Heavy haul rail inspection is the ultrasonic testing of rail that carries heavy axle loads and high annual tonnage. On Class 4 and 5 track, 49 CFR 213.237(c) caps the gap between internal rail tests at 370 days or 30 million gross tons (MGT), whichever is shorter. At 100 MGT a year, tonnage forces a test roughly every 110 days.
That one number explains most of what is different about heavy haul. On a light branch line the calendar sets the test date. On a coal or grain corridor running 286,000 lb cars, the tonnage limit arrives first, rail fatigue damage builds faster, and the surface condition of the rail decides whether an ultrasonic search is even valid. This guide covers what heavy axle loads do to rail, how the federal rules scale with tonnage, how the probes work on worn rail, where grinding and continuous testing fit, and what ultrasonic testing can't do.
Contents: Axle loads and rail damage · Test intervals · Service failure targets · How the probes work · Grinding · Continuous testing · Verification · Newer methods · Limits · FAQ
What do heavy axle loads do to rail?
Heavier axles wear rail faster, start rolling contact fatigue (RCF) sooner and shorten rail fatigue life. The best US evidence comes from the Facility for Accelerated Service Testing (FAST) at the Transportation Technology Center near Pueblo, Colorado, where a train of 315,000 lb cars (39-ton axle loads) has run on a test loop since 1988.
A North American freight car rides on four axles, so its gross rail load divides by four:
| Car gross rail load | Load per axle | Load per wheel | Where it is seen |
|---|---|---|---|
| 263,000 lb | 65,750 lb (about 33 tons) | 32,875 lb | Older standard car weight |
| 286,000 lb | 71,500 lb (about 36 tons) | 35,750 lb | Widely used in North American bulk unit trains |
| 315,000 lb | 78,750 lb (about 39 tons) | 39,375 lb | The FAST heavy axle load (HAL) research train |
Calculated from the gross rail load. "Tons" here are US short tons of 2,000 lb.
Wear and fatigue findings from FAST
FAST results show how sensitive rail is to the vehicles running on it. In TTCI Technology Digest TD-00-024 (LoPresti and Kalay, November 2000), early Phase V running with standard three-piece trucks under 39-ton cars showed "an increase of nearly 600 percent in rail wear" compared with Phase III, when the same cars ran on improved suspension trucks. The truck design, not just the axle load, drove that change.
RCF is the other half of the problem. TTCI reported in TD-13-001 (January 2013) that high-hardness premium rails at FAST had to be ground after 381 MGT "to remove RCF that was severe enough to interfere with ultrasonic rail-flaw detection." In a separate intermediate-hardness rail test (TD-14-010, May 2014), the high rail was removed at 390 MGT because of excessive gage-corner shelling.
Older field data points the same way. A 1983 British Rail and AAR seminar paper by Zarembski, Stone, Wells and Armstrong, Effect of Increasing Axle Loads on Rail Fatigue Life, reported a reduction in rail life of about 40% (measured in MGT) when car loads rose from 200,000 lb to 263,000 lb. That study predates 286,000 lb cars and used data from two mining railroads, so treat it as a direction, not a current rate.
Why heavy axle loads make defects grow faster
Defect growth is driven by tonnage, so a busy line accumulates damage on a shorter calendar. TTCI's shelling study at FAST (TD-10-004, March 2010) found gage-corner shells starting 1/4 to 3/8 in. below the surface of the high rail in 5° and 6° curves, and growing from initiation to critical failure in 20 to 40 MGT under 39-ton axle loads. On a line carrying 100 MGT a year, 20 MGT passes in about 73 days.
FRA made the same point in the 2014 Improving Rail Integrity rule (79 FR 4234). When it lowered the remedial-action threshold for detail fractures and similar defects from 80% to 60% of head area, it cited detail fracture growth of more than 5% of head area per MGT. Those are the defects shelling and head checks breed. Our article on rolling contact fatigue in rail covers how head checks and shells start.
Key takeaway: on heavy haul track, defect growth is counted in MGT, and a crack can go from small to critical inside a single test interval.
How often must heavy haul track be tested?
On Class 4 and 5 track, internal rail tests can be no more than 370 days or 30 MGT apart, whichever is shorter. The same applies to Class 3 track with regularly scheduled passenger trains or that is a hazardous materials route. The rule text in 49 CFR 213.237(c) on eCFR reads:
"Internal rail inspections on Class 4 and 5 track, or Class 3 track with regularly-scheduled passenger trains or that is a hazardous materials route, shall not exceed a time interval of 370 days between inspections or a tonnage interval of 30 million gross tons (mgt) between inspections, whichever is shorter."
Other Class 3 track follows the "whichever is longer" rule: once a calendar year or every 30 MGT, with a 5-year cap. Class 1 and 2 track has no 213.237 interval. The full table and its exceptions are in our guide to 49 CFR 213.237 rail flaw detection.
Worked example: tonnage sets the schedule
Days to reach 30 MGT = 30 ÷ annual MGT × 365. Tests per year = annual MGT ÷ 30.
| Segment (Class 4) | Annual tonnage | Days to reach 30 MGT | Intervals per year | What controls |
|---|---|---|---|---|
| Secondary freight line | 40 MGT | 273.8 days (about 9 months) | 1.33 | Tonnage, but only just |
| Busy main line | 60 MGT | 182.5 days (about 6 months) | 2.00 | Tonnage |
| Heavy haul corridor | 100 MGT | 109.5 days (about 3.6 months) | 3.33 | Tonnage |
| Very dense coal route | 150 MGT | 73.0 days (about 2.4 months) | 5.00 | Tonnage |
| Highest tonnage in TTCI data | 250 MGT | 43.8 days (about 6 weeks) | 8.33 | Tonnage |
The 250 MGT row isn't hypothetical. TTCI's premium insulated joint test (TD-15-003, February 2015) ran on Nebraska lines carrying 200 to 250 and 120 to 150 MGT a year. On those lines the 370-day limit never comes into play.
In practice, 3.33 intervals a year on the 100 MGT corridor means at least 4 test runs a year to stay inside 30 MGT between each. Owners usually plan below the limit, because a test car that slips a week on a 109-day cycle can push a segment past 30 MGT. Our article on how often rail must be tested works through lower-tonnage cases.
Rail testing is separate from visual track inspection
A 213.237 internal test does not replace visual inspection under 49 CFR 213.233. Class 4 and 5 track must be inspected visually twice weekly, at least one calendar day apart. Class 1 to 3 main track that carried more than 10 MGT the previous year moves from weekly to twice weekly. A visual inspector can see a shell or a broken bar but can't see a transverse defect growing inside the head.
Key takeaway: at 60 to 100+ MGT a year, the 30 MGT limit means a rail test every two to six months, not once a year.
What are the 213.237 service failure rate targets?
A service failure is a broken rail caused by a transverse fissure, compound fissure, detail fracture or vertical split head. Under 213.237(a), the owner must test often enough to keep the rate of these failures, per designated rail inspection segment, below a target:
| Track | Maximum service failures per mile per year |
|---|---|
| Class 4 and 5 | 0.10 |
| Class 3, 4 and 5 with regularly scheduled passenger trains or a hazmat route | 0.09 |
| Class 3, 4 and 5 with regularly scheduled passenger trains and a hazmat route | 0.08 |
The rate covers a 12-month period and is calculated within 45 days of the period's end. Paragraph (j) defines the four defects that count, and a hazmat route as track carrying at least 10,000 car loads or intermodal tank loads of hazardous materials a year (or 4,000 loads of the materials listed in 49 CFR 172.820).
Worked example: a 120-mile coal segment
A 120-mile Class 4 segment is also a hazmat route, so its target is 0.09.
- Allowed: 0.09 × 120 = 10.8 failures a year, so 10 complies and 11 does not.
- With 11 service failures: 11 ÷ 120 = 0.0917 per mile per year, over 0.09.
- The same 11 failures on a segment that isn't a hazmat route would meet the 0.10 target (0.10 × 120 = 12).
Missing the target triggers 213.237(d). The owner must tell FRA within 45 days of the end of the 12-month period and may submit an explanation and a remedial action plan. If the target is missed for two consecutive years, the owner must either cut the test interval to 10 MGT or reduce the area with the most service failures to Class 2 until the target is met.
On the 100 MGT corridor, a 10 MGT interval means 10 test runs a year, one every 36.5 days. The Class 2 option limits freight to 25 mph. On a heavy haul line, either is far more expensive than testing well in the first place, which is why the failure rate, not the 30 MGT cap, is the real performance standard. For national context on how often broken rails cause derailments, see our FRA rail failure statistics.
Key takeaway: the 30 MGT limit is the floor; the service failure rate tells the owner whether that floor is enough.
How does heavy haul rail inspection find internal defects?
A rail test vehicle sends ultrasound into the rail from several angles at once, because each defect lies in a different plane. Most systems use fluid-filled wheel probes (roller search units) that roll on the railhead. Transducers inside the tire fire through the fluid and tire wall into the steel. A water or water-glycol spray on the railhead couples the tire to the rail.
Coupling is the whole point of that spray. Ultrasound at the usual rail test frequencies (FRA's 2014 rule notes they are "normally" 2.25 to 5.0 MHz) does not cross an air gap. At a steel-to-air boundary about 99.996% of the sound is reflected, so any dry patch, grease smear or lift-off over a battered weld leaves that stretch untested. A 1999 Colorado State analysis for the Mountain-Plains Consortium (MPC 99-106) described wheel probes carrying 3 to 6 transducers per wheel, with 0°, 45°, 70° and side-looking beams.
| Beam | Wave type | What it is aimed at | Typical targets |
|---|---|---|---|
| 0° (straight beam) | Longitudinal | Straight down through head, web and base; watches the base echo | Horizontal split heads, head-web separation, piped rail, loss of base echo |
| 37° to 45° | Shear | Down and along the web, toward bolt holes and the web-base area | Bolt hole cracks, web defects |
| 70° forward and backward | Shear | Angled down into the head along the rail | Transverse fissures, detail fractures, the transverse part of compound fissures |
| Side-looking or skewed 70° | Shear | Toward the gage corner at an angle to the rail axis | Detail fractures under the gage corner, vertical split heads |
Exact probe sets vary by system. A Level III should confirm the angles in any system specification.
Getting the 70° geometry right
The 70° beam is measured from the vertical, the normal to the running surface. It runs down into the head at 20° below horizontal and along the rail, forward in one channel and backward in another. A transverse defect stands roughly vertical across the head, so the beam strikes its face within about 20° of head-on and reflects strongly. The beam does not travel parallel to the running surface; it enters from the top and angles down through the head.
Worked geometry: a transverse defect 0.75 in. below the running surface is reached at a sound path of 0.75 ÷ cos 70° = 2.19 in., at a surface distance of 2.19 × sin 70° = 2.06 in. ahead of the probe's beam entry point.
Inside a wheel probe, the beam angle is set by refraction from the fluid into the steel. With a fluid velocity of 1,480 m/s and a shear velocity in steel of 3,250 m/s, a 70° shear beam needs an incidence of about 25.3°.
On a worn railhead the surface the beam enters is no longer flat. A 2° tilt of the running surface one way drops the refracted angle to about 60°; a 2° tilt the other way passes the second critical angle (about 27.1°), and no shear wave enters the rail. TTCI warned in TD-16-030 that on worn rail "the wrong volume of material is inspected." Heavy haul curves, with their deep gage-face wear and field-side metal flow, are where this matters most.
Key takeaway: the 70° beam goes down and along the head, and worn rail profiles bend it away from where the operator thinks it is looking.
How does rail grinding interact with ultrasonic testing?
Grinding removes the shallow RCF damage that scatters ultrasound, so it often turns an invalid search into a valid one. Part 213 doesn't require grinding, and grinding earns no credit toward a 213.237 test, but on heavy haul lines it is part of keeping the rail testable.
TTCI Technology Digest TD-08-041 (October 2008) explained why RCF matters for inspection:
"…the poor rail surface conditions can diffuse the ultrasonic signal that rail flaw detection vehicles use to find rail defects…"
| Finding | Source | What it means for testing |
|---|---|---|
| Corrective grinding restored a valid UT test for 68 of 88 exceptions (77%) | TTCI TD-20-002 | Most no-test stretches can be recovered by grinding |
| 100% restored where at least 0.060 in. was removed | TTCI TD-20-002 | About 8 to 10 production passes at 0.006 to 0.008 in. per pass (calculated) |
| 90% of 880 RCF cracks and spalls were 0.040 in. deep or less | TTCI TD-18-027 (2018) | Shallow damage causes most low-confidence tests |
| Typical preventive grinding cycle about 25 MGT | TTCI digests and industry practice (hedged) | On a 100 MGT line, about 4 grinding cycles a year |
| FAST high rail removed at 390 MGT for shelling, with minor wear attributed partly to "the lack of grinding" | TTCI TD-14-010 (2014) | Skipping grinding trades wear life for fatigue damage |
On a 100 MGT corridor, a 25 MGT grinding cycle and a 30 MGT test limit run at about the same pace. Running the test car after a grinding pass, rather than just before one, gives the 70° probes a cleaner surface to couple through. Grinding has costs: it removes metal that would otherwise be service life, and poor practice can leave grinding burns. Our article on rail grinding and ultrasonic testing covers sequencing and the grinding standards in detail.
Key takeaway: on high-tonnage curves, a test without a grinding plan often produces no-test stretches, not data.
When does continuous rail testing under 213.240 make sense?
Continuous testing lets the test car keep running and field-verify suspect locations afterward, within 84 hours or 36 hours for the most serious suspects. It is an option under 49 CFR 213.240, added in 2020, and it can satisfy 213.237.
In a traditional stop-and-verify test, the car stops at each indication, the operator confirms it by hand, and the defect is marked before the car moves on. That ties up the track. The 1999 MPC 99-106 analysis found test cars capable of about 30 mph non-stop but averaging only 6 to 8 mph with stop-and-verify. On a corridor with trains every few minutes, track time is the scarcest resource.
| Stop-and-verify | Continuous (213.240) | |
|---|---|---|
| Verification | At the indication, before the car moves on | Field verification later by a 213.238-qualified person |
| Time limit to verify | 213.113(b): within 4 hours, or immediately for suspected A, A2 and B defects | 84 hours; 36 hours for suspects that would need remedial action A, A2 or B |
| 213.113(b) applies? | Yes | No: the owner is exempt from 213.113(b), but all other parts of 213.113 apply |
| Data review | Operator at the screen | Data interpreted by a separately qualified person; suspects listed so each can be located and verified |
| Unverified suspect | Not applicable | Gets the most restrictive remedial action until verified |
| Extra reporting | None | Annual report to FRA under 213.240(g) |
The owner must designate and record which type of test is being run before it starts (213.240(c)). A suspect location is not a defect until a 213.238-qualified person field-verifies it. The trade-off is that a suspect the car found may carry traffic for up to 84 hours before anyone puts hands on it, so the data review and the prioritization of suspects must be good.
Key takeaway: continuous testing buys track time on high-tonnage lines, in exchange for a stricter data and verification process.
What happens after a heavy haul test finds a defect?
Under 49 CFR 213.113(b), the owner must verify an indication within four hours, or immediately if the defect would require remedial action A, A2 or B. That rule applies unless the owner is running a continuous test under 213.240. The text in 49 CFR 213.113 on eCFR reads, in part:
"…the track owner must verify the indication within four hours, unless … a defect that requires remedial action A, A2, or B … in which case the track owner must immediately verify the indication."
The four-hour window is a 213.113 defect-response rule, not a 213.237 test interval, and it doesn't make the test schedule itself shorter.
Once verified, the remedial action table sets the response by defect type and size. For a transverse fissure or detail fracture, 5 to under 25% of head area is action C (joint bars within 10 days, with speed limits until then), 25 to under 60% is D, 60 to under 100% is A2 or E and H, and 100% is A or E and H. Remedial action A means a designated person visually supervises each movement over the rail. The 213.237(e) marking rule applies to every defective rail. Records go under 213.241 and are kept at least two years after the inspection and one year after the initial remedial action.
Are phased array, EMAT and onboard analytics ready for heavy haul?
They are promising, but in North American revenue testing they remain verification tools or research. The probe set described above is still the basis of 213.237 testing.
Phased array and FMC/TFM
Phased array ultrasonic testing (PAUT) steers and focuses a beam electronically across many angles from one multi-element probe. In rail it is used mainly for hand verification of indications and in development systems. Full matrix capture with the total focusing method (FMC/TFM) records every element pair and builds an image in software. Published rail work on FMC is research. The EU-funded SAFTInspect project, for example, developed an FMC array for inspecting high-manganese crossings from the top surface (CORDIS project 314949). We found no source showing FMC/TFM in routine revenue-car testing. PAUT sizing also needs care: ASTM E2700 warns that phased array can oversize.
EMAT
Electromagnetic acoustic transducers (EMATs) generate ultrasound inside a conductive part with a magnet and a coil, so they need no liquid couplant. ASTM E1774 lists the trade-offs: EMATs suit moving, hot or rough parts, but they have "very low efficiency," with insertion losses of 40 dB or more. Railhead EMAT work published to date is mainly lab research. EMAT does not replace the 0°/45°/70° probe set for a 213.237 search.
Onboard analytics: the Shuohuang case
China's heavy haul coal lines report gains from combined inspection vehicles. Kong, Li, Ding, Qi and Wang, writing in Railway Sciences in November 2025 (doi.org/10.1108/RS-09-2025-0043), describe an integrated inspection vehicle first used in 2014 and an "intelligent" version in 2024 that raised detection precision by 30%. They report annual track defects falling from 25,000 to 3,800 and maintenance savings of more than 40 million yuan a year on the Shuohuang Railway. These are one operator's results for all track defects, reported by its own staff, not an independent study of rail flaw detection. They show what integrated data can do; they don't prove the same return on a US line.
What heavy haul rail inspection can't do
Ultrasonic testing can miss defects hidden under surface damage, in closed cracks, in the rail base and in worn profiles. Stating these limits is part of a good test program.
| Limit | Why it happens | What helps |
|---|---|---|
| Masking by shells and head checks | Surface damage scatters or blocks the beam, so a detail fracture below it returns no echo. In TTCI's 1998 hi-rail benchmark (TD-99-006, 1999), transverse defects under shells were found about 62% of the time versus 77% elsewhere | Grinding; recording the stretch as untested; hand testing |
| Closed or tight cracks | Crack faces pressed together by residual or wheel stress transmit some sound instead of reflecting it, so the echo is weaker | Multiple angles; repeat tests; not relying on one channel |
| Rail base | Test vehicles scan the head and web. TTCI TD-00-013 (Kristan, 2000): the process scans "not the majority of the base" | Hand UT of field welds and the base where risk is high |
| Worn rail profiles | Refraction changes on a tilted or flattened head (above) | Profile-aware setups; grinding to a target profile |
| Coupling loss | Ultrasound does not cross air; grease, frost and battered welds lift the probe | Coupling monitors; recording untested track under 213.241(c)(6) |
| Detection is not perfect | The same benchmark found 72% of 49 known defects overall, and 33% of bolt hole cracks | Valid-search checks, qualified operators under 213.238, service failure tracking |
A hi-rail pickup is a light load by comparison. Dealer listings for hi-rail F-350 trucks give gross vehicle weight ratings of 10,000 to 14,000 lb, about 5 to 7 tons for the whole truck, against roughly 36 tons on a single axle of a 286,000 lb car. Whatever the test vehicle, crack closure is a known limit, not something a heavier test car is proven to solve.
Several of these figures are from TTCI research in 1998 to 2000 and describe equipment of that time. Current systems may perform better, but no public source gives a current national detection rate. FRA's 2011 review of rolling contact fatigue (DOT/FRA/ORD-11/24, Magel) concluded that "current best practice is to combine nondestructive evaluation with preventative rail grinding." Where a search is compromised, 213.237(g) says that stretch has not been inspected. Before the deadline the owner must retest it, run trains over it at no more than 25 mph, or replace the rail.
Who carries the responsibility for heavy haul rail testing?
The track owner is responsible for compliance, even when a contractor runs the test. Under 49 CFR 213.5, contractors who perform Part 213 functions must do so as the part requires, but FRA can hold the owner, the assignee or both responsible. Each operator of rail flaw detection equipment must be qualified under 49 CFR 213.238: a documented training program, written employer authorization, at least 160 hours of supervised experience and periodic reevaluation. FRA doesn't certify operators; the employer qualifies them.
FRA inspectors audit the owner's records, service failure rates and test intervals.
How EMA Industries applies this
EMA Industries performs ultrasonic rail flaw testing and rail weld inspection for transit agencies, short lines and industrial track owners from Southern California, and mobilizes nationwide. On high-tonnage track, we scope a test program from the segment's annual MGT, its service failure history and its grinding cycle, and we record every untested stretch so the owner can act on it under 213.237(g). Hand verification of car indications and UT of field welds and rail base are the areas where our work most often supports a heavy haul program. To plan testing for a segment, see our ultrasonic rail testing services or contact us.
Related reading: 49 CFR 213.237 rail flaw detection · How often must rail be tested? · Rail grinding and ultrasonic testing · Rolling contact fatigue in rail · Rail shelling and spalling · Detail fracture development
Frequently asked questions
How often must heavy haul track be ultrasonically tested?
On Class 4 and 5 track, at least every 370 days or 30 MGT, whichever comes first, under 49 CFR 213.237(c). At 100 MGT a year that is about every 110 days. The owner must also test often enough to meet the 213.237(a) service failure targets, so busy segments are often tested more often than the minimum.
Why do heavy axle loads make rail defects grow faster?
Higher wheel loads raise contact stress at the gage corner, which speeds up wear and rolling contact fatigue. At FAST, gage-corner shells went from initiation to critical failure in 20 to 40 MGT under 39-ton axle loads. Because damage accumulates per MGT, a high-tonnage line reaches that point in weeks or months.
Does rail grinding count as a rail flaw test?
Grinding is not an internal rail inspection and earns no credit under 213.237. It removes surface damage so the next ultrasonic test can see into the head. TTCI found that corrective grinding restored a valid test in 77% of 88 no-test locations.
What is the difference between continuous and stop-and-verify rail testing?
In stop-and-verify testing the car stops to confirm each indication, and 213.113(b) requires verification within four hours, or immediately for suspected A, A2 and B defects. In continuous testing under 213.240 the car keeps running, and a qualified person field-verifies suspects within 84 hours, or 36 hours for the most serious ones.
Can phased array or EMAT replace standard wheel probes on heavy haul track?
Not yet. In North American revenue testing, phased array is used mainly for hand verification, and EMAT and FMC/TFM rail work is research. The 0°, 45° and 70° wheel-probe set remains the basis of a 213.237 search.
Sources
- 49 CFR 213.237, Inspection of rail, eCFR (text read on Cornell LII, October 2026)
- 49 CFR 213.240, Continuous rail testing, eCFR
- 49 CFR 213.113, Defective rails, eCFR
- 49 CFR 213.233, Track inspections, 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 via govinfo
- TTCI TD-00-024, FAST Heavy Axle Load Testing Update (LoPresti and Kalay, November 2000), MxV Rail
- TTCI TD-08-041, Premium Rail Performance Under 39-Ton Axle Loads at FAST (October 2008), MxV Rail
- TTCI TD-10-004, shelling at FAST (Szablewski and Robles Hernández, March 2010), MxV Rail
- TTCI TD-13-001, 2012 Results at FAST (January 2013), MxV Rail
- TTCI TD-14-010, intermediate strength rail test at FAST (May 2014), MxV Rail
- TTCI TD-20-002, grinding to restore ultrasonic tests, MxV Rail
- TTCI TD-18-027, RCF and ultrasonic test confidence (Cummings, November 2018), MxV Rail
- TTCI TD-99-006, Benchmarking Results of North American Rail-Flaw Detection Technologies (Garcia and Reiff, February 1999), MxV Rail
- TTCI TD-00-013, thermite weld performance (Kristan, July 2000), MxV Rail
- DOT/FRA/ORD-11/24, Rolling Contact Fatigue: A Comprehensive Review (Magel, 2011), FRA via ROSA P
- MPC 99-106, Assessment of Rail Flaw Inspection Data (Jeffrey and Peterson, 1999), Mountain-Plains Consortium
- Zarembski, Stone, Wells and Armstrong, Effect of Increasing Axle Loads on Rail Fatigue Life (1983), University of Delaware repository (abstract)
- Kong et al., Current status and trends in the development of inspection technologies and equipment for heavy-haul railway infrastructure, Railway Sciences 4(6), 2025
- SAFTInspect project record, CORDIS, European Commission
- ASTM E1774, Standard Guide for Electromagnetic Acoustic Transducers, ASTM store page