Structural Health Monitoring (SHM): a complete guide

In short

• Monitoring means continuity, thresholds, alarm, and evidence. A sensor with a logger from which someone downloads a file once a quarter is periodic measurement, not monitoring.
• You measure what the designer considers representative for asset safety: strain (µε), inclination (mrad), deflection and displacement (mm), pore pressure, anchor force, temperature as background, vibration.
• Static measurement, with readings every 15 minutes by default, or more often, and dynamic measurement, with vibration events and FFT spectra, are two different regimes. Most assets need the first; some need both.
• The law rarely requires monitoring explicitly. It requires keeping the asset in proper condition, inspecting it periodically, and for geotechnical category II and III assets, defining the monitoring scope in the geotechnical design. Monitoring is the simplest way to prove compliance.
• The most sensible start is a pilot on one asset: a few well-chosen measurement points, thresholds from the design, an alarm procedure, and the first month of data.

What structural monitoring is, and what it is not

Structural monitoring is a system made up of four elements: sensors mounted on the asset, automatic collection of readings at a fixed rhythm, warning and alarm thresholds for each channel, and a procedure that turns a threshold breach into an action by a specific person. If any of those elements is missing, it is not monitoring. It is measurement.

SHM is the English term for the same practice. In the literature, it is used for engineering structures, most often bridges, while in Polish practice people also use structural monitoring or condition monitoring, and, where soil, excavation, and foundations are concerned, geotechnical monitoring. The boundary between those terms is conventional. A diaphragm wall for an excavation and the adjacent building form one physical system, so they are measured together.

What monitoring is not? It is not a one-off survey measurement, even a very accurate one. Surveying gives the condition at the moment of measurement, while monitoring gives the course over time. It is also not a periodic inspection under Article 62 of the Polish Construction Law. An inspection, colloquially a review, is a technical condition assessment performed by a person with building qualifications at least once a year, and for buildings with a built-up area over 2 000 m² and other structures with a roof area over 1 000 m², at least twice a year, by 31 May and by 30 November (Article 62(1)(1) and (3)). Inspection and monitoring do not compete with each other. An inspection answers the question “what condition is the asset in today”, while monitoring answers “what has happened since the last inspection, and is anything happening now”. We discuss this relationship further in the text on periodic inspections under Article 62 and continuous monitoring of halls and stadiums.

There is one more thing that is often called monitoring but is not monitoring: a logger on the asset from which a technician downloads data into a spreadsheet once a quarter. Technically, the sensors measure all the time. Practically, the asset is blind between visits, because nobody is looking at the numbers and nobody gets a notification when something crosses a threshold. The difference between “a reading every quarter” and “a reading at the default 15-minute interval” is pure arithmetic: 90 days is 8 640 quarter-hours, a weekly reading gives 672 quarter-hours, and a daily one 96. That is how much shorter the path from event to knowledge becomes. We are not comparing manufacturers here. We are comparing measurement rhythm. How to calculate your own multiplier and what it changes in operations is described in the text 100x faster - from event to decision.

A good working definition for a conversation with management is therefore this: structural monitoring is a service that ensures you learn about a safety threshold breach within 15 minutes, not after a quarter, and that you have a record that will stand up to an expert witness. How to turn that definition into an argument for finance, legal, and the CEO is shown in the text how to justify structural monitoring to management.

Four questions that monitoring answers

A well-designed system answers four questions and no more. First: is the asset behaving as the designer assumed, and do the measured values stay within the intended range? Second: is anything changing, and once temperature and service loads are removed, is there a trend? Third: is something happening right now that requires action? That is the alarm question, the only one that must be answered in minutes. Fourth: what exactly happened to the asset during a given period? That is the evidential question, asked after the fact by an expert witness, insurer, or regulator. If a monitoring system offer does not explain clearly how it answers each of these questions, it probably answers only the first.

What is measured and with what: quantities, sensors, units

You do not measure everything. You measure what the designer or monitoring engineer considers representative for the failure mechanism you fear. On a diaphragm wall for an excavation, the concern is rotation and loss of support, so you measure the displacement profile with depth, anchor and strut forces, and pore pressure behind the wall. On a hall roof beam, the concern is snow overload, so you measure strain in the zone of maximum moment. On a bridge, the concern is loss of stiffness and local damage, so you measure deflections, strains, support inclinations, and natural frequencies. You choose the quantity first, then the sensor. Never the other way round.

Table: quantity -> sensor -> unit -> typical threshold

The thresholds in the last column are only illustrative examples or a description of where a threshold comes from in the design. Real values for your asset come from the designer's calculations, from a standard, or from the baseline measurement period, never from a table in an article.

Four sensor families you will encounter most often

A vibrating wire sensor is a transducer in which the measured quantity is read from the vibration frequency of a tensioned steel wire. The frequency signal does not degrade over long cables, and the sensor design provides stability measured in decades. Manufacturer documentation refers to 20 years and more. That is why vibrating wire strain gauges, piezometers, and load cells remain the standard for long-term monitoring. Their limitation is reading speed. They are not suitable for dynamics. Why they are still the first choice almost a hundred years after the first installations is explained in a separate text on vibrating wire sensors and their stability.

An inclinometer is a sensor of angle of rotation relative to vertical or horizontal, reporting the result in milliradians. The conversion is simple: displacement [mm] = angle [mrad] × height [m]. Illustratively, a 0.5 mrad rotation at 20 m height gives 10 mm at the top. A chain inclinometer is a series of articulating inclinometer segments lowered into a casing in the ground or in a wall. From the angles of the individual segments, the full displacement profile is calculated by summing from the anchor point. We write about the difference between a probe and a chain in the text chain inclinometer or probe, and about where and how to place instruments in an excavation in the deep excavation geotechnical monitoring guide.

MEMS sensors, microelectromechanical accelerometers and inclinometers, are low-cost and fast electronics. They measure acceleration many times per second, so they are the basis of dynamic monitoring, and in inclinometer form, inclination. Their weak points are drift and temperature sensitivity, which must be compensated.

Resistive strain gauges and inductive sensors complete the range where a fast response or linear displacement measurement is needed. In practice, a good system uses several families in parallel, and the platform must be able to read them all. That is why the question “which sensors and loggers do you accept data from” is one of the first questions asked by an experienced specifier. For you, this means one thing: first define what you fear and what must be measured, and only then ask about sensors.

Static vs dynamic: a reading every 15 minutes by default, or more often and vibration events

Static monitoring, more precisely quasi-static monitoring, is the recording of quantities that change slowly, over hours, days, and seasons. Beam strain under snow, excavation wall rotation in the next digging stage, pore pressure after heavy rain, the inclination of an adjacent building, all of these are phenomena for which a reading at the default 15-minute interval, is dense enough, while still sparse enough for battery, transmission, and database systems to last for years. The standard quarter-hour rhythm is not accidental. It gives 96 points per day, which makes it possible to reconstruct the daily thermal cycle and separate it from the trend.

Temperature is the most important and most often neglected quantity in static monitoring. Every structure “breathes”. Steel and concrete expand during the day and contract at night, so the inclination of a column or the strain in a beam draws a daily wave on the chart. The wave itself is not the problem. The problem is that without temperature on the same chart, you cannot tell the wave from the trend, and a threshold set on an absolute value will be breached every afternoon. That is why the basic tool of the monitoring engineer is a multi-axis chart: the measured quantity on one axis, temperature on the other, same time base. Illustratively, if a wall returns to the same inclination every night, that is breathing. If the bottom of the wave is slightly higher every night than the previous one, that is a trend, and it is what should trigger a threshold.

Dynamic monitoring is the recording of vibration events: a heavy vehicle passing, a pile-driving strike, an explosion, vibratory compaction, wind. Here the sampling must be many times denser than one second, and the logger records not a single reading, but the full time history of the event. Analysis of such a record is no longer “value and threshold” only. It uses a set of tools: the spectrum (FFT), which shows at which frequencies the structure vibrates; analysis in third-octave bands, because those are the bands used by standards to assess the impact of vibration on buildings and people, in the 1-100 Hz range there are 21 bands; Butterworth filters to remove noise; and a dynamic envelope alarm, which triggers when the signal exceeds a defined limit curve in any band. Polish standard PN-B-02170:2016-12 assesses the harmfulness of vibrations transmitted through the ground to buildings, using an approximate method with SWD-I and SWD-II scales and five zones from negligible to failure, while PN-B-02171:2017-06 assesses the impact of vibration on people in buildings, in the 1-80 Hz bands, measured on the floor slab, with the criterion depending on room use and time of day. Details are in the text on building vibration monitoring according to PN-B-02170.

When do you need both? When there is a vibration source next to a static risk: a construction site with pile driving in dense urban fabric, a bridge where a change in natural frequency may reveal loss of stiffness, a hall with machinery. In such cases, MEMS accelerometers work alongside vibrating wire sensors, and the platform must be able to show both worlds on one asset. How this looks in bridge practice, from deflections to natural frequencies, is described in the bridge monitoring guide for asset managers.

Practical advice: do not buy dynamic monitoring “just in case”. Vibration events generate far more data than static monitoring and require someone to interpret them. If there is no vibration source and no question about stiffness, static monitoring with temperature as background is enough.

From sensor to SMS: system architecture

The architecture of any monitoring system can be described with one sequence, which is worth keeping in mind when reading proposals:

[sensor on the asset]
   → [cable or radio link]
   → [logger / hub: power, reading, buffer]
   → [transmission: GSM/LTE, Ethernet]
   → [platform: database, thresholds, alarms, dashboard, reports]
   → [SMS / e-mail to authorised person]
   → [confirmation: who took over, what they did]
   → [archive: raw data, communication log, export]

The scheme is deliberately textual. Logic matters more than drawing.

The sensor measures. The logger or hub powers the sensors, reads them at the defined rhythm, buffers data in case of communication loss, and sends them on. This is the layer where failures most often happen: a flat battery, a cut cable, a flooded penetration, a removed SIM card. That is why a good system treats missing data as a separate state (NO_DATA) and alarms on it just as it does for a threshold breach. Silence from the asset is not good news. It is the absence of news.

Transmission is usually via mobile network, or via wired network on assets with infrastructure. Data reaches the platform as frames in the logger manufacturer's format or as HTTP/JSON messages. This is where system openness is decided: if the platform accepts data from existing loggers and hubs, you can change software without replacing sensors. If it accepts only its own devices, you are locked to the manufacturer. How the path from an existing logger to an online platform looks, and how long it takes, is described in the text how to connect existing loggers to an online platform.

The platform does four things: it stores data, both raw and converted to engineering units, compares each reading with thresholds, sends notifications, and presents everything to people, to the engineer in a multi-axis chart and to management in a dashboard with three colors. How to turn such a dashboard into one page that management will read in a minute is described in the text on the monitoring report for management. The fifth thing, rarely mentioned in proposals, is auditability: the ability to check exactly what came in, when, and from which device. Without that, a chart is an image, not evidence.

The last layer is people. An SMS with no recipient assigned to act on it is worthless. The architecture therefore ends not at the server, but at the person who confirms the alarm and makes the decision, and at the record of that decision.

What this looks like in Inclify

The Inclify platform handles the automatic part of this chain: it accepts data from any measurement channel, from loggers and hubs via HTTP/JSON, including history from an existing system with original timestamps, stores raw frames in the communication log, by default for 7 days, compares each channel with WARNING and ALARM thresholds with hysteresis, distinguishes OK / WARNING / ALARM / NO_DATA states, sends SMS and e-mail to authorised people, and allows the person who takes over the case to confirm it, with a record of who and when, or silence it for a defined time. The team behind the platform has spent 15 years designing and maintaining measurement systems, from stadiums to hydraulic locks, and offers more than 40 sensor types for selection and installation. Details are on the platform page, and the vibration module is at /platform/vibration.

When monitoring is required, and when it simply pays

Management will ask: “is this mandatory at all?” The honest answer is that Polish law rarely uses the word “monitoring” and almost never requires it explicitly for a specific asset. It does, however, require things that are hard to prove without monitoring. The references below paraphrase the regulations as of 22 August 2026, not a direct quotation.

Construction Law: maintenance, inspections, responsibility

Article 61 of the Construction Law imposes on the owner or manager the duty to keep the asset in proper technical condition and to ensure, with due care, safe use in the event of external factors such as strong winds, intense precipitation, including snow, lightning, shocks, landslides, fires, or floods. Article 62(1) requires periodic inspections: at least once a year, at least once every five years, and for buildings with a built-up area over 2 000 m² and other structures with a roof area over 1 000 m², at least twice a year, by 31 May and by 30 November (item 3). This last obligation was introduced after the MTK hall disaster in 2006. Article 91a provides criminal liability for failure to comply with Article 61, with a fine of no fewer than 100 daily rates, restriction of liberty, or imprisonment of up to one year. The mere failure to carry out a required inspection under Article 62 is an offence (Article 93(8)). Article 22 lists the duties of the site manager, including securing the construction site, keeping construction documentation, and stopping works if there is a risk of danger and notifying the authority without delay. Protection of neighbouring property follows indirectly from these points and from Article 5(1)(9), which requires respect for the justified interests of third parties within the impact area of the structure.

None of these provisions says “install sensors”. But each one says: “you are responsible for the condition of the asset even when nobody is on it.” Monitoring is the tool that turns that responsibility into data.

Civil Code: the neighbour and strict liability

Article 147 of the Civil Code prohibits a landowner from carrying out earthworks in a way that threatens neighbouring properties with loss of support. Article 415 establishes general liability for damage caused by fault. Article 435(1) imposes liability on a risk basis, regardless of fault, on the operator of an enterprise set in motion by natural forces, and courts also classify general contractors using machinery as such enterprises, for example in the Supreme Court judgment of 17 March 2022, II CSKP 482/22, concerning damage to an adjacent building caused by construction works. In practice, this means that when cracks appear in the tenement next to an excavation, the contractor can defend itself only by proving force majeure or the exclusive fault of the injured party or a third party. Without an as-built inventory and continuous records of inclination and vibration, it is difficult even to show that the cracks existed earlier or that their growth did not coincide with the works. We devote a separate text to monitoring neighbouring buildings during construction and claims.

Geotechnics: categories and the observational method

The Regulation of the Minister of Transport, Construction and Maritime Economy of 25 April 2012 on determining geotechnical conditions for the foundation of structures introduces three geotechnical categories in Section 4. Excavations, retaining structures, embankments, and ground anchors are classified at least as category II. Category III includes, among others, structures in complex ground conditions, unusual structures whose execution or use may pose a serious hazard, high-rise buildings designed within existing urban fabric, tall structures founded deeper than 5 m or with more than one underground storey, heritage structures, and critical infrastructure. The designer assigns the category. For categories II and III, a geotechnical design is required (Section 7(2)), and it must define the scope of necessary monitoring of the built structure, adjacent structures, and surrounding ground during the works, as a result of the works, and during use (Section 10(10)).

PN-EN 1997-1 (Eurocode 7), clause 2.7, permits the observational method when prediction of geotechnical behaviour is difficult, and requires, before construction, acceptable limits of behaviour, a monitoring plan that will reveal exceedance in time, a contingency action plan, and, a sentence worth remembering, a response time for instruments and analysis procedures that is short enough in relation to the possible evolution of the system. Few standards make speed of data access a design requirement rather than an add-on. In the second generation of Eurocodes, EN 1997-1:2024, the observational method remains one of the recognised verification methods, but the clause numbering is different. Before citing it in documentation, check which edition applies to your project. We write about this in the text on the observational method under Eurocode 7 in practice.

Contracts, insurers, client requirements

The most common real source of obligation is not statute, but contract: a public client specification, contract conditions, insurer requirements, or an administrative decision with conditions. If monitoring is contractually required, its documentation is usually required as well, and that brings us back to the question of evidence.

And when it simply pays

When the cost of a blind period is greater than the cost of sensors. Halting an excavation for a week, a neighbour's lawsuit, closing a bridge pending expert review, evacuating a hall because of snow loading, each of these events costs more than a few measurement points with a platform. Monitoring also pays when it allows a more ambitious design, through the observational method, or extends the service life of an asset that calculations say should already be strengthened, but measurements show is performing properly. We break down the seven costs that are absent from the budget but paid after the fact in the text what happens when you do not monitor a structure.

Which assets are monitored most often

Bridges and viaducts

A bridge lives daily and seasonally. It expands during the day, contracts at night, and bearings and expansion joints work to the rhythm of temperature. That is normal. What is abnormal is a trend change after temperature is removed, or a drop in natural frequency, or increased deflection under the same load. You therefore measure span deflections, beam strains, support and pylon inclinations, bearing displacements, temperature, and, where it makes sense, vibration. Monitoring is especially valuable during works such as launching, cable-staying, and proof loading.

Deep excavations and neighbouring buildings

This is the most common case in cities. A diaphragm wall or Berlin wall, anchors or struts, and a tenement with residents a few metres away. You measure the wall displacement profile with chain inclinometers, anchor and strut forces, pore pressure, benchmark settlements, and, on neighbouring buildings, inclinations, crack opening with crack gauges, and vibration. The measurement frequency changes with the construction stage: during excavation, every day without data is a day of risk.

Halls, stadiums, warehouses

Buildings with a built-up area over 2 000 m² and other structures with a roof area over 1 000 m² are inspected under Article 62(1)(3) of the Construction Law at least twice a year, by 31 May and by 30 November, but snow falls between inspections. Vibrating wire sensors on beams measure strain and therefore indirectly roof loading, while inclinometers on columns measure inclination. When beam strain exceeds the warning threshold set in the design, the manager receives an SMS, not a call from the fire brigade.

Dams, locks, hydraulic structures

Here piezometers dominate, for pore pressure and seepage, together with displacement sensors on joints and expansion joints, inclinometers in the structure and subsoil, and strain gauges in concrete. The phenomena are slow, the horizon spans decades, and sensor stability matters especially here.

High-rise buildings and industrial facilities

High-rise buildings are monitored for inclination and wind-induced vibration. Industrial facilities are monitored for foundation strain under machines, settlement, and vibration. The common denominator is a long horizon and the need to distinguish normal operation from a trend.

Where point sensors are not enough: hybrid systems

A point sensor tells you what is happening over time at the place where it is installed. It does not tell you what is happening between points. If a crack appears one metre away from a strain gauge, the gauge may not see it. The answer to that limitation is hybrid systems, described in a 2025 industry publication on hybrid systems using bridges as an example. Automatic, continuous point-sensor measurements, such as vibrating wire, thermistors, accelerometers, and inclinometers, are combined with periodic, geometrically continuous DFOS measurements, in which a fibre laid along the member is read by a dedicated interrogator during a measurement session.

The division of roles is clear: point sensors are responsible for trends, early warning, and alarms. DFOS is responsible for strain distribution along the length, direct crack detection and location, and verification. An alarm from point sensors can trigger an additional DFOS session. The literature describes, among others, a five-span prestressed bridge about 260 m long, where more than 1 000 m of fibre optic sensor for periodic measurements was complemented by 12 point strain and temperature sensors at control sections. After the first year, operation dominated by temperature changes was confirmed, with no concerning trends and no local cracking. The full discussion, arguments, limitations, and installation stages are in the text on hybrid systems: point sensors and DFOS.

Thresholds, alarms, procedure

An alarm that nobody cares about is the worst alarm, worse than no alarm, because it teaches people to ignore notifications. That is why threshold and procedure design matters more than sensor choice.

Where threshold values come from

The warning threshold (WARNING) is the value after which someone must look at the data and assess the situation. The alarm threshold (ALARM) is the value after which someone must act on the asset. There are four sources of values: the design, expressed as a percentage of calculated displacement or strain; a standard, such as SWD scales for vibration; a contract or administrative decision; and baseline statistics, when the asset exists but there is no design, in which case the threshold is derived from the spread of the first weeks of measurement. Illustratively, if the designer allows a wall top displacement of 30 mm, WARNING can be set at 70% of that value, or 21 mm, and ALARM at 90%, or 27 mm, leaving room for reaction. This is a calculation example, not a recommendation.

Value, trend, rate

A threshold on an absolute value is the minimum. In geotechnics, the rate of change is equally important. Illustratively, a wall that moved 5 mm in three months and a wall that moved 5 mm in three days are two different assets, even though both show the same number. Platforms handle this differently. Some have a separate alarm type for growth in a time window. Others, including Inclify, where alarms are threshold-based, missing-data based, and dynamic, show the rate of change on the trend chart and in the 7-day and 30-day risk assessment report. Ask the provider directly how you will see the rate of change, not just the value. In dynamics, the equivalent is the envelope alarm: a limit curve as a function of frequency.

The third state: NO_DATA

Missing data is not an OK state. If a sensor goes silent, you need to know that just as you need to know about a threshold breach. A good procedure defines after how much silence an alarm is sent and who checks the logger.

Procedure: who, what, when

The response matrix answers three questions for each state: who gets the notification, what they must do and within what time, and to whom they hand it over if they cannot act themselves. It is written down before the system starts, not after the first alarm. Two post-alarm actions are equally important: confirmation, meaning who took over the case and when, and silencing with an end date for known works, after which the alarm returns. Without an end date, silencing becomes a permanent shutdown.

False alarms and tuning

Are you worried the alarm will be false and after the third SMS nobody will read it? That is a valid concern and it can be controlled. Most false alarms have four sources: temperature, a daily cycle without compensation, thresholds set too low “just in case”, electrical interference on long cables, and work on the sensor without silencing. After the first month of data, thresholds must be reviewed and tuned based on spread, not intuition. A good platform will suggest thresholds from history percentiles and show how many events per week they will generate. Statistical anomaly detection, such as z-scores relative to a baseline window, can help, provided it is treated as a hint for the engineer, not an oracle. What artificial intelligence actually does in monitoring, and what is marketing, is discussed in the text AI in structural monitoring: what works. The full guide to threshold setting, with seven reasons for false alarms and a who-what-when matrix, is in the text on warning and alarm thresholds.

Data as evidence

A dispute, with a neighbour, contractor, insurer, or building authority, usually starts with the question “how do you know that?”. A spreadsheet chart prepared after the fact is a weak answer. A record that stands up has several features.

Zero reading: the initial value of each channel, documented with a date, before works or loading begin. Without it, every change is “there from the start”. Sensor calibration: certificates and dates, and awareness of which sensors have overdue calibration, the calibration debt. Timestamps: unambiguous, preferably in one time zone, assigned on the device or upon receipt. Raw data: not only the value converted to millimetres, but also the raw logger values, wire frequency, ADC code, temperature, and the equations used to calculate the value. Communication log: a record of what came in, when, and from which device, including gaps. Platforms usually keep raw frames only briefly. If you need them longer, specify that in the contract. Human decision trail: who confirmed the alarm and when, who silenced it and until when. Immutability and export: the ability to show data to a third party in a form that cannot be challenged as altered. Retention: a contractual record of how long data are stored and who has access to them after the contract ends.

From an expert witness perspective, the most valuable thing is not the chart, but continuity: a 15-minute series from the zero reading to the event, with no gaps, and with temperature as background, showing that the neighbouring building's inclination did not change beyond its daily breathing. That is either the contractor's alibi or the injured party's proof, depending on what the numbers show. Monitoring is impartial, and that is where its value lies. We describe the features of reliable records and what to include in the contract in the text monitoring data as evidence in a dispute.

Cost: what it consists of

We will not give a price list here. Quotation is prepared on request, because two assets with the same number of sensors can differ in cost by several times depending on access, cabling, and whether traffic must be stopped. We can, however, show the structure so that a quotation can be read with understanding.

The cost components are usually these: sensors, with price depending on the family, because vibrating wire instruments and chain inclinometers are more expensive than MEMS but last longer; cabling or radio links; loggers and hubs with power supply; installation and asset access, such as scaffolding, lifts, lane closures, divers on hydraulic structures; data transmission; the platform, usually on subscription; calibration and service; and finally dismantling, which is often forgotten in proposals. Three things drive cost: the number of measurement points, the presence of dynamics, which means more data and more analysis, and access difficulty.

There are three cooperation models. Turnkey monitoring: sensor selection, installation, and platform from one team. This is the simplest option for a manager without an in-house measurement service. A platform only for existing sensors: when the asset already has loggers and needs only thresholds, alarms, and an online view. In that case, connection takes a few days. Pilot on one asset: limited scope, limited time, decision after data.

When comparing monitoring with periodic measurement, remember the costs that do not appear on the invoice: technician travel, labour time spent re-entering data, the risk of a blind period between readings, and the cost of a decision made without data. A five-year total cost of ownership template, without amounts and ready to fill in with your own figures, is in the text what structural monitoring costs consist of and how to calculate TCO.

How to start: a pilot on one asset

The biggest mistake when deploying monitoring is to start with the entire portfolio. Start with one asset where the risk is real and the question is specific: “is the excavation wall moving faster than the design assumed”, “does the hall beam have enough reserve under snow”, “is the bridge after reconstruction behaving as it did before”. The criteria for choosing such an asset and the plan for the first weeks are described in the text how to choose an asset for a structural monitoring pilot. A pilot has five steps.

1. Define the question and the failure mechanism. What exactly must not happen, and how you will know it is approaching.
2. Choose quantities and points. A few well-placed sensors are worth more than dozens scattered around. Temperature always as background.
3. Set thresholds and procedure before start-up. From the design, a standard, or the baseline period; the who-what-when matrix in writing.
4. Take the zero reading and start collection. If the asset already has loggers, connection to the platform takes a few days; full deployment with installation takes from several to a dozen weeks, depending on the asset and access.
5. After the first month of data, review the thresholds and decide. Expand the scope, change the points, or stop.

What to prepare before talking to a provider

The more you know about your own asset, the shorter the quotation and the fewer surprises during installation. Before the first conversation, gather: design or as-built documentation, plans, sections, structural scheme, because that defines the measurement points; what you fear and over what time horizon; a list of existing sensors and loggers with type and manufacturer; access conditions, such as height, traffic, power supply, and mobile network coverage; the list of people who should receive notifications and their roles; and formal requirements, meaning the clauses in the contract, decision, or specification that monitoring must satisfy. If you have documents from earlier periodic measurements, add them. They form the baseline from which thresholds can be estimated at an early stage.

How to choose a system

There are four classes of solutions on the market: a logger with a spreadsheet, the cheapest at the start and the most expensive to operate; the sensor manufacturer's software, for example Geokon Agent or Campbell Scientific LoggerNet; general platforms such as Vista Data Vision from Bentley, Encardio Proqio, Worldsensing, or Sixense Beyond/Geoscope; and Polish online platforms, including Inclify. No class is inherently bad. There is only a class that does not fit the question. A logger with a spreadsheet will not send an SMS, and a platform tied to one manufacturer will not read the sensors you already have. We have gathered a list of 30 questions to ask a provider, about data intake, export, alarms, procedure, dynamics, inclinometers, roles, audit, and total cost, in the text how to choose a structural monitoring platform: checklist, and an honest comparison of two approaches to data in the text Vista Data Vision and Inclify.

A good pilot ends not with a presentation, but with one of three decisions: expand, change, or stop. Any of them is good if it is based on data.

Glossary: 25 structural monitoring terms

The definitions are intentionally one sentence each. This is the terminology canon for the whole series.

1. Structural health monitoring (SHM) is the continuous, automated measurement of selected physical quantities of a structure, compared against thresholds and converted into an alarm and an evidential record.
2. Geotechnical monitoring is monitoring focused on soil, excavation, foundations, and their influence on adjacent structures, using inclinometers, piezometers, anchor load sensors, and benchmarks.
3. Vibrating wire sensor is a transducer in which the measured quantity is read from the vibration frequency of a tensioned steel wire, which provides resistance to long cables and stability measured in decades.
4. Inclinometer is a sensor of the angle of rotation relative to vertical or horizontal, reporting the result in milliradians (mrad).
5. Chain inclinometer (in-place) is a series of articulated inclinometer segments permanently installed in a casing, from which the displacement profile with depth is calculated automatically.
6. Strain gauge is a sensor of linear strain, vibrating wire or resistive, measuring the relative elongation of a member in microstrain.
7. MEMS accelerometer is a microelectromechanical acceleration sensor used to record vibration and, after processing, inclination.
8. Piezometer is a sensor of pore water pressure in soil or groundwater level, most often vibrating wire.
9. Crack gauge is a displacement sensor measuring the change in crack width or joint opening relative to the zero reading.
10. Strain (µε) is the relative change in member length expressed in millionths, where one microstrain is an elongation of one millionth of the length.
11. Inclination (mrad) is the angle of rotation expressed in milliradians, converted into linear displacement by multiplying by the height in metres.
12. Zero reading is the documented initial value of each measurement channel, against which all later changes are calculated.
13. Temperature compensation is the correction of a sensor reading for the effect of temperature, necessary to distinguish the daily “breathing” of the structure from a permanent trend.
14. Warning threshold (WARNING) is the value after which the authorised person must assess the data and the situation.
15. Alarm threshold (ALARM) is the value after which the previously recorded action procedure is activated on the asset.
16. NO_DATA is the state of a channel from which data do not arrive at the expected rhythm, treated as a separate reason for alarm.
17. Logger (hub) is a device on the asset that powers the sensors, reads them at the defined rhythm, buffers the data, and transmits them to the platform.
18. Dynamic monitoring is the recording and analysis of vibration events with dense sampling: time history, spectrum, third-octave bands, and envelope alarm.
19. Spectrum (FFT) is the decomposition of a vibration signal into component frequencies, showing at which frequencies and with what amplitude the structure vibrates.
20. Third-octave bands are the division of the spectrum into bands one third of an octave wide, typically 21 bands in building monitoring, used when assessing the impact of vibration on buildings and people.
21. Dynamic envelope alarm is an alarm triggered when the event spectrum exceeds a defined limit curve in any band.
22. Observational method is an Eurocode 7 design approach in which assumptions are verified by measurements during construction, with a predefined range of acceptable behaviour and a contingency plan.
23. Geotechnical category is the class of an asset under the 2012 regulation on foundation geotechnical conditions (I-III), which determines the scope of investigation and documentation, including monitoring.
24. DFOS (distributed fibre optic sensing) is a geometrically continuous measurement technique in which a fibre optic cable laid along a member acts as a strain and temperature sensor over the entire length, read by an interrogator.
25. Hybrid system is a combination of automatic, continuous point-sensor measurements with periodic, geometrically continuous DFOS measurements, in which an alarm from point sensors can trigger an additional fibre optic session.

Sources and further reading

• Act of 7 July 1994 - Construction Law (Articles 5, 22, 61, 62, 91a, 93), Journal of Laws 1994 No. 89 item 414, as amended - isap.sejm.gov.pl.
• Act of 10 May 2007 amending the Act - Construction Law and certain other acts (large-area inspections, Article 91a), Journal of Laws 2007 No. 99 item 665 - isap.sejm.gov.pl.
• Act of 23 April 1964 - Civil Code (Articles 147, 415, 435), Journal of Laws 1964 No. 16 item 93, as amended - isap.sejm.gov.pl.
• Judgment of the Supreme Court of 17 March 2022, II CSKP 482/22 (liability of a general contractor under Article 435 of the Civil Code for damage to a neighbouring building) - sn.pl.
• Regulation of the Minister of Transport, Construction and Maritime Economy of 25 April 2012 on determining geotechnical conditions for the foundation of structures, Journal of Laws 2012 item 463 - isap.sejm.gov.pl.
• PN-EN 1997-1:2008 Eurocode 7: Geotechnical design - Part 1: General rules (clause 2.7 Observational method; chapter 4) - Polish Committee for Standardization, sklep.pkn.pl. Second generation: EN 1997-1:2024 (CEN).
• PN-B-02170:2016-12 Assessment of the harmfulness of vibrations transmitted through the ground to buildings - PKN, sklep.pkn.pl.
• PN-B-02171:2017-06 Assessment of the impact of vibration on people in buildings - PKN, sklep.pkn.pl.
• Peck R.B., “Advantages and Limitations of the Observational Method in Applied Soil Mechanics”, Géotechnique 1969, vol. 19, no. 2, pp. 171-187.
• Dunnicliff J., “Geotechnical Instrumentation for Monitoring Field Performance”, Wiley, New York 1988.
• Industry publication (2025) “Hybrid systems - introduction using bridges as an example” and “Hybrid systems - case study: implementation on a prestressed bridge”.
• Technical documentation of vibrating wire sensor and logger manufacturers: geokon.com, campbellsci.com.
• Monitoring platform documentation: help.vistadatavision.com (Bentley), worldsensing.com, encardio.com, sixense-group.com.

What next

If after reading this you have one asset where the safety question is specific, such as an excavation next to a tenement, a hall before winter, or a bridge after reconstruction, the most sensible next step is a pilot: a few measurement points, thresholds from the design, an alarm procedure, and the first month of data, after which you decide what to do next. The best time is before the event you want captured in data: before excavation starts, before the first snow, before the periodic inspection you want to supplement with records. If you already have sensors and loggers, connecting them to a platform takes a few days.

Let's talk about a pilot on your asset - describe the asset and the question, and we will reply within 24 hours and propose a pilot scope, or honestly tell you that monitoring is not needed.