Rail Shelling and Spalling: Surface Defects That Lead to Internal Failures

Rail shelling and spalling are two different rail head conditions. Shelling is rolling contact fatigue cracking that starts about 1/4 to 3/8 inch below the gauge corner and breaks out as a shell. Spalling is shallow chipping of the running surface. Neither is a 49 CFR 213.113 defect on its own, but both can hide one from ultrasonic testing.

That last point is why the two conditions matter to anyone who buys or runs rail flaw testing. The rule names shelled spots as a starting point for the detail fracture, one of the four defects that count as service failures under 49 CFR 213.237. The same shells also scatter the ultrasonic beam that is meant to find it. This guide separates the terms, shows where each one sits in the federal rules, and works through what a track owner must do when shelling makes a test invalid.

What is the difference between rail shelling and spalling?

Shelling grows from a crack below the surface; spalling is material breaking off at the surface. The two are often written as one phrase, but they start in different places and carry different risks.

Shelling is a fatigue crack that forms inside the steel at the gauge corner of the rail head, grows roughly parallel to the corner surface, and eventually breaks out so that a shell-shaped piece of metal separates. AAR/TTCI's scanning electron microscope study of shells at the Facility for Accelerated Service Testing (FAST) found that they "originate as cracks in the subsurface of the high rail gage corner 1/4 to 3/8 inch from the gage corner surface" in 5° and 6° curves (TTCI Technology Digest TD-10-004, March 2010). Transit track inspection training describes the same condition as "a progressive horizontal separation that may crack out at any level on the gauge side, generally at the upper gauge corner" (MARTA track inspection training, Section 7). A small shell may show no surface sign at all until the head is broken or sectioned.

Spalling is shallow chipping or flaking of the running surface. It usually comes from a damaged or brittle surface layer rather than a deep crack. Two common sources:

• Engine burns. A slipping locomotive wheel overheats the running surface. The burned area does not grow as a crack, but it "may gradually chip out and roughen under repeated traffic" (MARTA training, Section 7).
• White etching layers (WEL). Frictional heat from wheel slip or hot grinding can form a hard, martensitic surface layer. A thermocouple study found WEL starts to form at a grinding temperature of about 400 °C (Zhou et al., International Journal of Heat and Mass Transfer, 2021), and rolling-sliding tests linked WEL with crack and defect formation (Mesaritis et al., Wear, 2023). The brittle layer cracks and pieces spall out.

Terminology varies. Some railroad manuals call very shallow shells "spalls", and FRA lists shells, spalling and flaking as separate surface conditions (see the quotation below). For testing purposes the useful line is depth: a shell undercuts the gauge corner at roughly 1/4 inch or more, while a spall is a surface breakout.

Key takeaway: a shell is a subsurface fatigue crack that has reached the surface; a spall is surface material that has chipped away.

How do shelling, spalling, head checks, flaking and detail fractures compare?

Four of these are surface conditions; the detail fracture is the internal defect that some of them start. The table puts them side by side.

Sources: 49 CFR 213.113(d); TTCI TD-10-004; MARTA track inspection training, Section 7. For the surface cracks in more depth, see EMA Industries' guides to head checks in rail and rolling contact fatigue.

Key takeaway: classify the surface condition, then ask whether a detail fracture is growing beneath it.

Is shelled rail a defect under 49 CFR 213.113?

No. Shelling by itself is not in the 213.113 remedial action table; it appears only in the definition of a detail fracture. The table lists compound fissures, transverse fissures, detail fractures, engine burn fractures, defective welds, split heads, split webs, piped rail, head-web separations, bolt hole cracks, broken bases, ordinary breaks, damaged rail, and flattened rail or crushed heads. There is no row for shelled rail, spalling, flaking or head checks. A full-text search of the 2025 Part 213 text finds "shell" only in the two detail fracture definitions, 213.113(d)(7) and the Class 6+ version in 213.337.

The definition in 49 CFR 213.113(d)(7) reads:

"Detail fracture means a progressive fracture originating at or near the surface of the rail head. These fractures should not be confused with transverse fissures, compound fissures, or other defects which have internal origins. Detail fractures may arise from shelled spots, head checks, or flaking."

So a shell becomes a regulatory matter the moment it turns down into a transverse crack. At that point it is a detail fracture, sized by the percentage of the head cross-section it weakens:

Source: 213.113(c) remedial action table, as published in the 2014 final rule (79 FR 4234, January 24, 2014). The full notes are in EMA Industries' guide to FRA defect codes.

Canada's rules are similar. A 1999 Transportation Safety Board investigation noted that Canada's Railway Track Safety Rules "do not require any remedial action for 'shelly rail' although they do make reference to fractures which may arise from shelly defects" (TSB R99M0046). In FRA accident data, a broken rail of this kind is coded T207, detail fracture from shelling or head check.

Key takeaway: cite 213.113(d)(7) for shelling. The remedial action table applies only once a detail fracture has formed.

How does shelling form in the rail head?

Repeated wheel loads create a high shear stress zone a short distance below the gauge corner, and a fatigue crack grows there. TD-10-004 found that the FAST shells grew in the plane parallel to the gauge corner surface "outward in all directions until they reach the head surface", that growth direction followed the direction of traffic, and that a shell took 20 to 40 MGT from initiation to critical failure under 39-ton axle loads. Inclusions of manganese sulfide and aluminum oxide, 50 to 100 micrometers across, sat in the fatigue marks, and the authors recommended cleaner rail steel. FAST runs heavier and faster than most revenue track, so treat the 20 to 40 MGT figure as an accelerated test-track value.

The conditions that raise the risk are the ones that raise gauge-corner contact stress:

• Curves, especially the high rail, where the wheel flange root bears on the gauge corner.
• Heavy axle loads. Shell growth at FAST was measured under 39-ton axles.
• Profile mismatch and gauge face wear, which move the contact band toward the corner. See gauge face wear in rail.
• Missed preventive grinding, which lets the cracked surface layer survive long enough to grow.
• Rail cleanliness. Inclusions act as stress raisers, according to TD-10-004.

Shelling and head checks often appear together on the same high rail, because both come from rolling contact fatigue at the gauge corner. Corrugation is a different mechanism. Part 213 excludes corrugation from flattened rail and sets no limit for it.

Key takeaway: shells are a curve, tonnage and profile problem, and they grow in tens of MGT, not years of calendar time.

Why do shells and head checks mask ultrasonic testing?

A shell or a cluster of head checks is a set of crack faces sitting right where the beam enters the rail, so it reflects and scatters the sound before the sound reaches what is underneath. This is the key NDT point about rail shelling and spalling. FRA described it in the preamble to the 2014 rail integrity rule:

"As with any non-destructive test method, these technologies are susceptible to physical limitations that allow poor rail head surface conditions to negatively influence the detection of rail flaws. The predominant types of these poor, rail head surface conditions are shells, engine driver burns, spalling, flaking, corrugation, and head checking."

— FRA, Track Safety Standards; Improving Rail Integrity, 79 FR 4234 (2014)

The same preamble says FRA research was targeting "the masking effect of transverse cracks by surface shelling". Three pieces of evidence show how large the effect is:

The TD-99-006 figures come from 1999 equipment, so read them as a historical baseline. The mechanism hasn't changed.

TD-18-027 calls the 0° symptom loss-of-bottom: "If the reflected signal strength decreases, it can be a sign that anomalies at or near the rail surface, such as RCF or spalling, may be scattering a portion of the ultrasonic signal." A drop in the back-wall echo is a testability warning, not a clean result. It is also not a false call. EMA Industries' article on false calls in rail UT explains why untested rail should be recorded separately.

Masking is mainly a detail fracture problem. A detail fracture starts under the shell, so the damage that starts the defect is the damage that hides it. A transverse fissure starts from a nucleus inside the head and can sit under clean rail, although a shelled surface over one will weaken its echo too. For how the defect grows, see how a detail fracture develops in rail.

A Canadian case makes the same point. In TSB investigation R99M0046 (Bedford, Nova Scotia, October 9, 1999), a CN train derailed on a 4.5° curve after a rail fractured from microcracks under shelly damage. The rail had been ultrasonically tested in June, and the TSB found: "Visual track inspection cannot identify developing fractures from shelly damage."

Key takeaway: on shelled or head-checked rail, a clean UT record is weaker evidence than the same record on clean rail.

How is shelled rail hand-verified?

The qualified operator stops, looks at the surface, and hand-scans the location from positions and angles the test vehicle can't use. Under 49 CFR 213.237(g), the qualified operator decides whether a valid search was made. Hand verification is how that decision is backed up. A typical sequence, set by the owner's or contractor's procedure:

1. Look first. Mark the extent of shelling, head checks and spalls, and note dark spots, gauge-side separations and any bleeding rust.
2. Check the 0° back-wall echo along the stretch. A steady loss-of-bottom over the shelled zone tells you how much of the rail the vehicle couldn't see.
3. Scan with the 70° probe from more than one position: centered, offset toward the field side and, where the profile allows, skewed toward the gauge corner. A beam entering away from the shelled band can reach the head volume beneath it.
4. Compare both directions. Detail fractures under shells are often inclined, and one direction may return a much stronger echo.
5. Use a second angle (often 45°) on any indication to confirm it and estimate its extent.
6. Record the result: a defect with its type and size, a verified clean stretch, or a stretch that still couldn't be tested.

Probe angles and offsets are typical industry practice, not Part 213 requirements. Part 213 sets no UT method. The test that matters is whether the search was continuous and "has not been compromised due to environmental contamination, rail conditions, or equipment malfunction", which is one of the qualified operator's duties under 49 CFR 213.238(b)(4). Heavy shelling is a "rail condition" in that sense. If hand probes can't get sound past the shell either, the stretch is still untested.

Key takeaway: the hand test either finds the defect, clears the rail, or confirms that the rail can't be tested as it is.

Worked example: shelling makes a search invalid

When a stretch can't be validly searched, 49 CFR 213.237(h) gives the owner three options, and they must be completed before the time or tonnage limit runs out. The rule says the owner shall, before expiration of the limits:

"(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."

— 49 CFR 213.237(h)

The inputs. A Class 4 freight main line carries 24 MGT a year, and its maximum freight speed is 60 mph. The last valid search was on day 0.

• Tonnage limit: 30 MGT ÷ 24 MGT a year = 1.25 years, about 456 days.
• Calendar limit: 370 days.
• The shorter applies, so the next valid search is due by day 370.

The test runs on day 300, after about 24 × 300 ÷ 365 = 19.7 MGT. On the high rail of a 5° curve between MP 112.40 and MP 112.62, about 1,160 ft (0.22 mile × 5,280 ft), the 0° channel loses its back-wall echo and the 70° channels show heavy surface noise. A visual check confirms continuous shelling. The operator decides a valid search was not made. The stretch can't count as inspected (213.237(g)), and its location must be recorded under 213.241(c)(6).

The deadline. The owner has 370 − 300 = 70 days. By day 370 the segment will have carried about 24 × 370 ÷ 365 = 24.3 MGT, so the calendar limit still controls.

What if the hand test finds something? Suppose the hand test on day 301 finds a detail fracture from shelling estimated at 15% of the head area. That falls in the 5 to less than 25% band, which is remedial action C. Because the defect was found during a 213.237 search on Class 4 track, note C allows 50 mph (lower than the 60 mph Class 4 maximum) for up to 4 days. If the rail isn't removed or permanently repaired by then, speed drops to 30 mph until joint bars are applied, and the bars must go on within 10 days or speed drops to 10 mph. The rest of the 1,160 ft stretch still needs its valid search.

The numbers are illustrative. The 0.060 in. grinding result is from the railroads TTCI studied, and the train delay ignores braking. The rule logic is exact.

Key takeaway: find invalid searches early in the interval, because a shelled curve tested at day 365 leaves only the speed restriction or new rail.

Which methods find which rail head defects?

Ultrasonic testing is the standard method for internal rail defects, including bolt hole cracks; magnetic particle and visual methods see only the surface. The original version of this article listed bolt hole cracks under "visual plus magnetic particle". That was wrong for rail in track. Bolt holes sit behind the joint bars, and 49 CFR 213.237(f) requires rail flaw equipment to be capable of finding defects between joint bars. MT would need the bars removed and still finds only surface-breaking cracks. Bolt hole cracks are found by angled UT through the head and web, then hand-verified. Detection is not perfect: in TD-99-006 the hi-rail systems found only 33% of the bolt hole cracks. See EMA Industries' guide to bolt hole cracks in rail.

For the method itself, see how ultrasonic testing works for railways.

Key takeaway: surface methods describe the shell; only UT looks for the detail fracture under it.

How does grinding control rail shelling and spalling?

Grinding removes the cracked surface layer before shells and head checks grow, and corrective grinding makes masked rail testable again. The TSB's Bedford report put it simply: "The recommended remedial action for shelly damage is rail grinding." After that derailment, the railroad began grinding the identified rail twice a year.

Intervals are set by the owner, not by Part 213, and grinding earns no credit under 213.237. TTCI's survey of seven North American railroads (TD-99-004, 1999) found that most used or wanted preventive grinding. It put the minimum practical interval at about 10 to 15 MGT and suggested aiming for about 25 MGT "unless special factors apply". EMA Industries' article on rail grinding and ultrasonic testing covers interval tables, metal removal per pass and how to time grinding against the 213.237 test.

Grinding has two side effects. Hot grinding can form the white etching layer that later spalls (Zhou 2021), and a grind can remove the surface part of a shallow defect, so the post-grind record can look better than the rail was. Test before grinding as well as after, where the schedule allows.

Key takeaway: grinding is the main control for shelling, but it doesn't replace the internal test.

What should an inspector record?

Record the location, extent and type of surface damage, because the next UT run needs it. Visual signs of shelling, from transit training material that follows the industry defect manuals:

• dark spots irregularly spaced on the gauge side of the running surface
• longitudinal separations at one or more levels in the upper gauge corner, with discoloration from bleeding
• on transposed rail, shelling on the field side with an overhanging lip of metal

Signs of the other conditions:

• Spalling: chipped cavities in the running surface, often at an engine burn
• Flaking: small slivers scaling off the tread near the gauge corner, often on high rails and switch points
• Head checks: fine parallel cracks at an angle across the gauge corner

A track inspector's 213.233 visual inspection does not cover internal defects. Passing the surface record to the rail flaw contractor lets the operator plan hand verification for the shelled curves before the run, instead of discovering them in the data.

Key takeaway: the visual record tells the UT crew where the rail is least testable.

What surface inspection and UT can't tell you

Neither method proves that shelled rail is defect-free. Know these limits before relying on a report:

• Small shells can be invisible. A shell may show no surface evidence until the head is broken or sectioned.
• Visual inspection can't see a detail fracture forming. The TSB said so directly in R99M0046.
• UT can be masked. Detection under shells was about 62% in TD-99-006, and loss-of-bottom cut 70° response to an average of 67% in TD-18-027.
• Masked rail is untested rail. It must be handled under 213.237(h), not counted as clean.
• Growth rates vary. TD-10-004's 20 to 40 MGT for shells is a FAST test-track figure under 39-ton axles. The NTSB's Superior analysis used about 2.5% of head area per MGT for detail fracture growth, and noted that growth is often not linear.
• Worn profiles move the beam. On gauge-face-worn rail a fixed-angle probe can inspect "the wrong volume of material" (TTCI TD-16-030).
• Definitions differ. FRA, railroad manuals and research papers use "spall", "flake" and "shell" slightly differently. Describe what you see, depth included, rather than relying on the label.

Key takeaway: treat shelling as a reason to test harder, not as a finding in itself.

How EMA Industries applies this

EMA Industries performs ultrasonic rail flaw testing for transit agencies, short lines and industrial track owners. It is based in Southern California and mobilizes nationwide. On curves with shelling or head checks, the report should separate verified defects from stretches where loss-of-bottom prevented a valid search, give mileposts for each, and state which hand-verification steps were used. That lets the owner choose between retesting, grinding, a 25 mph restriction or new rail while there is still time inside the interval. To scope testing on shelled track, see our ultrasonic rail testing services or contact EMA Industries.

Related reading: 49 CFR 213.237 rail flaw detection requirements · Detail fracture development in rail · Transverse fissure in rail · False calls in rail UT · Rail grinding and ultrasonic testing

Frequently asked questions

What is the difference between shelling and spalling in rail?

Shelling starts as a fatigue crack about 1/4 to 3/8 inch below the gauge corner, grows parallel to the corner and breaks out as a shell. Spalling is shallow chipping of the running surface, often from engine burns or a brittle white etching layer. Shelling is the bigger testing problem because it is deeper and can hide a detail fracture.

Does FRA require remedial action for shelled rail?

Not for the shell itself. The 49 CFR 213.113 remedial action table has no row for shelling, spalling, flaking or head checks. Remedial action applies once a detail fracture forms, and 213.113(d)(7) says detail fractures may arise from shelled spots, head checks or flaking.

Why does shelling cause missed rail defects?

The shell's crack faces sit where the beam enters the rail, so they reflect and scatter the sound before it reaches the head volume below. TTCI's 1999 trial found transverse defects under shells about 62% of the time, against 77% elsewhere. At Superior, Wisconsin, in 1992, overlapping shelling cracks blocked the sound from detail fractures six weeks before the derailment.

What must a railroad do if shelling prevents a valid rail test?

Under 49 CFR 213.237(h), before the time or tonnage limit expires, the owner must conduct a valid search, limit speed to 25 mph until one is made, or replace the untested rail. The location must also be recorded under 213.241(c)(6). Corrective grinding followed by a retest is the usual way to make a valid search possible.

How often should shelled rail be ground?

Part 213 sets no grinding interval, so the owner decides. TTCI's 1999 survey put the practical minimum preventive interval at about 10 to 15 MGT and suggested aiming for about 25 MGT unless special factors apply. Corrective grinding on already-shelled rail is set by depth: in TD20-002, grinds that removed at least 0.060 inch restored a valid test every time.

Sources

• 49 CFR 213.113, Defective rails, eCFR (text checked on Cornell LII and the 2025 govinfo CFR XML, October 2026)
• 49 CFR 213.237, Inspection of rail, eCFR
• 49 CFR 213.238, Qualified operator, eCFR
• 49 CFR 213.241, Inspection records, eCFR
• Track Safety Standards; Improving Rail Integrity, 79 FR 4234 (January 24, 2014), Federal Register (preamble on surface conditions; remedial action table and notes)
• Szablewski, D., and Robles Hernández, F. C., Shell Defect Analysis Using Scanning Electron Microscopy Techniques, AAR/TTCI Technology Digest TD-10-004, March 2010
• Garcia, G., and Reiff, R., hi-rail rail flaw detection system evaluation, AAR/TTCI Technology Digest TD-99-006, February 1999
• Cummings, S., Testability and Corrective Grinding Depth of Rails with Poor Surface Condition, AAR/TTCI Technology Digest TD-18-027, November 2018
• Cummings, S., Effectiveness of Corrective Grinding, AAR/TTCI Technology Digest TD20-002, April 2020
• Sawley, K., North American Rail Grinding on Curves in Track, AAR/TTCI Technology Digest TD-99-004, January 1999
• Witte, M., and Poudel, A., false indications in rail UT, AAR/TTCI Technology Digest TD-16-030, July 2016
• NTSB Safety Recommendation R-94-1 and -2, Superior, Wisconsin derailment of June 30, 1992; full report NTSB/HZM-94/01 (TRID record)
• TSB Railway Investigation Report R99M0046, CN derailment, Bedford, Nova Scotia, October 9, 1999
• MARTA Track Inspection Training, Section 7: Rail Defects, Part 1, Metropolitan Atlanta Rapid Transit Authority (training material; definitions follow industry defect manuals)
• Zhou, K. et al., Temperature field and material response as a function of rail grinding parameters, International Journal of Heat and Mass Transfer 175, 2021
• Mesaritis, M. et al., Assessment of rail grinding maintenance surface quality and damage propagation in subsequent loading cycles for premium rail grades, Wear 530–531, 2023