Rail joint inspection is the check of the joint bars, bolts and rail ends at a bolted joint. Under 49 CFR 213.121(c), a joint bar cracked between the middle two bolt holes must be replaced. Yet the rail flaw ultrasonic test that track owners run under 213.237 examines the rail, not the bars. Finding a hidden joint bar crack needs a dedicated UT setup or a close on-foot inspection.
That gap is the subject of this guide. It covers what the federal rules require at a joint, why ordinary rail ultrasonic testing (UT) doesn't cover the bars, how dedicated joint-bar UT works, what the research shows it can find, and where it still falls short.
What does a rail joint inspection have to find under FRA rules?
Four sections of the FRA Track Safety Standards apply at a bolted joint, and each covers a different part. Section 213.121 governs the joint bars and bolts. Section 213.113 governs defects in the rail itself, including bolt hole cracks. Section 213.237(f) requires the rail flaw test to reach the rail inside the joint. Section 213.119(h) requires CWR owners to inspect their joints on a schedule.
| Rule | What it covers at the joint | What it requires |
|---|---|---|
| 49 CFR 213.121(b) | Joint bars on Class 3–5 track | Replace a bar that is cracked, broken, or worn so that either rail moves vertically with all bolts tight |
| 49 CFR 213.121(c) | Joint bars on any class | Replace a bar cracked or broken between the middle two bolt holes |
| 49 CFR 213.121(d), (e) | Bolts | At least 2 bolts per rail at each joint on Class 2–5 jointed track (1 on Class 1); at least 2 per rail at every CWR joint |
| 49 CFR 213.121(g), (h) | Torch work | No torch-cut or burned bolt holes on Class 2–5 track; no torch-cut reconfiguring of joint bars on Class 3–5 track |
| 49 CFR 213.113 | Rail defects, including bolt hole cracks and head-web separation in the joint | Remedial action table by defect size |
| 49 CFR 213.237(f) | The rail flaw test equipment | Must detect defects in the rail enclosed by the joint bars |
| 49 CFR 213.119(h) | Joints in CWR track | A written procedure and a minimum number of joint inspections each year |
Section 213.121(c) is short and absolute:
"If a joint bar is cracked or broken between the middle two bolt holes it shall be replaced." (49 CFR 213.121(c))
It applies on every track class. On Class 3 to 5 track, 213.121(b) goes further: a crack or break anywhere in the bar means replacement.
Rail defect or joint bar defect: why the difference matters
A bolt hole crack is a rail defect. Section 213.113(d) defines it as "a crack across the web, originating from a bolt hole", and multiple cracks in one rail end count as a single defect. It is sized by length, and the remedial action table sets what happens next:
| Bolt hole crack length | Remedial action under 213.113 | What that means |
|---|---|---|
| More than 1/2 in. to 1 in. | H and F | 50 mph limit (or the class maximum, if lower); reinspect within 90 days |
| More than 1 in. to 1½ in. | H and G | 50 mph limit; reinspect within 30 days |
| More than 1½ in. | B | Speed of 30 mph or less, set by a designated person |
| Break out in the rail head | A | Visual supervision of each train movement |
A crack in the joint bar isn't on that table. The bar is a track component under 213.121, and the rule there is replacement, not a speed restriction by size. Bolt hole cracks in the rail are found by the rail flaw test. Cracks in the bars are found by joint inspection. One test doesn't stand in for the other. The defect sizes and actions are covered in more depth in our guide to bolt hole cracks.
Key takeaway: 213.113 governs the rail inside the joint, 213.121 governs the bars, and a complete rail joint inspection checks both.
How often must CWR joints be inspected?
Section 213.119(h)(6) sets a minimum number of joint inspections per year in continuous welded rail (CWR) track, by class, traffic type and annual tonnage. The rule requires each CWR plan to include procedures "to detect cracks and other indications of potential failures in CWR joints", including at least periodic on-foot inspections. Under 213.119(h)(2), the owner must "identify joint bars with visible or otherwise detectable cracks and conduct remedial action pursuant to § 213.121."
| Track class | Freight, < 40 MGT | Freight, 40–60 MGT | Freight, > 60 MGT | Passenger, < 20 MGT | Passenger, ≥ 20 MGT |
|---|---|---|---|---|---|
| Class 5 and above | 2 | 3 | 4 | 3 | 3 |
| Class 4 | 2 | 3 | 4 | 2 | 3 |
| Class 3 | 1 | 2 | 2 | 2 | 2 |
| Class 2 | 0 | 0 | 0 | 1 | 1 |
| Class 1 and excepted | 0 | 0 | 0 | 0 | 0 |
Source: 49 CFR 213.119(h)(6). The rule also sets minimum spacing: 4 inspections a year must be at least 60 days apart, 3 at least 90 days, 2 at least 120 days, and 1 at least 180 days. Where freight and passenger trains share the track, the more frequent count applies. Extreme weather can extend an interval by up to 30 days. CWR joints in switches, turnouts, track crossings and lift rail assemblies on movable bridges are inspected on foot at least monthly under 213.235.
Under 213.119(h)(3), each inspection must look at least for loose, bent or missing bolts; rail end batter or mismatch; and signs of excessive longitudinal rail movement, such as a wide joint gap or displaced anchors. Records are made on the day, signed, and must locate each problem precisely enough to find the joint again (213.119(h)(7)).
The rule doesn't say how a crack is to be found. "Visible or otherwise detectable" leaves room for UT, cameras or a close look on foot. Joints in jointed (non-CWR) track have no equivalent federal table; they're covered by the visual track inspection schedule in 213.233. Our guide to continuous welded rail explains the rest of the CWR plan.
Worked example: joint inspections on a mixed-traffic line
A Class 4 CWR line carries 50 MGT a year, mostly freight, plus a commuter service.
- Freight column: Class 4, 40–60 MGT gives 3 inspections a year.
- Passenger column: Class 4, 20 MGT or more gives 3 a year.
- Mixed traffic: the more frequent figure applies, so 3 a year, at least 90 days apart. Without the commuter trains the answer would still be 3, from the freight column.
Now compare that with how fast a joint bar crack can grow. In a 2008 finite element study, the Transportation Technology Center, Inc. (TTCI) estimated that a joint bar flaw can grow from detectable size to failure in 10 to 50 MGT in heavy axle load service (TTCI Technology Digest TD-08-040). On this line, 50 MGT a year is about 0.14 MGT a day:
| Crack growth window (TTCI model) | Time at 50 MGT a year |
|---|---|
| 10 MGT | about 73 days |
| 50 MGT | about 12 months |
At the fast end, a crack that becomes detectable just after one inspection could fail before the next one, 90 or more days later. TTCI notes its model assumed all-loaded cars and constant support, so real growth is likely slower. The point stands, though: the legal minimum is a floor, and the inspection has to be able to find cracks while they are still small.
Key takeaway: count the joint inspections from the 213.119(h)(6) table, then ask whether that interval is short enough for the tonnage on the line.
Why doesn't standard rail flaw UT find joint bar cracks?
Rail flaw test probes are aimed at the rail, so their sound paths run through the rail head, web and base, not through the bars bolted to its sides. Section 213.237(f) says the equipment "shall be capable of detecting defects between joint bars, in the area enclosed by joint bars." That means the rail ends inside the joint. It says nothing about the bars themselves.
The Transportation Safety Board of Canada (TSB) reached the same conclusion in its 2006 safety issue investigation of broken-rail derailments in western Canada (TSB SII R05-01):
"Current rail defect detection equipment or geometry cars are unable to identify joint bar defects."
The same report noted that there was "no known production method for volume field testing of joint bars" at the time, although bars can be tested with a hand-held transducer, MT or PT. It also found that joint bar fatigue cracks small enough to matter are "impossible to see" from a moving hi-rail vehicle.
The rail inside the joint is a different story. On the six CN subdivisions in that study, 1,164 of the 1,495 rail defects detected ultrasonically in 2001–2003 (78%) were bolt hole cracks or breaks in the joint area; on seven CPR subdivisions the share was 35%. The TSB noted that detecting fatigue cracking around bolt holes "requires ultrasonic inspection". Rail flaw test wheels typically carry angled probes (often around 37° to 45°) for this, alongside 70° and 0° probes for other defects.
| Rail flaw UT (213.237) | Joint bar UT | |
|---|---|---|
| Component examined | The rail, including rail ends between the bars | The joint bars |
| Typical targets | Bolt hole cracks, head-web separation, transverse defects | Fatigue cracks at the top center of the bar |
| Required by FRA rule | Yes, at the 213.237 intervals | No federal requirement to use UT on bars |
| Typical sound path | Down and along the rail from the running surface | Across the bar from its outside face |
| Who does it | Most rail flaw test programs | A few specialized systems and hand UT |
Key takeaway: a clean rail flaw test report says the rail ends in the joint were tested; it says nothing about the joint bars.
Where do joint bar cracks start?
Most joint bar cracks start at the top center of the bar, on the surface that bears against the underside of the rail head. That is the hardest place to see.
TTCI's 2008 feasibility study analyzed joint bar defect reports filed with FRA and a sample of 50 bars removed from service. It found that:
- about 97% of reported joint bar defects were at the top center of the bar, within the middle 4 bolt holes
- about two-thirds of reported joint bar defects were service failures (broken bars), and only one-third were cracks found before failure
- cracked bars usually had flaws that started at the top center, but broken bars more often had flaws that started at the bottom of the bar or at a bolt hole
The TSB's 2015 investigation of the Gladwick, Ontario derailment shows the mechanism (TSB R15H0013). An insulated joint in 136 lb CWR broke under a crude oil unit train at −31 °C, and 29 cars derailed. The TSB laboratory found fatigue cracks in both bars starting at the upper fishing surface, the mating surface between the top of the bar and the underside of the rail head. The cracks grew through about 20% of the bar's cross-section before the bars broke. Secondary fatigue cracks sat below the bolt hole closest to the fracture.
The same report links joint bar cracks to the condition of the joint. Poor support under the joint raises wheel impact forces and rail deflection. That loosens the assembly, causes rail end batter, and drives fatigue cracking in the bars and at the bolt holes. In that case a rail flaw test had recorded 3.5 mm of rail end batter at the joint about a month before the failure. Cracked insulated joints are covered in more detail in our guide to insulated rail joint inspection.
Key takeaway: rail end batter, a low joint or loose bolts are the warning signs that a joint bar may be cracking where no one can see it.
How does ultrasonic testing detect joint bar cracks?
Joint bar UT sends sound across the bar from its exposed outside face toward the hidden top inside corner, and reads the echo from a crack there. A crack is a steel-to-air boundary, so it reflects most of the sound that reaches it. The time of flight places the reflector, and the probe position shows where along the bar it is.
The TSB described the system that TTCI and Herzog developed (R15H0013, citing TTCI and a 2010–2011 trade article):
"This system uses ultrasonic transducers mounted into a sliding fixture or roller search unit to scan along the outside of a joint bar while introducing pulsed sound waves across the bar in order to detect flaws and cracks located at the top inside surface of the joint bar." (TSB R15H0013)
The system targets the part of the bar "masked by the railhead to web radius", which cameras can't see. Herzog holds US patents on in-track joint bar UT. One describes a transducer on an actuator that engages the vertical face of the joint bar head, positioned with a laser profile system (US 8,365,604). The same patent notes that joint bars are typically 24 or 36 in. long, with 4 or 6 bolt holes.
What has been shown in testing:
| Study | Approach | Reported result | Limits of the result |
|---|---|---|---|
| TTCI TD-08-040 (2008) | Contact UT on the top of the bar | Nonvisible flaws 0.5 in. long on the top of the bar "can be found reliably" | Laboratory tests on artificial and service flaws; full cross-section coverage not shown |
| MxV Rail and University of Missouri, TD23-017 (2023) | Surface waves on unattached joint bars | A 0.5 MHz transducer was the best of 3 tested; best signal at 30° yaw and 1.5 in. offset | Laboratory, simulated cracks |
| Agbede et al., Engineering Failure Analysis 167 (2025) | Surface waves on bars with implanted cracks of 3.175, 6.35, 12.7 and 25.4 mm (1/8 to 1 in.) | 0.5 MHz most suitable for cracks in the inaccessible head region | Laboratory, implanted cracks |
| MxV Rail and University of Missouri, TD24-016 (Phase III, 2023 work) | Surface waves on bars bolted to rail | Results similar to unattached bars; crack signals about 20 dB (10% amplitude) below the calibration notch | "Further work is needed to advance the technology readiness level" |
These figures belong to the systems and specimens tested. None is a field probability of detection for joint bar UT, and none should be quoted as one.
Why surface waves?
A surface (Rayleigh) wave travels along the steel surface, so it can be launched from the part of the bar head that a probe can reach and carried around into the region against the rail head that it can't. A lower frequency gives a longer wavelength that penetrates deeper below the surface and is less disturbed by rust and roughness. That fits the 0.5 MHz result, but at the cost of resolution for very small cracks. The MxV Rail work reports that signal-to-noise ratio across varying bar profiles is still the main obstacle.
Key takeaway: contact UT across the bar is proven in the laboratory for top-center cracks; surface-wave methods for the hidden head region are still research.
What does a joint bar UT inspection look like in practice?
A hand UT check of a suspect joint follows the same discipline as any contact UT procedure: a written technique, a reference standard, and documented scanning. The specific angles and settings come from the procedure, which a Level III should write and approve.
- Identify the joints. Use the CWR joint list, rail flaw test reports (rail end batter, bolt hole cracks) and geometry data (low joints) to pick which joints to test first.
- Clean the scanning surface. Remove scale, dirt and loose rust from the outside face of the bar so the probe couples consistently.
- Calibrate on a reference block. Set range and sensitivity on a block with a known reflector, ideally a notch in a section of the same joint bar profile.
- Scan the high-risk zone first. Cover the top of the bar across the middle bolt holes, where most cracks start, from both the field and gauge sides.
- Separate geometry from cracks. Bolt holes, bar ends and the bar's own corners all return echoes. Plot each indication against the bar's known geometry before calling it.
- Confirm and record. Confirm suspect indications from a second angle or with MT where the surface is accessible. Record the joint's location, which bar, the indication position and the action taken.
- Hand off to remedial action. A confirmed crack between the middle bolt holes means replacement under 213.121(c), on any class of track.
Key takeaway: an indication at a bolt hole may be geometry, not a crack, so each echo has to be checked against the bar's known shape.
What can't joint bar UT detect?
Joint bar UT can miss cracks that are tilted away from the beam, too small for the setup, or hidden by poor coupling. It also tells you nothing about why the joint is cracking.
- Crack orientation. A crack lying parallel to the beam returns little sound. Bottom-of-bar and bolt-hole cracks, which TTCI found more common in broken bars, may need a different probe position than top-center cracks.
- Coverage. TTCI noted that, as with rail, a system contacting the bar from one face may not inspect the whole cross-section.
- Geometry echoes. Bolt holes, bar ends and section changes produce signals that can mask or imitate a crack.
- Surface condition. Heavy corrosion pitting, scale and paint weaken coupling. The TSB found corrosion pits at the crack origins in the Gladwick bars.
- Embedded joints. Joints in road crossings or other structures may not be reachable without removing material; 213.119(h)(4) requires a procedure for them.
- Joint condition. UT doesn't measure support, bolt tension, joint gap or rail end batter. An on-foot inspection is still needed for those.
- No regulatory driver. In 2015 the TSB noted that, "unlike track geometry inspection and rail flaw testing, the use of high-speed cameras and ultrasonic testing to inspect joint bars is not required by regulation, and is not yet widespread throughout the industry."
A clean UT result means no reflector was found along the paths that were scanned. It doesn't prove the bar is sound.
How does UT compare with other rail joint inspection methods?
No single method covers a whole joint, so a sound rail joint inspection program combines them. Each method answers a different question.
| Method | What it finds at the joint | What it misses |
|---|---|---|
| On-foot visual inspection | Exposed cracks, loose or missing bolts, rail end batter, joint gap, pumping | Cracks hidden under the rail head; small cracks |
| Machine vision from a moving vehicle | Cracks that reach the exposed face of the bar; missing bolts | Cracks on surfaces facing the rail (the TSB notes "only cracks that extend to the outside (exposed) surface" are visible) |
| Rail flaw UT (213.237) | Bolt hole cracks, head-web separation and other rail defects in the rail ends | The joint bars themselves |
| Joint bar contact UT | Top-center cracks hidden by the rail head-to-web radius | Some orientations; full cross-section not shown |
| Surface-wave UT | Research stage: cracks in the inaccessible head region | Not yet a field method |
| MT or PT on removed or exposed bars | Surface-breaking cracks on the surfaces that can be reached | Cracks on hidden faces while the bar is in track |
The machine-vision system the TSB describes was developed by FRA's research office and ENSCO. It uses 4 line-scan cameras at up to 70 mph, and some rail flaw test vehicles carry it. It is a real gain over walking, but it only sees what reaches the outside of the bar.
The most useful link in practice is between the rail flaw test and the joint inspection. A rail flaw report that flags rail end batter, a bolt hole crack or a poor test through a joint is a reason to look hard at the bars at that joint. Our article on computer vision for track inspection covers camera systems, and rail joints covers joint types.
Key takeaway: use the rail flaw test to find rail defects in the joint, and use joint inspection, by eye, camera or UT, to find cracks in the bars.
How EMA Industries applies this
EMA Industries performs ultrasonic rail flaw testing for transit agencies, short lines and industrial track owners, plus rail weld inspection, MT and PT, and NDT Level III services. We're based in Southern California and mobilize nationwide. On a joint, the rail flaw test should report bolt hole cracks and rail end batter by joint location, so the owner can feed those joints straight into the 213.119(h) joint inspection list. For suspect joint bars, a Level III can write the hand UT or MT procedure and its reference standard. To scope rail flaw testing or a joint bar check, see our ultrasonic testing services or contact us.
Related reading: How ultrasonic testing works for railways · 49 CFR 213.237 rail flaw detection requirements · Rail fatigue crack detection · Surface wave UT for rail
Frequently asked questions
How are joint bar cracks detected?
Most are found by eye, either on foot or by machine-vision cameras, but only once the crack reaches the exposed face of the bar. Cracks at the top center of the bar, under the rail head, need joint bar ultrasonic testing that sends sound across the bar from its outside face. Standard rail flaw UT doesn't examine the bars.
Does a rail flaw test inspect the joint bars?
No. Under 49 CFR 213.237(f), rail flaw equipment must detect defects in the rail enclosed by the joint bars, such as bolt hole cracks. The bars themselves are a separate component under 213.121, and the TSB has found that rail defect detection equipment can't identify joint bar defects.
What must happen when a cracked joint bar is found?
Under 49 CFR 213.121, a bar cracked or broken between the middle two bolt holes must be replaced on any class of track. On Class 3 to 5 track, a bar cracked or broken anywhere must be replaced. In CWR track, 213.119(h)(2) requires remedial action under 213.121.
How many bolts does a rail joint need?
At least 2 bolts per rail at each joint on Class 2 to 5 jointed track, and 1 on Class 1, under 49 CFR 213.121(d). Every CWR joint needs at least 2 bolts per rail on any class (213.121(e)).
What are the main types of rail joints?
Standard bolted joints join two rails of the same section with joint bars. Insulated joints add an insulating layer to separate signal track circuits, and are often bonded with epoxy. Compromise joints join rails of different sections. In CWR track, bolted joints remain at plugs, temporary repairs, switches and circuit boundaries.
Sources
- 49 CFR 213.121, Rail joints, eCFR (text read on Cornell LII, October 2026)
- 49 CFR 213.119, CWR plan contents, eCFR, paragraph (h) joint inspection procedures and table
- 49 CFR 213.113, Defective rails, eCFR, bolt hole crack definition and remedial action table
- 49 CFR 213.237, Inspection of rail, eCFR, paragraph (f)
- 49 CFR 213.233, Track inspections, eCFR
- TSB Rail Transportation Safety Issue Investigation Report SII R05-01, Transportation Safety Board of Canada, 2006
- TSB Railway Investigation Report R15H0013, main-track derailment, Gladwick, Ontario, 14 February 2015
- Davis, Akhtar and Garcia, Evaluation of the Feasibility of Automated Joint Bar Inspection, TTCI Technology Digest TD-08-040, October 2008
- Poudel, Washer and Shrestha, Non-Contact Ultrasonics for Joint Bar Defect Detection, Phase II, MxV Rail TD23-017, 2023
- Poudel and Washer, Optimizing Surface Waves for Joint Bars Defect Detection, MxV Rail TD24-016, 2025
- Agbede, Yadav, Washer and Poudel, Study on surface waves for detection of fatigue cracks in railway joint bars, Engineering Failure Analysis 167, 109001 (2025)
- US Patent 8,365,604, Apparatus for and method of detecting defects in a rail joint bar, Herzog Services, 2013