Rail Flaw Detection Methods: What Is Used in Track and What Is Still Research

Rail flaw detection in North America is built on multi-probe ultrasonic testing (UT), used to meet the internal rail inspection rule in 49 CFR 213.237. On Class 4 and 5 track that means a test every 370 days or 30 million gross tons, whichever comes first. Visual inspection, induction, eddy current and guided waves fill specific gaps.

This page is about the methods: what each one finds, whether it is in routine field use or still in the lab, and how a track owner layers them into one program. For the defects themselves (how a transverse fissure or a vertical split head forms, and what 213.113 requires once one is found), see the rail defect library.

What does rail flaw detection have to find?

Rail flaw detection has to find the defects listed in 49 CFR 213.113, because those are the defects a track owner must act on. The rule names them and prescribes a remedial action for each by type and size. Each method on this page is useful only to the extent that it finds one of these defects, or the surface damage that grows into one.

The FRA classes fall into three groups by where they start and which way the crack runs. That orientation, more than anything else, decides which method and which probe angle can see a defect.

Sources: 213.113(d) definitions; probe assignments are typical practice, not a rule (see the UT section).

Two of the defect descriptions in common use need correcting.

A transverse fissure starts from a nucleus inside the head, not just "from the center". 213.113(d)(15) defines it as "a progressive crosswise fracture starting from a crystalline center or nucleus inside the head". Historically that nucleus was usually a shatter crack, a tiny internal tear left by hydrogen in the steel. AREA rolling-load tests reported in 1941 that only shatter-cracked rails developed fissures, and control cooling and later vacuum degassing largely removed the cause (AREA Bulletin abstracts, Med confidence). That history is why modern transverse defects are more often detail fractures, which start at or near the surface from shells or head checks.

A squat is its own rolling contact fatigue (RCF) defect, not a deeper head check. FRA report DOT/FRA/ORD-11/24 describes the squat as "a surface defect most commonly associated with high-speed rail and areas of high tractive effort": a dark depression near the center of the head, with a short leading crack and a longer trailing crack. Head checks, by contrast, form at the gauge corner, mostly on the high rail of curves. Squats are not a row in the 213.113 table. If a squat's branch crack turns down into the head, the result is sized and handled as the transverse defect it becomes.

Key takeaway: pick the method by the defect's crack plane and depth, and use the FRA class names, because the rule and the report both use them.

Which rail flaw detection methods are in field use today?

Only ultrasonic testing and visual inspection are routine, rule-linked methods on North American track; induction and eddy current are complementary, and the rest are niche or research. The table below states each method's status honestly, with the rule or report that supports it.

Key takeaway: if a method has no rule basis and no named field deployment, treat it as research, whatever a brochure says.

How does multi-probe ultrasonic testing find internal rail defects?

Rail UT sends several beams into the head at once, typically a 0° beam, 37–45° beams and 70° beams plus side-looking probes, and reads the echoes from each. No single probe can see every defect plane, so the channels are chosen to cover them between them. No public source names one standard North American channel set, so treat any probe layout, including this one, as typical rather than fixed.

A 1999 Mountain-Plains Consortium report on North American practice (MPC 99-106, Colorado State University) described transducers mounted in fluid-filled wheels, three to six per wheel, usually two wheels per rail, and as many as 12 transducers per rail. Coupling fluid is water or oil based. The same multi-probe idea runs on three platforms: rail-bound test cars, hi-rail trucks with retractable test carriages, and hand-pushed carts or walking sticks for short sections and verification.

FRA's 2014 rule preamble describes four ways the testing is run:

• Portable, pushing a test device at walking pace
• Start/stop, testing at a slow speed "normally not exceeding 20 mph" and stopping to verify each indication
• Chase car, where a lead vehicle tests and a following crew verifies
• Continuous, where a vehicle tests non-stop, the data is analyzed centrally and suspects are verified later

Continuous testing is governed separately by 49 CFR 213.240. The two modes are compared in continuous vs stop-and-verify rail testing.

The same preamble sets the coverage limit. The waves "normally scan the entire rail head and web, as well as the portion of the base directly beneath the web". The outer base flanges are mostly outside the beam. Whatever the platform, the equipment must also meet 213.237(f): "Inspection equipment shall be capable of detecting defects between joint bars, in the area enclosed by joint bars."

Part 213 never uses the word "ultrasonic". It requires "internal rail inspections" with rail defect detection equipment, run by an operator qualified under 213.238. UT is the industry's choice for meeting that requirement, not a rule. For the step-by-step process see how ultrasonic testing works for railways.

Phased array UT is used in rail mainly to verify and size indications and in research, not as routine test-car output. Test cars show channel maps and A-scans to the operator rather than images.

Key takeaway: rail UT is a set of beams aimed at known crack planes, and its coverage stops at the part of the base directly under the web.

Is induction testing still used on rail?

Induction testing was the original rail flaw detection method, and FRA said in 2014 that it was still in use as a complement to UT on U.S. railroads. The first commercial detector car test ran in November 1928 on the Wabash. Ultrasonic test cars followed: MPC 99-106 dates the first to 1959, built for the New York City Transit Authority, and many cars then carried both systems.

FRA's 2014 preamble explains the principle and states the balance between the two methods:

"The main technologies utilized for non-destructive testing on U.S. railroads are the ultrasonic and induction methods. Ultrasonic technology is the primary technology used, and induction technology is currently used as a complementary system." — FRA, Track Safety Standards; Improving Rail Integrity, 79 FR 4234, 4237 (January 24, 2014)

Induction "requires the introduction of a high-level, direct current into the top of the rail and establishing a magnetic field around the rail head", in the preamble's words. A transverse flaw distorts the current, and a search unit passing through the field picks up that distortion.

That statement is from January 2014. Whether a given contractor's car carries induction today is a question for that contractor, so ask for the system description in the proposal rather than assuming either way. Induction is aimed mainly at transverse defects in the head. It adds nothing for horizontal or web defects, which is one reason it has stayed a complement rather than an alternative to UT.

Key takeaway: describe induction as historical and still complementary on some cars, and confirm it per contractor.

What can eddy current and ACFM tell you about rolling contact fatigue?

Eddy current testing (ET) measures shallow surface-breaking RCF cracks such as head checks, and gives an estimated depth that depends on an assumed crack angle; ACFM does a similar job but remains a research and trial method in rail. Neither one is a 213.237 internal inspection. Their job is to tell the owner how deep the surface damage is, which drives grinding and helps explain UT no-tests. For the defect background see rolling contact fatigue in rail.

Eddy current: in field use, non-contact but lift-off sensitive

ET is non-contact, but it is far from indifferent to the gap. FRA's RCF review (ORD-11/24, 2011) notes that eddy current probes need a standoff of about 1–2 mm. The FHWA's eddy current guidance states that "uniform and very small lift-off values are preferred for achieving better flaw detection sensitivity". Uneven surfaces or inconsistent probe movement can produce signals that look like defects.

Three field examples show where ET is really used:

• Grinding trains. ORD-11/24 describes a grinding-train head-check scanner with four eddy current probes per rail at about 1 mm standoff, recording head-check position and depth before and after grinding.
• Dedicated ET vehicles. TTCI tested a commercial eddy current Draisine on Norfolk Southern heavy-haul track (TD-14-007, 2014). Depth errors ran 5–33% against sectioned rail at the default 25° crack angle. The report states: "The Draisine relies on the user to enter an assumed crack angle, as the actual angle cannot be determined through nondestructive testing."
• Contract scope. An MTA procurement for the Long Island Rail Road and Metro-North (RFP issued January 2023) bundled rail flaw testing, joint bar inspection and eddy current testing in one five-year contract.

In Europe the method has its own standard: EN 16729-2 (2020) covers eddy current testing of rails in track from inspection trains, reprofiling machines and manual systems. ET's depth reach is limited to roughly the skin depth (FHWA gives "typically less than 1–3 mm" in steel). ORD-11/24 adds that "it is difficult with eddy current to know the actual depth of damage". The eddy current vs ultrasonic testing page compares the two in more detail.

ACFM: crack depth sizing, still in trials

Alternating current field measurement (ACFM) induces a uniform current in a very thin skin (ORD-11/24 gives under 0.1 mm) and measures how a crack disturbs the field. Its appeal for RCF is a direct depth signal: ORD-11/24 notes that "the deeper the crack, the larger the response".

The field basis for ACFM in rail is thin. A 2008 review of rail NDE by Papaelias, Roberts and Davis listed ACFM among the techniques "under development for application in the rail industry". The same group's 2009 high-speed tests used a rotary test piece with machined notches, detected at 121.5 km/h, and then a spinning rail rig at lower speed. The sensor found most notches, but lift-off variation affected the signal. Treat ACFM as a research and trial method for rail.

Surface-wave UT is sometimes offered for RCF depth too. It is also mostly at the research stage, as set out in surface-wave UT for rail.

Key takeaway: eddy current sizes surface RCF where it is used, ACFM is still being trialled, and neither replaces the internal search.

What does visual and imaging inspection cover under 213.233?

Visual track inspection under 49 CFR 213.233 is a separate, more frequent requirement from internal testing, and imaging systems may support it but not replace the inspector. On Class 4 and 5 track the inspection is due twice weekly, at least one calendar day apart. The person doing it must be designated under 213.7.

The rule allows technology in a supporting role. 213.233(b) says inspections are made on foot or from a vehicle, and that "mechanical, electrical, and other track inspection devices may be used to supplement visual inspection." Vehicle speed over crossings and turnouts is limited to 5 mph.

Visual inspection finds what shows on the surface: broken rails and bars, crushed or flattened heads, missing bolts, engine burns, heavy shelling and the rust streak that can mark a vertical split head. Camera and machine-vision systems record the same surface at speed and give a repeatable record. ORD-11/24 cites Canadian work (NRCC, 2004) finding that fine RCF cracks are an order of magnitude harder for vision systems than spalls. Neither the eye nor a camera can see an internal transverse defect. That is why 213.237 opens: "In addition to the inspections required by § 213.233, each track owner shall conduct internal rail inspections".

Key takeaway: 213.233 visual inspection and 213.237 internal testing are two different duties, and imaging only supplements the first.

Where do guided waves fit in rail flaw detection?

Guided waves in rail are used for long-range broken-rail monitoring, not for the periodic internal search. Low-frequency waves travel along the rail as several waveguide modes over long distances. A complete break interrupts or reflects them.

The best documented installation is on South Africa's Sishen–Saldanha heavy-haul iron ore line. A permanently installed ultrasonic broken rail detection system there "detects complete rail breaks at long-range using guided wave ultrasound" (Ramatlo et al., MATEC Web of Conferences 347, 2021). The same paper notes that real damage signals in service are very hard to obtain, so detection of partial defects is being developed through simulation.

Guided-wave research in North America has also aimed at the masking problem. FRA's 2014 preamble describes work on "minimizing the masking effect of transverse cracks by surface shelling". None of this produces a 213.237 valid search, which the rule defines as continuous inspection with equipment interpreted by a qualified operator. Guided wave and long-range UT are two names for the same family of methods, explained on our guided wave ultrasonic testing page.

Key takeaway: guided waves can tell you a rail has broken; they cannot tell you it passed its internal inspection.

Which methods are still research or niche?

Magnetic flux leakage, inductive thermography, acoustic emission and laser or air-coupled UT all have published rail research, but none is routine North American practice. Each entry below gives the evidence that does exist, so a reader can judge a vendor claim against it.

Magnetic flux leakage (MFL). MFL magnetizes the rail and senses the field that leaks out at a flaw. Papaelias and co-authors (2008) wrote that "magnetic induction sensors have been extensively used by the rail industry" for in-service inspection, a phrase that covers the DC induction cars described above. Dedicated MFL rail work is reported mainly from Russia: Antipov and Markov (Russian Journal of Nondestructive Testing 55, 2019), writing from a university and a flaw-detector maker, analyzed how well MFL detects rail head defects. Outside that context, treat MFL for rail as research, and don't present it as a corrosion or "deep flaw" tool for rail.

Inductive thermography. An induction coil heats the rail surface for a fraction of a second and an infrared camera records how cracks disturb the heat flow. Tuschl, Oswald-Tranta and Eck (Applied Sciences 11, 2021) imaged head checks and a squat on 1 m rail pieces in the laboratory, both stop-and-go and while moving. That is promising, but it is laboratory work.

Acoustic emission (AE). AE sensors listen for the stress waves a growing crack releases under load. Bruzelius and Mba titled their 2004 paper "An initial investigation on the potential applicability of Acoustic Emission to rail track fault detection" (NDT&E International 37). That remains a fair description. AE locates active sources but does not size them, so any AE finding would still need UT to evaluate it.

Laser and air-coupled UT. These generate or receive ultrasound without contact, which would remove the wheel probe and its couplant. Kenderian and co-authors reported a laser-air hybrid technique for dynamic rail inspection in Insight (2005, 2006). Mariani and Lanza di Scalea modeled the defect detection of an air-coupled rail system in Structural Health Monitoring (2017). Both lines of work are prototypes, and the 2008 review listed laser ultrasonics and electromagnetic acoustic transducers (EMATs) as still under development.

Key takeaway: cite the paper, call it research, and don't give these methods a 213.237 role.

How do the methods fit together? A worked example for a 50-mile line

A rail flaw detection program layers the methods by job: visual inspection for the surface, UT for the legal internal search, surface sizing for RCF, and verification and records to close each finding. The example below shows the arithmetic for one line, using the rule values and stating each assumption.

Assumptions: 50 miles of single main track (100 rail-miles), Class 4, 35 MGT a year, regularly scheduled passenger trains, not a hazardous materials route, designated as one rail inspection segment. About 6 miles of the line is in curves with known head checking.

Step 1: visual inspection (213.233)

Class 4 track needs visual inspection twice weekly, at least one calendar day apart. That is 2 × 52 = 104 inspections a year, or 104 × 50 = 5,200 miles of track inspected visually each year.

Step 2: internal UT interval (213.237(c))

For Class 4 track the test interval is 370 days or 30 MGT, whichever is shorter. At 35 MGT a year, 30 MGT accumulates in 30 ÷ 35 × 365 = 312.9 days, so tonnage controls and the latest legal interval is about 312 days. Any line above about 29.6 MGT a year is tonnage-controlled under this rule.

A 312-day maximum leaves little room for a lost possession or bad weather. A sensible plan is two tests a year, about 182 days apart, which is 35 × 182 ÷ 365 = 17.5 MGT between tests.

Step 3: test time

MPC 99-106 reported that stop-and-verify testing averaged 6–8 mph once stops are included (1999 figure; treat as an order of magnitude). For 50 miles that is 50 ÷ 8 = 6.3 hours to 50 ÷ 6 = 8.3 hours of testing per run, before travel and setup. Two runs a year need about 12.5–16.7 hours of track time for testing.

Step 4: no-test follow-up (213.237(g) and (h))

Suppose the first run records 0.4 miles (2,112 ft) of high rail in the curves where heavy head checking blocked a valid search. Under 213.237(g) that length doesn't count as inspected. Before the interval expires the owner must get a valid search over it, reduce speed to 25 mph until one is made, or replace the rail. A walking-pace hand test over the 2,112 ft is the usual first choice.

Step 5: RCF sizing and grinding

On the 6 miles of curve, an eddy current survey of the 6 rail-miles of high rail before and after grinding shows whether the damage is shallow enough to grind out. This layer has no Part 213 basis. Its value is fewer UT no-tests on the next run.

Step 6: verification, disposition and records

Each suspect from a stop-and-verify run must be verified within four hours under 213.113(b), or immediately for a defect that requires remedial action A, A2 or B. Defects are then handled under the 213.113 table. 213.241 records are kept at least two years after the inspection and one year after the remedial action, including the location of track not tested.

Step 7: service failure check (213.237(a))

With passenger trains, the service failure limit is 0.09 per mile per year. For 50 miles that is 0.09 × 50 = 4.5. So 4 service failures in a 12-month period (4 ÷ 50 = 0.08) stays within the limit, and 5 (5 ÷ 50 = 0.10) exceeds it.

Key takeaway: the rule sets the floor (104 visual inspections and a test within ~312 days here); the program adds margin, RCF sizing and no-test recovery on top. For how the interval rules apply to other track classes see how often must rail be tested.

What are the limits of rail flaw detection?

No rail flaw detection method finds every defect, and the rule doesn't pretend otherwise: it sets a service failure rate, not a detection guarantee. The limits below are the ones that matter when planning a program.

• Surface masking. Shells, head checks and engine burns can block UT from what lies beneath them. In TTCI's 1999 benchmark of six hi-rail systems on 49 known defects (TD-99-006), transverse defects under shells were found about 62% of the time (7 defects), against 77% where no surface anomaly was present (33 defects). The overall hit rate was 72%.
• Small and awkward defects. The same benchmark found bolt hole cracks only 33% of the time and defective welds 54%, while horizontal split heads reached 92%. These are 1999 figures for the systems tested then.
• Base coverage. UT covers the head, the web and only the part of the base directly beneath the web. Broken bases starting at the flange edge are largely outside the beam.
• Depth from surface methods. Eddy current depth depends on an assumed crack angle, and TTCI's tests found errors of 5–33% at the default. Accurate measurement of crack angle and depth remains "a primary limitation" for both ET and UT, according to TTCI TD-14-008 (2014).
• No-tests are untested rail. Poor coupling, lubrication or RCF scatter can prevent a valid search. That length is not inspected until it is re-searched, and 213.238(b)(4) makes the qualified operator responsible for judging when that has happened.
• Monitoring is not inspection. Guided-wave and acoustic systems can flag a break or an active crack, but they don't produce the valid search 213.237 requires.

Key takeaway: plan around masking and coverage gaps with grinding, hand testing and records, because no layer is complete on its own.

How EMA Industries applies this

EMA Industries provides ultrasonic rail flaw testing for transit agencies, short lines and industrial track owners, from Southern California and mobilizing nationwide. In practice that means multi-probe UT run by qualified operators, hand verification of every suspect, and reports that record defect type, size and location as well as any track where a valid search could not be made. To scope a testing program like the 50-mile example above, see our ultrasonic rail testing services or contact us.

Related reading: Rail defect library · How ultrasonic testing works for railways · Rolling contact fatigue in rail · Surface-wave UT for rail · Top railway inspection techniques · 49 CFR 213.237 explained

Frequently asked questions

What is the main rail flaw detection method in the United States?

Ultrasonic testing. FRA's 2014 rule preamble calls it "the primary technology used", with induction as a complementary system. Test cars, hi-rail vehicles and hand carts all use several probe angles, typically 0°, 37–45° and 70° plus side-lookers.

Does eddy current testing count as a 213.237 internal rail inspection?

Not on its own. Eddy current measures shallow surface cracks, reaching only about the skin depth, so it cannot search the head and web for internal defects. It is used alongside UT to size rolling contact fatigue and plan grinding.

How fast do rail flaw detection vehicles test?

FRA describes stop-and-verify testing as normally not exceeding 20 mph. A 1999 university research report (MPC 99-106) found that average speeds including stops ran about 6–8 mph. Continuous testing runs faster because suspects are verified later under 213.240.

Can guided-wave monitoring replace periodic rail testing?

No. Installed guided-wave systems detect complete breaks over long distances. The 213.237 interval requires a valid search with equipment interpreted by a qualified operator, so monitoring adds warning but doesn't reset the clock.

Which rail defects are hardest to detect?

Defects hidden under shells or heavy head checking, bolt hole cracks, defective welds and defects in the outer base. In TTCI's 1999 benchmark, bolt hole cracks were found 33% of the time and transverse defects under shells about 62%.

Sources

• 49 CFR 213.237, Inspection of rail, eCFR (text read on Cornell LII, October 2026)
• 49 CFR 213.233, Track inspections, eCFR
• 49 CFR 213.113, Defective rails, eCFR
• 49 CFR 213.238, Qualified operator, eCFR
• 49 CFR 213.240, Continuous rail testing, eCFR
• 49 CFR 213.241, Inspection records, eCFR
• Track Safety Standards; Improving Rail Integrity, 79 FR 4234 (January 24, 2014), Federal Register, preamble p. 4237
• Magel, Rolling Contact Fatigue: A Comprehensive Review, DOT/FRA/ORD-11/24 (2011), FRA
• Garcia and Reiff, Benchmarking Results of North American Rail-Flaw Detection Technologies, TTCI TD-99-006 (1999)
• Baillargeon, Tournay and Kerchof, eddy current RCF depth measurement, TTCI TD-14-007 (2014)
• Jeffrey and Peterson, Assessment of Rail Flaw Inspection Data, MPC 99-106 (1999)
• FHWA, Eddy current testing
• Papaelias, Roberts and Davis, "A review on non-destructive evaluation of rails", Proc. IMechE Part F 222 (2008)
• Papaelias, Lugg, Roberts and Davis, "High-speed inspection of rails using ACFM techniques", NDT&E International 42 (2009)
• Ramatlo, Loveday, Long and Wilke, guided-wave broken rail detection, MATEC Web of Conferences 347 (2021)
• Antipov and Markov, "Detectability of rail defects by magnetic flux leakage method", Russian Journal of Nondestructive Testing 55 (2019)
• Tuschl, Oswald-Tranta and Eck, "Inductive thermography as non-destructive testing for railway rails", Applied Sciences 11 (2021)
• Bruzelius and Mba, acoustic emission for rail track fault detection, NDT&E International 37 (2004)
• Kenderian et al., "Dynamic railroad inspection using the laser-air hybrid ultrasonic technique", Insight 48 (2006)
• Mariani and Lanza di Scalea, air-coupled ultrasonic rail inspection, Structural Health Monitoring 17 (2017)
• EN 16729-2:2020, Railway applications – Infrastructure – Non-destructive testing on rails in track – Part 2: Eddy current testing of rails in track (CEN; scope from public catalog listings)