Deep excavation geotechnical monitoring: what to measure and how often

Geotechnical monitoring of a deep excavation is the continuous measurement of retaining structure movements, groundwater level, loads in struts and anchors, and the response of nearby structures, compared with design values. It is measured with inclinometers in the wall, piezometers, strain gauges and load cells, benchmarks, and crack gauges. The measurement frequency increases during excavation and anchoring phases, and warning and alarm thresholds trigger a predefined response plan.

In brief

• An excavation is designed on assumptions about ground and water; monitoring checks whether reality stays within those assumptions before you see the effects on the wall of the building next door.
• The MTBiGM Regulation of 25.04.2012 classifies excavations and anchors as at least geotechnical category II, and for categories II and III it requires a geotechnical design with a monitoring scope; Eurocode 7 makes the observational method depend on a monitoring plan and an emergency action plan.
• Minimum set: inclinometers in the retaining structure, piezometers behind the wall, load sensors in struts or anchors, benchmarks on the crest and on neighbouring buildings, crack gauges on existing cracks.
• A weekly reading during excavation is too infrequent; between readings the excavation is blind, and retaining structure movements happen over days, not quarters.
• Record thresholds separately for each phase as a percentage of design values and assess the rate of change as well; every threshold must have an assigned person and action.

Why measure an excavation if it is already designed

An excavation support design is a model. A good model, but based on a few boreholes, soil parameters from laboratory tests, and the assumption that the contractor will do everything in the order shown on the drawings. In practice, a weak soil lens may occur between boreholes, the groundwater level may stand half a metre higher than in the documentation, and the second level of struts may be installed a week late because steel was still awaited. Each of these changes the movements of the support, and together with them the ground settlements behind the wall and what happens to the foundations of neighbouring buildings.

If you design or accept an excavation monitoring system, your usual concern is one: that wall movement will occur between readings and you will learn about it from the neighbour, not from the chart. Geotechnical monitoring is the only way to find out while something can still be done: add a strut, halt excavation, lower the water level, or prestress the anchors further. After the fact, only a damage report and a dispute over fault remain. From the site manager's point of view, the difference is simple: if you measure, you have data and a procedure; if you do not, you have an opinion and hope.

There is a second reason, mentioned less often. Excavation support design in existing urban fabric is sometimes overdesigned out of caution: more struts, deeper wall embedment, more conservative parameters. Eurocode 7 allows the observational method, meaning less conservative design, provided that monitoring confirms the assumptions during the works. Without measurements, that path simply does not exist. We cover it in more detail in a separate post on the observational method under Eurocode 7.

This guide explains how to plan measurements on a deep excavation so that they answer three questions: is the support working as designed, is the water behaving as assumed, and is the surrounding area affected. If you are just starting with structural monitoring, begin with the complete SHM guide, then return here for excavation-specific details. Every sensor in this text should answer one of those three questions; if it does not answer any of them, it is not needed.

Legal and standard framework: the 2012 regulation and Eurocode 7

Geotechnical categories

The Regulation of the Minister of Transport, Construction and Maritime Economy of 25 April 2012 on determining geotechnical conditions for the foundation of building structures (Journal of Laws 2012, item 463) introduces three geotechnical categories in § 4, depending on the complexity of the ground conditions and the structure, including how severe the consequences of failure may be. Category I covers small structures in simple conditions (for example, excavations up to 1.2 m and bracing with a level difference of up to 2.0 m). Category II covers structures in simple and complex ground conditions that require a quantitative assessment of geotechnical data, and the regulation explicitly lists excavations, embankments, retaining structures and ground anchors. The category is determined by the designer and changed if the encountered conditions differ from those assumed.

Category III includes, among others, all structures in complex ground conditions (landslides, quicksand, expansive soils, mining damage), unusual structures whose construction or use may create a serious hazard, high-rise buildings designed in existing urban development, tall structures founded directly deeper than 5.0 m or with more than one basement level in the ground, heritage structures and critical infrastructure. The regulation does not use the term "deep excavation": an excavation as such falls into category II, and it moves into category III because of the ground conditions or the character of the structure it serves. A garage pit for a tower building with several levels between tenement houses, with groundwater above the excavation bottom, often meets these criteria in practice. The designer makes the decision.

Geotechnical design and monitoring scope

For category II and III structures, the regulation requires, in addition to the geotechnical opinion and subsoil investigation documentation, a geotechnical design (§ 7(2)); for category III and for complex ground conditions in category II, geological-engineering documentation is also required (§ 7(3)). The geotechnical design must comply with PN-EN 1997-1 and PN-EN 1997-2 and, under § 10 point 10, must define the scope of necessary monitoring of the completed structure, neighbouring structures and the surrounding ground, needed to identify hazards during the works, as a result of the works, and during service.

Two conclusions follow. The monitoring scope is created at the geotechnical design stage, not a week before the first excavation, and it is endorsed by the designer, not only by the contractor. This already applies to category II, so when the investor asks "is it mandatory at all?", the answer is: the monitoring scope is a required element of the geotechnical design, and the number of sensors and the frequency come from that design.

Observational method

PN-EN 1997-1 (Eurocode 7, Part 1; Polish edition PN-EN 1997-1:2008) describes the observational method in clause 2.7 as a design approach when the prediction of geotechnical behaviour is difficult. Before construction starts, one must set the limits of acceptable behaviour, assess the range of possible behaviour and show that the actual behaviour will remain within the limits with acceptable probability, develop a monitoring plan that will reveal any exceedance early enough for effective response, ensure that the instruments and data analysis procedures have a sufficiently short reaction time relative to the possible evolution of the system, and develop an emergency action plan. During construction, monitoring is carried out according to the plan, results are assessed at appropriate stages, and actions are triggered after the limits are exceeded; equipment that does not provide reliable data must be replaced or supplemented. Monitoring is therefore not an addition to the observational method, but a condition for its existence.

The same Eurocode, in Chapter 4, requires supervision of execution, monitoring of structural behaviour during and after construction, and maintenance; monitoring is meant to verify the accuracy of design predictions and the continued proper performance of the structure. In the second generation of Eurocodes (EN 1997-1:2024), the observational method remains one of the recognised verification methods, but the clause numbering differs from the 2008 edition. Before referencing it in documentation, check which edition applies.

General obligations also apply. The basic duties of the site manager under Article 22 of the Construction Law include, among others, securing the construction site, keeping construction records, directing the works in accordance with the design and regulations, and stopping the works if a potential hazard is identified and notifying the relevant authority without delay. In practice, the response plan on an excavation is the tool used to carry out that last point. Liability for damage to neighbouring structures is regulated by the Civil Code (Article 147 on earthworks threatening loss of support, Articles 415 and 435 on liability for damage); we discuss how monitoring protects the site manager from claims in a post on monitoring neighbouring buildings.

For you, the conclusion is simple: the monitoring scope belongs in the geotechnical design, the thresholds and action plan belong in the monitoring plan, and you must be able to show that you acted according to them.

What to measure: instrument -> quantity -> location -> purpose

The measurement set for a deep excavation answers the three questions from the first section: how the support is working, what the water is doing, and how the surroundings are reacting. Every instrument in the table answers one of them.

Definitions of terms from the table

An inclinometer is an instrument that measures inclination relative to the vertical; on an excavation, the inclinations of successive segments in the inclinometer tube are summed from the bottom, or from the anchor, to obtain the horizontal displacement profile of the support. An in-place chain inclinometer is a set of sensors permanently installed in the tube and read automatically, unlike a probe that the operator lowers into the tube for each measurement. Manufacturers typically offer chain segments from 0.5 to 3 m (some systems from 0.25 m), and the system accuracy of probes is given at about ±2 mm over 25 to 30 m of tube (datasheets, for example Geokon, Sisgeo, Measurand). A piezometer is a sensor for pore pressure or groundwater level in a given layer. A benchmark is a stabilised geodetic point whose position is measured periodically relative to reference points outside the excavation influence zone. A crack gauge is a sensor for crack width or joint opening, mechanical or electronic.

Vibrating wire sensors are transducers in which the measured quantity is read from the frequency of a tensioned wire's vibration; they are used in piezometers, strain gauges and load cells because of their long-term stability.

How to convert inclination and strain

Converting inclination to displacement is simple: segment displacement ≈ segment length × inclination in radians. As an illustration: a 2 m segment and an inclination of 0.5 mrad give 2 000 mm × 0.0005 = 1 mm. The profile is obtained by adding segment displacements from the anchor upward, which is why it is so important that the lower end of the tube actually sits in immovable ground. We explain how to read such a profile and when to choose a chain system instead of a probe in the text chain inclinometer or probe.

On struts, strain is usually measured and force is calculated from the cross-section and the steel modulus. As an illustration, ignoring temperature effects: a steel pipe strut with a cross-sectional area of 190 cm² and a strain of 200 µε carries a force of F = E·A·ε = 210 GPa × 0.019 m² × 0.0002 ≈ 800 kN. We return to temperature in the section on errors, because that is where interpretation most often fails. Remember one thing from this section: measure what you can compare with the design number, displacement with calculated displacement, force with design force, water level with the assumed level.

Layout: measurement sections

Instruments are not distributed uniformly around the perimeter. Measurement sections are planned, meaning places where, on one line perpendicular to the wall, you have: an inclinometer in the support, a piezometer behind the wall, benchmarks on the crest and on the ground, and if a building is within range, its benchmarks, crack gauges and tilt sensor. A section gives the full picture: wall deflection, water level, ground settlement and building response in one place, at the same time.

Where should sections be placed? Where the excavation is deepest, where the ground is weakest or the water is highest, where a sensitive structure is closest (an old building, a utility tunnel, metro tunnel, rail tracks), and at corners and at changes in support type, because there the displacement pattern differs from plane-strain assumptions. Usually this means at least one section on each side of the excavation and additional ones near sensitive neighbours. The number is determined by the designer based on analysis, not budget. If budget forces a reduction, it is better to have three complete sections than ten isolated sensors without context.

A few practical notes on individual instruments:

• The inclinometer tube should extend clearly below the excavation bottom and below the potential rotation zone of the wall toe so that the lower point can be treated as immovable. If it does not, the profile calculated from the anchor has a systematic error that grows as excavation deepens. An alternative is geodetic control of the tube head and calculating the profile from the top.
• Piezometers are installed in specific aquifers, not "somewhere in the ground"; in a system with a confined water table beneath the excavation bottom, a separate piezometer below the bottom is mandatory because of the risk of hydraulic heave.
• Struts are instrumented so that there is at least one element from each level in the section; strain gauges are mounted in pairs or in sets of four around the circumference to average bending and leave pure compression.
• Anchors - load cell on several anchors in each row, with emphasis on those in the most critical section; anchors are also checked by test stressing, but the test tells you about a moment in time, while the sensor tells you about the trend.
• Neighbouring buildings - benchmarks on the corners and at mid-span of the walls facing the excavation, tilt sensors on the wall closest to the excavation, crack gauges on cracks documented in the zero-condition survey.

The extent of the excavation influence zone depends on the ground and the type of support. It is assessed in the design; monitoring should cover structures within this zone, and at least one benchmark should be placed farther away, outside it, as a reference point. If you are accepting the system, start here: is every section complete, and is there a reference point outside the influence zone?

Measurement frequency by work phase

An excavation does not work uniformly. The largest increases in support displacement occur at each excavation stage, before the next level of struts or anchors is installed, and during dewatering. After the base slab and the structural floors are completed, the structure stabilises and movements decrease. Measurement frequency should follow that rhythm. The table below is practical guidance, not a standard and not deployment data; the designer sets the final frequency in the monitoring plan.

Two notes on this table. First, for manual reading, "daily" practically means "once a day, if someone has time and the tube is not buried"; with continuous monitoring, the table stops being a compromise between safety and logistics. Second, the point of dense measurement follows arithmetic: a weekly periodic measurement is one point, a reading at the default 15-minute interval, is 672 points in the same week. Excavation down by one level usually takes several days. With weekly reading, you see the whole increase in displacement from that phase after the fact, as a single number, without knowing whether it jumped on Wednesday after cutting through a lens of weak soil or rose evenly.

Geodetic measurements remain periodic because a benchmark is measured by a surveyor. Their role, however, differs from automatic sensors: they provide absolute reference and control what the inclinometer does not see, namely displacement of the tube head and ground settlements. The results of both measurement types are compared with each other; if the inclinometer shows 10 mm of deflection and the crest moved 25 mm according to the surveyor, it means the wall toe is rotating and the assumption of an immovable anchor is no longer valid. For you: during excavation and anchoring phases, you need data from every work shift, not every week. Use that criterion when assessing an offer for the system.

Thresholds and response plan

Measurement without thresholds is an archive. Thresholds turn numbers into decisions. On an excavation, two levels are usually used: warning, after which the data must be reviewed and actions prepared, and alarm, after which actions are carried out. Some projects add a third, intervention level. For each measurement channel, the threshold should be recorded in the monitoring plan together with the responsible person and the action.

Where threshold values come from

The best source is the support design: calculated displacement for each phase, design force in struts and anchors, assumed water level. Thresholds are often taken as a percentage of design values. As an illustration: a warning threshold at around 70 to 80% of the calculated displacement for a given phase, an alarm threshold at around 100%, and if there is a third level, then above the calculated value at which the designer still allows safe operation of the structure. This is not a code requirement, only design practice. The designer sets the actual values for their excavation and support.

It is important that thresholds be phase-specific. A displacement of 15 mm in the cantilever phase, when the design predicts 8 mm, is an alarm; the same 15 mm after two levels of struts have been installed, when the design predicts 25 mm, is within assumptions. One threshold for the entire construction period either misses problems at the start or triggers false alarms at the end. Worried that the alarm will be false and after the third one nobody will take it seriously? Phase-specific thresholds and hysteresis, meaning the alarm clears only after the value drops clearly below the threshold, are the answer.

The second source is allowable impact on neighbouring assets: building settlement and tilt (from technical condition reports), allowable crack width, allowable displacements of underground infrastructure according to utility owner requirements. These thresholds are sometimes stricter than those resulting from the support capacity and, in city centres, they usually govern.

Rate of change

Alongside the absolute threshold, the rate of change must be assessed. A displacement of 12 mm that built up over three weeks and 12 mm of which 6 mm arrived in the last day are two different situations. The first is normal, the second requires a response. As an illustration: an increase above 2 mm per day in a stable phase can be a warning criterion in the monitoring plan, regardless of how far it is from the design value. Rate is especially important for piezometers beneath the excavation bottom and for anchor forces, where a sudden jump says more than the absolute level. Not every platform calculates such a threshold automatically, so the rate criterion is recorded in the plan as a duty of the person reviewing the trends after each work shift. We discuss threshold design, hysteresis and avoiding false alarms in the text on warning and alarm thresholds.

Response plan

A threshold without an assigned action is decoration. The response plan for an excavation is a table: channel, threshold level, who receives the notification, what they do, within what time, and who confirms it. Illustrative framework:

No data deserves its own row. If the inclinometer is silent for six hours during excavation, that is not a neutral state. Someone must know about it and someone must deal with it just like an alarm. The plan should also say what happens if the alarm recipient does not confirm it within the set time: who calls whom. Phase-specific thresholds, a rate criterion, a "no data" row, and a named person for each level. Missing any one of those four should stop approval of the monitoring plan.

Checklist before the first excavation level

Before the excavator goes below the first level, you should be able to tick off:

• monitoring scope described in the geotechnical design and monitoring plan signed by the designer;
• complete measurement sections at the deepest, weakest and closest-to-neighbour locations, plus a reference point outside the influence zone;
• inclinometer tubes concreted into the support, piezometers in the aquifers, benchmarks and crack gauges on neighbouring buildings, all before any ground interference;
• documented zero reading from a series of several days, with date and conditions;
• warning and alarm thresholds recorded separately for each phase, with hysteresis and a rate criterion;
• response plan with names, phone numbers, reaction times and a path for non-confirmation;
• data visible online for supervision and the site manager, not only recorded in the logger;
• zero-condition survey of the surroundings with crack photographs.

Most common errors: installation and zero reading

These errors recur on excavations regardless of project scale. Each can be eliminated at the planning stage.

Installation too late. An inclinometer inserted into the wall after the first excavation level has not seen the cantilever deflection, which is usually one of the largest components of the total displacement. Its zero is shifted and nobody knows by how much. Inclinometer tubes are cast into the diaphragm wall or secant pile wall during concreting, piezometers are installed before dewatering, and benchmarks on neighbouring buildings are installed before any ground interference. This error cannot be fixed later, which is why the decision on the monitoring system must be made before the first excavation level, not after it.

No proper zero reading. One reading on the day of installation is not zero. After installation, the sensor stabilises, grout in the tube sets, and ambient temperature changes in a daily cycle. Zero should be the average of a series of readings over several days, taken before the start of works affecting the instrument, and documented with date and conditions. In a dispute with a neighbour, the first question will be: "what are you measuring from?".

Inclinometer without an immovable anchor. Tube too short, profile calculated from the bottom, wall toe rotates. Result: a neat profile that consistently underestimates displacement. Periodic geodetic measurement of the tube head is the check.

No zero-condition survey of the surroundings. A crack gauge on a crack that nobody has previously described and photographed proves only that the crack is moving; it does not prove that it existed before construction. See also the post on monitoring neighbouring buildings. The common denominator of these four errors is this: all are made before the works begin, and none can be undone after the fact.

Most common errors: interpretation and alarms

No temperature compensation for struts. A steel strut in the sun heats up and wants to expand; the walls do not allow it, so the force inside it rises. As an illustration: the coefficient of thermal expansion of steel is about 1.2·10⁻⁵ 1/K, a strut with a cross-section of 190 cm², temperature difference between night and afternoon of 20 K. Under full restraint, the force increase is ΔF = E·A·α·ΔT = 210 GPa × 0.019 m² × 1.2·10⁻⁵ × 20 ≈ 960 kN. In reality the restraint is not complete, so the increase is smaller, but the order of magnitude shows that without a temperature sensor on the strut, the force threshold will trigger every afternoon and clear at night. Measure temperature next to the strain gauge, plot force against temperature, and learn to distinguish daily "breathing" from trend.

One set of thresholds for the whole project. As described above: either false alarms or a missed problem.

Alarm without an addressee. The notification goes to a mailbox that nobody checks on Saturday. The alarm must reach specific people by phone and must be confirmed; the response plan must say who calls whom when there is no confirmation.

Data in the logger, not in view. Automatic reading every 15 minutes by default, or more often, but the result downloaded from the logger only once a week to the site is, in practice, weekly reading with good historical resolution. Data must be visible in real time to supervision and the site manager, not only stored. These four errors do not damage the measurement, but they strip it of meaning. You have numbers, but no decisions from them in time.

What it looks like in Inclify

Inclify handles the automatic part of what we described: in-place chain inclinometers, piezometers and vibrating wire strain gauges, load sensors, crack gauges and temperature sensors connected to loggers that send readings every 15 minutes by default, or more often. The platform accepts any interval and calculates data completeness against the actual cadence. For the chain inclinometer, the platform calculates the displacement profile from the anchor upward in axes A and B and the resultant based on the segment lengths configured by you, provides the maximum resultant displacement (the depth at which it occurs is read from the profile), and after the first valid frame it automatically creates a dashboard with the vertical profile and trend. Supervision sees the shape of the wall deflection after each shift, not after the logger is downloaded. Manual probe readings are not supported by the platform; if you have a probe on the excavation, you compare its results with the chain profile outside the platform.

Each channel has a warning threshold and an alarm threshold with hysteresis, and the states OK / WARNING / ALARM / NO_DATA. Missing data from an inclinometer during excavation is shown on the dashboard and notified after the expiry of the window that you set yourself (default 60 minutes), rather than being a blank place on the chart. Notifications are sent by SMS and e-mail, plus in the app, to project users according to their preferences; you confirm the alarm (with the record of who and when) and can silence it only for a defined period. The platform does not have a threshold for rate of change or automatic escalation to additional people. The rate criterion and the path of "who calls when nobody has confirmed" are recorded in the response plan, and trend assessment is supported by a 7/30 day risk report calculated from the data.

Strut force is plotted against steel temperature on a two-axis chart, which solves the afternoon alarm problem described above. Measurements are stored in the database in UTC without automatic deletion, and you can export them to CSV; raw frames from devices are kept in the communication log for the configured period (default 7 days). If the site already has loggers, data can be received via HTTP/JSON. We describe how this works in the post on connecting existing loggers, and the details of inclinometers are on the inclinometer solutions page.

Sources and further reading

• Regulation of the Minister of Transport, Construction and Maritime Economy of 25 April 2012 on determining geotechnical conditions for the foundation of building structures (Journal of Laws 2012, item 463) - ISAP: https://isap.sejm.gov.pl/isap.nsf/DocDetails.xsp?id=WDU20120000463
• Act of 7 July 1994 - Construction Law (consolidated text) - ISAP: https://isap.sejm.gov.pl/isap.nsf/DocDetails.xsp?id=WDU19940890414
• Act of 23 April 1964 - Civil Code - ISAP: https://isap.sejm.gov.pl/isap.nsf/DocDetails.xsp?id=WDU19640160093
• PN-EN 1997-1 Eurocode 7: Geotechnical design - Part 1: General rules (including clause 2.7 observational method, Chapter 4 supervision and monitoring) - Polish Committee for Standardization, sklep.pkn.pl
• PN-EN 1997-2 Eurocode 7: Geotechnical design - Part 2: Ground investigation and testing - Polish Committee for Standardization, sklep.pkn.pl
• Borecka A., Stopkowicz A., Sekuła K., "Metoda obserwacyjna i monitoring geotechniczny w świetle przepisów prawa...", Przegląd Geologiczny 2017, vol. 65, no. 10/2
• Nicholson D., Tse C., Penny C., "The Observational Method in ground engineering: principles and applications", CIRIA Report R185, London 1999
• Dunnicliff J., "Geotechnical Instrumentation for Monitoring Field Performance", Wiley, 1988/1993
• Datasheets for chain inclinometers and probes: Geokon (https://www.geokon.com/6180), Sisgeo (https://sisgeo.com/products/ipi-in-place-inclinometers/), Measurand (ShapeArray)
• Related posts in this series: complete guide to structural monitoring, chain inclinometer or probe, observational method under Eurocode 7, monitoring neighbouring buildings, warning and alarm thresholds, connecting existing loggers

What next

If you are planning a deep excavation in an urban setting, the best time to decide on a monitoring system is now, before wall concreting and before the first excavation level, because inclinometer tubes and zero readings cannot be added later. If you are already running an excavation where sensor data end up in a logger and a weekly spreadsheet, we can show you on a site similar to yours what excavation monitoring looks like in Inclify: inclinometer profile from the anchor, strut forces against temperature, channel states and SMS alerts for supervision. You can start with a pilot on one excavation or by connecting the sensors and loggers you already have, which usually takes a few days. Write to us. We reply within 24 hours: let's talk about monitoring your excavation.