Grade crossing rail inspection is the testing of rail where a road crosses the track at grade. Under 49 CFR 213.237, crossing rail is internally tested on the surrounding track's schedule, such as 370 days or 30 MGT on Class 4 track. A test car scanning from the rail head cannot see web corrosion hidden under the crossing surface.
That gap is not theoretical. In 2019 a VIA Rail passenger train derailed at the Lakeville Road crossing in New Brunswick after a corroded rail web broke under the train. The rail had passed an ultrasonic test about three months earlier.
This guide explains why ultrasonic testing (UT) coverage is limited at road crossings, which Part 213 rules apply, what the Lakeville Road investigation found, and what a track owner should put in a testing specification so crossings are not the weak link in the program.
What is grade crossing rail inspection?
It is the combined visual, ultrasonic and, when the surface is lifted, direct examination of running rail inside a road crossing. The crossing structure (timber planks, rubber or concrete panels, asphalt, flangeway fillers) usually leaves only the rail head exposed. The web and base sit below the road surface.
A road crossing therefore differs from open track in three ways that matter for inspection:
| Factor | Open track | Inside a road crossing |
|---|---|---|
| Rail visible to a track inspector | Head, web, base, fastenings | Usually the head only |
| Rail head surface for UT coupling | Normally clean | Often coated with road grit, sand, mud or grease |
| Corrosion environment | Open atmosphere, drains and dries | Panels hold moisture, road salt and debris against the web and base |
| Typical way to see the web and base | Walk the track | Remove the crossing surface |
Key takeaway: at a road crossing, the inspector and the test car both work through a narrow window at the top of the rail.
A note on terms
"Crossing" means different things in railroad usage, and articles often mix them up:
- A highway-rail grade crossing (a "level crossing" in UK and Canadian usage) is where a road crosses the track at the same level. This article is about these.
- A track crossing or diamond is where two tracks cross. In 49 CFR 213.235, "track crossing" sits in the same list as switches and turnouts.
- In UK usage, a crossing can also mean the cast frog (crossing nose) in a turnout.
Research on cast manganese "crossings" is mostly about frogs and diamonds, not road crossings. It still matters when a turnout or diamond sits inside or beside a road crossing, which happens in yards, street running and industrial track.
What happened at Lakeville Road (TSB R19M0018)?
A corroded rail web, hidden by the crossing surface, broke under a passenger train. The Transportation Safety Board of Canada (TSB) published its investigation report R19M0018 on 30 September 2020. The key facts from the report:
| Item | Finding in TSB R19M0018 |
|---|---|
| Date and train | 4 April 2019, VIA Rail train 14, about 60 mph |
| Location | Lakeville Road public crossing, Mile 15.27, CN Newcastle Subdivision, New Brunswick |
| Outcome | Last 2 cars derailed upright; 3 passengers with minor injuries |
| Crossing construction | Lumber planks, 100-pound rail, rubber mud guard and asphalt surface; about 100 ft from a 2-lane highway that is frequently salted and sanded in winter |
| Last inspections | Hi-rail visual inspection the same morning (no defects); ultrasonic rail flaw test on 28 December 2018 (no internal defects noted); geometry test on 4 October 2018 (no anomalies) |
| Rail condition | About 119 in. of rail head missing at the point of derailment; web corroded to about 0.16 in. (4 mm) at its thinnest, against an original 0.56 in. (14.28 mm) |
| Lab findings | No pre-existing internal defects; overstress fracture surfaces with no signs of fatigue; corrosion scale contained sodium, chlorine and other elements found in road salt and sand |
| Rail age | In service for about 30 years at a low-tonnage crossing |
The TSB's finding as to causes is direct:
"The corrosion of the rail web at the crossing had not been identified by the visual or ultrasonic track inspections, because the web of the rail was hidden by the crossing surface." — TSB, Rail transportation safety investigation report R19M0018 (2020), Findings 3.1
The report also explains why low-tonnage crossings are exposed. On busy lines, rail at crossings tends to wear out and get replaced before corrosion weakens it. On light-tonnage lines the same rail can stay in place for decades. The TSB noted about 6,500 federally regulated crossings in Canada on track averaging 1.1 million gross tons a year or less.
Worked example: how much web was left?
Using the TSB's own measurements:
| Location on the broken rail | Web thickness | Share of original 0.56 in. (14.28 mm) left |
|---|---|---|
| Original section | 14.28 mm | 100% |
| Eastern end of the broken length | about 11 mm | 11 ÷ 14.28 ≈ 77% |
| Western end | about 7 mm | 7 ÷ 14.28 ≈ 49% |
| Thinnest point, near the center of the crossing | about 4 mm | 4 ÷ 14.28 ≈ 28% |
Two points follow. First, the loss was worst near the middle of the crossing, the part that is hardest to see and reach. Second, the web lost more than 70% of its thickness without producing any internal crack for a flaw detector to find. The rail failed by overload of a thin section, not by a fatigue defect growing from the head.
Key takeaway: a clean UT result at a crossing tells you the head had no reportable internal flaw. It doesn't tell you how much web is left.
Why is UT coverage limited under crossing surfaces?
Rail flaw UT sends sound down from the rail head, and it is built to find cracks, not to measure wall loss on the sides of the web. The TSB report puts it plainly: rail flaw inspection "is not currently designed to identify rail web and base corrosion at crossings".
The physics and the field conditions both work against it.
The beams are aimed at cracks in the head
A typical rail test carriage uses a set of probes on the running surface. Exact configurations vary by supplier, but they typically include:
- a 70° shear-wave probe for transverse defects in the head, such as detail fractures and transverse fissures
- a 0° straight beam for horizontal defects and loss of back-wall echo, such as horizontal split heads and some web defects
- side-looking or 37–45° probes for the web and bolt-hole area
Corrosion on the outside faces of the web is a different kind of target. It removes metal gradually from a surface that runs parallel to the 0° beam. A straight beam travelling down the web toward the base can still return a base echo from a thinned web, because the sound path down the center of the web stays intact. The flaw detector sees no crack and may show no loss of back-wall, so the rail is reported clear. (Level III to confirm this description against the beam sets used on the owner's track.)
The TSB adds that UT "is less effective at detecting defects located deeper in the web or in the base of the rail".
The coupling surface is dirty
UT needs clean, continuous contact (couplant) between the probe and the rail head. At road crossings the rail head often carries sand, grit, mud, de-icing residue and grease from road traffic. The TSB notes that grease or dirt on the rail head, "which is common at crossings", can affect detectability. Poor coupling weakens or interrupts the signal, and over a long crossing it can leave stretches of rail with no valid test data at all.
The web and base can't be reached from the side
A thickness reading on the web needs a probe placed on the side of the web. Inside a crossing, the web faces sit behind panels, planks or pavement. The TSB states that current UT "cannot be used to detect material loss in the web of the rail at a crossing without removing the crossing surface". It also says UT could detect web material loss by repositioning the scanners, for example where one side of the web is accessible, but is not designed to do so today.
The material may not be ordinary rail steel
If a turnout frog or diamond sits in or next to the crossing, the cast austenitic manganese steel in those parts is coarse-grained and scatters sound. An FRA-sponsored study (Research Results RR 22-25, September 2022) reports that revenue-service examination of these special trackwork components currently uses only visual inspection. TWI's RailSAFT project reached a similar conclusion for UK cast crossings.
| Limitation | Effect on a crossing test | What helps |
|---|---|---|
| Beams aimed at head cracks | Web and base wall loss not reported | Direct thickness measurement when the web is exposed |
| Contaminated rail head | Weak signals or no-test stretches | Clean the head; hand-test the crossing |
| Panels cover the web | No side access for a probe | Lift the surface at rehabilitation; guided wave from outside the crossing |
| Cast manganese frogs or diamonds | High attenuation, noisy signals | Visual inspection; specialist low-frequency methods (still research) |
Key takeaway: at a crossing, a flaw test answers "is there a crack in the head?" Only an exposed web can answer "how thick is the web?"
Which 49 CFR Part 213 rules apply at grade crossings?
Part 213 has no separate testing rule for road crossings. Rail inside a crossing is ordinary running rail, so the general rules apply to it in full. The sections that bear most on crossings are:
| Section | What it requires | Why it matters at a crossing |
|---|---|---|
| 49 CFR 213.237(c) | Internal rail test intervals, e.g. 370 days or 30 MGT (whichever is shorter) on Class 4–5 track | The crossing gets the same deadline as the line around it |
| 213.237(e) | Mark defective rail on both sides of the web and base; where they are "inaccessible because of permanent features", mark on or next to the head | Recognizes that crossings can hide the web |
| 213.237(g) | Track where the qualified operator decides a valid search could not be made "may not be considered as internally inspected" | A lost-coupling crossing is untested track |
| 213.237(h) | Before the time or tonnage limit expires: re-test, limit speed to 25 mph, or replace the rail | Sets the deadline for following up a missed crossing |
| 49 CFR 213.241(c)(6) | Rail inspection records must show the "location of any track not tested" under 213.237(g) | Untested crossings must appear in the record |
| 49 CFR 213.113 | Remedial action for defects found, including split web and head and web separation | Applies to any defect found when a crossing is exposed |
| 49 CFR 213.233 | Visual track inspection intervals; devices may "supplement visual inspection" | The visual inspector sees only the head at a crossing |
| 49 CFR 213.33 | Drainage under or next to the roadbed must be kept free of obstruction | Standing water in a crossing speeds corrosion |
The valid-search definition in 213.237(j) is the one to keep in mind at crossings:
"Valid search means a continuous inspection for internal rail defects where the equipment performs as intended and equipment responses are interpreted by a qualified operator as defined in § 213.238." — 49 CFR 213.237(j)
What Part 213 does not cover
None of the Part 213 sections listed above mentions corrosion, and Part 213 sets no minimum rail web thickness. The TSB found the same gap in Canada's Track Safety Rules. A web-thinning limit, an inspection interval for lifting crossing surfaces and a method for measuring the web are therefore owner decisions, made under the general authority in 213.1(a) to adopt requirements stricter than the federal minimum.
Key takeaway: compliance with 213.237 at a crossing means a valid search for internal defects on time. It doesn't mean the web has been checked for corrosion.
What happens when the test car can't get a valid search?
The crossing counts as untested, and the owner has until the 213.237 deadline to re-test it, slow trains to 25 mph, or replace the rail. The deadline is the interval that started with the last valid search, not a new clock from the day of the missed test.
Worked example: deadline for a missed crossing
A Class 4 freight segment carries 36 MGT a year. Its last valid search, including the crossing, was on day 0. On day 200 the test car runs again and the operator flags 90 ft of rail through a paved crossing as a no-test because of lost coupling.
- The 30 MGT limit is reached at 30 ÷ 36 × 365 ≈ 304 days. The 370-day limit comes later, so tonnage controls.
- The rest of the segment now has a new valid search dated day 200. The 90 ft through the crossing does not; under 213.237(g) it is still on its day-0 clock.
- Before day 304, the owner must do one of the 213.237(h) options for that 90 ft: get a valid search (for example a hand test after cleaning the head), limit speed to 25 mph, or replace the rail.
- The day-200 record must list the 90 ft as track not tested, per 213.241(c)(6).
| Day | Event | Status of the crossing rail |
|---|---|---|
| 0 | Valid search | Tested |
| 200 | Test car loses coupling through crossing | Untested; still on day-0 interval |
| Before 304 | Hand test, 25 mph limit or replacement | Must be resolved |
| 304 | 30 MGT reached | Deadline for the crossing rail |
Even a valid re-test only satisfies 213.237. On a low-tonnage line with old rail, the owner may also want a web check, which the rule does not require.
Which methods can fill the gaps?
No single method covers the whole rail section at a crossing, so owners combine them. The choice depends on whether the concern is a head crack, web and base corrosion, or a cast special trackwork component.
| Method | What it can show at a crossing | Limits |
|---|---|---|
| Rail test car or cart UT | Internal defects in the head, some web defects, with a valid search | Web wall loss; dirty heads cause no-tests |
| Hand UT after cleaning the head | A valid search over short no-test lengths | Same beam limits as the car; slower |
| UT thickness gauge on the exposed web | Remaining web thickness at known points | Needs the surface lifted; spot readings only |
| Visual and dimensional check with the surface removed | Corrosion pattern, pitting, base condition, fastenings | Only possible during rehabilitation or planned lifts |
| Guided wave testing from outside the crossing | Screening for section loss along the embedded rail | Needs validation on the owner's rail and crossing type; results are indications, not measured thickness |
| Low-frequency UT on cast manganese parts | Internal flaws in coarse-grained frogs or diamonds | Research stage (FRA RR 22-25 used a test sample with drilled holes) |
Guided wave testing is the method most often proposed for crossings because it can be applied from an accessible point outside the crossing. The EU-funded MONITORAIL project (2011–2013) worked on long-range ultrasonic monitoring of the rail foot for this reason, noting that "defects in the foot are more likely where corrosion is more severe, such as at level crossings or in tunnels". Treat published detection figures for guided wave rail systems as specific to the systems tested. See guided wave ultrasonic testing for how the method works.
The FRA special trackwork study is useful but narrow. It showed "the potential" for low-frequency UT to characterize internal flaws in a crossing-diamond sample. It did not show field detection rates.
What should an owner specify for testing at crossings?
Ask for crossing-by-crossing coverage reporting, a hand-test follow-up within the 213.237 deadline, and web measurement whenever a surface is lifted. These items fit into a standard rail flaw testing contract or RFP.
| Specification item | What to ask for | Rule or reason |
|---|---|---|
| Crossing list | Every crossing in the test limits by milepost and name, from the owner's records | Lets the report be checked crossing by crossing |
| Coverage report | Each crossing marked "valid search" or "no test", with length and cause (coupling, debris, equipment) | 213.237(g); 213.241(c)(6) |
| Head preparation | Rail head cleaning at known dirty crossings before the test run, where the owner allows it | Reduces no-tests |
| Follow-up | Hand UT of every no-test crossing, with a completion date before the 213.237 time or tonnage limit | 213.237(h) |
| Operator qualification | Operators qualified under 49 CFR 213.238, with records available | 213.237(j) valid search definition |
| Defect marking | Marking on or next to the head where the web is covered | 213.237(e) |
| Exposed-rail inspection | When a surface is removed: visual check of web and base, UT thickness readings at set spacing across the crossing, photographs | Owner standard; TSB R19M0018 |
| Action threshold | The owner's web-thickness or percent-loss limit and who decides disposition | Not set by Part 213; owner engineering decision |
| Risk ranking | Priority for crossings on low-tonnage lines, old rail, near salted highways, or with poor drainage | TSB R19M0018; 213.33 |
A field procedure for one crossing
- Before the run, list the crossing, its length, rail section and last surface renewal date.
- Clean the rail head through the crossing if contamination is visible.
- Run the test car or cart and record whether a valid search was achieved over the full crossing length.
- If not, hand-test the no-test length and record the result against the same milepost limits.
- At the next crossing rehabilitation, lift the surface, clean the web and base, and take thickness readings across the full crossing, with extra readings near the center.
- Compare the readings with the original section and the owner's limit, and send results to the engineer who decides disposition.
- Keep the crossing's coverage history so the next team knows what was and wasn't examined.
A clean test result and an untested crossing are different things in a record. The report should never let one look like the other.
What UT at crossings can't tell you
Ultrasonic rail flaw testing at a crossing is a crack search on the rail head, with real limits.
- It doesn't measure web or base thickness. Corrosion like that at Lakeville Road can pass a flaw test.
- It doesn't find base corrosion or base cracks well from the head.
- It can't give a valid search through a heavily contaminated head; the result is a no-test, not a pass.
- It doesn't inspect crossing panels, ties, fastenings or drainage. Those belong to the visual inspection and crossing maintenance program.
- It doesn't reliably test cast austenitic manganese frogs or diamonds with standard probes.
- A 213.237-compliant test doesn't satisfy any corrosion requirement, because Part 213 has none.
These limits are why web condition at crossings needs its own line in the maintenance plan, usually tied to crossing rehabilitation. The TSB noted that removing the crossing structure during rehabilitation "provides an opportunity to examine and assess the condition of the rail and the rail web".
How EMA Industries applies this
EMA Industries provides ultrasonic rail flaw testing for transit agencies, short lines and industrial track owners. The company is based in Southern California and mobilizes nationwide. For owners with road crossings in their test limits, the specification items above (crossing coverage reporting, hand-test follow-up of no-test lengths, and UT web thickness readings when a surface is lifted) can be built into the testing scope, and EMA Industries' NDT Level III services can help write the procedure. To plan crossing coverage for your next test cycle, contact EMA Industries.
Related reading
- 49 CFR 213.237 rail flaw detection inspection requirements
- Railroad switch and crossing inspection
- Ultrasonic thickness testing in railway inspection
- Ultrasonic couplant for rail inspection
- Tunnel rail inspection with ultrasonic testing
- Rail inspection RFP guide
- How ultrasonic testing works for railways
Frequently asked questions
Does 49 CFR 213.237 require a separate test at grade crossings?
No. Rail inside a crossing is part of the rail inspection segment, so it follows the same 213.237(c) interval as the track around it. What the rule does require is a valid search over that rail. If the operator can't get one through a crossing, the crossing counts as untested and must be re-tested, slowed to 25 mph or replaced before the deadline.
Can a rail test car measure web thickness at a crossing?
Not as currently designed. Rail flaw test cars scan from the head to find internal cracks. The TSB's Lakeville Road report found that current UT cannot detect web material loss at a crossing without removing the crossing surface. Web thickness is measured with a UT thickness gauge on the exposed web, or screened from outside the crossing with methods such as guided wave testing.
How often should crossing surfaces be lifted to inspect the rail?
Part 213 sets no interval, so the owner decides. A practical approach ties the web inspection to crossing rehabilitation, when the surface comes off anyway, and brings it forward for higher-risk crossings: low-tonnage lines where rail stays in place for decades, crossings near salted highways, and crossings with poor drainage.
What causes rail web corrosion at road crossings?
Crossing structures hold moisture against the rail, and road traffic carries in salt, sand and grit. In the Lakeville Road case, the TSB's lab found sodium, chlorine and other road-salt and sand elements in the corrosion scale, and concluded that winter road salt caused the web to corrode at an accelerated rate.
Which defects does UT find reliably at a crossing?
With a valid search, UT at a crossing finds the same internal head defects it finds on open track, such as transverse defects with the 70° probe and horizontal split heads with the 0° beam. Detection still depends on defect size and orientation. It is weakest for defects deep in the web or base, and it doesn't measure corrosion wall loss.
Sources
- Transportation Safety Board of Canada, Rail transportation safety investigation report R19M0018, VIA Rail train 14 derailment at Lakeville Road, published 30 September 2020
- 49 CFR 213.237, Inspection of rail, eCFR (read via Cornell LII, October 2026)
- 49 CFR 213.241, Inspection records, eCFR
- 49 CFR 213.113, Defective rails, eCFR
- 49 CFR 213.233, Track inspections, eCFR
- 49 CFR 213.235, Inspection of switches, track crossings, and lift rail assemblies, eCFR
- 49 CFR 213.33, Drainage, eCFR
- Federal Railroad Administration, Feasibility of Special Trackwork Inspection Using Ultrasonics, Research Results RR 22-25, September 2022 (R. Wilson and A. Poudel, TTCI)
- European Commission CORDIS, MONITORAIL final report summary, project 262194, 2011–2013
- TWI, RailSAFT: ultrasonic inspection of cast austenitic manganese steel railway crossings