Ultrasonic Testing vs Other NDT Methods: What's Best for Each Railway Component?

Ultrasonic testing vs other NDT methods on a railway comes down to the component. Under 49 CFR 213.237, Class 4 and 5 track must be internally inspected at least every 370 days or 30 million gross tons, and ultrasonic testing (UT) is how owners do it. Magnetic particle, penetrant, eddy current and visual methods cover what UT reads poorly.

No single method is "best for railways". Rail in track, field welds, cast manganese frogs, switch points, joint bars and bolts each fail in different ways, and each sits in a different place for an inspector to reach. This article matches methods to components, gives a decision matrix and a worked turnout example, and is honest about where each method falls short.

For how each method works in general (physics, equipment, codes across industries), see EMA's comparison of non-destructive testing methods. This page stays on the railway.

Contents: Quick answer · What 213.237 requires · Rail in track · Welds · Frogs and switch points · Joint bars and bolts · Wheels and axles · Decision matrix · Cost and time · Worked example · Limits · FAQ

Which NDT method fits which rail component?

UT is the primary method for anything that can fail from the inside: rail in track, rail welds and bolt holes behind joint bars. Surface methods are chosen where the defect breaks the surface, where the material defeats UT, or where the part is off the track and on a bench.

Key takeaway: pick the method by where the defect starts and whether the part can be reached, not by which method is "most advanced".

Does 49 CFR 213.237 require ultrasonic testing?

49 CFR 213.237 requires internal rail inspection, but the rule text never names a method. It requires each track owner to "conduct internal rail inspections" at set intervals, using "rail defect detection equipment" run by a qualified operator. The word "ultrasonic" does not appear in 213.237, 213.238 or anywhere else in Part 213.

What the rule does say about the test and its equipment is short and specific:

• Interval (213.237(c)). Class 4 and 5 track, and Class 3 track with regularly scheduled passenger trains or a hazardous materials route: no more than 370 days or 30 MGT between inspections, whichever is shorter. Other Class 3 track follows a longer schedule with a 5-year cap. There is no 213.237 interval for Class 1 or 2 track.
• Joint bars (213.237(f)). The equipment must see inside the joint.
• Valid search (213.237(j)(4)). "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."
• Untested rail (213.237(g)). If the operator decides a valid search could not be made over a length of track, that length doesn't count as inspected.
• Equipment check (213.238(b)(2)). The qualified operator must be authorized to "determine that such equipment is performing as intended". The rule sets no calibration method, sensitivity or test speed.

The joint-bar sentence is the only equipment performance requirement in the section:

"Inspection equipment shall be capable of detecting defects between joint bars, in the area enclosed by joint bars." — 49 CFR 213.237(f)

In practice, UT is the method that meets these duties on a running railroad. Magnetic particle and penetrant testing can't produce a continuous internal search of miles of rail, so they can't satisfy 213.237. They are used on welds, switches and frogs alongside it, which no FRA rule mandates. The full rule, including service failure rates, is explained in EMA's 49 CFR 213.237 guide and the FRA track safety standards overview.

Key takeaway: 213.237 is performance-based; UT is the industry's way of meeting it, not a method the rule names.

How is rail in track inspected, and where do other methods help?

Rail in track is inspected internally by UT, from a test vehicle or a hand-pushed cart, with visual inspection under 213.233 and eddy current on surface damage as supplements. UT reads the head and web; the other methods read the running surface that UT can't interpret.

Rail test systems typically combine a 70° shear-wave beam for transverse defects (transverse fissures, detail fractures) with a 0° beam for horizontal splits and loss of back-wall, plus side-looking or 37–45° channels. Configurations vary by system and owner. The physics behind these angles is covered in the science behind ultrasonic testing.

Vehicle UT has a known blind area at the bottom of the rail. A TTCI study of thermite welds put it plainly:

"The ultrasonic rail flaw detection process only scans the head and web of the rail, and not the majority of the base." — TTCI Technology Digest TD-00-013 (2000)

Why do MT and eddy current complement UT on the rail head?

UT reads the first few millimeters under the running surface poorly, and that is where rolling contact fatigue (RCF) cracks live. The reason is not a simple probe dead zone. Rail systems couple through fluid-filled wheels, sleds or delay paths, and the 70° path enters the head at an angle, so the classic contact-probe dead zone (explained in EMA's dead zone article) is largely engineered out.

The real near-surface limits are these:

• Gating. FRA's 2011 RCF review notes that UT analysis ignores "the first few millimeters" because the surface returns noise there (DOT/FRA/ORD-11/24).
• Shallow cracks. In TTCI TD-18-027 (2018), 90% of 880 RCF cracks and spalls examined were 0.040 in. (1 mm) deep or less (TD-18-027).
• Masking. Heavy head checks and shelling scatter the beam, so UT can't see a detail fracture growing beneath them. The same FRA review says surface fatigue "prevents the successful transmission of ultrasonic waves to the base of the rail and back", which forces a no-test and a hand test. EMA's rail shelling and spalling article covers this in depth.

Eddy current is the in-track answer for head checks. A 2014 TTCI trial ran a four-probe eddy current inspection vehicle (a draisine) over the top and gauge corner and tracked crack depth falling from 0.14 in. to under 0.04 in. across five grinding passes (TD-14-007). Eddy current needs a close, steady standoff (about 1–2 mm in the FRA review) and sizes depth only approximately. EMA's eddy current vs ultrasonic testing article compares the two in detail.

Key takeaway: UT finds the defects that break rail; eddy current, visual inspection and grinding manage the surface damage that hides them.

What are the practical limits of MT and PT on rail in track?

Magnetic particle (MT) and penetrant testing (PT) are spot methods. They are useful on a weld, a switch point or a crack someone has already seen, and impractical as a way to inspect running rail. A detector car covers miles in a shift; an MT technician covers inches per setup.

Neither method finds a transverse fissure inside the head, and neither produces a record that counts toward 213.237. For a side-by-side of UT and MT on rail, see ultrasonic testing vs magnetic particle testing in rail and EMA's magnetic particle examination page.

Which methods inspect thermite and flash-butt rail welds?

Hand UT of the full weld (head, web and base), with a visual geometry check and MT or PT on the finished surface, is the usual combination for rail welds. Radiography (RT) is possible, but it is rarely the routine choice in track.

Thermite welds. The same TTCI digest broke down 265 thermite weld service failures: slag and porosity caused 40% (56% of those at the web-to-base fillet), run-out fatigue 32% and hot tears 15% (TD-00-013). Most of those start in the web and base, the region a detector car mostly doesn't scan. So each new field weld needs its own hand UT, including the base, plus a check of crown height, alignment and the finish of the ground surface. Details are in EMA's aluminothermic weld inspection guide.

Flash-butt welds. Plant and mobile flash-butt welds fail mainly through planar defects: lack of fusion and "cold" welds from too little heating before upset. UT is the method that can find a planar, internal flaw; MT or PT only checks the trimmed upset for surface cracks. See flash-butt weld inspection.

Once in track, a bad weld is a rail defect. 49 CFR 213.113(d)(6) defines a defective weld as a flaw from incomplete penetration, lack of fusion, slag or sand, under-bead or shrinkage cracking, or fatigue cracking, starting in the head, web or base. The remedial action then follows the 213.113 table by size.

Where does radiography fit for rail welds?

RT gives a permanent image of volumetric flaws like porosity and slag, but in track it needs a radiation-controlled area, a licensed crew and track time, and it is weak on tight planar cracks that are not aligned with the beam. For those reasons it is usually reserved for weld-process qualification, investigations and research rather than routine acceptance.

The railroad industry has long questioned how well a radiograph predicts weld life. A 1970 study compared radiographs of thermite welds in track with laboratory rolling-load tests and concluded that "more study is needed on the effectiveness of the radiographic examination technique" for thermite welds in service (AREA Bulletin 626, 1970, via TRID).

Field radiography carries its own rules. Under the NRC's 10 CFR 34.41, radiography away from a permanent installation requires the radiographer to be accompanied by at least one other qualified individual. Agreement States, including California, apply equivalent rules, and the crew must also set and control a radiation boundary while workers and trains stay clear.

Which code sets acceptance for each method: AWS D1.1 or ASME V?

Neither one governs rail welds in track. The acceptance criteria for a thermite or flash-butt rail weld come from the track owner's specification and the weld-kit or plant procedure, often built on AREMA Manual for Railway Engineering Chapter 4 (Rail) practice. Check what your contract names.

The two codes do matter around a rail project, in different ways:

Within D1.1, RT and UT acceptance criteria differ for statically and cyclically loaded connections, and cyclic criteria are stricter. Cracks are rejectable regardless of size. The specific clause and figure numbers vary by edition, so confirm them in the edition your contract cites. D1.1 also doesn't require any NDT beyond visual inspection by itself; the contract documents set the extent of testing. EMA's AWS D1.1 weld inspection criteria article covers this in detail.

Key takeaway: UT is the volumetric method for rail welds; RT is a specialist tool, and AWS D1.1 belongs to the structural steel around the track, not the rail.

How are frogs and switch points inspected?

Frogs and switch points are inspected mainly by eye, monthly on foot under 49 CFR 213.235, with PT or MT used to confirm cracks and hand UT used only where the steel and geometry allow it. Turnout inspection is covered in full in EMA's railroad switch and crossing inspection article.

Cast manganese frogs. Most mainline frogs and crossing diamonds are cast austenitic manganese steel (AMS). FRA's 2022 research report describes AMS as "highly attenuative" and "coarse-grained, nonhomogeneous, and anisotropic", and records current practice:

"Revenue service examinations of these special trackwork components currently use only visual inspections." — FRA RR 22-25, Wilson and Poudel

The coarse grain scatters ultrasound into noise, which is why conventional rail UT struggles on frogs. Research systems using low frequencies and synthetic aperture focusing have shown promise, but they are not routine field tools. AMS is austenitic and not suited to MT, so PT is the practical surface method on a manganese casting. Rail-bound manganese frogs also contain carbon rail sections, where MT and UT work normally.

Switch points. Switch points are machined rail steel, so MT and PT both work on them. Damage is mostly at the surface: chipped or worn points and cracks near the tip. The thin, tapered section limits hand UT near the point, while the thicker heel can be tested normally. Under 49 CFR 213.135(b), cracked switch point rails follow the 213.113 remedial actions, except that remedial action B governs where joint bars can't be placed. 213.237's text never mentions frogs or switch points.

Key takeaway: in a turnout, UT is strongest on the plain rail and welds; frogs rely on visual inspection and PT.

How are joint bars and bolts inspected?

Joint bars are inspected visually in service, while the rail inside the joint, including the bolt holes, is inspected by UT through the bars. The two are easy to confuse: 213.237(f) is about defects in the rail between the bars, not cracks in the bars themselves.

Joint bars. Under 49 CFR 213.121(c), on any track class, a joint bar "cracked or broken between the middle two bolt holes" must be replaced. About 97% of reported joint bar defects are at the top center of the bar, and research by TTCI and MxV Rail found that UT can find 0.5 in. top-of-bar flaws, with 0.5 MHz working best. That work is research, not standard practice; in routine maintenance, bars are checked by eye and by MT or PT once removed. See EMA's rail joint inspection article.

Bolt holes. Bolt hole cracks grow from the hole edge into the web, hidden behind the bars. UT finds them without removing the bars, and that is exactly what 213.237(f) asks the equipment to do. Assigning bolt holes to "visual plus MT" is a common error, because the inspector can't see or magnetize a hole covered by a bar. The defect and its remedial bands are in bolt hole cracks in rail.

Track bolts. Track bolts are checked visually and for tightness during track inspection, and 213.121(d) requires at least two bolts per rail at each joint on Class 2–5 jointed track (one on Class 1). A cracked or stretched bolt is normally just replaced. MT or UT of bolts is a shop or failure-investigation task on removed bolts, or applies to structural anchor rods, not routine track work.

What about wheels and axles?

Wheels and axles are rolling stock, not track, so different rules apply: the freight car standards in 49 CFR Part 215 and AAR specifications. Part 215 doesn't name a test method either. Under 49 CFR 215.103(d), a car may not be placed in service if a wheel "rim, flange, plate, or hub area on the car has a crack or break", and 215.105(a) applies the same to an axle that "has a crack or is broken".

In practice, new wheels and axles are ultrasonically tested at manufacture to AAR specifications (commonly cited as M-107/M-208 for wheels and M-101 for axles), and shops use UT and MT on axles during overhaul. Wayside systems screen wheels in service. This is a separate discipline with its own codes, so this article treats it only briefly.

Ultrasonic testing vs other NDT methods: the decision matrix

The matrix below rates each method for each rail component: P = primary, S = supplement, L = limited or special cases, X = not suitable. It reflects common practice and the sources cited in this article. Owner procedures and contracts take precedence.

Phased array (PAUT) appears as L almost everywhere because, in rail, it is mainly a verification and research tool. Rail test cars produce B-scan style channel maps and A-scans, not routine phased array images. Where PAUT does pay off is in the detailed sizing of a weld indication; EMA's UT vs PAUT article explains the trade-off and the sizing caveats.

How do the methods compare on cost and time?

Per mile of rail, vehicle UT is by far the cheapest way to search for internal defects; per spot, MT, PT and visual are cheapest; RT is the most expensive per item because of radiation controls and crew size. No prices are given here, because rates depend on mobilization, track access and volume.

The hidden cost on a railway is track time. Every stop needs protection under Part 214, so a method that is cheap per spot but needs 30 stops can cost more in lost possession time than one test-car pass. EMA's rail testing cost article looks at what drives the price of a UT program.

Worked example: choosing methods for a turnout renewal

Scenario. A track owner renews one No. 10 turnout on Class 4 continuous welded rail main track. The new turnout has a rail-bound manganese frog and new switch points. It is joined into the existing CWR with 4 thermite welds (2 at the switch end, 2 at the frog heel end), and 2 more thermite welds join the closure rails. The turnout uses no bolted joints except the frog's own bolted assembly.

Step 1: List the components and how each fails.

Step 2: Count the examinations.

• Hand UT: 6 welds × 1 examination each = 6 UT weld reports.
• Surface method on welds: 6 welds × 1 MT examination each = 6 MT reports.
• Switch points: 2 points × 1 visual and fit check = 2 checks, plus PT or MT only where a mark is found.
• Frog: 1 records review and visual check, plus PT on suspect marks.
• Frog bolted assembly: 1 torque and visual check.

That gives 12 weld examinations (6 UT + 6 MT), 3 visual checks (2 points + 1 frog) and 1 torque check, all done in the same work window after the welds have been ground and cooled.

Step 3: Decide what was left out, and why.

• No RT. It would need a radiation-controlled area and a two-person crew on a live main line, for welds that UT covers volumetrically.
• No MT on the manganese casting. It's austenitic; PT covers it.
• No standalone 213.237 test. The rule has no turnout-specific provision, and the turnout's rail enters the owner's normal cycle (370 days or 30 MGT on Class 4).

Result. The owner gets a UT report for every new weld, including the base that a detector car won't cover, and a documented surface check on the parts UT can't read.

What no NDT method can do on rail

Every method on this page misses something, and an inspection plan should say what it doesn't cover.

• UT can't reliably read through shelling or heavy head checks, scans little of the base from a vehicle, and returns a no-test (untested rail under 213.237(g)) where coupling is lost.
• MT and PT find only surface-breaking or very shallow cracks, need clean, dry surfaces and track time, and can't inspect miles of rail.
• Eddy current sizes crack depth only approximately and is sensitive to standoff, so it guides grinding rather than replacing UT.
• RT is weak on tight planar cracks not aligned with the beam, and its radiation controls make it slow on a live railroad.
• Visual inspection sees nothing internal: a transverse fissure is invisible until the rail breaks.
• Cast manganese frogs have no proven routine volumetric field method yet; FRA research describes visual inspection as current practice.

No rule requires 100% detection, and no method achieves it. Testing intervals, remedial action under 213.113 and owner maintenance practices exist because some defects will be missed.

How EMA Industries applies this

EMA Industries provides ultrasonic rail flaw testing for transit agencies, short lines and industrial track owners, plus rail weld inspection and AWS D1.1 structural weld UT, MT and PT. On a turnout renewal or a weld program, EMA can test each new field weld through head, web and base and add MT or PT where UT can't read the surface. EMA is based in Southern California and mobilizes nationwide. To plan the methods for your next project, contact EMA Industries.

Frequently asked questions

Is ultrasonic testing required by FRA for rail?

The rule doesn't name UT. 49 CFR 213.237 requires "internal rail inspections" with equipment that can detect defects between joint bars, and a valid search interpreted by a qualified operator under 213.238. UT is the method the industry uses to meet that requirement.

Can magnetic particle testing be used on rail in track?

Only as a spot check, on welds, switch points or a crack that's already visible. MT needs clean, dry steel, a powered yoke and track time for every setup, and it finds only surface and near-surface cracks. It can't provide the continuous internal search that 213.237 requires.

Why can't standard UT inspect manganese frogs well?

Cast austenitic manganese steel is coarse-grained, so it scatters ultrasound into noise. FRA research report RR 22-25 says revenue-service examinations of these components use only visual inspection. PT is the practical surface method, since the austenitic steel is not suited to MT.

Should rail welds be radiographed or ultrasonically tested?

For routine acceptance in track, hand UT covering head, web and base is the usual choice, because it is fast and finds the planar flaws that cause weld failures. RT is used mainly for weld-process qualification and investigations, because field radiography needs a controlled area and a two-person crew under 10 CFR 34.41 or the equivalent Agreement State rule.

What NDT method finds head checks and gauge corner cracking?

Eddy current, visual inspection and, after grinding, MT or PT are the methods for surface RCF. UT is still needed to find detail fractures growing under the damage, but heavy head checks can mask it, which is why owners combine testing with rail grinding.

Sources

• 49 CFR 213.237, Inspection of rail, eCFR (text read on Cornell LII, October 2026)
• 49 CFR 213.238, Qualified operator, eCFR
• 49 CFR 213.235, Inspection of switches, track crossings and lift rail assemblies, eCFR
• 49 CFR 213.113, Defective rails, eCFR
• 49 CFR 213.121, Rail joints, eCFR
• 49 CFR 213.135, Switches, eCFR
• 49 CFR 215.103, Defective wheel and 215.105, Defective axle, eCFR
• 10 CFR 34.41, Conducting industrial radiographic operations, Cornell LII
• TTCI Technology Digest TD-00-013, thermite weld failures and detection (2000), MxV Rail
• TTCI Technology Digest TD-18-027, RCF and UT confidence (Cummings, 2018), MxV Rail
• TTCI Technology Digest TD-14-007, eddy current rail inspection trial (2014), MxV Rail
• DOT/FRA/ORD-11/24, Rolling Contact Fatigue: A Comprehensive Review (Magel, 2011), ROSA P
• FRA RR 22-25, ultrasonic inspection of cast manganese special trackwork (Wilson and Poudel), ROSA P
• AAR Studies of Butt Welded Rail Joints, AREA Bulletin 71 (626), 1970, TRID record
• ASME BPVC Section V, Nondestructive Examination, 2025 edition (AFNOR store listing)
• AWS D1.1/D1.1M, Structural Welding Code – Steel, American Welding Society