What a T-bar actually measures
Did you know that the load recorded by a T-bar pushed into seabed clay is not itself the clay’s undrained shear strength? A T-bar is a horizontal cylindrical crossbar driven vertically through sediment, typically on the end of an instrumented rod. Soil flows around it. The instrument measures the force necessary to continue penetration at a specified speed, while the interpretation supplies a resistance factor that turns the measured resistance into an estimate of strength. Understanding that difference is the central skill in reading an offshore T-bar report: identify the observation, identify the correction and identify the model-dependent inference rather than presenting them all as identical facts.
A translation can fail even when every numerical digit survives. Suppose a report says “net T-bar resistance was 90 kPa and the interpreted undrained strength was 7.5 kPa using a factor of 12.” A careless summary might state that the penetrometer measured 7.5 kPa directly, concealing the assumption that made the value possible. Another version might write “the soil supports 90 kPa”, confusing a penetration resistance with the allowable pressure of a foundation. Neither conveys what the test demonstrates. A reliable explanation names the method, the depth, the direction of movement, the selected factor and the limitation on how strength is used.
This guide treats translation as the transfer of scientific meaning from a machine record into intelligible, trustworthy language. We begin with the geometry and physics; proceed through net resistance, rate, sensitivity and cycles; test the arithmetic in examples; then ask what a reader can legitimately infer about a pipeline, anchor or foundation. All numerical worked records here are invented teaching datasets rather than field measurements. They show calculation procedures, not site-specific design values. Professional ground investigation, calibration and geotechnical judgement remain essential before using the results for a real structure.
What makes the seabed test different from a conventional cone
A cone penetration test and a T-bar test both involve pushing an instrument through soil, but their measured resistances are not interchangeable. A cone concentrates resistance at a pointed tip and reports cone resistance together with sleeve friction and, when equipped, pore-water pressure. A T-bar is a full-flow device: sufficiently deep in soft clay, soil can move around the crossbar and close behind it. The resulting resistance is more closely linked to undrained shear strength through a full-flow mechanism. This is why offshore programmes may use T-bars alongside CPTu rather than merely treating them as alternative names for one probe.
The phrase “sufficiently deep” matters. Near a sediment surface, the soil displaced above a T-bar may open into a cavity instead of flowing completely around the cylinder. A cyclic test started before a dependable full-flow regime develops can therefore mix genuine material softening with a change in geometry and boundary conditions. A report that lists a dramatic reduction in resistance should not automatically be translated as “the clay lost the same percentage of its strength”. The operator must establish whether the cavity closed, whether the motion was symmetric and whether the rod and instrument were behaving as intended.
A full-flow interpretation also depends on drainage conditions. In very soft, fine-grained soils under a sufficiently rapid penetration rate, movement may be effectively undrained on the timescale of each push. A more permeable or partially drained layer may not behave that way. A crossbar encountering shells, sand lenses, fissures, roots or a stiff crust can produce spikes that are physical observations but poor representatives of homogeneous soft-clay strength. Accurate reading begins with the material description and depth profile before any generalisation about how the entire seabed will respond.
Dimensions and the meaning of projected area
For a cylindrical T-bar with diameter D and horizontal length L, its nominal projected area normal to vertical motion is D multiplied by L. If D is 40 millimetres and L is 250 millimetres, the area is 0.040 times 0.250, or 0.0100 square metres. This is a projected silhouette, not the curved surface area of the cylinder and not the plan area of the surrounding seabed. Resistances often appear in kilopascals because an instrument force in newtons is divided by the relevant area in square metres. The correct area definition must match the device’s calibrated reporting convention.
Imagine a net vertical force of 900 newtons on that bar. Dividing 900 N by 0.0100 m² gives 90,000 pascals, or 90 kilopascals. If someone accidentally enters D and L in millimetres without converting the units, the arithmetic may still look neat but the resistance becomes nonsensical. Equally, if they mistake 0.0100 m² for 0.100 m², they reduce every reported pressure by a factor of ten. A careful transfer writes the dimensional conversion visibly and checks that the output carries force per area rather than force per length.
Instrument makers may already report a calibrated resistance after applying device-specific tare and correction procedures. When that happens, it would be wrong to divide the already corrected resistance by the projected area a second time. Ask which channel is being read: measured axial load, corrected net force, or calculated net bearing resistance. The metadata should also identify whether shaft effects were corrected or considered negligible, whether the sign convention changes on extraction, and whether reported depths refer to crossbar centre, leading edge or another datum.
Net resistance and the corrections behind it
The phrase “net resistance” means that a raw load has been adjusted to represent the resistance attributed to soil acting on the penetrating tool, according to the method’s correction rules. The raw load may contain hardware weight, sensor offset, pressure effects and forces carried by parts of the push system. Offshore guidance often notes that a deep full-flow geometry reduces the need for the same overburden correction applied to cone resistance, but that does not mean every recorded T-bar force is automatically clean. The loading arrangement and instrument instructions decide which corrections belong in a particular calculation.
Consider a fictional instrument that registers 1,045 N while the operator’s documented correction ledger accounts for 105 N of non-soil contributions, leaving 940 N net. A nominal projected area of 0.0100 m² yields 94 kPa. It would be misleading to describe 1,045 N as the corrected force, and equally misleading to subtract an additional arbitrary 105 N because another job used a similar rig. Real correction items must be independently identified, justified and signed. The example only demonstrates that a transparent ledger leads from raw observation to reported resistance.
Differences between push and pull values deserve attention. A positive compressive push signal and a negative extraction signal are often plotted on opposite sides of the axis. For comparing resistance magnitudes, an analyst may take the absolute value of the extraction force, but must preserve which half-cycle it belongs to. Reporting the extraction reading as simply “negative soil strength” would be meaningless. The sign describes a direction and instrumentation convention; the magnitude describes resistance. A good plot labels both and states the depth interval over which any average was calculated.
Converting resistance to undrained strength
An idealised reading often takes the form sᵤ = qₜ/Nₜ, where sᵤ denotes an interpreted undrained shear strength, qₜ the net T-bar resistance and Nₜ the chosen dimensionless resistance factor. The equation is useful precisely because it makes the assumption visible. For qₜ = 96 kPa and Nₜ = 12, the inferred strength is 8.0 kPa. A different, defensible factor of 10 would give 9.6 kPa for the same observation. Neither conclusion can be selected honestly without explaining which strength reference and calibration justify the factor.
The factor is not a universal constant stamped onto every sea floor. It can depend on whether the target comparison is triaxial compression, direct simple shear, an average reference strength or another calibrated measure, and on the material’s rate sensitivity, anisotropy and strain softening. Research in soft-clay offshore investigations commonly discusses resistance factors around an order of ten, but the relevant programme should report its chosen value and basis rather than importing a convenient number from an unrelated site. If the factor is uncertain, a range of interpreted strengths is often more informative than one falsely exact decimal.
Worked example: a report gives qₜ = 108 kPa and states an adopted Nₜ = 12.0. Divide 108 by 12.0 to obtain 9.0 kPa. If a review compares Nₜ = 10 and 14 as hypothetical sensitivity bounds, the corresponding estimates are 10.8 kPa and 7.71 kPa. Notice that the larger factor produces the smaller strength: the relationship is inverse. Therefore saying that a “more conservative factor” is always bigger or smaller without identifying the design check is not defensible. Conservatism depends on whether a low or high operative strength creates the more adverse case.
Depth, vertical datum and geotechnical units
An offshore strength profile is a function of depth, not simply a list of numbered rows. The depth may be measured below seabed, referenced to a geodetic or tidal datum, or given as rod penetration before correcting for vessel motion and seabed settlement. A value at 4 m below seabed is not equivalent to a point at elevation −4 m in a chart unless the seabed elevation is exactly zero in that chart’s reference system. Always carry both the number and the datum wording when transferring the data into another language, table or diagram.
A robust depth table retains crossbar centre depth, local seabed position, penetration direction and whether the interval includes entry disturbance. If the T-bar crosses from a stiff upper crust into soft clay, the earliest readings can reflect transition effects before steady soil flow develops. Moving one value by half the bar diameter can change which geological layer it appears to represent. A machine log may be sampled several times per second while a client summary uses half-metre intervals; explain how the downsampling was done and whether it suppresses thin weak layers or isolated instrument spikes.
Values of force, pressure and stress should not be silently mixed. Newtons, kilonewtons and kilopascals are separate quantities. A bar force of 1.2 kN over 0.010 m² represents 120 kPa; this is not 1.2 kPa. For a very soft clay, inferred shear strengths of a few kilopascals can be physically plausible even while the corresponding penetration resistance is much higher. The apparent difference is not a contradiction. It comes from the geometry and resistance factor that connect a full-flow field response to a reference laboratory strength.
Why repeated cycling changes the reading
A cyclic T-bar experiment repeatedly pushes and extracts the crossbar over a controlled short interval. The objective is to observe how soil resistance evolves as a limited region of clay experiences repeated large deformation. Marine clay may initially retain a fabric formed by deposition, consolidation and chemical bonding. Disturbance can break down that structure and reduce resistance. When the force decreases from cycle to cycle, the curve tells a story about changing response under that protocol; it does not automatically reveal a unique intrinsic “sensitivity” without additional interpretation.
The cycle index should identify each penetration and extraction half-cycle. A first downward traverse may be called initial penetration, while the upward traverse is initial extraction; later traverses are numbered consistently. If one dataset calls the first downward pass cycle zero and another calls it cycle one, a graph can appear shifted even though the measurements match. State the numbering convention before comparing degradation curves. Also state the stroke length and reversal depths, because a cycle that repeatedly disturbs one zone is not comparable to a longer traverse entering previously untouched sediment.
Suppose the first push gives a net average resistance of 100 kPa and a late-cycle push gives 40 kPa in the same depth interval. The resistance degradation ratio is 40/100 = 0.40, or 40% of initial resistance retained. Resistance has fallen by 60%. But the ratio of intact to remoulded shear strength may differ from 100/40 = 2.5 because intact and remoulded strengths can require different resistance factors. A useful report shows measured resistance ratios first, then any separately calibrated interpretation of strength sensitivity, preventing the reader from confusing an observed machine response with a derived material index.
The difference between remoulding and sensitivity
Soil sensitivity in a geotechnical sense is commonly defined as a ratio of an intact or otherwise specified reference undrained strength to an appropriately determined remoulded strength. The numerator and denominator need matching soil, conditions and consistent definitions. A T-bar resistance degradation ratio resembles that concept but is not guaranteed equal to it. The crossbar changes the soil and then measures a resistance influenced by the already disturbed flow field. The same remoulding path, measurement strain rate and factor calibration are essential when comparing interpretations.
For a teaching calculation, suppose initial qₜ is 120 kPa with N_intact = 12 and late-cycle qₜ,rem is 36 kPa with N_rem = 18. The interpreted intact strength is 120/12 = 10 kPa; remoulded strength is 36/18 = 2 kPa. The corresponding strength sensitivity is 10/2 = 5, while the raw resistance ratio is 120/36 = 3.33. These are deliberately hypothetical calibration factors, not a recommendation for a field project. The demonstration proves why replacing strength sensitivity with the raw resistance ratio can materially understate or overstate the inferred parameter.
Before reporting a remoulded value, ask whether the late cycles had approached a stable resistance, whether the penetration direction and extraction direction gave similar magnitudes, whether the surrounding soil was actually fully remoulded and whether pore-pressure effects might still be evolving. A statement such as “the strength reaches 2 kPa after ten cycles” overclaims if the tenth cycle merely marks the end of the selected laboratory protocol rather than a physical end state. Prefer “the strength interpreted from the tenth cycle is approximately 2 kPa under the specified calibration” when that is what the evidence shows.
Penetration rate, drainage and the competition between effects
Rate can influence a T-bar result through at least two competing mechanisms. At high shearing rates, viscous or rate-dependent components of soil strength can raise resistance. At slower rates, partial drainage and pore-pressure redistribution can change the effective stresses and therefore the resistance. Their effects are not interchangeable and may even offset one another over part of a response curve. One speed should not be described as inherently more accurate without considering the sediment permeability, crossbar dimension, consolidation timescale and purpose of the test.
A report that doubles the penetration speed should record whether the observed resistance increased, decreased or remained within scatter. It should not automatically label the difference “higher strength” because the test condition changed. A normalised velocity may combine actual speed, a characteristic dimension and a consolidation coefficient, but the definition used must be given before cross-project comparison. Rate sensitivity inferred from a sequence of tests also needs a stable state of remoulding; otherwise rate effects are mixed with continued structural damage from the preceding passes.
For example, if two first-pass tests in apparently comparable clay give 110 kPa at one rate and 121 kPa at another, the second is 10% higher in measured resistance. That observation alone does not prove that the intact undrained strength is 10% higher. Check specimen comparability, depth, stress history, equipment and corrections. If a source says the rate effect was isolated by tests at a common remoulded state, preserve that qualification. If it was not isolated, frame the pattern as a combined test response and keep the possibility of drainage changes open.
Reading a complete example profile
Consider four fictional depths below seabed: 1, 3, 5 and 7 metres. Suppose their corrected first-push resistances are respectively 60, 84, 108 and 132 kPa. Using the same explicitly chosen factor 12 for this classroom exercise produces interpreted strengths of 5, 7, 9 and 11 kPa. The profile suggests an increase of 2 kPa per 2 m interval, equivalent to an average gradient of 1 kPa per metre over the illustrated range. Do not call that a proven geologic law: it is a pattern in an invented four-point dataset that should be tested on genuine high-resolution observations.
At the 5 m test level, imagine the net resistance falls across repeat passes from 108 to 78 to 59 to 50 kPa and then settles around 48 kPa. The consecutive changes are not uniform: the first drop is 30 kPa, the second 19 kPa, then 9 kPa, then about 2 kPa. If the report plots only the first and last numbers, the reader loses the shape of degradation and the evidence that resistance may be approaching a plateau. An explanation should preserve the complete curve or at least enough intermediate points to show whether the end-state claim is supported.
Suppose a field summary rounds the 5 m initial resistance to 0.11 MPa while a spreadsheet uses 108 kPa. The difference is only 2 kPa due to rounding, but conversions must be consistent. One megapascal is one thousand kilopascals. Thus 108 kPa is 0.108 MPa. A table that mistakenly equates 0.11 MPa to 11 kPa introduces an order-of-magnitude error. The answer is to label units visibly at the column header and at any transformation, not to insist that every cell display more decimal digits than the measurement supports.
The uncertainty budget and honest rounding
Three broad sources of uncertainty deserve separate treatment: measurement uncertainty in the force and depth systems, correction uncertainty in the calculated net resistance, and model uncertainty in the adopted resistance factor. A precise load cell does not cancel uncertainty in the interpretation factor. Likewise, a strongly justified factor cannot restore data lost through uncertain depth reference or signal drift. Readers should learn to see each uncertainty at the step where it enters, rather than being handed a single percentage whose origin is invisible.
In a simplified independent-error exercise, take q = 96 ± 4 kPa and N = 12 ± 2, where the ± values are illustrative standard uncertainties and correlation is ignored. For s = q/N, a first-order relative combination gives sqrt[(4/96)^2 + (2/12)^2] ≈ 0.172. The strength estimate is 8.0 kPa with an approximate uncertainty of 1.38 kPa under those assumptions. The model term dominates. This is not a universal uncertainty prescription; its meaning depends on how the factor bounds were obtained, whether distributions are appropriate and whether the inputs are correlated.
A simpler scenario analysis avoids presenting uncertain probability assumptions as established. If measured net resistance might be between 92 and 100 kPa and an acceptable factor range is 10 to 14, the smallest ratio is 92/14 = 6.57 kPa and the largest is 100/10 = 10.0 kPa. These are extreme combinations, not a statistical confidence interval. They communicate that an interpretation centered at 8 kPa has meaningful model spread. In a design review, describe the conditional nature of those bounds and seek the project-specific parameter selection procedure.
What a contractor, scientist or translator should never infer
A T-bar result is not an allowable bearing pressure, an anchor capacity, a pipeline embedment prediction or a guarantee of stability. Each of those decisions needs a separate physical model that considers geometry, loading direction, duration, drainage, cyclic history, soil variability and appropriate factors of safety or reliability. Penetrometer results can inform the soil strength inputs to those models, but the model connecting strength to structural performance must be shown. A sentence that converts an interpreted strength directly into a safe installation load skips the critical engineering step.
The test also does not resolve all geological uncertainty at a point. A thin sand seam, gas-charged pocket, cemented lens or highly variable fabric may violate the assumptions used for a smooth soft-clay interpretation. Even a perfect resistance profile measures the response along the penetrated trajectory, not every metre of a proposed foundation footprint. Where data are sparse, the distinction between observed local behaviour and inferred site-wide conditions must survive translation. If an engineer has treated a zone as uncertain, the public-facing explanation should not erase that uncertainty in the interests of tidy prose.
Readers should distinguish “intact” from “undisturbed” as well. A sample taken offshore can experience stress relief, vibration and transportation damage even if it is described as relatively intact. A T-bar first pass may encounter soil disturbed by prior tools or by drilling, especially where the spacing is small. Neither adjective means pristine, guaranteed natural state. The correct statement identifies how a strength reference was obtained and whether disturbance has been evaluated. Likewise, “remoulded” describes the state created by a specified procedure; it is not a universal zero-structure condition independent of method.
When the penetrometer disagrees with other tests
A sensible investigation compares the T-bar with CPTu, vane, triaxial and direct simple shear results without forcing all numbers to match exactly. The tests load soil in different directions, at different strain rates and along different deformation paths. The desired reference strength for an offshore design can therefore differ from the strength that best reproduces a laboratory compression result. Begin by plotting comparable depths and separating each test’s observed resistance from its interpreted strength. Then identify which stress paths and calibration factors explain the residual differences.
Suppose a T-bar interpretation yields 8 kPa and a nearby DSS test gives 7 kPa. Calling one “wrong” would be premature. Did the tests sample the same stratum? Was the DSS specimen disturbed? Was the T-bar factor calibrated to DSS or to compression strength? Did one test include a different consolidation stress? A reasoned interpretation might conclude that both are consistent within uncertainty, or that the difference reflects anisotropy. Only after those questions does it make sense to decide which parameter to adopt for a particular limit state.
An offshore project can use paired tests as a calibration network. For each depth interval, show test method, reference strength, measured net resistance, interpreted factor, uncertainty and sample condition. Record the data excluded from calibration and why: sensor saturation, shallow cavity response, missing depth control or a demonstrably different soil. Otherwise a reported mean factor may look objective but conceal an untraceable choice of points. The purpose of translation is to let a reader reconstruct the reasoning, not simply repeat a final average.
Worked example: an apparent contradiction solved
A fictional field log gives a first push of 132 kPa at 7 m below seabed. A review table lists sᵤ = 11 kPa, but a later project note lists 12 kPa. First divide 132 by 11 to infer that the review used N = 12. Next divide 132 by 12 to infer that the later note used N = 11. The two entries can therefore be arithmetically consistent while reflecting different calibration choices. The detective question is whether the documents state the factor revision, why it was made and which design stage each interpretation belongs to.
Now suppose the note instead claims N = 12 but sᵤ = 12 kPa. Multiplying 12 by 12 returns 144 kPa, not the stated 132 kPa. This is an internal numerical inconsistency rather than merely a change of judgement. The correct repair is not to choose the convenient number. Recheck the measured channel, the selected depth interval, any new force correction, the sign of extraction, and whether the note uses a different averaged value. Raise the discrepancy explicitly if the source information is insufficient.
Finally, imagine a graph whose strength axis is labelled kPa but whose plotted values look like 0.006 to 0.014. They may be in MPa, or the plotting code may have scaled the data incorrectly. Test one known point: 11 kPa should plot as 11 on a kPa axis and as 0.011 on an MPa axis. The axis title, trace and exported spreadsheet must agree. Do not silently rewrite the graph label from intuition; check the original data export and obtain a corrected plot or a note documenting the discrepancy.
Independent transfer problems with checked answers
Problem one: a device has projected area 0.0120 m² and corrected net axial force 1.08 kN. Find resistance, then interpreted strength for N = 12. Answer: 1.08 kN = 1,080 N; q = 1,080/0.0120 = 90,000 Pa = 90 kPa; sᵤ = 90/12 = 7.5 kPa. The area must not be entered as 0.0120 mm². The result is a conditional estimate, not an allowable load. A reader can reverse-check by multiplying 7.5 kPa by 12 to recover the measured 90 kPa resistance.
Problem two: first-push net resistance is 125 kPa and late-cycle resistance is 45 kPa. Find percentage resistance retained and percentage drop. Answer: retained = 45/125 = 0.36 = 36%; drop = (125 − 45)/125 = 0.64 = 64%. You cannot infer strength sensitivity of 125/45 from this calculation unless the interpretation deliberately uses equal factors for initial and remoulded states and that assumption is justified. Write “resistance degradation ratio” for the direct measurement and reserve “sensitivity” for the defined strength ratio.
Problem three: if a result of 84 kPa is interpreted using factors 10.5 and 12.0, calculate both candidate strengths. Answer: 84/10.5 = 8.0 kPa; 84/12.0 = 7.0 kPa. The interpretation is lower under the larger factor. If the report uses 7 kPa, it should show N = 12 or identify an equivalent calibrated approach. The numbers alone cannot tell whether either calibration matches the actual clay and design stress path.
Problem four: an instrument reads −510 N on upward extraction. Its correction convention yields a net magnitude of 470 N and projected area is 0.010 m². The physical resistance magnitude is 47 kPa, not negative 47 kPa as a material property. The extraction direction and the negative recorded sign still need to be shown. This problem tests the distinction between a signed instrument channel and the unsigned resistance magnitude used in a degradation plot.
Practical review and error repair
Before accepting an interpreted profile, reconstruct one row from the original instrument channel to the final shear-strength estimate. Confirm crossbar dimensions, area, load units, force zero, correction ledger, sign convention, depth reference and the actual factor. If the calculation cannot be reproduced, identify which missing variable prevents it. This one-row audit often catches scaled columns, wrong resistance factors and mismatched datasets more quickly than reviewing a long narrative in isolation. Once the row passes, repeat the check at a shallow point, a deep point and a cyclic level where the sign reverses.
Next inspect the shape of the full profile. Look for abrupt jumps at known lithological boundaries, a progressive depth trend that is plausible, and anomalies coincident with rod changes or sensor limits. Compare first-pass readings to repeat passes at the same depth rather than comparing unrelated sections. A curve described as remoulding may actually include varying soil layers if the stroke positions are not aligned. For averages, state the exact depth window and whether the statistic is arithmetic mean, median or another definition; different choices can move the reported parameter.
A high-quality translation should make uncertainty visible without burying the reader in caveats. State what was measured, what was calculated and what was inferred. Give the chosen denominator and factor. Separate findings supported by the particular test from possible consequences for design. Use consistent symbols and define the first use of every abbreviation. Preserve whether the published source speaks about a comparison, a correlation, a recommended interpretation or a proven site-specific value. That discipline improves both specialist communication and a non-specialist reader’s ability to ask better questions.
Frequently asked questions
Is a T-bar the same as a cone test? No. Their probe geometries, penetration mechanisms and resistance interpretations differ; a project may use both. Can a T-bar measure remoulded strength? Repeated penetration and extraction can provide data to interpret remoulded behaviour, but the inference depends on cycling protocol and calibration. Is a decrease in resistance necessarily a loss of the same percentage of shear strength? No. The instrument’s ratio and the strength ratio may involve different factors and different physical histories.
Why is the T-bar useful offshore? It can characterise soft sediments in situ, generate continuous resistance profiles and allow controlled cyclic disturbance without relying only on recovered samples. What does the first push represent? It is the first measured resistance under the specified conditions; calling it perfectly intact requires care. Does a large net resistance mean the sediment is safe for an anchor? Not by itself. Anchor performance depends on design geometry, load history, soil response and calculation method.
What should a translation preserve when space is limited? At minimum, keep the test type, depth datum, net versus raw resistance, units, resistance factor, inferred strength, cycling state and stated uncertainty. If the summary must be shorter, shorten context before deleting those distinctions. Where the original paper uses an unfamiliar symbol, add a parenthetical explanation rather than silently replacing it with a familiar but different parameter. Technical clarity is not the same as replacing technical meaning with confident generalities.
Sources and connected reading
For methodological background, see the peer-reviewed discussion of full-flow penetrometers and cyclic softening in Refined Prediction of Strain-Softening Parameters of Marine Soft Clay Using Cyclic T-Bar Penetration Tests, and the International Journal of Geomechanics analysis of cyclic soil-strength degradation. For the wider offshore field context, consult the International Society for Soil Mechanics and Geotechnical Engineering. These links are methodological references; none supplies the fictional figures in the examples above.
Related eduKate explanations include direct simple shear and anisotropy, offshore CPTu interpretation and ring shear and residual strength. These are different tests, not substitutes for T-bar calibration. Read them side by side to learn how the same soil can produce different observations under different mechanical conditions.