Rail Grinding and Its Effect on Ultrasonic Inspection

Rail grinding removes a thin layer of steel from the rail head, typically 0.006 to 0.008 inch per pass of a large production grinder at normal speed (AAR/TTCI, 2018). For ultrasonic inspection it cuts both ways: it clears the shelling and head checks that block the beam, but it also changes the surface and profile the probes ride on.

That matters because a test only counts if it is a valid search. Under 49 CFR 213.237(g) on eCFR, rail the qualified operator could not search properly "may not be considered as internally inspected". Grinding is often what turns an untestable stretch back into a testable one.

This guide covers how much metal grinding removes, how preventive and corrective cycles differ, what grinding does to probe coupling and beam entry, whether to test before or after the grinder, and how to fit both into 213.237 intervals.

Contents: Metal removal · Preventive vs corrective · Coupling · Masking · Before or after · Scheduling · Limits · FAQ

How much metal does rail grinding remove?

A production grinder removes roughly 0.006 to 0.008 inch (0.15 to 0.20 mm) per pass at typical travel speeds, and up to about 0.014 inch (0.36 mm) per pass at slow speeds. Those figures come from pre- and post-grind profile comparisons in AAR/TTCI Technology Digest TD-18-027 (2018), repeated in TD20-002 (2020).

The total removed in one grinding cycle depends on the number of passes, the stone count, motor power, travel speed and rail hardness. A 2003 best-practice draft for AREMA Committee 4 by Peter Sroba of the National Research Council of Canada gave these typical preventive targets per cycle for premium rail (340 to 420 BHN):

Source: Sroba, Rail Grinding Best Practice (draft, 2003; figures "in 2002"), hosted by Loram, a grinding contractor. The draft says to double most of these figures for standard rail of 260 to 320 BHN, and lists 10 MGT cycles with 0.004 to 0.012 inch removal for passenger track. Treat them as one engineer's 2002 summary of North American practice, not a specification.

Corrective grinding removes far more. In TD20-002, corrective grinding on 88 rail exceptions ranged from 1 to 17 passes and an estimated 0.006 to 0.157 inch of metal at the center of the head.

Key takeaway: one pass removes thousandths of an inch; clearing deep fatigue damage takes several passes and tenths of an inch of head height.

What is the difference between preventive and corrective rail grinding?

Preventive grinding is frequent, light grinding (ideally a single pass) to stop surface cracks before they grow. Corrective grinding is less frequent, uses multiple passes, and removes damage that has already formed. AAR/TTCI's survey of North American railroads (TD-99-004, 1999) of seven railroads found that most used, or wished to use, preventive grinding, and that the minimum interval was in the range of 10 to 15 MGT.

The 2003 Sroba draft listed typical preventive cycles for premium rail:

Test-track results support these ranges but show they are site-specific. At FAST, a joint AAR and FRA program, the preventive zone of a 6° curve with standard rail was ground every 17.5 MGT, later extended to 25 MGT, and had no fatigue defects after 292 MGT (TD-01-020, 2001). It also lost more total metal than the unground zone. TD-99-004 concluded that no single interval fits every curve, but suggested an aimed-for minimum of about 25 MGT "unless special factors apply".

Key takeaway: for planning, assume preventive cycles of roughly 15 to 25 MGT in sharp curves and much longer on tangent track, then adjust from your own pre- and post-grind surveys.

How does rail grinding change probe coupling and beam entry?

Grinding changes three things a rail UT probe depends on: surface finish, transverse profile and the position of the running band. Each can raise or lower the signal reaching the steel.

The Sroba draft gives sample finish limits that railroads accepted after preventive grinding: facet widths of about 0.16 to 0.47 inch (4 to 12 mm) depending on position, and an average surface roughness of 10 to 12 microns on passenger track. Low-rail crowns are commonly reground to a radius of about 7 to 10 inches (TD-99-004).

Coupling is the main reason a freshly ground surface usually tests better than a damaged one. Couplant only works if it fills the gap between the probe and the rail; see our guide to ultrasonic couplant for rail inspection. On corrugated rail the probe rides over a wave of peaks and troughs, which is why grinding out corrugation can improve test quality. The wavelengths and frequencies involved are covered in corrugation on rail.

A changed profile is a setup issue, not a defect. When a rail has been reprofiled, the operator should check probe alignment against the new running band and confirm system sensitivity before relying on the results; see ultrasonic testing calibration for rail inspection.

Key takeaway: a ground surface usually couples better, but swarf, facets and a moved contact band can still cost signal.

Why does surface damage mask internal defects from UT?

Shelling, head checks, spalls and deep RCF cracks scatter the ultrasonic beam at the running surface, so less energy reaches the defects inside the head and less returns to the probe. The detector car sees this as a weak or missing back-wall signal. AAR/TTCI describes it this way in TD-18-027:

"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. This is commonly referred to as 'loss-of-bottom.'"

This matters most for the detail fracture, which 49 CFR 213.113(d) defines as "a progressive fracture originating at or near the surface of the rail head", arising from shelling, head checks or flaking. The same surface damage that starts the defect can hide it from the 70° probes meant to find it. More on the surface conditions themselves: rail shelling and spalling, head checks and rolling contact fatigue.

How much signal is lost? TTCI drilled 1/8-inch holes in five rails pulled from service for low-confidence UT and compared them with the same hole in a clean control rail. Using a 70° hand probe, the damaged rails returned on average 67% of the control response, and never less than 15%. Optical microscopy of 880 cracks and spalls in those rails found that 90% were no deeper than 0.040 inch (TD-18-027).

Worked example: passes needed to clear typical surface damage

Using the TTCI figures, how many passes would remove damage 0.040 inch deep?

That matches TTCI's own conclusion that three to five production-grinder passes would likely have improved UT confidence on those rails. It is an estimate at the center of the head; gage-corner damage and the remaining head height decide what is practical.

Two more TTCI findings change how a field team should think about it:

• Not every crack has to go. TD20-002 found that "complete elimination of all cracks is not necessary for acceptable ultrasonic testing (UT) response in rail."
• Deeper grinds succeed more often. Corrective grinding restored a valid test at 77% of sites that lost at least 0.005 inch, 88% at 0.020 inch, and 100% of sites where at least 0.060 inch was removed. TTCI declined to set a blanket depth because the data varied widely.

Key takeaway: grinding is how most loss-of-bottom stretches are made testable again, but it is a depth question, and one light pass may not be enough.

Should rail be tested before or after grinding?

Neither order is always right: test before grinding to record what is in the rail and to target corrective work; test after grinding when the goal is a valid search on rail that previously could not be searched. Part 213 does not mention grinding and sets no order, so the sequence belongs in the owner's procedures.

Two practical points follow:

1. Grinding can uncover damage. In TD20-002, observers saw that removing open spalls sometimes exposed spalls "previously hidden beneath the surface". A post-grind test, or a visual check between passes, is the only way to know where the rail ended up.
2. Keep the records together. A drop in indications after grinding may be a real removal of defects, not a loss of sensitivity. Comparing test records without grinding dates, pass counts and locations invites the wrong conclusion. Unexplained or non-repeating indications are covered in false calls in rail UT.

TTCI's practice in its case studies was to inspect before, during and after grinding, with photographs and cross-section profiles. For grinding works let under European contracts, acceptance of the finished rail (longitudinal and transverse profile and related measurements) is set by EN 13231-2:2020, Acceptance of reprofiling rails in plain line, switches, crossings and expansion devices. It replaced EN 13231-3:2012 and EN 13231-4:2013, which were withdrawn. A grinding acceptance check confirms the profile; it is not an internal rail inspection.

Key takeaway: if a stretch failed its last valid search because of surface damage, grind first and test second; otherwise, keep a pre-grind test record so changes can be explained.

How do rail grinding cycles fit 49 CFR 213.237 intervals?

Grinding never counts as an internal rail inspection, and it does not restart the 213.237 clock; only a valid search does. On Class 4 and 5 track, and on Class 3 track with scheduled passenger trains or a hazmat route, internal tests can be no more than 370 days or 30 MGT apart, whichever is shorter (49 CFR 213.237(c)). The full interval rules are explained in our guide to 49 CFR 213.237.

If the operator decides a valid search was not made, 213.237(h) gives the owner three options, which must be completed before the time or tonnage limit runs out:

1. conduct a valid search;
2. reduce speed to a maximum of 25 mph until a valid search can be made; or
3. replace the rail that was not inspected.

Corrective grinding is the usual way to make option 1 possible. That only works if the grinder can get to the site, and the retest can be run, before the deadline.

Worked example: aligning a grinder cycle with a 30 MGT test interval

A Class 4 freight line carries 40 MGT a year. Its sharp curves are on a 25 MGT preventive grinding cycle. The last valid search was at 0 MGT.

Running the test shortly after the grinder gives the probes a clean surface and leaves about 3 to 4 MGT, or roughly four to five weeks, to regrind and retest any stretch that still fails before the 30 MGT limit. Testing at 29.5 MGT would leave almost no margin, and an invalid search at that point leaves only the 25 mph restriction or rail replacement.

On a lighter line the calendar controls instead. A Class 4 line at 12 MGT a year reaches 30 MGT after 2.5 years, so the test is due at 370 days, while a 25 MGT grinding cycle would come round only about every two years. Grinding and testing will then fall at different times, and the pre-grind test record matters more.

The rule does not tell you when to grind; it tells you when the test must be valid. The 213.237(a) service failure rate (for example, 0.10 per mile per year on Class 4 and 5 track) is the outcome measure, and rail breaks from detail fractures count toward it.

Key takeaway: schedule the grinder ahead of the detector car, with enough tonnage margin left for one regrind and retest before the 213.237 limit.

Can rail grinding damage the rail or create false indications?

Yes, if the grinding runs too hot. Excess heat can burn the surface and form a white etching layer (WEL), a hard surface layer that researchers attribute to martensite formation. A rail grinding experiment with thermocouples (Zhou et al., International Journal of Heat and Mass Transfer, 2021) found that "with a grinding temperature of around 400 °C, a WEL starts to appear on the rail surface."

WEL matters because it can crack. Rolling-sliding tests on ground premium rail samples (Mesaritis et al., Wear, 2023) reported "a correlation between White Etching Layer and the formation of cracks and defects", and harder grades retained more WEL.

For UT, a shallow burn layer is a surface condition, not an internal reflector. Fine cracks in WEL are surface features, but they can scatter sound and weaken coupling like any other surface damage. Grind marks, facets and burns should be checked visually and recorded with the grinding data, so the test operator can tell surface effects from a real indication.

Key takeaway: good grinding practice limits heat; a burned or cracked ground surface is something to note at the next test, not something the test can measure.

What rail grinding and UT cannot do

How EMA Industries applies this

EMA Industries performs ultrasonic rail flaw testing for transit agencies, short lines and industrial track owners, from Southern California and mobilizing nationwide. When a segment has recently been ground, or has stretches that failed their last valid search, the test plan should record grinding dates and pass counts, list every stretch that could not be validly searched, and leave time inside the 213.237 window for a retest. To plan testing around a grinding program, see our ultrasonic rail testing services or contact us.

Related reading: How ultrasonic rail flaw detection works · Detail fracture development in rail · Hi-rail truck vs inspection cart

Frequently asked questions

Does rail grinding affect ultrasonic testing?

It usually improves it. Grinding removes the shelling, spalls and RCF cracks that scatter the beam and cause loss-of-bottom. It also changes the profile and surface finish, so probe alignment, coupling and swarf need checking before the results are relied on.

How much metal does one grinding pass remove?

About 0.006 to 0.008 inch per pass for a large production grinder at typical speed, and up to about 0.014 inch at slow speed, based on AAR/TTCI profile measurements published in 2018. A preventive cycle is often one pass; a corrective visit can be many passes and more than 0.1 inch.

Does grinding remove rail cracks?

It removes shallow surface cracks. TTCI found that 90% of the cracks and spalls in its sample rails were no deeper than 0.040 inch, within reach of a few passes. Internal defects such as transverse fissures, and fractures that have already grown deep into the head, stay in the rail and still need ultrasonic testing.

Should ultrasonic testing be done before or after grinding?

It depends on the goal. Test after grinding to get a valid search on rail that surface damage previously masked, and before grinding to record a baseline and target corrective work. Either way, keep the grinding records with the test records.

Does rail grinding reset the 49 CFR 213.237 test interval?

No. Only a valid search for internal defects restarts the 370-day or 30 MGT clock. Grinding can make that valid search possible, but the test still has to be completed before the limit expires, or the owner must reduce speed to 25 mph or replace the rail.

Sources

• 49 CFR 213.237, Inspection of rail, eCFR (text checked on Cornell LII, October 2026)
• 49 CFR 213.238, Qualified operator, eCFR
• 49 CFR 213.113, Defective rails, eCFR
• 49 CFR 213.240, Continuous rail testing, eCFR
• 49 CFR Part 213, 2024 annual edition, govinfo (full-text check that Part 213 does not mention grinding)
• 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
• LoPresti, J., Grinding Tests on Standard Rail in a 6-degree Curve at FAST, AAR/TTCI Technology Digest TD-01-020, October 2001
• Sroba, P., Rail Grinding Best Practice, draft for AREMA Committee 4, National Research Council of Canada, 2003 (hosted by Loram)
• EN 13231-2:2020, Railway applications – Track – Acceptance of works – Part 2, EVS catalog listing; see also BSI Knowledge
• 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