Railway Infrastructure

Broken Rail Causes: Why Rails Break and How Owners Prevent It

Updated

broken rail causes

The main broken rail causes are growing internal defects, failed welds, bolt-hole cracks, wheel impacts, cold-weather tension in continuous welded rail and corrosion. FRA data show broken rails and welds caused 384 of 1,145 track-caused main-line derailments in 2016–2025, and 49 CFR 213.237 caps service failures at 0.1 per mile per year on Class 4–5 track.

This guide is about the rail itself: why a rail breaks, what the FRA accident cause codes show, and which levers an owner controls. It isn't a general derailment guide. For every cause of derailment, including gage, equipment and human factors, see what causes a train to derail.

What is a broken rail, and what is a service failure?

A broken rail is any partial or complete fracture of the rail in track. A service failure is the narrower regulatory term: a broken rail caused by one of four internal defects. The difference matters because only service failures count toward the federal performance target for rail testing.

The definition sits in 49 CFR 213.237(j) on eCFR:

"Service failure means a broken rail occurrence, the cause of which is determined to be a compound fissure, transverse fissure, detail fracture, or vertical split head." 49 CFR 213.237(j)(3)

A broken base, a failed weld, a bolt-hole crack that breaks out or an ordinary break is still a broken rail. It just isn't a service failure. A service failure also counts whether or not a train derails, while FRA's accident data only see the broken rails that cause a reportable accident.

TermWhat it coversWhere it is recorded
Broken railAny fracture of the rail in track, from any causeOwner's records; FRA accident data only if it causes a reportable accident
Service failureA broken rail from a TF, CF, DF or VSHOwner's 213.237 records, kept under 213.241
Detected defectA flaw found by testing or inspection before the rail breaksOwner's 213.241 test records, with the 213.113 remedial action
Broken-rail derailmentA reportable accident whose primary cause is a "Broken Rail" T2xx codeFRA Form 6180.54, public data

Key takeaway: every service failure is a broken rail, but most broken rail causes outside the four internal defects never enter the federal service-failure count.

Which broken rail causes lead to the most derailments?

Detail fractures lead, followed by transverse or compound fissures, vertical split heads and broken bases. Together, the three service-failure codes (T207, T220 and T221) caused 222 of the 384 main-line broken-rail derailments from 2016 to 2025, or 58%.

FRA cause codeDefectMain-track derailments, 2016–2025Cause family in this guide
T207Detail fracture from shelling or head check110Internal defect growth (surface-initiated)
T220Transverse or compound fissure57Internal defect growth
T221Vertical split head55Internal defect growth
T202Broken base53Base flaws, corrosion, impact
T204Field weld34Weld failure
T210Head and web separation, outside joint bars34Head-web fatigue
T201Bolt hole crack or break21Bolt-hole and joint
T212Horizontal split head10Internal defect growth
T211Head and web separation, within joint bars5Bolt-hole and joint
T203Plant weld2Weld failure
T208Engine burn fracture2Wheel damage
T218Piped rail1Manufacturing
Total384

Source: EMA Industries query of FRA's unique train accident dataset (derailments on main track, primary cause beginning "Broken Rail"), run October 24, 2026, with data through July 31, 2026. Counts change as railroads file late reports.

Main-line broken-rail derailments ran between 31 and 44 a year over the decade, with 38 in 2025 and no clear downward trend. Our rail failure statistics article covers the national rates, severity and the wider cause groups. For what the codes mean, see FRA defect codes explained.

Where on the network do rails break?

The same query by the track class reported on each accident shows something the national totals hide:

Reported FRA track classMain-track broken-rail derailments, 2016–2025Share
Excepted (X)4210.9%
Class 19925.8%
Class 28421.9%
Class 36416.7%
Class 47519.5%
Class 5205.2%
Total384

Excepted, Class 1 and Class 2 track account for 225 of the 384 (58.6%). Section 213.237 only requires internal rail testing on Class 3, 4 and 5 track, so most main-line broken-rail derailments happened on track with no federal testing interval at all. These are counts, not rates: FRA's public data have no track-mile denominator by class, and lower-class track is often older, lighter rail at lower speeds.

Key takeaway: the four internal defects drive most broken-rail derailments, and much of the risk sits on lower-class track that the owner tests only by choice.

How do internal rail defects grow until the rail breaks?

Most rails don't break from a single overload. A fatigue crack starts inside the head or at its surface, grows with each wheel pass, and the rail fractures when the remaining section can no longer carry the load. Part 213 defines each defect by where it starts and how it grows.

  • A transverse fissure is "a progressive crosswise fracture starting from a crystalline center or nucleus inside the head" (213.113(d)(15)). See our transverse fissure guide.
  • A compound fissure starts as a horizontal split head that turns up or down, then develops at a right angle to the rail length (213.113(d)(3)). FRA's T220 code combines it with the transverse fissure.
  • A detail fracture originates "at or near the surface of the rail head" and "may arise from shelled spots, head checks, or flaking" (213.113(d)(7)). Rolling contact fatigue at the gage corner is the usual starting point. See detail fracture development.
  • A vertical split head is a vertical split through or near the middle of the head; "a crack or rust streak may show under the head close to the web" (213.113(d)(16)). See the vertical split head guide.
  • A horizontal split head starts inside the head, "usually 1/4 inch or more below the running surface", and grows horizontally (213.113(d)(11)).

FRA's 2014 rail integrity rule explains what speeds growth up. Transverse defect initiation is hard to predict, but "once initiated, transverse defect development is influenced by tonnage." The preamble adds:

"Rapid defect growth rates can also be associated with rail where high-tensile residual stresses are present in the railhead and in CWR in lower temperature ranges where the rail is in high longitudinal tension." FRA, Track Safety Standards; Improving Rail Integrity, 79 FR 4234 (January 24, 2014)

Three things therefore drive internal defect failures: the tonnage since the defect started, the tension in the rail, and the time between tests. The 2014 rule also tightened the response to large head defects. Transverse fissures, detail fractures and defective welds at 60% or more of the head area now get the most restrictive remedial actions, a threshold lowered from 80% because detail fractures had been seen growing more than 5% of head area per MGT.

Key takeaway: a defect that a test finds at 10% of the head can reach failure before the next test if tonnage is high and the rail is in winter tension.

Why do rail welds fail?

Welds fail when a casting or fusion defect, such as slag, porosity, lack of fusion or a hot tear, starts a fatigue crack under traffic. Field thermite welds carry most of the risk: FRA data show 34 main-line derailments from field welds (T204) against 2 from plant welds (T203) in 2016 to 2025.

Section 213.113(d)(6) defines a defective weld as a field or plant weld with discontinuities "exceeding 5 percent of the rail head area individually or 10 percent in the aggregate, oriented in or near the transverse plane." The causes it lists are incomplete penetration, lack of fusion, slag or sand entrainment, under-bead or shrinkage cracking, and fatigue cracking. A weld defect that progresses lengthwise through the weld is treated as a split web.

TTCI's survey of 265 thermite weld failures in revenue service (Kristan, TD-00-013, July 2000) found:

Thermite weld failure cause (TTCI, 265 revenue failures)Share
Slag or porosity40%
Run-out fatigue32%
Hot tears15%
Columnar grains5%
Hot pull-aparts3%
Grinding burns3%
Sand2%

TTCI judged "approximately 91 percent" of those failures preventable by following or improving installation practice. Where tonnage data existed, the welds averaged 169 MGT to failure. A later TTCI digest (TD-21-027, 2021) reported that about 95% of thermite weld breaks at FAST, its accelerated-service test loop, were in the web, the web-base fillet and the base.

That location matters for prevention. TD-00-013 states that "the ultrasonic rail flaw detection process only scans the head and web of the rail, and not the majority of the base." A weld defect in the base can grow through a test car's blind spot. For inspection methods, see aluminothermic weld inspection.

Key takeaway: weld failures are mostly an installation-quality problem, and they start where routine rail testing sees least.

How do bolt-hole cracks break rails?

A bolt-hole crack starts at a hole in the web at a joint and grows upward toward the head or downward toward the base until a piece of the rail end can break out. Section 213.113(d)(1) defines it as "a crack across the web, originating from a bolt hole," and notes that fully developed cracks "may progress into and through the head or base to separate a piece of the rail end from the rail."

Joints concentrate load. Wheels strike the rail-end gap, the bolt holes remove web section, and loose bolts or batter make the impacts worse.

Bolt-hole cracks (T201) caused 21 main-line derailments in 2016 to 2025, and head-web separations inside joint bars (T211) caused 5 more. In Canada, a Transportation Safety Board safety issues investigation (SII R05-01) reported that bolt-hole cracks made up 78% of the rail defects found by ultrasonic testing on six CN subdivisions, which shows how common they are where jointed rail survives.

Two rules target the joint directly:

  • 213.237(f): "Inspection equipment shall be capable of detecting defects between joint bars, in the area enclosed by joint bars." The holes sit behind the bars, so a visual check can't see the crack until it breaks out.
  • 213.119(h): CWR plans must schedule joint inspections and name the warning signs, including "loose, bent, or missing joint bolts" and "rail end batter or mismatch that contributes to instability of the joint."

The 213.113 remedial action table sizes bolt-hole cracks by length: over 1/2 to 1 inch, over 1 to 1-1/2 inches, over 1-1/2 inches, and break-out, each with a stricter action. See the bolt-hole crack inspection guide.

Can wheel impacts and wheel flats break rails?

Yes. A flat, shelled or out-of-round wheel hammers the rail at each revolution, and that impact can start or grow a rail crack, or break a rail that already has one. Part 213 recognizes this in its definition of damaged rail: "any rail broken or otherwise damaged by a derailment, broken, flat, or unbalanced wheel, wheel slipping, or similar causes" (213.113(d)(5)).

Wheel slip is a separate cause. An engine burn fracture is "a progressive fracture originating in spots where driving wheels have slipped on top of the rail head" (213.113(d)(8)), and it can resemble a transverse or compound fissure. FRA data show only 2 main-line derailments from engine burn fractures (T208) in 2016 to 2025.

Railroads measure the impacts with wheel impact load detectors (WILDs), strain gauges on the rail web. The NTSB explained the stakes in its brief on a 2015 BNSF crude oil train derailment at Heimdal, North Dakota: "High-impact wheel forces can damage cargo and track infrastructure, primarily rails" (NTSB RAB-17/12). At the time, BNSF's action levels were a maintenance advisory at 65 kips, a wheel change at the next shop visit at 80 kips, and condemning the wheel at 90 kips.

Track conditions can also create impacts. At a bridge approach on a 230 MGT coal line, TTCI measured impacts "at least twice as high in magnitude as static wheel loads," and a rail broke in a mud-pumping hole about 1.5 years into the study, "likely" from those impacts (TD-07-042, 2007).

Key takeaway: the track owner doesn't control the wheels, but it does control the dips, battered joints and weak spots that turn wheel impacts into broken rails.

Why do more rails break in cold weather?

In continuous welded rail, cold weather puts the rail in tension, because the rail wants to shorten and the fastenings stop it. That tension adds to the bending stress at every crack tip, so existing defects grow faster and a weak section can snap. FRA data show 214 of 384 main-line broken-rail derailments (55.7%) happened from November to March in 2016 to 2025.

Month (main track, 2016–2025)Broken-rail derailments
November48
December40
January38
February46
March42
April to October (7 months)170

Source: EMA Industries query of FRA's unique train accident dataset, October 24, 2026. A Czech study of buckles and broken rails found the same seasonal split: 78% of broken rails from November to March, and 83% of buckles from June to August (Nezval et al., European Transport Research Review, 2025, as summarized by CDV).

The key term is rail neutral temperature (RNT). Under 49 CFR 213.119(l), it "is the temperature at which the rail is neither in compression nor tension." Below the RNT the rail is in tension, and above it the rail is in compression. FRA research (DOT/FRA/ORD-13/16, 2013) reports that RNT is typically set at about 90 to 115 °F at installation but can fall to 60 °F or lower through rail movement and maintenance.

Worked example: tension in 136 lb CWR on a cold morning

Thermal force in restrained rail is P = E × A × α × (RNT − T), where E is about 30,000,000 psi, α is 0.0000065 in./in./°F, and A is the rail's cross-section area. These are the values in FRA's training module Track Buckling Prevention and Improved Rail Stress Management (ROSA P dot/92839, 2023).

Input or resultValue
Rail section136 lb, A = 13.35 in.²
E × α30,000,000 × 0.0000065 = 195 psi per °F
E × A × α195 × 13.35 ≈ 2,600 lb per °F
RNT as laid95 °F
Rail temperature at dawn5 °F
Temperature difference95 − 5 = 90 °F
Tensile force in each rail2,600 × 90 = 234,000 lb
Tensile stress195 × 90 = 17,550 psi

That stress sits on top of the bending stress from every passing wheel. If the RNT has drifted down to 65 °F, the same morning gives 60 °F of difference, 156,000 lb and 11,700 psi. A lower RNT eases winter tension, but it raises summer compression and the risk of a buckle, which is why 213.119 plans aim for a set installation range rather than simply laying rail cold.

The rule's own definition of that range names both failures. The desired rail installation temperature range is the range "at which forces in CWR should not cause a buckling incident in extreme heat, or a pull apart during extreme cold weather." Cold-weather testing brings its own equipment issues; see cold weather UT testing and our guide to continuous welded rail.

Key takeaway: winter doesn't create most defects, but it loads the ones already in the rail, so the test before winter matters most.

How does corrosion lead to broken rails?

Corrosion removes steel from the web and base, where a test car looks least, until the remaining section breaks under load. It is most likely where water, salt and debris sit against the rail: highway-rail crossings, tunnels, and wet or poorly drained track.

The clearest case is a 2019 VIA Rail derailment at the Lakeville Road crossing in New Brunswick. Canada's Transportation Safety Board found the rail web had corroded from 14.28 mm to about 4 mm under the crossing surface.

The corrosion "had not been identified by the visual or ultrasonic track inspections, because the web of the rail was hidden by the crossing surface" (TSB R19M0018). The rail had been ultrasonically tested about three months earlier with no defects found.

Part 213 sets no rail corrosion or web thickness limit; the word "corrosion" doesn't appear in it. Corrosion control is therefore the owner's maintenance program: drainage, crossing surface removal at set intervals, and visual and thickness checks of the web and base where they can be reached.

Broken bases (T202) caused 53 main-line derailments in 2016 to 2025, though FRA's code doesn't record whether corrosion, a base seam or impact started each one. See grade crossing rail inspection for the inspection approach.

How do track owners prevent broken rails?

The levers are test frequency under 213.237, the CWR plan under 213.119, rail grinding, weld quality, and analysis of every service failure. Each one targets a different cause family.

Prevention leverRule or sourceCauses it targets
Internal rail testing at the right interval49 CFR 213.237TF, CF, DF, VSH, HSH, bolt-hole cracks, weld defects in head and web
CWR plan: RNT, installation range, pull-apart and joint inspections49 CFR 213.119Cold-weather breaks, pull-aparts, joint failures
Preventive and corrective grindingOwner program; FRA ORD-11/24Head checks and shelling that start detail fractures and mask defects
Weld procedure, inspection and recordsOwner specification; TTCI TD-00-013Field weld failures
Service-failure analysis213.237(a), (d); 213.241All causes, by showing where and why rails break
Remedial action on found defects49 CFR 213.113Prevents a found defect becoming a break

Lever 1: test often enough under 213.237

Section 213.237(c) caps the interval on Class 4 and 5 track, and on Class 3 track with scheduled passenger trains or a hazmat route, at "370 days between inspections or a tonnage interval of 30 million gross tons (mgt) between inspections, whichever is shorter." Other Class 3 track is tested at least once a calendar year or every 30 MGT, whichever is longer, and never more than 5 years apart.

The interval is a legal maximum, not a target. FRA described how railroads set real frequencies in the 2014 rule: "inspection frequency is established annually based on several factors, including the total detected defect rate per test, the rate of service failures between tests, and the accumulated tonnage between tests." On a 45 MGT line, the 30 MGT limit arrives after 30 ÷ 45 × 365 ≈ 243 days, well before 370 days. For the full schedule, see how often must rail be tested and our guide to 49 CFR 213.237.

A test only counts where it was a valid search. Where the operator determines that a valid search couldn't be made, the owner must retest, hold trains to 25 mph, or replace the untested rail before the interval runs out (213.237(g) and (h)).

Lever 2: manage CWR under 213.119

A CWR plan must set a desired rail installation temperature range, de-stressing procedures, anchoring patterns, joint procedures and inspection schedules. Three provisions bear most directly on broken rails:

  • Repairs in cold weather: rail must not be added below the designated installation temperature "without provisions for later adjustment" (213.119(d)(2)). Rail added cold lowers the RNT, which shifts the risk to summer.
  • Pull-apart prone conditions: the plan must say when inspections look for "locations where pull-apart or stripped-joint rail conditions are likely to occur" and what remedial actions follow (213.119(g)).
  • CWR joints: at least two bolts per rail at each joint, and a joint that fails with a rail gap must be welded, re-bolted with extra bolts and anchors, or reinforced as the rule lists (213.119(c)). Joint inspections are set by class and tonnage, up to four a year on Class 4 and 5 freight track over 60 MGT (213.119(h)(6)).

Lever 3: grind before cracks turn into fractures

Grinding removes the surface cracks that start detail fractures, and it lets the test see into the head. FRA research report ORD-11/24 lists one purpose of grinding as "to improve the capability of ultrasonic testing (signals cannot pass through cracks)" (ROSA P dot/23669). The same report describes a 25-year program in which preventive grinding with improved rail metallurgy doubled system rail life and quadrupled fatigue life, as reported by that railroad.

Grinding also fixes untested rail. In TTCI's field study (TD-20-002), corrective grinding restored a valid ultrasonic test at 68 of 88 locations (77%) that had failed to test because of surface damage, and at all locations where at least 0.060 in. was removed. See rail grinding and ultrasonic testing.

Lever 4: control weld quality

With 91% of the thermite weld failures in TTCI's survey judged preventable, weld quality is mostly a process discipline: correct gap, alignment, preheat, mold fit and finishing, done by trained welders to a written procedure. AWS D15.2/D15.2M:2022, Specification for Joining Railroad Rail and Related Rail Components, and AREMA's Manual for Railway Engineering Chapter 4 give the industry framework, and the owner's specification sets acceptance. Because test cars miss most of the base, a new field weld needs its own inspection, including a hand ultrasonic scan of the base, and a weld record that ties each weld to its location and welder.

Lever 5: act on found defects under 213.113

A detected defect isn't a broken rail yet. Section 213.113(b) requires the owner to verify an indication within four hours, or immediately for defects that need remedial action A, A2 or B, and then to replace or repair the rail or apply the table's action.

Typical actions are speed limits, joint bars, visual supervision of each train and replacement within a set time. Our slow order and rail defect guide covers how restrictions work in practice.

Key takeaway: testing finds defects, but CWR management, grinding and weld quality reduce how many defects there are to find.

How should owners analyze service failures?

Record every broken rail, not just the four service-failure types, by defect, location, date, rail temperature, rail age and weld or joint ID, then compare each break with the last test. Section 213.237(a) requires the service failure rate to be calculated per rail inspection segment within 45 days of the end of each 12-month period, and 213.241 requires the records to be kept.

The federal rate tells you whether the interval is working for internal defects. A full analysis also asks why each break happened, so that the fix matches the cause.

Worked example: one segment's year of broken rails

A 25-mile Class 4 rail inspection segment carries regularly scheduled passenger trains but isn't a hazmat route. Its limit is 0.09 service failures per mile per year, which allows 0.09 × 25 = 2.25 in 12 months.

Broken rail foundCounts as service failure?What the analysis points to
Detail fracture, curve, JanuaryYesGage-corner RCF; grinding interval
Detail fracture, same curve, FebruaryYesSame curve; masking by shelling at the last test
Vertical split head, tangent, DecemberYesHead defect missed or grown since the last test
Broken base, at a crossingNoCorrosion under the crossing surface
Broken base, open trackNoBase seam or impact; check rail age and batch
Field thermite weld breakNoWeld installation; check the welder's other welds

Service failure rate: 3 ÷ 25 = 0.12 per mile per year, over the 0.09 limit. The owner must inform FRA within 45 days and may submit an explanation and a remedial action plan (213.237(d)).

If the target is missed again the next year, then for the area with the most service failures the owner must either cut the test interval to 10 MGT or reduce the track to Class 2 until the target is met (213.237(d)(1)). On this line at 20 MGT a year, a 10 MGT interval means a test about every 10 × 365 ÷ 20 ≈ 183 days.

The three breaks that don't count still need action. Two point to corrosion and rail condition, which testing frequency won't fix, and the weld break points to an installation process. A segment can meet its federal target and still break rails.

Short segments have a separate rule. On a 6-mile Class 5 segment, one service failure gives 1 ÷ 6 ≈ 0.17, over 0.10, but under 213.237(d)(2) the segment is treated as compliant "unless a second such failure occurs within a designated 12-month period."

Key takeaway: the service failure rate is a compliance test; the cause of each break is what tells you which prevention lever to pull.

What can't broken rail prevention guarantee?

No testing interval or maintenance program reduces broken rails to zero. These are the main limits an owner should plan around:

LimitWhy it matters
Vehicle UT doesn't cover most of the rail baseTTCI: test cars scan "the head and web of the rail, and not the majority of the base", so base and weld-foot defects need other checks
Surface damage masks defectsShells and head checks shadow detail fractures beneath them; the 2014 FR notes current technology's limits "beneath surface conditions and in the base flange area"
Detection is not 100%A defect can be missed, or start and grow to failure between tests at high tonnage
Corrosion hides under crossingsTSB R19M0018: the corroded web was hidden by the crossing surface from both visual and ultrasonic inspection
Part 213 sets no interval below Class 3Most main-line broken-rail derailments in 2016–2025 were on excepted, Class 1 or Class 2 track
FRA accident data see only accidentsBroken rails that don't derail a train, and service failures, stay in owners' records
T220 combines two defectsFRA data can't separate transverse from compound fissures
Reported class and causeEach accident's track class and primary cause are as reported by the railroad

Rail transit systems not connected to the general railroad system fall outside Part 213 and these FRA datasets. Their rail testing rules come from the agency's own standards and state safety oversight, such as the CPUC in California.

Related reading: rail inspection after a derailment (rail damaged by a derailment is "damaged rail" under 213.113) · how to prevent train derailments (detectors, train handling and geometry, beyond the rail itself)

How EMA Industries applies this

EMA Industries provides ultrasonic rail flaw testing and rail weld inspection for transit agencies, short lines and industrial track owners. We're based in Southern California and mobilize nationwide. For a segment with rising broken rails, the useful starting point is the owner's own break records sorted by cause, then a test plan that covers internal defects, joints and field welds, including the weld base areas that a test car misses. To scope testing or weld inspection, see our ultrasonic rail testing services or contact us.

Frequently asked questions

What is the most common cause of broken rails?

In FRA main-track derailment data for 2016 to 2025, detail fractures (T207) were the largest single cause, with 110 of 384 broken-rail derailments. Transverse or compound fissures (57) and vertical split heads (55) came next. All three are fatigue defects that grow in the rail head.

Why do rails break more often in winter?

Continuous welded rail is restrained, so when it cools below its neutral temperature it goes into tension. That tension adds to wheel-load stresses at existing cracks, so defects grow faster and weak rail fractures. From 2016 to 2025, 55.7% of main-line broken-rail derailments were in November to March.

Is every broken rail a service failure?

No. Under 49 CFR 213.237(j), a service failure is a broken rail caused by a compound fissure, transverse fissure, detail fracture or vertical split head. Broken bases, weld failures, bolt-hole cracks and ordinary breaks are broken rails, but they don't count toward the service failure rate.

How often must rail be tested to prevent breaks?

On Class 4 and 5 track, and on Class 3 track with scheduled passenger trains or a hazmat route, at least every 370 days or 30 MGT, whichever comes first (49 CFR 213.237(c)). Many owners test more often, setting the interval from their defect and service-failure rates and tonnage.

Can ultrasonic testing prevent all broken rails?

It can't. Testing finds most internal head and web defects in time for remedial action, but test cars don't scan most of the base, surface damage can mask defects, and corrosion under crossings can be hidden. Grinding, weld quality, CWR management and corrosion control cover the gaps.

Sources

← All articles