Detecting Rail Defects: A Complete Guide for Inspectors

Detecting rail defects under FRA rules runs on two tracks: visual inspection under 49 CFR 213.233 (twice weekly on Class 4 and 5 track) and internal rail testing under 213.237 (at least every 370 days or 30 MGT on Class 4 and 5). Once a defect is verified, 213.113 sets the remedial action by defect type and size.

This guide is the hub of the EMA Industries rail defects library. It names each defect in the 213.113 table, says how it is found and which ultrasonic channel normally sees it, gives the remedial action row and the FRA accident cause code, and links to the detailed article for each defect. It ends with a worked example of a broken rail found next to a second suspect.

How does detecting rail defects work under FRA rules?

Three separate rules do three separate jobs. Visual inspection finds what shows on the surface, internal testing finds what is inside the rail, and 213.113 says what the owner must do about either. Mixing them up is the most common error in rail defect writing.

Two things follow from this split. A clean visual inspection says nothing about a transverse fissure inside the head, and a clean ultrasonic search says nothing about a cracked joint bar, because 213.237(f) covers the rail between the bars, not the bars. The wider rulebook is summarized in our guide to FRA track safety standards.

Key takeaway: detection is two duties (213.233 and 213.237), and disposition is a third (213.113).

Which rail defects does 213.113 name, and how is each one found?

The 213.113 remedial action table names 16 rail conditions, and each one has its own row, size bands and lettered actions. The table below groups them by where they start. Percentages are of the rail head cross-section weakened by the defect; lengths are in inches. The probe channels are typical industry practice, not a rule: Part 213 sets no ultrasonic method.

Source: 49 CFR 213.113 and the remedial action table as printed in the 2014 final rule (79 FR 4234). Cause codes from FRA's Appendix C track cause codes. Each deep article covers the full limits; for how the lettered actions and cause codes fit together, see FRA defect codes explained.

Transverse defects: TF, DF, compound fissure and engine burn

A transverse fissure grows inside the head from a nucleus, so nothing shows on the surface until the rail breaks. A detail fracture starts at or near the surface, and 213.113(d)(7) says it "may arise from shelled spots, head checks, or flaking". The rule warns that detail fractures "should not be confused with transverse fissures, compound fissures, or other defects which have internal origins." The distinction matters for detection: a detail fracture is the defect most often hidden under surface damage.

A compound fissure starts as a horizontal split head that turns transverse. Its row has no joint-bar option at all, which is why bars are never the remedy for a compound fissure.

Longitudinal head defects: horizontal and vertical split heads

Both are sized by length, not head area. The 213.113 definitions give the visual clues: a horizontal split head is "generally accompanied by a flat spot on the running surface", and for a vertical split head "a crack or rust streak may show under the head close to the web or pieces may be split off the side of the head."

Joint, web and base defects

Bolt hole cracks sit behind the joint bars, so a walking inspector rarely sees them until a piece breaks out. That is why 213.237(f) requires flaw detection equipment that works "in the area enclosed by joint bars." Broken bases, split webs and head-web separations can often be seen, but only after they have grown.

Welds

A defective weld under 213.113(d)(6) contains discontinuities over 5% of the head area individually or 10% in total. Weld defects of the head and web are found by the same channels as rail defects. Vehicle UT covers the head and web, not most of the base, so field welds need hand UT of the base (TTCI TD-00-013). Weld defect families are covered in weld defect types.

Surface conditions that are not table rows

Shelling, spalling, flaking, head checks and corrugation are not rows in the 213.113 table, and shelling alone carries no remedial action. Shelling appears in Part 213 only inside the detail fracture definition, (d)(7). Their importance for detection is that they block ultrasound from the rail beneath them. See rail shelling, spalling and surface defects, head checks and rolling contact fatigue.

Key takeaway: name the defect by its 213.113(d) definition first, because the definition picks the row and the row picks the action.

Which ultrasonic probe channel finds which defect?

Each channel looks for a flaw at a particular orientation. An angle beam finds planar flaws tilted to the running surface, and a 0° straight beam finds flaws parallel to it. The angles below are typical of North American test equipment and should be confirmed against the equipment manufacturer's procedure.

Vehicle systems carry several transducers per rail in fluid-filled wheels or sleds, often around 2.25 MHz. When an indication appears, a stop-and-verify run stops, and a qualified operator confirms it by hand. Under a continuous test under 213.240, a suspect isn't a defect until a 213.238-qualified person field-verifies it, within 84 hours, or 36 hours for suspected A, A2 or B defects. The physics behind each channel is in how ultrasonic rail flaw detection works.

Key takeaway: 70° channels find the transverse defects that cause service failures; 0° finds what lies flat.

Where does eddy current fit?

Eddy current finds surface-breaking cracks in the rail head, such as head checks and gauge-corner cracking, but it is sensitive to lift-off and surface condition, and it doesn't see deep internal defects. The original version of this article said eddy current is "less affected by surface contamination". That was wrong.

The Federal Highway Administration's NDE guidance on eddy current testing says "uniform and very small lift-off values are preferred for achieving better flaw detection sensitivity", that irregular surfaces or inconsistent probe movement can produce defect-like signals, and that in steel the depth of penetration is "typically less than 1–3 mm" (FHWA). FRA's rolling contact fatigue review notes that rail eddy current systems run with a 1 to 2 mm standoff (DOT/FRA/ORD-11/24). Debris, flaking, lubricant, worn profiles and probe bounce all change that gap.

TTCI found eddy current depth estimates differed from cut sections by 5% to 33% (TD-14-007, 2014), so treat eddy current depth as an estimate. In practice it supports grinding decisions and complements UT. It never replaces the internal search. How a rail eddy current system works, how accurate its depth readings are and how a depth map turns into a grinding pass and a 213.237(h) re-test is covered in eddy current rail inspection.

What can an inspector see during a 213.233 visual inspection?

A visual inspector can see the outward signs of several defects, and each sign is a reason to call for hand UT before the next train if it suggests an internal defect. Section 213.233 sets how often and how fast to inspect; it doesn't list defect signs. The signs below come from the 213.113(d) definitions and from industry training material.

The limits of what the eye can do are well documented. The Transportation Safety Board of Canada concluded, after a 1999 derailment from a fracture under shelly damage, that "Visual track inspection cannot identify developing fractures from shelly damage" (TSB R99M0046). For a fuller comparison, see ultrasonic vs visual inspection.

When to call for UT: any sign of a split head, a crack in the web or at a bolt hole, a burn on the running surface, or a rail next to a break. Call immediately when the sign could be an A, A2 or B defect, because 213.113(b) then requires immediate verification.

How often must rail be tested for internal defects?

On Class 4 and 5 track, and on Class 3 track with scheduled passenger trains or a hazardous materials route, internal rail inspections must be no more than 370 days or 30 MGT apart, whichever is shorter. Other Class 3 track is tested once a calendar year (no more than 18 months apart) or every 30 MGT, whichever is longer, and never more than 5 years apart.

The "40 MGT or once a year" figure that still circulates is the pre-2014 rule. The 2014 Improving Rail Integrity rule replaced it with the 30 MGT limit and a performance target: service failures (broken rails from a transverse fissure, compound fissure, detail fracture or vertical split head) may not exceed 0.10 per mile per year on Class 4 and 5 track, with lower rates for passenger and hazmat routes. Worked tonnage examples are in how often must rail be tested.

A search only counts if it is a valid search, defined in 213.237(j) as one where "the equipment performs as intended and equipment responses are interpreted by a qualified operator". Rail that couldn't be tested, for example under heavy shelling, must be re-searched, run at 25 mph or less, or replaced, under 213.237(h).

What does 213.113 require once a defect is found?

The owner must verify the indication, then either replace or repair the rail or start the lettered remedial action. A person designated under 213.7 decides whether the track stays in use. The rule reads:

"When an owner of track learns that a rail in the track contains any of the defects listed in the table contained in paragraph (c) of this section, a person designated under § 213.7 shall determine whether the track may continue in use. If the designated person determines that the track may continue in use, operation over the defective rail is not permitted until— (1) The rail is replaced or repaired; or (2) The remedial action prescribed in the table contained in paragraph (c) of this section is initiated." — 49 CFR 213.113(a)

Under 213.113(b), the owner must verify an indication within four hours, or immediately if it points to a defect needing action A, A2 or B. An owner testing under 213.240 uses that section's windows instead.

Source: remedial action notes, 79 FR 4256–4257 (2014), as codified in 213.113(c). Section 213.9(a) lists 213.113(a) as an exception to the class speeds, which is why each note caps speed at the lower of its own figure and the class maximum. The practical side of these restrictions is covered in slow orders for rail defects.

Worked example: a broken rail with a suspect next to it

The scenario. On a twice-weekly 213.233 inspection of Class 4 main track (maximum 60 mph freight, 80 mph passenger under 213.9), a track inspector finds a complete break in the south rail with a small gap. About 12 ft away on the same rail, a rust streak shows under the head next to the web.

Step 1: protect the break. A broken rail is already verified; it is visible. A person designated under 213.7 decides whether the track may stay in use. Until the rail is replaced or repaired, the designated person must visually supervise each move (action A). If only a person qualified under 213.7(d) is on site, that person may pass trains over the break at no more than 10 mph, watching each move, while a fully qualified person is sent out.

Step 2: classify the break. The fracture face shows a smooth oval with a crystalline center inside the head. That is a transverse fissure at 100% of the head. Its row is "A, or [E and H]". The owner can keep action A, or apply joint bars bolted per 213.121(d) and (e) (E) and limit speed to 50 mph (H). Because a transverse fissure caused the break, it is also a service failure under 213.237(j) and counts toward the segment's failure rate.

Step 3: verify the suspect. The rust streak matches the 213.113(d)(16) description of a vertical split head. A VSH over 4 in. is action B, so 213.113(b) requires the owner to verify the indication immediately, not within four hours. A 213.238-qualified operator runs hand UT. The 0° and side-looking checks show a split 3 in. long.

Step 4: apply the row. A VSH of more than 2 in. but not more than 4 in. takes "I and G": 30 mph and reinspection within 30 days.

Step 5: choose the repair. A train passes both spots, so the 30 mph VSH restriction controls the whole move. The cleaner fix is one plug rail long enough to remove both defects. That plug must itself be tested: under 213.237, plug rail installed after March 25, 2014 must not exceed 30 MGT since its last internal test, or trains run at no more than Class 2 speed over it (25 mph freight, 30 mph passenger).

Step 6: record it. Under 213.241 the owner records each defect's location, type, size and initial remedial action with its date. The records are kept for two years after the inspection and one year after the remedial action.

Which FRA cause codes cover rail defects?

FRA's numbered rail codes are accident cause codes (the T2xx group), entered on accident reports. They are not defect codes in 213.113, which names defects and uses lettered actions only.

On main track in 2016 to 2025, broken rails and welds caused 384 of 1,145 track-caused derailments. The top codes were T207 (110), T220 (57), T221 (55) and T202 (53) (FRA data on data.transportation.gov, queried October 24, 2026; counts change with late filings). T220 combines transverse and compound fissures, so FRA data can't separate them.

What are the limits of rail defect detection?

No inspection method finds every defect, and Part 213 doesn't ask for 100% detection. It asks for a valid search on schedule and a service-failure rate under the limit. The known gaps:

• Masking. Shells, head checks and spalls block ultrasound from the rail below. That mainly hides detail fractures, which start under that damage. In TTCI's 1998 tests of six hi-rail systems on 49 known defects, transverse defects under shells were found about 62% of the time against 77% for those not under shells (TD-99-006, 1999).
• Bolt holes and welds. The same tests found 33% of bolt hole cracks and 54% of defective welds, against 72% overall. Those 1998 figures are dated, but they show where the hard cases are.
• The base. Vehicle UT "only scans the head and web of the rail, and not the majority of the base" (TTCI TD-00-013, 2000). Broken bases are found by eye or by hand UT.
• Hidden rail. At road crossings the web and base sit under the crossing surface. TSB found web corrosion that "had not been identified by the visual or ultrasonic track inspections" (R19M0018).
• Frogs. Cast manganese frogs are too coarse-grained for routine UT; in-service checks are visual (FRA RR 22-25).
• Worn rail. On worn profiles, fixed-angle probes can inspect "the wrong volume of material" (TTCI TD-16-030).
• Eddy current depends on lift-off and gives estimated depths only.

Grinding can restore a testable surface, but grinding is not in Part 213 and earns no 213.237 credit.

How EMA Industries applies this

EMA Industries provides ultrasonic rail flaw testing for transit agencies, short lines and industrial track owners, plus rail weld inspection and NDT Level III support. Based in Southern California and mobilizing nationwide, EMA Industries works so that each test report names every verified defect by its 213.113 definition, size and location, which is what the owner's designated person needs to apply the right remedial action. To plan a test or follow up a suspect, contact EMA Industries.

Related rail defect guides

• Transverse fissure in rail
• Detail fracture development
• Compound fissure in rail
• Horizontal split head and vertical split head
• Bolt hole cracks
• Rail shelling, spalling and surface defects
• Weld defect types
• FRA defect codes explained
• Broken rail causes and prevention

FAQ

What is the most common rail defect behind broken-rail derailments?

On US main track in 2016 to 2025, detail fractures from shelling or head checks (cause code T207) led the broken-rail causes with 110 derailments, ahead of transverse and compound fissures (T220, 57) and vertical split heads (T221, 55). Those counts come from FRA accident data queried in October 2026.

Does 213.113 require any action for shelled rail?

Shelling is not a row in the 213.113 table, so shelling alone carries no remedial action. It matters because a detail fracture, which is a row, often starts under a shell, and the shell can block the ultrasonic search. Grinding is the usual treatment for shelly rail (TSB R99M0046), and any track the search could not test is recorded under 213.241(c)(6).

How fast must a suspected defect be verified?

Within four hours under 213.113(b), or immediately when the indication points to a defect that needs remedial action A, A2 or B. An owner using continuous testing under 213.240 must field-verify suspects within 84 hours, or 36 hours for suspected A, A2 or B defects.

Can a visual inspection replace ultrasonic testing?

It can't. Visual inspection under 213.233 and internal testing under 213.237 are separate requirements. Transverse fissures grow inside the head with no surface sign, and the TSB found that visual inspection cannot identify fractures developing under shelly damage.

Who may run rail flaw detection equipment?

An operator qualified under 213.238: trained under a documented program, authorized in writing by the employer, with at least 160 hours of supervised experience and periodic reevaluation. FRA certifies no one; the employer qualifies the operator.

Sources

• 49 CFR 213.113, 213.7, 213.9, 213.233, 213.237, 213.238, 213.240, 213.241, 213.339 (eCFR; text read on Cornell LII, October 2026)
• FRA, Track Safety Standards; Improving Rail Integrity, final rule, 79 FR 4234 (January 24, 2014): remedial action table and notes
• FRA, Appendix C, track cause codes
• FRA accident data, data.transportation.gov dataset byy5-w977, queried October 24, 2026
• FHWA, Eddy current testing, NDE web manual
• Magel, Rolling Contact Fatigue: A Comprehensive Review, DOT/FRA/ORD-11/24 (2011)
• TTCI Technology Digests TD-99-006 (1999), TD-00-013 (2000), TD-14-007 (2014), TD-16-030 (2016)
• Transportation Safety Board of Canada, R99M0046 and R19M0018
• FRA, Inspection of Cast Manganese Frogs, RR 22-25, ROSA P dot/64412