated with temperature, and so is the structure’s own thermal deformation. One temperature channel removes both at once, which is why temperature should be logged on the same clock as the tilt rather than reconstructed afterwards from a weather station several kilometres away.
The zero point deviation of plus or minus 0.05 deg is a different animal. It is a fixed offset rather than a random error, so against a baseline it subtracts out and it does not accumulate. What cannot be averaged away or subtracted away is drift, which is why the drift term rather than the accuracy term decides whether a slow trend is provable.
Mounting deserves the last word in this section, because on a real structure it is normally the largest error in the chain. The published installation instructions for these digital inclinometers set two rules: the mounting surface of the sensor must sit completely against the mounting surface of the object being measured, with no angle between them, and the sensor axis must be parallel to the axis of the surface being measured, with no angle between the two axes. The same instructions state that improper installation can produce large angle measurement errors. On a curved pier face or a rough concrete crest the bracket, not the sensor, sets the error floor, and an over long bracket arm adds its own thermal expansion to the measurement.
What the Deployment Records Show About Configuration
Two deployments from the company’s own project record show what these choices look like on real structures.
The first is the Hangzhou Bay Bridge in Zhejiang province, which carries the G15 expressway between Jiaxing and Ningbo. It opened in May 2008 as a 36 km route of which 35.7 km is bridge, with six lanes and a design speed of 100 km per hour. ZCT330M-SWP-N-YK wireless tilt sensors were deployed to monitor the verticality of the piers on a schedule and to raise alarms in real time, so that the owner obtains pier verticality data on a schedule and can analyse the structural trend instead of relying on a walk round.
The company’s account of that project states both the engineering reason and the practical constraint. The engineering reason is that bridge problems are often rooted in uneven settlement of piers, where a small inclination affects stability and a larger one affects safety. The practical constraint is that the tilt detection equipment available in that class uses RS232 or RS485 connection with an external power supply, and applying it across a bridge of that length would require comprehensive cabling and mains power on the structure, increasing deployment difficulty and adding an electrical safety risk. That is the case for a wireless instrument with an internal battery on a bridge: it is a structural and safety requirement, not a convenience.
The second deployment is at an operating railway bridge near the high speed rail station in Fuzhou, Fujian province. LoRa wireless tilt sensors were used, published in the company’s account as the ZCT1000ML and later offered as the ZCT1000ML-S230. The characteristics quoted for that deployment are LoRaWAN compliance with a radio range of 3 to 5 km, one gateway serving multiple sensors, an internal battery with no cabling, and a dual axis measuring module with accuracy stated at 0.005 deg. The point of the deployment is the constraint it satisfied: the bridge is in service, so the installation had to be completed without disturbing the railway.
Placement is where the two cases share a lesson. The points that carry information are the ones where the mechanism appears. On a pier that means the pier head and the bearing, because both the rotation and the bearing movement are expressed there. On a span it means the ends and mid span, where the change in profile is largest. Spacing sensors evenly along a parapet because the parapet is easy to reach produces a tidy drawing and a weak data set.
What Continuous Tilt Monitoring Does Not Replace
The limits are worth stating in the words of the agency that writes the bridge inspection rules, because they are more conservative than most proposals are. In evidence to Congress on bridge inspection technology, the Federal Highway Administration observed that the most practical monitoring systems are used by owners following an in depth or special inspection, to monitor the performance of an element or the bridge when a specific concern has been raised but the concern is not considered a short term safety hazard. That is the correct use case, and it is narrower than continuous monitoring as a general policy.
The same source adds three constraints that any proposal should be able to answer. Monitoring systems do not eliminate the need for regular visual inspections. In many circumstances it is more effective to increase the inspection frequency, repair or retrofit the areas of concern, or replace the structure. And on longevity: bridges are now built to last 75 to 100 years, and installing monitoring systems and expecting them to remain durable and serviceable for that long has never been done before, so monitoring hardware requires routine maintenance and continuous assessment to confirm that it is still working.
For scour specifically there is one further caveat from the same guidance. A real time monitoring system does not erase scour critical status. It is a tool for earlier detection, for better response time and for deciding which countermeasure to apply. It does not change the condition of the bridge.
The honest summary for anyone specifying a tiltmeter on a dam or a bridge is therefore short. The instrument buys time resolution at a point, without cabling, without an external reference monument and without taking the structure out of service. It does not buy absolute displacement, it does not replace a deformation survey, and it does not remove the obligation to inspect. On a dam it also does not replace the piezometers, the seepage measurements and the internal instruments that a surveillance programme already relies on. What it adds is a continuous record of rotation at the locations where a mechanism will show up first.
Frequently Asked Questions
- Can a tilt sensor detect bridge scour?Not directly. It records the rotation that scour causes at the pier, so it detects the consequence rather than the erosion itself. Scour happens underwater at the bed, and the Federal Highway Administration treats monitoring as one of three countermeasure families alongside hydraulic and structural measures. The value a tiltmeter adds is time resolution: a flood lasts hours and the pier rotation happens during the event rather than after it, so an event driven alarm configuration catches what a daily heartbeat misses. In the Texas Department of Transportation review of fixed scour instrumentation, the tiltmeter is listed for exactly this reason, as a device that measures movement of the structure rather than bed elevation.
- What accuracy does dam tilt monitoring actually need?Two different questions are usually hiding in that one. For an overturning type alarm, the factory default of 3 deg is far coarser than a dam needs, and the alarm accuracy of plus or minus 0.01 deg in the central band is what matters at the trigger point. For trend separation, what matters is resolution and drift rather than headline accuracy: the published resolution is 0.001 deg and the zero point temperature drift is plus or minus 0.001 deg per degree Celsius typical, against an annual irreversible trend on a large arch dam of around 0.1 mm. Choose the instrument against the smallest trend you intend to prove, not the largest angle you can imagine.
- How often should a dam tilt sensor report?Monthly is the classical cadence for dam deformation observations, and a 24 hour heartbeat already delivers about thirty times that. Battery life is published against the 24 hour heartbeat: six years on a 19,000 mAh D size lithium thionyl chloride battery for the LoRaWAN model. Shortening the heartbeat trades years of service life for data the reversible term separation does not need, because the hydrostatic and seasonal components are slow. Keep the fast path for the alarm channel, where the alarm delay parameter controls how long the tilt must persist before the device wakes and reports.
4. Can a tilt sensor replace a plumb line or a deformation survey?No, and the reason is a matter of definition rather than quality. A tilt sensor is referenced to gravity, so it needs no external monument, but it measures rotation at one point. Absolute displacement requires reference points outside the structure and beyond its influence, and the reference network itself has to be monitored to confirm the reference points have not moved. Dam engineering also reports that the most sensitive device installed in a dam is the pendulum, at a potential accuracy of 0.5 mm that is often degraded by wind in the shaft, condensation and physical disturbance, while the geodetic survey remains the most reliable form of displacement monitoring. Use tilt for continuous trend and event detection, and keep the survey for absolute position.
TAG: