Eddy current rail inspection is a surface method that detects and sizes surface-breaking cracks in the rail head, mainly head checks at the gauge corner, to a depth of roughly 2 to 3 mm. It gives no valid search under 49 CFR 213.237. It gives a depth map that says how much to grind before the ultrasonic re-test.
This page covers eddy current on rail only: how the probe sees a crack, what a system reports, how accurate its depths are, where the probes ride, how a system is accepted, and how a reading turns into a grinding pass and a 49 CFR 213.237 re-test. The defects themselves are covered in rolling contact fatigue in rail and head checks in rail, the grinding side in rail grinding and ultrasonic testing, and the library hub is detecting rail defects.
Contents: What it is · How the probe sees a crack · What it measures · Depth accuracy · Where the probes ride · System acceptance · Grinding and 213.237(h) · Worked example · With UT · Limits · FAQ
What is eddy current rail inspection?
Eddy current rail inspection runs a small alternating-current coil along the rail head and reads the change in the coil's impedance when a surface-breaking crack disturbs the currents induced in the steel. In rail it has one job: finding and sizing rolling contact fatigue cracks, above all head checks on the gauge corner of the high rail.
The European standard for the method on track defines the scope in one sentence: "This document is applicable to testing of rails installed in track for detecting rail surface cracks." It covers "inspection-trains or reprofiling machines and manual systems" and is "not concerned with production testing of rails in a production plant" (EN 16729-2:2020 scope, published August 17, 2020, confirmed September 3, 2025). Its only annex is normative and is titled "System acceptance of eddy current systems for detecting head checks".
In the United States there is no rule for it. Part 213 names no inspection method; 213.237 asks for "internal rail inspections", and eddy current does not inspect the inside of the rail. It is a maintenance-planning measurement beside the regulatory search, which Part 213 alone requires.
How does an eddy current probe see a crack in the rail head?
The coil induces currents in a thin skin of the rail surface, and a crack that cuts across those currents forces them to detour, which changes the coil's impedance. In the Federal Highway Administration's NDE manual, the coil's alternating field induces eddy currents in the steel, those currents make a secondary field that opposes the first, and "the presence of a discontinuity, defect, or irregularity in the grain structure of the material disrupts the flow of the current in the material, and thus, changes in the impedance of the coil are produced" (FHWA, Eddy Current Testing).
The currents live near the surface. Their density falls exponentially with depth, and the standard depth of penetration, where it has fallen to about 37% of the surface value, is δ = 503 √(ρ / (μr f)) meters, with ρ the resistivity in ohm-meters, μr the relative permeability and f the frequency in hertz (Evident, Depth of Penetration in Eddy Current Testing). Rail steel is ferromagnetic, so μr is large and the skin is "typically less than 1–3 mm (0.04–0.11 inches) depending on the frequency used", which is why "ECT in steel can detect only surface or near-surface cracks" (FHWA).
| Frequency | Standard depth δ at μr = 100 | Standard depth δ at μr = 500 |
|---|---|---|
| 1 kHz | 0.71 mm (0.028 in.) | 0.32 mm (0.013 in.) |
| 10 kHz | 0.22 mm (0.009 in.) | 0.10 mm (0.004 in.) |
| 100 kHz | 0.07 mm (0.003 in.) | 0.03 mm (0.001 in.) |
Computed from the formula above with ρ = 2.0 × 10⁻⁷ Ω·m. The permeability of a work-hardened rail surface is not a catalog value, so this is a range, not a rail-steel fact.
If the currents run in a fraction of a millimeter, how does a rail system report depths of 2 to 3 mm? Because it does not measure depth. It measures how much current the crack diverts, which grows with the length of the crack face: "Eddy current systems provide a return signal that is proportional to the length of the surface breaking crack" (DOT/FRA/ORD-11/24, Magel, 2011). Depth follows from an assumed angle between the crack and the running surface.
Key takeaway: the probe senses the crack face, not the crack tip; depth is a calculation from signal amplitude and an assumed crack angle.
What does a rail-head eddy current system measure?
A rail eddy current system reports where cracks are along the track and across the head, how dense they are, and an estimated depth per channel; it does not report the crack angle, and it does not by itself tell a crack from a spall. TTCI's survey of what a rolling contact fatigue measurement needs scored eddy current against six requirements (TTCI TD-14-008, Baillargeon and Tournay, May 2014):
| Requirement (TD-14-008) | Eddy current delivers it? | How a rail system does it |
|---|---|---|
| Location along the track | Yes | Distance encoder, GPS and track markers tied to each reading |
| Lateral location on the head relative to the gauge corner | Yes | Several probes side by side, each covering a strip of the head, so each channel is a lateral position |
| Distinguish spalls, shells and cracks | No | A spall edge and a crack both disturb the currents; the operator or a vision system separates them |
| Crack density | Yes | Counts per unit length per channel |
| Crack depth | Yes, estimated | Amplitude converted to depth with an assumed crack angle |
| Crack angle | No | Entered by the user; TD-14-007 says the angle "cannot be determined through nondestructive testing" |
The lateral coverage makes the output a map rather than a trace. On the Eurailscout test trains used in Germany and the Netherlands, four sensors per rail cover "an area of approximately 0.2 to 25 mm on the running edge of the rail", the probes follow the wear profile independently of the ultrasonic sleds, and depth comes "from the recorded amplitude of the eddy current signals. By taking into consideration the initiating crack angles, a crack depth of approximately 2–3 mm can be measured" (INNOTRACK D4.4.1, Rail Inspection Technologies, University of Birmingham, 2008).
The report is a per-channel, per-block table: for every 100 ft or 100 m of rail, the count and the maximum estimated depth in each lateral strip, often colored into depth classes. The Polish Railway Institute's four-probe hand device displays 0 to 0.5 mm, 0.5 to 1.5 mm, 1.5 to 2.7 mm, 2.7 to 5 mm and over 5 mm, with the caution that "measurements of defects with a depth of more than 3 mm are subject to a large error" (Stencel, Analysis of the occurrence of headchecks in rails, Transportation Overview 6/2019); the bands are one institute's settings, not a standard.
How accurate is eddy current crack depth on rail?
In the only published North American validation, readings at the default 25° crack angle were 5% to 33% off the depths measured on cut sections, and the error scaled at up to 0.003 in. (0.08 mm) per degree of angle error. TTCI and Norfolk Southern ran a four-probe hand trolley on the FAST High Tonnage Loop and on Norfolk Southern coal lines in Virginia, then sectioned rail to check it (TTCI TD-14-007, Baillargeon, Tournay and Kerchof, May 2014).
| Finding (TD-14-007, 2014) | Value |
|---|---|
| Probes | Four, in ceramic casings that follow the profile and keep the coils off the rail, covering the gauge corner and the top |
| Crack angles measured on sections from a heavy-haul high rail | 18° to 28° |
| Default angle assumed by the device | 25° |
| Depth error at 25°, across the Glen Lyn samples | 5% to 33% |
| Example | A sectioned crack 0.071 in. deep read as about 0.083 in., a 16.7% difference |
| Error per degree of angle error | Less than 0.003 in. (0.08 mm) per degree |
| Residual error after correcting with the measured angle | Up to 0.0071 in. (0.18 mm) |
| Grinding check at Wabun, Virginia (100 ft of low rail, 6° curve) | 0.14 in. (3.6 mm) before grinding to less than 0.04 in. (1.0 mm) after five passes |
| Effect of grinding-stone marks on the readings | "did not appear to interfere" |
Two things in that table matter for planning. The error is systematic: a crack steeper than the assumed angle is deeper than its reading, and a shallower one is less deep. And correcting the angle fixes most of it but not all; TTCI still saw up to 0.18 mm left over. ORD-11/24 reached the same conclusion, that "it is difficult with eddy current to know the actual depth of damage", and recommended that "Validations for North American steels and correlations of crack length to depth are required."
Deutsche Bahn's 2008 to 2012 tests are reported, secondhand in the trade press, as "within 10% of the actual crack depth" (Interface Journal, McWilliams and Eberius, May 2014).
The practical ceiling is about 3 mm, which is why the Railway Institute flags readings over 3 mm and the Eurailscout trains quote "approximately 2–3 mm". A head check that has turned down into a detail fracture is a job for the ultrasonic 70° channels; see detail fracture development.
Key takeaway: plan with the reading plus an angle margin, treat anything over about 3 mm as "deep, size it another way", and ask what angle the device assumed.
Where do the probes ride: hand trolleys, grinders and test vehicles?
The same four-probe arrangement is carried three ways: pushed by hand at walking pace, mounted on a rail grinder to measure before and after each pass, and carried on a test vehicle beside the ultrasonic sleds. Standoff is the design problem in all three:
"From a practical consideration, the eddy current system is very sensitive to changes in the distance between the coils and the target, so a consistent standoff is required. Typically, that distance is 1–2 mm, which is difficult to achieve at high speed on rough rail and with continually varying rail head shapes." — DOT/FRA/ORD-11/24, Rolling Contact Fatigue: A Comprehensive Review, section 3.1.1 (2011)
| Carrier | Example in the public record | Probes and coverage | Speed | Standoff |
|---|---|---|---|---|
| Hand trolley | Four-probe trolley evaluated by TTCI and Norfolk Southern (TD-14-007, 2014); four-probe Railway Institute device (Stencel, 2019) | Four channels across the gauge corner and top | Walking pace; INNOTRACK describes manual systems at "a few metres per minute" | Ceramic casings hold the coils off the rail |
| Grinder-mounted | Speno HC Grinding Scanner (ORD-11/24, citing Pohl, Krull and Meierhofer, 2006) | Four eddy current probes per rail, 25 mm measurement width, focused on the gauge corner | Grinding speed; measures before and after passes | 1 mm on a UIC60 head; the sensors work from −1 to +2 mm as the head wears |
| Test vehicle, combined with UT | Eurailscout UST-02 trains (INNOTRACK, 2008); the Linke & Rühe combined unit (ORD-11/24) | Four sensors per rail over 0.2 to 25 mm of the running edge; the combined unit pairs four eddy current sensors on the gauge-side 25 mm with 10 ultrasonic probes at +70° to −70° | Trains at 72 km/h (45 mph); the combined unit "up to 60 km/h (35 mph)", with marketing literature quoting surface-fatigue measurement at up to 30 km/h (18 mph) | Probes follow the wear profile independently of the UT heads |
| Vendor-rated rail systems (not independently verified) | A current manufacturer's product page | "at least 2 and up to 8 sensors" per rail | "up to 20 km/h" (12 mph) for the on-track unit; higher for the train version | Not stated |
On a combined vehicle the two methods read each other: BAM reported in 2007 that an inspection train carrying both could re-label an ultrasonic "group 2" indication when the eddy current channels showed head checking at the same spot (Thomas, Heckel and Hanspach, Insight 49(6):341–344, 2007). In North America the method is still a trial and planning tool.
Key takeaway: the hand trolley sizes a known problem, the grinder-mounted scanner proves the grind, and the test-vehicle array surveys the network.
How is a rail eddy current system accepted and checked?
A rail eddy current system is proven on cracks of known depth and angle, then checked in service against sectioned rail and against the lift-off it will meet on worn heads. There is no FRA rule for this; the references are the European acceptance annex and the published validations.
- Reference cracks of known geometry. EN 16729-2's Annex A is normative but paywalled. The NURail study cut cracks 4, 8 and 12 mm deep at 90°, 75°, 45° and 15° into a steel bar and showed that both depth and angle move the signal, so a block with one depth cannot calibrate a device that will meet several angles.
- Validation against sections. TTCI read the rail, cut it, measured the deepest crack under a microscope and compared.
- Lift-off and surface condition. The FHWA warns that "Irregular surfaces or inconsistent movement of a probe (e.g., lifting the probe off the surface or tilting the probe) can produce signals that may be interpreted as defects". INNOTRACK requires eddy current sensors "at a close (<2mm) and constant distance" and notes the signal falls with the cube of lift-off. A system accepted on a new profile should be re-checked on a worn high rail.
- Operator settings. The assumed crack angle, the depth classes and the channel positions are user settings, and a report that does not state them cannot be compared with the next one.
None of this is a 213.238 qualification, which covers the internal search; an eddy current operator's competence is set by the owner's program, or by EN 16729-4 where that series is used.
Key takeaway: accept the system on cracks of known depth and angle, validate it on cut rail, and record the angle and class settings on every report.
How does an eddy current reading drive a grinding decision and a UT re-test?
The reading sets how much metal the grinder must remove to clear the cracks; the ultrasonic re-test afterward is what Part 213 requires, and 213.237(h) sets the clock. The chain runs through five rule paragraphs, none of which mentions eddy current or grinding.
| Step | Rule | What it says |
|---|---|---|
| 1. The test car reports a no-test | 213.237(g) | If the qualified operator "determines that a valid search for internal defects could not be made over a particular length of track, that particular length of track may not be considered as internally inspected" |
| 2. Why it was a no-test | 213.238(b)(4) | The operator must determine that each search "has not been compromised due to environmental contamination, rail conditions, or equipment malfunction". Heavy head checks are a rail condition |
| 3. It goes on the record | 213.241(c)(6) | Records must specify the "location of any track not tested pursuant to § 213.237(g)", kept two years |
| 4. The owner's three options | 213.237(h) | Before the 370-day or 30 MGT limit runs out: "(1) Conduct a valid search for internal defects; (2) Reduce operating speed to a maximum of 25 m.p.h. until such time as a valid search can be made; or (3) Replace the rail that had not been inspected" |
| 5. What counts | 213.237(j)(4) | "Valid search means a continuous inspection for internal rail defects where the equipment performs as intended and equipment responses are interpreted by a qualified operator" |
Eddy current enters between steps 3 and 4. Option (1) usually means grinding the surface fatigue off and testing again, and the depth map is the only measurement that says how much to grind and where. TTCI's corrective-grinding study gives the removal figures: a production grinder takes off 0.006 to 0.008 in. (0.15 to 0.20 mm) per pass at typical speed, up to 0.014 in. at slow speed; corrective grinding restored a valid test at 68 of 88 rail exceptions (77%); and the success rate was 77% where at least 0.005 in. was removed, 88% at 0.020 in. and 100% at 0.060 in. (1.5 mm) (TTCI TD-20-002, 2020). Efforts ran from one pass to 17.
In five rails pulled for low ultrasonic confidence, microscopy found 880 cracks and spalls, 90% of them no more than 0.040 in. (1.0 mm) deep, and TTCI judged that "three to five passes with a production grinder would likely have improved UT confidence" (TTCI TD-18-027, Cummings, November 2018).
Grinding itself earns nothing under Part 213; the credit comes only when a qualified operator completes a valid search over the ground rail. The interval arithmetic is in 49 CFR 213.237 rail flaw detection inspection and how often must rail be tested.
Key takeaway: eddy current tells you how much to grind; only the ultrasonic valid search closes the 213.237(h) item.
Worked example: from a depth map to a grinding pass
This example turns a six-block eddy current depth map on one high rail into a grinding plan, a block that needs hand UT first, and a post-grind check. The track and readings are illustrative; the error figures and removal rates are TTCI's.
Situation. The high rail of a 5° curve on Class 4 track, 600 ft long, carrying 30 MGT a year. The test car recorded a no-test over the curve, logged under 213.237(g). A four-channel hand trolley then measured the rail in 100 ft blocks, channel 1 on the gauge corner and channel 4 at the crown side of a 25 mm strip, at an assumed 25° crack angle. Readings are the maximum estimated depth per block, in millimeters.
| Block | Ch 1 (gauge corner) | Ch 2 | Ch 3 | Ch 4 (toward crown) | Governing reading |
|---|---|---|---|---|---|
| A, 0–100 ft | 0.6 | 0.4 | 0.2 | 0.1 | 0.6 mm |
| B, 100–200 ft | 1.4 | 1.1 | 0.5 | 0.2 | 1.4 mm |
| C, 200–300 ft | 2.2 | 1.6 | 0.7 | 0.3 | 2.2 mm |
| D, 300–400 ft | 2.9 | 2.0 | 0.9 | 0.3 | 2.9 mm |
| E, 400–500 ft | 1.2 | 0.8 | 0.4 | 0.1 | 1.2 mm |
| F, 500–600 ft | 0.5 | 0.3 | 0.1 | 0.0 | 0.5 mm |
Step 1: set the angle margin. TTCI's sections ran from 18° to 28° against the 25° default. A steeper crack is deeper than its reading, so the planning risk is on the 28° side: 3° × 0.003 in. per degree = 0.009 in. (0.23 mm). Add TTCI's residual of up to 0.0071 in. (0.18 mm). The removal target for each block is the reading plus 0.23 + 0.18 = 0.41 mm (0.016 in.).
Step 2: sort the blocks. Channel 1 governs everywhere, so the metal comes off the gauge corner; channels 3 and 4 are in the light band and need only the profile passes. Block D reads 2.9 mm, above the 2.7 mm "heavy" band and close to the 3 mm reliability limit. It gets hand ultrasonic testing with the 0° and 70° probes from both sides before any grinding, so that a detail fracture is handled under 213.113 rather than ground over.
Step 3: convert the targets to passes at TTCI's 0.006 to 0.008 in. per pass:
| Block | Reading | Target (reading + 0.41 mm) | Target in inches | Passes at 0.008 in. | Passes at 0.006 in. |
|---|---|---|---|---|---|
| A | 0.6 mm | 1.01 mm | 0.040 in. | 5 (4.97) | 7 (6.63) |
| B | 1.4 mm | 1.81 mm | 0.071 in. | 9 (8.91) | 12 (11.88) |
| C | 2.2 mm | 2.61 mm | 0.103 in. | 13 (12.84) | 18 (17.13) |
| D | 2.9 mm | 3.31 mm | 0.130 in. | 17 (16.29) | 22 (21.72), after hand UT |
| E | 1.2 mm | 1.61 mm | 0.063 in. | 8 (7.92) | 11 (10.56) |
| F | 0.5 mm | 0.91 mm | 0.036 in. | 5 (4.48) | 6 (5.97) |
Checks: 0.6 + 0.41 = 1.01 mm; 1.01 ÷ 25.4 = 0.0398 in.; 0.0398 ÷ 0.008 = 4.97, rounded up to 5; 0.0398 ÷ 0.006 = 6.63, rounded up to 7. Block D: 3.31 ÷ 25.4 = 0.1303 in.; 0.1303 ÷ 0.008 = 16.29 → 17; 0.1303 ÷ 0.006 = 21.72 → 22.
Step 4: read the plan against the evidence. Blocks A and F sit at TTCI's "three to five passes" for typical surface damage. Blocks B and E need about the 0.060 in. (1.5 mm) at which every exception in TD-20-002 recovered. Blocks C and D need 13 to 22 passes, at or beyond the 1 to 17 passes and 0.060 to 0.157 in. (1.5 to 4.0 mm) TTCI recorded at its case-study sites. If Block D's hand UT finds a detail fracture, it is a 213.113 defect; if not, the owner weighs 22 passes against replacing 100 ft of rail.
Step 5: grind the corner, re-measure, re-test. The grinder works the gauge-corner facets hardest, because channel 1 governs. After the visit the trolley re-runs all six blocks; a block still above the light band has steeper cracks than assumed or a lift-off problem on the new profile, and gets more passes before the test car returns. Only then does the valid search under 213.237(j)(4) close the no-test.
Step 6: mind the clock. At 30 MGT a year, 30 MGT accumulates in 365 days, shorter than 370, so tonnage controls: the owner has 365 days from the last valid search, or the 25 mph limit of 213.237(h)(2) applies to the curve.
Key takeaway: the depth map decides where the stones work and which block gets a hand test first; the TTCI removal figures decide how many passes; the valid search decides when the item is closed.
Where does eddy current fit with ultrasonic rail testing?
Eddy current and ultrasonic testing see different volumes of the rail, so on a test vehicle they are complementary and on a maintenance plan they are sequential. The ultrasonic analysis gates out the surface, because the near-surface noise is "so large and complex that the analysis systems generally ignore the signal associated with the first few millimeters of the rail surface" (ORD-11/24). Eddy current lives in those millimeters.
| Eddy current on rail | Ultrasonic rail testing | |
|---|---|---|
| Sees | Surface-breaking head cracks, gauge-corner lipping, squats (location), welds and joints (location) | Internal defects of the head and web: transverse and compound fissures, detail fractures, split heads, bolt hole cracks |
| Depth | About the first 2 to 3 mm, by inference | The head and web; most of the base is not covered |
| Output | Count and estimated depth per lateral channel | Indications by channel, verified by hand as defects under 213.113 |
| Rule status | None in Part 213 | The 213.237 internal inspection, by a 213.238 qualified operator |
| Hurt by | Lift-off, worn or rough heads, grinding marks on some systems, lubricant | Surface fatigue that scatters the beam (a no-test), poor coupling, worn profiles |
| Best use | Deciding how much to grind and proving the grind | The valid search before and after |
The probe physics of each ultrasonic channel is in how ultrasonic rail flaw detection works.
What eddy current rail inspection cannot see
Eddy current rail inspection cannot see below the skin, cannot see the web or base, cannot measure the crack angle, and cannot give a valid search. The specific gaps:
- Internal defects. "Only surface-breaking cracks are detectable" in steel (FHWA). A transverse fissure, a turned-down detail fracture, a split head or a bolt hole crack is invisible to the head probe; the hub's 213.113 table shows which ultrasonic channel finds each.
- Depth over about 3 mm, where the amplitude-to-depth conversion stops being reliable.
- The crack angle. It is entered, not measured, and TTCI's 5% to 33% spread came from that one assumption.
- Spalls versus cracks. TD-14-008 scores eddy current "N" on this; see rail shelling, spalling and surface defects.
- Lift-off. Signal falls with the cube of the gap; worn heads, debris, lubricant and probe bounce all change it. INNOTRACK lists grinding marks as an adverse factor, while TTCI found stone marks did not interfere; the answer is system-specific.
- Speed. The hand trolley is a walking-pace tool, and train-mounted arrays are limited by holding a 1 to 2 mm standoff.
- Regulatory credit. No eddy current result makes rail "internally inspected" under 213.237, and no amount of grinding does either.
Claims that eddy current reaches about 10 mm into rail steel, or is less sensitive to surface condition than ultrasonic testing, have no support in the FRA or FHWA sources.
How EMA Industries applies this
EMA Industries provides ultrasonic rail flaw testing for transit agencies, short lines and industrial track owners, with capabilities that include hyrail ultrasonic rail flaw detection and manual rail scanning, supported by an ASNT NDT Level III. When an owner's eddy current survey or grinding program leaves a curve that still needs a valid search, that re-test is ultrasonic work: a hyrail or manual search over the ground rail, and hand testing of any block whose surface readings suggest a crack may have turned down. Based in Southern California and mobilizing nationwide, EMA Industries can schedule the re-test around the grinder's visit so the 213.237(h) clock is met. To plan one, contact EMA Industries.
Frequently asked questions
How deep can eddy current measure head checks in rail?
To roughly 2 to 3 mm (0.08 to 0.12 in.). The currents run in a skin of a fraction of a millimeter; the depth figure is calculated from signal amplitude and an assumed crack angle, and two European sources put the reliable limit near 3 mm.
Is eddy current rail inspection accepted by the FRA?
Part 213 neither accepts nor rejects it, because it names no inspection method. What 213.237 requires is a valid internal search by a 213.238 qualified operator. Eddy current inspects the surface, so it cannot produce one; its results serve grinding and maintenance planning alongside the ultrasonic search.
Why do eddy current depth readings differ from cut sections?
Because the device reads the length of the crack face and converts it to depth with an angle the user enters. TTCI's 2014 validation found 5% to 33% differences at the default 25° when the real angles ran 18° to 28°, up to 0.003 in. per degree, with a residual of up to 0.0071 in. after the angle was corrected.
What is an eddy current array on a rail test vehicle?
A row of probes mounted side by side across the gauge corner and head, each one a channel reporting its own strip, so the output is a lateral map rather than a single trace. Published examples use four sensors per rail covering about 25 mm of the running edge, and current vendors offer up to eight per rail.
Does grinding after an eddy current survey count toward the 213.237 interval?
It does not. Grinding is not in Part 213 and earns no credit. The no-test block is closed only when a qualified operator completes a valid search over it within the 370-day or 30 MGT limit; otherwise 213.237(h) requires a 25 mph limit or replacement of the rail.
Sources
- 49 CFR 213.237, 213.238, 213.241 and 213.113 (eCFR; text read on Cornell LII, November 2026)
- Magel, Rolling Contact Fatigue: A Comprehensive Review, DOT/FRA/ORD-11/24, November 2011, sections 1.2, 3.1.1, 3.1.2 and 3.2.1
- TTCI Technology Digests TD-14-007 (Baillargeon, Tournay and Kerchof, May 2014), TD-14-008 (Baillargeon and Tournay, May 2014), TD-18-027 (Cummings, November 2018) and TD-20-002 (2020)
- FHWA InfoTechnology, Eddy Current Testing, NDE web manual
- Evident Scientific, Depth of Penetration in Eddy Current Testing
- EN 16729-2:2020, Railway applications. Infrastructure. Non-destructive testing on rails in track. Part 2: Eddy current testing of rails in track, scope and status; annex titles from the DIN EN 16729-2:2020-05 table of contents
- INNOTRACK, D4.4.1 Rail Inspection Technologies, University of Birmingham, November 2008 (public deliverable)
- Thomas, Heckel and Hanspach, Advantage of a combined ultrasonic and eddy current examination for railway inspection trains, Insight 49(6):341–344, 2007 (abstract)
- Stencel, Analysis of the occurrence of headchecks in rails, Transportation Overview 6/2019, DOI 10.35117/A_ENG_19_06_04
- Wang, Dai and Lautala, Real-time rail defect detection with Eddy Current (EC) technique, NURail2020-MTU-R18, Michigan Technological University, August 2020
- McWilliams and Eberius, Measuring With Eddy Current: Gage Corner Cracking and the Magic Wear Rate, Interface Journal, May 2014 (trade press)