Ground-penetrating radar for ground and slab assessment.

GPR is an indirect geophysical method. A strong or repeatable response may justify further investigation, but signal amplitude alone must not be converted into a confirmed void, defect severity, affected volume, safe load capacity, repair quantity or repair design.

1. What GPR measures

A GPR system transmits electromagnetic energy and records returned energy as a function of travel time. Responses occur where electromagnetic properties change, at discrete objects, interfaces, reinforcement, edges, joints or other contrasts. The displayed image is a processed representation of recorded waveforms—not a literal cross-section.

Material conductivity, dielectric behaviour, moisture, reinforcement, surface coupling, antenna characteristics, target geometry and processing all affect what can be detected and how it appears. Similar radar patterns may have different physical causes; different conditions may also produce weak or ambiguous responses.

2. Begin with the engineering decision

The survey objective should state the decision that the evidence must inform. “Scan the floor” is not sufficient. Better questions include:

• Is there a repeatable response change beneath the rocking slab zone compared with reference ground?
• Where should targeted physical verification be considered?
• Can the survey identify reinforcement or interfaces that constrain drilling?
• Does the suspected disturbance repeat on adjacent lines and intersecting cross-lines?
• Is monitoring or another investigation method needed before a treatment scope is considered?

The decision determines required coverage, line spacing, direction, positioning accuracy, antenna selection, processing and verification. It also determines which uncertainties are acceptable.

3. Survey geometry controls the conclusion

Every line needs a stated origin, direction and coordinate convention tied to visible or surveyed references. Record offsets, line spacing, grid dimensions, surface type, obstructions, shortened lines and inaccessible areas. Distinguish:

• surveyed area—where usable data were acquired;
• area of interest—where a response or reported symptom warrants attention;
• investigation buffer—an advisory zone for follow-up; and
• confirmed affected area—which requires evidence beyond an interpreted radar response.

Parallel lines may show repeatability along one direction, but a closed plan boundary should not be inferred without adequate intersecting coverage. A cross-line that does not physically reach the strongest response cannot corroborate its lateral extent.

4. Frequency and resolution involve trade-offs

Higher-frequency data can provide finer detail at shallower ranges under suitable conditions; lower-frequency data may retain broader or later responses with less detail. Actual penetration and resolution depend on the complete site and equipment configuration, not frequency alone.

Where complementary bands are available, compare them as different views of the same ground—not as automatic confirmation. Acquisition settings must be separated from later display choices such as gain, contrast and displayed time window.

5. Source-data integrity comes before interpretation

Before reading patterns, verify that the dataset is complete and internally consistent. For each file or line, record the identifier, band or antenna, trace count, sample interval, recorded time window, trace spacing, direction and length. Check for truncated, duplicated, reversed, missing or non-intersecting lines.

When SEG-Y or another structured export is available, confirm that trace records decode completely, samples are finite and metadata agree with field records. Screenshots can support communication, but they do not replace traceable source data.

6. Use a controlled processing sequence

Start with minimal processing so that the original behaviour remains visible. A defensible sequence commonly includes time-zero review, dewow or moving-mean removal, and conservative gain. Then create a processing-control variant—for example, background subtraction—while keeping axes and coordinate ranges consistent.

A feature that persists across reasonable variants supports confidence that a waveform change was recorded. Persistence does not prove that the feature is an open cavity or any other single material condition. A feature visible only after strong subtraction, filtering or contrast adjustment is processing-sensitive and should receive lower confidence.

Document whether migration, calibrated time-zero correction, velocity modelling, topographic correction and depth conversion were applied. Aggressive processing should not be used merely to make a presumed target look cleaner.

7. Control the visual comparison

Like-for-like panels should use consistent axes, coordinate direction, anomaly bands and, where practical, display scaling. If each panel is contrast-scaled independently, say so. Colour intensity should never be described as defect severity, density or certainty without an independently established relationship.

Mark reported surface symptoms separately from interpreted radar responses. Caption an anomaly band as an area of interest or investigation reference—not as a cavity outline.

8. Interpret through an evidence hierarchy

1. Data integrity: are the files, geometry and coordinates reliable?
2. Waveform repeatability: does the change persist on the line and across reasonable processing?
3. Spatial corroboration: is it repeated on adjacent and genuinely intersecting lines?
4. Context: how does it differ from reference ground on both sides?
5. Competing explanations: which physical causes remain plausible?
6. Specific-cause confidence: what evidence supports or limits identification?

Do not reverse this order by assuming a void first and processing toward it.

9. Keep competing explanations visible

Depending on the site, a response may be influenced by disturbed or variable backfill, moisture variation, drainage infrastructure, pavement or slab/base changes, utilities, chambers, reinforcement, joints, surface coupling, ringing, multiples or an unresolved void. Drainage leakage, washout or pipe failure should remain hypotheses unless independently verified.

A weak or visually featureless lower radargram does not demonstrate intact ground. Failure to identify a clear target does not prove the absence of a defect.

10. Treat depth as a calibrated output

Radar records two-way travel time. Converting that time to depth requires an appropriate velocity assumption or calibration. Moisture and material variability can change velocity across a site. If no known-depth target, velocity analysis or physical correlation is available, preserve time on the vertical axis or clearly label depth as approximate and assumption-dependent.

Do not infer a void roof, base, thickness or volume from a single reflector or colour band without adequate geometry, calibration and corroboration.

11. Grade confidence separately

A report should avoid one overall confidence label when different parts of the conclusion have different support. Grade at least:

• confidence that a repeatable radar-response change exists;
• confidence in its recorded coordinate;
• confidence in lateral continuity or closure;
• confidence in displayed depth; and
• confidence in a specific cause.

For example: confidence may be moderate that a local variation exists, moderate in its along-line coordinate, low in a closed plan boundary because cross-lines do not intersect it, low in depth without calibration, and low that it specifically represents an open void.

12. Applications and limits

13. Physical verification is decision-dependent

Verification may include records correlation, opening or inspection of accessible chambers, survey control, probing, coring, drilling, borescope inspection, excavation or another method selected for the target and risk. Verification locations should be chosen from the evidence and completed under the client’s engineering, permit and safety controls.

Physical verification can confirm a local condition at the tested point. It does not automatically validate every interpolated area, so the relationship between sample points and survey coverage must remain explicit.

14. What a defensible deliverable contains

• Objective, decision context and scope boundaries.
• Survey plan with origins, directions, line spacing and inaccessible areas.
• Equipment and acquisition settings separated from display settings.
• Data inventory and integrity findings.
• Processing sequence and exclusions.
• Representative minimally processed and control-variant profiles.
• Results tied to exact dataset, line and recorded coordinate.
• Integrated interpretation with competing explanations.
• Separate confidence grades for existence, location, extent, depth and cause.
• Suggested verification options and unresolved questions.
• Limitations and a clear statement that GPR does not determine repair design.

15. GPR technical resource library

16. Questions to ask before commissioning GPR

1. What decision must the survey support?
2. What target is expected, at what approximate scale and range?
3. What surface, access, reinforcement and operating constraints apply?
4. How will coordinates be controlled and tied to site references?
5. Will adjacent and intersecting lines provide adequate coverage?
6. What source data and processing record will be retained?
7. How will depth be calibrated or qualified?
8. Which competing explanations must be tested?
9. What physical verification is feasible if the decision requires it?
10. Who will review the findings in the wider structural, geotechnical or asset context?

Engineering conclusion

GPR is most useful when it narrows uncertainty rather than pretending to remove it. Its engineering value comes from disciplined geometry, complete source data, controlled processing, spatial corroboration, explicit alternatives and proportionate verification. The result should be a traceable basis for the next decision—not an unsupported picture of a presumed subsurface defect.

Đọc thư viện kỹ thuật GPR bằng tiếng Việt →

Place GPR within a wider investigation plan →