NDT Inspection Services for Rail Welds, Frogs, Switches and Rail Bridges

Component-level NDT inspection services examine the steel around the running rail: field welds, frogs, switch points, bolted joints and bridge steel. Each part has its own acceptance basis. Thermite and flash-butt welds go to AWS D15.2:2022 Tables 8.1 and 8.2 where the owner adopts it, bolt hole cracks to 49 CFR 213.113, and bridge welds to AWS D1.5 or D1.1.

This page is about those components. For the defects inside the running rail itself (transverse fissures, detail fractures, split heads) and which probe finds each, see how rail defects are detected with NDT methods. For choosing and contracting a rail inspection provider, see the buyer's guide to professional rail inspection services.

What do component-level NDT inspection services cover?

They cover the parts a rail flaw test car doesn't fully examine: the weld base, turnout castings, joints and the structures that carry the track. A test car is built to search the head and web of plain rail. Welds, frogs, switch points and bridge members need a different method, a different acceptance document, or both.

The table below is the short answer for each component. The sections that follow give the reasoning and the code references.

Key takeaway: every component needs a named acceptance document before testing starts; a method with no acceptance criterion produces a report nobody can act on.

Which codes and rules apply to each rail component?

AWS D15.2:2022 and AREMA MRE Chapter 4 cover rail welds, 49 CFR Part 213 sets in-service limits and remedial actions, and AWS D1.5 or D1.1 governs bridge and building steel. None of these documents applies itself. An AWS standard, AREMA practice or owner procedure becomes binding when a contract, specification or regulation invokes it.

AWS rewrote D15.2 for the 2022 edition. Its title is now Specification for Joining Railroad Rail and Related Rail Components, and its foreword says the most significant revision changed it "from a recommended practice to a specification" with mandatory language. The scope names crossings and turnouts, "which further include switch points, stock rails, switch point guards, plates, frogs, and frog components." The abstract is short and clear on inspection:

"This document specifies the minimum standards for the welding of rails and related rail components... Inspection methods and acceptance criteria are specified." — AWS D15.2/D15.2M:2022, Abstract (public preview, pubs.aws.org)

One point trips up specifications: ASTM E1444 is not the right MT document for rail or structural work. The current E1444/E1444M-26 is the aerospace practice and refers industrial users to E3024/E3024M. Specify E3024 (with E709 as guidance) unless the code you work to names another document.

Key takeaway: name the code, its edition and the acceptance table in the purchase order; "inspect per AREMA" isn't an acceptance criterion.

How are thermite field welds inspected?

A thermite weld is inspected visually after the mold is removed and the weld is ground, then by hand UT of the head, web and base, and the results are judged against D15.2 Tables 8.1 and 8.2 or the owner's equivalent. D15.2 Clause 8.10 covers thermite acceptance, 8.12 alignment (Table 8.3) and 8.13 rail head finish (Table 8.4).

Hand UT of the base is the step that matters most, because a rail flaw test car doesn't reach it. TTCI's thermite weld failure survey (Kristan, July 2000) put the gap plainly: "the ultrasonic rail flaw detection process only scans the head and web of the rail, and not the majority of the base." Failures concentrate there. In TTCI's FAST data (TD-21-027, Poudel, 2021), about 95% of thermite weld breaks were in the web, the web-base fillet and the base.

TD-21-027 also describes the state of practice after installation: methods "are limited to visual inspection, which does not reveal purely internal defects." That is the case for specifying UT of each new field weld instead of relying on the next test car run.

What does a thermite weld UT examination include?

A typical owner procedure (check yours) runs a 0° straight beam over the head for volumetric defects, then angle beams (commonly 45° and 70°) over the head, web and base from both sides of the weld. TTCI's high-hardness weld trials (TD-19-010, 2019) tested every weld with 0°, 45° and 70° probes. The examiner has to separate real indications from the weld collar and riser geometry; TTCI's 2000 survey attributes false calls to "various surface finishes" from thermite welds responding to 45° and 70° beams.

The defects to target are the ones the survey found in failed welds: slag and porosity (40% of 265 revenue failures, more than half of those at the web-base fillet), run-out and fatigue cracks, hot tears, shrinkage and lack of fusion. Thermite is a casting process, so porosity comes from moisture and gas entrapment, not shielding gas, and field welds receive no post-weld heat treatment.

Visual inspection covers the collar, riser removal, surface cracking and the finish of the running surface. Alignment is measured with a straightedge against the D15.2 Table 8.3 limits or the owner's. The D15.2 preview has no MT or PT acceptance table, so if a specification asks for MT on the running surface, it has to supply its own acceptance criteria.

Key takeaway: specify hand UT of the base for every new thermite weld; a test car will not see most of it.

How are flash-butt welds inspected?

A flash-butt weld is inspected visually after upset trimming and grinding, checked for alignment and finish, and examined by UT, with acceptance under D15.2 Clause 8.11 and AREMA MRE 3.10 quality assurance. Most flash-butt welds are made in the plant when strings of continuous welded rail are built; mobile flash-butt welders also work in track.

Flash-butt welds fail less often than thermite welds; TTCI's TD-06-008 (2006) says so directly ("EFB fail at a lower rate than thermite welds"). They aren't defect-free, though. Their typical flaws are flat, planar regions of incomplete bond on the weld plane: cold welds from too little heat before upset, and silicate "matt spots". These lie across the rail, so an angle beam aimed along the rail is the right tool, and visual checks can't find them.

Two practical points affect flash-butt UT. Un-ground flashing gives geometric reflections that the examiner must recognize. And a conventional flash-butt weld can consume as much as 1.5 in. of rail (TD-06-008), which matters when the record must tie a weld to a rail string and its neutral temperature.

For the full procedures, see EMA's pages on aluminothermic (thermite) weld inspection and flash-butt weld inspection.

How should frogs and manganese castings be inspected?

Cast austenitic manganese steel (AMS) frogs are inspected visually, with penetrant testing to trace crack extent; magnetic particle testing doesn't work on them, and UT needs a special low-frequency procedure. AMS is essentially non-magnetic. ASTM E3024's scope says MT is "not applicable to non-ferromagnetic metals and alloys such as austenitic stainless steels", and AMS behaves the same way for MT. A yoke on an AMS casting produces no useful field.

FRA Research Report RR 22-25 (Wilson and Poudel) explains why UT struggles. Most mainline frogs and crossing diamonds are AMS castings, and AMS "is highly attenuative" and coarse-grained, with a mean grain size of about 0.037 in. (0.94 mm). In its trials a 0.75 in. (19 mm) deep hole gave only a 26% echo at 2.25 MHz. The report states current practice directly:

"Revenue service examinations of these special trackwork components currently use only visual inspections." — FRA RR 22-25 (ROSA P, dot/64412)

Research systems have shown more. TWI's RailSAFT project used low-frequency transducers with synthetic aperture focusing from the top surface. Treat that as research, not a field method you can specify from a standard.

Which methods work on which frog parts?

FRA's limits for frogs are about wear and geometry, not cracks. Under 49 CFR 213.137, a frog point worn more than 5/8 in. down and 6 in. back, or a casting tread worn more than 3/8 in. below its original contour, limits speed to 10 mph. For spring rail frogs, 213.139 requires a frog with a bolt hole defect or head-web separation to be replaced.

Weld repair of AMS castings has its own rules. D15.2:2022 added inspection procedures for AMS (Clause 8.9, with qualification in 7.3 to 7.5 and guidance in Annex B), and AREMA MRE Chapter 5 Section 5.9 covers welding of manganese steel castings. In practice, cracks are ground out until they disappear, and RR 22-25 calls that judgment "highly subjective." PT after each grinding pass gives the repair crew an objective stopping point. For turnout inspection in general, see railroad switch and crossing inspection.

Key takeaway: if a specification calls for MT on a cast manganese frog, it is wrong; ask for PT and a visual record.

How are switch points, bolt holes and joint bars inspected?

Switch points, bolt holes and joint bars are carbon or alloy steel, so VT, MT, PT and UT all work on them, and Part 213 sets the disposition. The method depends on whether the joint bars are on, and on how much of the part you can reach.

Switch points

A switch point is planed to a thin taper, so a carriage probe on a test car loses coupling and its fixed angles stop matching the section. Switch points are therefore mainly a visual and hand-NDT task. MT or PT shows the length of a surface crack on a point or stock rail, and a hand angle beam can follow it into the section where the geometry allows.

Disposition follows 213.113, with one substitution in 49 CFR 213.135(b):

"Broken or cracked switch point rails will be subject to the requirements of § 213.113, except that where remedial actions C, D, or E require the use of joint bars, and joint bars cannot be placed due to the physical configuration of the switch, remedial action B will govern..." — 49 CFR 213.135(b)

Section 213.135(h) also requires unusually chipped or worn points to be repaired or replaced. Arc-weld build-up on points falls under the owner's procedure and, where adopted, D15.2 Clause 8.8 for arc welds.

Bolt holes

Bolt hole cracks are UT territory, because the holes sit behind the joint bars. 49 CFR 213.237(f) requires inspection equipment "capable of detecting defects between joint bars, in the area enclosed by joint bars," and angle beams do that with the bars in place. MT is useful only when the bars are off, for example during joint renewal, to confirm the length of a crack. Detection of bolt hole cracks in plain rail is covered in depth on the bolt hole crack inspection page.

Worked example: disposing of a bolt hole crack

A hand UT examination on Class 3 freight track sizes a bolt hole crack at 1-1/4 in.

1. 1-1/4 in. is more than 1 in. and not more than 1½ in., so the row is H and G.
2. Note H limits speed to 50 mph or the class maximum, whichever is lower. Class 3 freight is already limited to 40 mph under 213.9, so 40 mph stands.
3. Note G requires the rail to be reinspected 30 days after the decision to keep it in service.
4. A second crack in the same rail end counts as the same defect; a crack in the adjacent rail end of the same joint is reported as a separate defect (213.113(d)).

The report must therefore give the crack length to the nearest 1/8 in., the rail end (leaving or receiving), and the remedial action letters, so the owner can schedule the 30-day reinspection.

Joint bars

Joint bars aren't covered by 213.237(f), which concerns the rail inside the joint. Cracked bars fall under 49 CFR 213.121: on any class, "If a joint bar is cracked or broken between the middle two bolt holes it shall be replaced." Most bar cracks start at the top center; TTCI's TD-08-040 found about 97% of reported joint bar defects there. In-service checks are visual (and for CWR joints, part of the 213.119 joint inspection plan). MT on removed bars is a sound shop check, and UT of bars in place is still a research method. For joint UT in general, see rail joint inspection with ultrasonic testing.

How is structural steel on rail bridges and facilities inspected?

Welds on steel railroad bridges designed to AREMA are normally fabricated and inspected to AWS D1.5, and welds on other structures (shops, platforms, supports) to AWS D1.1; each code sets UT, MT and PT rules, but the contract decides the extent of testing. AREMA MRE Chapter 15 (Steel Structures) acknowledges AASHTO and AWS for making the Bridge Welding Code available "for use by reference".

D1.5 is more prescriptive than D1.1 about which welds get which method. FHWA's report FHWA-HRT-04-124 (2005), describing the 2002 edition, summarizes it this way: RT is required for complete joint penetration (CJP) welds in tension or stress reversal, CJP corner joints require UT, either RT or UT is allowed for CJP welds in compression or shear, and MT is required for fillet and partial joint penetration welds. Inspection moved from Clause 6 to Clause 8 in the 2020 renumbering, so confirm each clause and table number in the edition your contract names.

D1.1 doesn't by itself require UT. Its NDT beyond visual inspection must be specified in the contract, and the common triggers come from AISC 360 Chapter N and AISC 341 for seismic work. D1.1 evaluates MT and PT indications against its visual acceptance criteria (Table 8.1 in the 2020 and 2025 numbering; confirm in your edition), and its Clause 8 UT method rates indications against a 0.06 in. side-drilled hole, not a DAC curve. EMA's AWS D1.1 weld inspection criteria page covers those limits.

Why bridge UT calibration needs extra care

In December 2024 FHWA issued a memorandum on a weakness in bridge weld UT. D1.5 requires reference blocks "acoustically equivalent" to the steel being tested, but research found that term undefined and rarely enforced. Some plate, especially thermo-mechanically processed (TMCP) steel, has a shear velocity different from the block, or one that changes with rolling direction.

"Furthermore, the inspector has no indication this is happening and believes the weld is defect free." — FHWA Office of Bridges and Structures, memorandum of December 13, 2024 (fhwa.dot.gov)

FHWA recommends contract special provisions to measure shear velocity in the reference block and the test plate, in both directions. The memo is written for highway bridges, but the physics is the same on a railroad bridge built to D1.5, so a rail bridge owner can adopt the same provision.

Bridge inspection under 49 CFR Part 237 is a separate duty. It requires an annual inspection of each bridge, no more than 540 days apart, by a designated bridge inspector, and it covers the structure. Rail flaw testing of the rail on the deck and approaches stays under 213.237. See railway bridge inspection for the inspection program itself.

Worked example: rating a UT indication on a bridge weld

The D1.1 and D1.5 UT methods both rate an indication as d = a − b − c, where a is the indication level in dB, b the reference level on the 0.06 in. side-drilled hole, and c the attenuation factor, 2 × (sound path in inches − 1) dB. A lower (more negative) d means a stronger reflector.

Inputs: a 1 in. CJP butt weld in tension, 70° probe, reference level b = 46 dB, indication found at a sound path of 2.5 in.

1. Attenuation factor: c = 2 × (2.5 − 1) = 3 dB.
2. Depth check: 2.5 × cos 70° = 0.855 in. That is less than the 1 in. thickness, so the indication is on the first leg. (The first leg ends at a sound path of 1 ÷ cos 70° = 2.924 in.) The surface distance from the beam exit point is 2.5 × sin 70° = 2.349 in.
3. Indication 1 at a = 52 dB: d = 52 − 46 − 3 = +3.
4. Indication 2 at a = 59 dB: d = 59 − 46 − 3 = +10.

FHWA's illustration of D1.5:2002 Table 6.3 for 1 in. plate in tension gives Class A at +8 or less (rejectable under all conditions), Class B at +9 (rejectable if longer than about 3/4 in., 20 mm), Class C at +10 (rejectable if longer than about 2 in., 50 mm) and Class D at +11 or more (acceptable). On those values, indication 1 is Class A and rejectable at any length. Indication 2 is Class C, so a 1-1/2 in. long indication is acceptable and a 2-1/2 in. one isn't.

The class bands change with thickness, probe angle, code, edition and tension or compression, so these numbers illustrate the method only. Take the band values from the acceptance table of the edition named in your contract.

Worked example: which yoke lift check applies?

An AC/DC yoke is checked by lifting a test weight before use, and the weight depends on the document you work to. Using DC on a bridge weld with the poles set 5 in. apart:

At a 5 in. spacing, D1.5 calls for the 50 lb DC check, while an ASME-based procedure would call for 40 lb at its maximum spacing. Name the document beside the figure in the procedure, and record the check weight on the report. (The D1.5 values come from secondary copies of the code; confirm them in your edition.)

What should an NDT inspection report show for each component?

A component report must let a third party repeat the examination and reach the same disposition: the part, the acceptance document and edition, the technique, the coverage, each indication and the decision. D15.2 has its own documentation clause (8.7), and owner specifications usually add weld numbering and location formats.

The coverage line is the one most often left blank. A report that says "no recordable indications" without saying the base was hand-scanned can't be used to accept a thermite weld.

Key takeaway: if the report doesn't name the acceptance table and the coverage, the owner can't rely on it.

What can component NDT not do?

Each method has blind spots, and several components can't be fully examined with field NDT at all. State these limits in the procedure and the report.

• Thermite weld collars give geometric echoes that can mask small defects near the fusion line, and the cast structure scatters sound. A clean UT result is not proof of a sound weld; it is proof that nothing exceeded the reporting level in the zones covered.
• Cast manganese frogs can't be examined internally by conventional UT, and MT doesn't work on them. Field examination is visual plus PT, which finds only cracks open to the surface.
• MT finds surface and near-surface discontinuities only. Colorado DOT's MT procedure for D1.5 work warns that cracks more than 1/64 in. below the surface can't be reliably detected. It also needs two magnetizations at about 90° to each other.
• PT finds only defects open to the surface, and it is unreliable on rough, painted or contaminated surfaces.
• Flash-butt bond defects are thin and planar; a defect that is very tight may reflect little sound, so UT sensitivity has to be proven on a reference with a suitable reflector.
• Bridge weld UT can be defeated by acoustic differences between the plate and the reference block, which the operator can't see on screen (FHWA, 2024).
• No NDT replaces Part 213 inspection. A component examination doesn't satisfy 213.233 visual inspection or 213.237 internal rail testing, and it doesn't satisfy the Part 237 bridge inspection.

How EMA Industries applies this

EMA Industries provides rail weld inspection, ultrasonic rail flaw testing, AWS D1.1 structural weld UT, MT and PT, and NDT Level III services. It is based in Southern California and mobilizes nationwide. On component work, EMA's first step is to agree the acceptance document, edition and coverage with the owner, so each report ends in a disposition. See EMA's ultrasonic examination services, or contact EMA Industries with your weld, turnout or structure scope.

Frequently asked questions

Can magnetic particle testing be used on manganese frogs?

Not on the cast manganese body. Austenitic manganese steel is essentially non-magnetic, and ASTM E3024 excludes non-ferromagnetic alloys from MT. Use penetrant testing for surface cracks on the casting. MT still works on the carbon steel rails of a rail-bound frog and on exposed bolt holes.

What standard sets acceptance criteria for thermite and flash-butt welds?

AWS D15.2/D15.2M:2022, Specification for Joining Railroad Rail and Related Rail Components, sets them in Clauses 8.10 and 8.11. Table 8.1 gives visual allowances, Table 8.2 UT acceptance, and Tables 8.3 and 8.4 alignment and finish. It binds only when the owner's specification adopts it; many owners use their own procedures alongside AREMA MRE Chapter 4 Part 3.

Does AWS D1.1 apply to rail welds?

AWS D1.1 doesn't cover rail welds in track. D1.1 is the structural steel code, for buildings, platforms and other structures. Rail joining is covered by AWS D15.2, AREMA MRE Chapter 4 and owner specifications, and steel railroad bridges designed to AREMA are normally welded to AWS D1.5.

Why can't a rail test car inspect thermite welds fully?

Test cars search the head and web, not most of the base, and the base and web-base fillet are where thermite welds usually fail. TTCI found about 95% of thermite breaks at FAST were in the web, fillet or base. New field welds need a hand UT examination that covers the base.

Is ASTM E1444 the right MT standard for rail and bridge work?

ASTM E1444 is not the right choice for rail or bridge work. The current edition, E1444/E1444M-26, is the aerospace MT practice. For general industrial and structural MT, specify ASTM E3024/E3024M-22a with E709 as guidance, unless the welding code you work to names a different document.

Sources

• AWS D15.2/D15.2M:2022, Specification for Joining Railroad Rail and Related Rail Components, public preview (title page, abstract, foreword, scope, table of contents): https://pubs.aws.org/Download_PDFS/D15.2-D15.2M-2022_PV.pdf
• AWS, AASHTO/AWS D1.5M/D1.5:2025 Bridge Welding Code, product page: https://pubs.aws.org/p/2269/aashtoaws-d15md152025-bridge-welding-code
• AWS, D1.1/D1.1M Structural Welding Code – Steel: https://www.aws.org/standards-and-publications/codes-and-standards/d1-1/
• AREMA, Manual for Railway Engineering 2025, Chapter 4 table of contents: https://publications.arema.org/publications_toc/MRE_006/Chapter%2004%20TOC.pdf
• 49 CFR 213.113, Defective rails: https://www.ecfr.gov/current/title-49/subtitle-B/chapter-II/part-213/subpart-D/section-213.113
• 49 CFR 213.121, Rail joints: https://www.ecfr.gov/current/title-49/subtitle-B/chapter-II/part-213/subpart-D/section-213.121
• 49 CFR 213.135, Switches: https://www.ecfr.gov/current/title-49/subtitle-B/chapter-II/part-213/subpart-D/section-213.135
• 49 CFR 213.137, Frogs: https://www.ecfr.gov/current/title-49/subtitle-B/chapter-II/part-213/subpart-D/section-213.137
• 49 CFR 213.237, Inspection of rail: https://www.ecfr.gov/current/title-49/subtitle-B/chapter-II/part-213/subpart-F/section-213.237
• 49 CFR Part 237, Bridge Safety Standards: https://www.ecfr.gov/current/title-49/subtitle-B/chapter-II/part-237
• FRA Research Report RR 22-25, cast manganese frog inspection (Wilson, Poudel): https://rosap.ntl.bts.gov/view/dot/64412
• TTCI TD-21-027, thermite weld inspection after installation (Poudel, 2021): https://mxvrail.com/wp-content/uploads/2023/01/TD21-027.pdf
• TTCI TD-00-013, Thermite Weld Failure Survey (Kristan, July 2000): https://mxvrail.com/wp-content/uploads/2023/01/TD00-013.pdf
• TTCI TD-06-008, reduced-consumption flash-butt welding (Kristan, LoPresti, Garcia, 2006): https://mxvrail.com/wp-content/uploads/2023/01/TD06-008.pdf
• TTCI TD-19-010, high-hardness thermite welds (Brice, 2019): https://mxvrail.com/wp-content/uploads/2023/01/TD19-010.pdf
• FHWA, memorandum "Recommendations to Account for Acoustic Dissimilarities in Steel Ultrasonic Weld Inspection", Dec 13, 2024: https://www.fhwa.dot.gov/bridge/inspection/Memo_Dec2024_Acoustic_Dissimilarities_Steel_Ultrasonic_Weld_Inspection.pdf
• FHWA-HRT-04-124, Laboratory and Field Testing of Automated Ultrasonic Testing Systems for Steel Highway Bridges, Section 2 (2005): https://www.fhwa.dot.gov/publications/research/infrastructure/structures/04124/02.cfm
• ASTM E3024/E3024M-22a, MT for General Industry: https://store.astm.org/e3024_e3024m-22a.html
• ASTM E1444/E1444M-26, MT for Aerospace: https://store.astm.org/e1444_e1444m-26.html
• ASTM E709-21, Guide for MT: https://store.astm.org/e0709-21.html
• TWI case study 564, RailSAFT: ultrasonic inspection of cast austenitic manganese steel railway crossings: https://www.twi-global.com/news-events/case-studies/railsaft-ultrasonic-inspection-of-cast-austenitic-manganese-steel-railway-crossings-564/