Aluminothermic Weld Inspection: A Thermite Weld Procedure by Zone

Aluminothermic weld inspection is the visual, profile, hardness and ultrasonic examination of a thermite rail weld after it is poured and finished. Under 49 CFR 213.113(d)(6), a weld in service is defective when transverse discontinuities exceed 5% of the rail head area individually or 10% in aggregate, and new-weld acceptance comes from the owner's specification.

TTCI testing at its FAST loop found that about 95% of thermite weld breaks occurred in the web, the web-base fillet and the base (TD21-027, 2021). Those are the zones that rail flaw detector cars cover least. A useful thermite weld inspection therefore has to name how each zone is reached, with which probe, from where and at what angle.

This guide is the deep dive on thermite welds only. The pillar article on rail weld inspection covers all weld types and the FRA background. The welding processes themselves are covered in advanced types of rail welding.

What does an aluminothermic weld inspection cover?

An aluminothermic weld inspection covers five checks in order: visual condition of the collar and running surface, profile and alignment, hardness where the owner requires it, ultrasonic testing of each zone, and a written disposition against a named acceptance document. Each check finds things the others miss, so skipping one leaves a gap.

A thermite weld is a small casting: molten steel from an aluminothermic reaction fills a mold around a gap TTCI gives as "1 to 1.5 inches" (TD21-027). The stripped weld keeps a raised collar around the web and base, and the head is ground to profile. TTCI's TD-04-011 (2004) put the thermite and flash-butt welds installed in North American track at "several hundred thousand" a year.

Key takeaway: treat the thermite weld inspection as five checks with five records, not one ultrasonic scan.

What should the visual and profile checks look for?

The visual check looks at the collar, the fillets and the running surface for cracks, run-out and unfused areas; the profile check measures the joint against the owner's alignment and finish tolerances with a straightedge. Both happen before any probe touches the rail.

Visual check of the collar and head

The collar is where many thermite failures start. In TTCI's Thermite Weld Failure Survey (TD-00-013, 2000), slag or porosity caused about 40% of 265 revenue-service failures, with 22.6% at the collar edge. Fatigue from weld metal run-out at the web/base fillet or base caused another 32%. Run-out is weld metal that leaked between the mold and the rail and left a thin fin or notch. That notch is a fatigue starter, whatever the UT shows.

A visual check of a finished thermite weld should cover:

• Collar toe at the base and web-base fillets: run-out fins, notches, cracks at the toe, sand or luting marks.
• Collar surface: unfilled areas, cold laps (a rounded, unfused fold of metal), porosity breaking the surface.
• Riser and feeder stubs: removed cleanly, with no gouging of the rail. Torch-cut excess metal is a red flag; the US Department of Defense guide specification for thermite rail welds says "Do not use a cutting torch to remove excess weld metal."
• Running surface and gauge corner: grinding burns (blue or dark discoloration), unground metal, low spots and visible porosity.

The same DoD specification (UFGS 34 11 19.00 20, August 2018) tells the inspector to "Pay particular attention to surface cracking, lack of fusion and other surface irregularities." It also says no finishing is required on the web and base. That matters to the UT technician: the collar stays as cast, so its shape returns echoes.

Profile and alignment

The profile check uses a straightedge and taper gauge once the rail has cooled to ambient temperature. The UFGS specification sets the pre-weld peak with a 36 in. (900 mm) straightedge centered on the joint, with 1/16 in. (2 mm) under each end. The head is set high because it contracts more than the web and base as it cools.

After finishing, the check measures head vertical offset and crown (peak or dip), horizontal offset and kink, base offset, and head finish deviation. AWS D15.2/D15.2M:2022 covers these in 8.12 Weld Alignment Inspection Procedures and 8.13 Rail Head Finish Inspection Procedure, with figures for each measurement and tables of alignment and finish tolerances (titles from the public preview). The numbers sit behind a paywall and vary by owner, so take them from the specification your contract cites.

Key takeaway: a run-out fin at the base fillet is a reason to act even when the UT is clean, so record the visual result separately.

How is hardness and other weld QC checked?

Hardness on a thermite weld is checked with a portable Brinell-type tester on the finished head, at the weld center, in the heat-affected zone and on the parent rail, where the owner's specification asks for it. It confirms that the weld metal suits the rail grade; it says nothing about internal soundness.

The UFGS guide specification gives one public example of a hardness rule. It calls for Brinell testing to ASTM E10 and ASTM E110 (portable testers). The weld and the rail for 6 in. (150 mm) each side of the joint must exceed 250 BHN. In the heat-affected zone, within 1 in. (25 mm) of the joint, the value may be up to 20 points lower, "except for the top of the rail, which must not be less than 250". Hardness is checked on at least 10% of welds. That is a guide specification for crane and railroad rail on DoD projects, not a universal rule.

TTCI's TD19-010 (July 2019) tested a full-section high-hardness thermite weld in 136RE head-hardened rail with a "top-of-rail hardness tester having a tungsten carbide ball indenter". The weld centers read 341 to 363 BHN.

Other QC items an owner may require:

Key takeaway: hardness, chemistry and bend tests qualify the procedure and the kit; field UT and visual checks accept each weld.

What reference rail is used to calibrate thermite weld UT?

Thermite weld UT is set up on a reference rail: a piece of the same rail section with machined reflectors at the depths and positions the probes will look at. There is no AREMA "step-wedge" calibration block for rail.

Range and zero are set on a standard block such as an IIW-type block; sensitivity is set in rail of the same section, because a flat block can't reproduce the beam paths to the fillets. AREMA's Manual for Railway Engineering lists a section titled Recommended Calibration Rails for Rail Flaw Detection System in Chapter 4 (4.6, 2025 table of contents). In Europe, the French adoption of EN 16729-1 includes an annex on artificial reflectors in reference rails. Neither text is public, so this guide doesn't reproduce their reflector sizes.

A practical reference rail for thermite weld UT might carry:

Some owners drill the reflectors into a sample thermite weld, so the reference echo travels through cast weld metal. Your procedure governs the reflector type and size; record the reference gain for each probe and recheck it at the end of the shift. The ultrasonic calibration for rail inspection article covers the wider calibration topic.

Key takeaway: set range on a block, but set sensitivity on rail of the same section, with a reflector at each zone you will report on.

What is the UT probe plan for each zone of a thermite weld?

A thermite weld UT plan normally uses a 0° probe and 70° probes on the head, 45° to 60° probes from the head into the web and web-base fillet, and 45° to 70° probes on the top of the base flanges, each scanned from both sides of the weld. The angles below are typical industry practice; your written procedure governs.

The research behind this plan is TTCI's beam modeling study, Phased Array Ultrasonic Beam Modeling for Rail Thermite Welds (TD21-027, Poudel, September 2021). Its recommended configuration was:

• Base: sectorial shear-wave scan at 45° to 70°, with probes 40 mm and 125 mm from the weld center
• Web: 30° to 70°, probes further back from the weld, scanning from the rail head
• Head-web and web-base fillets in the heat-affected zone: a fixed 45° shear wave
• Head: a sectorial scan across ±45° with the probe turned to look across the rail

Two findings matter to every hand-UT plan. First, TTCI found that full coverage of the base "could not be achieved without using a 'full skip'", and that the full skip works only off the rail, not the weld, "due to the uneven underside of the welded area". So base-flange probes must sit far enough back that the skip point lands on parent rail. Second, the study concluded that full volumetric inspection needs "several transducers to be placed on both sides of the rail simultaneously".

TD21-027 is a phased array (PAUT) modeling study, not a field procedure, but hand UT with single-element probes follows the same zone logic. The PAUT vs TOFD for rail weld inspection article explains where each advanced method helps and where it doesn't.

Immersion probes and lab-style "echo mapping" have no place in this field procedure. The rail is tested in place with contact probes and couplant, and every indication is plotted by sound path and probe position, then confirmed from the opposite side.

Key takeaway: if the plan doesn't say where the base-flange probe sits and where its skip lands, the base isn't covered.

What do thermite weld defects look like on the screen?

Each thermite defect has a typical location and echo pattern: shrinkage in the web gives a broad echo and loss of back wall, porosity gives clusters of small echoes, lack of fusion gives a sharp planar echo at the fusion boundary, and hot tears give a jagged echo near the fillet. These patterns are guides for a trained technician, not rules for calling defects.

The geometry echo is the main source of false calls. TTCI suggested that the "various surface finishes introduced from thermite welds" may explain why about a quarter of detector-car weld removals in TD-00-013 had no observable defect. TD21-027 adds that the weld's variable geometry "generates indications that could be misinterpreted as defects". Plotting every indication from both sides is the step that separates the two.

Detection depends on the plane of the flaw, not the wave type, as the detecting rail defects guide explains for plain rail.

Key takeaway: a defect plots to one point from both sides; collar geometry plots to the collar edge and moves with the probe.

Where do acceptance criteria for thermite welds come from?

Acceptance of a new thermite weld comes from the track owner's specification, which in the US often adopts AWS D15.2/D15.2M:2022 and AREMA's Manual for Railway Engineering, and in Europe from EN 14730-2. AWS D1.1 doesn't apply. The specification named in the contract is the one that counts.

The D15.2 statement on the use of AWS standards explains the hierarchy: these standards "carry the full legal authority of the contract or other document that invokes the AWS standards." In practice, a thermite weld report should name the document and edition used, for example "UT to D15.2:2022 clause 8.10 as amended by the owner's welding instructions."

Key takeaway: a thermite weld can't be "accepted" until the report names the acceptance document and edition.

What do FRA rules require for thermite welds?

FRA rules define a defective weld in service under 49 CFR 213.113 and require CWR records under 213.119(j). They don't set welding procedures, welder qualification or new-weld acceptance criteria.

213.113: defective weld in service

The definition in 49 CFR 213.113 on eCFR is the FRA test for a weld in track:

"Defective weld means a field or plant weld containing any discontinuities or pockets, exceeding 5 percent of the rail head area individually or 10 percent in the aggregate, oriented in or near the transverse plane, due to incomplete penetration of the weld metal between the rail ends, lack of fusion between weld and rail end metal, entrainment of slag or sand, under-bead or shrinkage cracking, or fatigue cracking." — 49 CFR 213.113(d)(6)

The rule adds that weld defects "may originate in the rail head, web, or base", and that a defect running longitudinally through the weld is treated as a split web for remedial action. When an owner learns of an indication, 213.113(b) requires it to be verified within four hours, or immediately where the suspected defect calls for remedial action A, A2 or B. A person designated under 213.7 then decides whether the track can stay in use.

How an owner sizes a base or fillet flaw against "percent of the rail head area" is a procedure question. Check the owner's instructions before reporting a base indication as a 213.113 defect.

213.119: CWR plans and field-weld records

Under 49 CFR 213.119 on eCFR, the owner's CWR plan must address "in-track welding" when keeping the rail installation temperature range during cutting and repairs (213.119(d)). For records, 213.119(j) requires an "adequate history of track constructed with CWR", including:

• rail temperature, location and date of CWR installations, kept at least one year (j)(1)
• a record of CWR installation or maintenance work that doesn't conform to the written procedures, kept until it is brought into conformance (j)(2)

The rule doesn't require a separate inspection record for each field weld. Per-weld logs come from the owner's specification and, where adopted, D15.2's documentation clause (8.7).

Welds in track are also tested as part of the rail under 49 CFR 213.237, but TTCI wrote that the detector process "only scans the head and web of the rail, and not the majority of the base" (TD-00-013).

Key takeaway: 213.113 judges a weld once it is in service; the owner's specification judges it the day it is poured.

Worked example: inspecting a fresh thermite weld

This example follows one new thermite field weld in 136RE rail from the visual check to UT by zone and the disposition of a base-fillet indication. The rail geometry and the arithmetic are real. The tolerances and the acceptance rule are illustrative owner rules, not quoted from D15.2, AREMA or EN 14730-2.

Inputs. 136RE rail: height 7-5/16 in. (7.3125 in.), base thickness under the web 1-3/16 in. (1.1875 in.), so the top of the base under the web is 6.125 in. below the running surface. Longitudinal velocity 0.2320 in./µs; shear velocity 0.1280 in./µs (about 3,250 m/s). Times are round trip with the wedge delay zeroed out. Collar half-width about 0.78 in., from the 39.4 mm collar Chen et al. modeled for a 25 mm gap.

Step 1: visual. The finish-ground weld has clean riser stubs, no torch marks, no burns and no collar cracks or unfilled areas. A small run-out fin sits at the collar toe on the field side of the base; it is noted because it will return a geometry echo.

Step 2: profile. With a 36 in. straightedge centered on the weld, the head reads 0.010 in. high at the center (a slight peak) and 0.008 in. out on the gauge side. Against an illustrative owner tolerance of 0 to +0.030 in. vertical and ±0.020 in. horizontal, both pass.

Step 3: hardness. A portable Brinell tester on the head reads 298 BHN at the weld center, 271 and 266 BHN about 3/4 in. either side, and 312 and 308 BHN at 6 in. Under the UFGS rule (over 250 BHN on the top of the rail throughout), all five readings pass.

Step 4: calibration. Range is set on a block. Sensitivity for each probe is set on the reference rail: the 45° fillet hole is brought to 80% of full screen height and the gain recorded. Scanning is done 6 dB above that reference.

Step 5: head, 0° and 70°. The 0° base echo should arrive at 7.3125 × 2 ÷ 0.2320 = 63.04 µs. It stays steady across the weld, so there is no large web shrinkage cavity in the 0° path. The 70° probes, scanned toward the weld from both sides, show nothing above the evaluation level.

Step 6: web and web-base fillet, 45° from the head. A 45° beam reaches the top of the base at a sound path of 6.125 ÷ 0.70711 = 8.662 in., or 2 × 8.662 ÷ 0.1280 = 135.3 µs. From side A, a clear indication peaks at 133.8 µs, slightly shallower, which places it in the fillet radius.

The two plots agree to 0.019 in. horizontally and 0.009 in. in depth. At 0.33 to 0.35 in. from the centerline, the reflector sits inside the 0.78 in. collar half-width, in weld metal. An echo from the run-out fin would plot to the toe, about 0.78 in. out, and from one side only, so this is treated as a real reflector.

Step 7: base flanges. Probes on the top of each base flange, aimed at the weld at 45° and 70°, are set back so the full skip lands on parent rail, not under the weld. They show no further indication. MT on the fillet and collar toe shows no surface-breaking crack, so the reflector is internal.

Step 8: disposition. Under the illustrative owner rule ("any indication in the web, fillet or base at or above reference level that plots inside the weld from both sides is rejectable"), this +5 dB / +3 dB, two-side-confirmed fillet indication rejects the weld. The weld is cut out and replaced by the owner's repair method, often a plug rail with two new welds. Because that is CWR maintenance, the rail temperature at cutting and the location go on the record, and the work follows the owner's CWR procedures under 213.119(d).

If the same reflector were found in service during a rail test, the path would differ. The owner would verify the indication within four hours under 213.113(b), a 213.7 designated person would decide whether the track stays in use, and the owner's instructions would decide how a fillet flaw maps onto "percent of head area".

Check: a depth of 6.05 in. in a 7.3125 in. rail is possible, and it sits 0.07 in. above the 6.125 in. top of base, where the fillet radius is. At 45°, depth and surface distance are equal, and both tables agree.

Key takeaway: the both-side plot, not the amplitude, is what turned this echo from "collar noise" into a rejected weld.

What aluminothermic weld inspection can't do

Aluminothermic weld inspection lowers the risk of weld failure; it doesn't prove that a weld is sound.

• The collar limits access. The as-cast collar can't be used as a scan surface, and the weld underside is too uneven to skip off (TD21-027). Some volume near the collar toe is reached only at steep angles or by MT.
• Signal-to-noise is poor at the collar line. TTCI reported that a favorable signal-to-noise ratio "is especially challenging to maintain in the weld collar line" (TD21-027).
• Geometry mimics defects. About one in four detector-car weld removals in TD-00-013 had no observable defect. Both-side plotting reduces this, but doesn't remove it.
• Tight planar flaws reflect little. A cold lap held closed by residual compression can return a weak echo from one side and almost nothing from the other.
• Access is limited. Anchors, tie plates, crossings and embedded track restrict probe positions on the base, and detector cars don't cover most of the base flanges.
• Acceptance isn't universal. D15.2, AREMA and EN 14730-2 limits are paywalled and owner-adopted, so one owner's reject can be another's accept.

As TTCI put it in the 2021 digest:

"Existing practices for flaw detection and characterization in thermite rail welds after installation are limited to visual inspection, which does not reveal purely internal defects." — TTCI, TD21-027 (2021)

How EMA Industries applies this

EMA Industries provides rail weld inspection for transit agencies, short lines and industrial track owners, including visual and profile checks, hand UT of thermite welds by zone with the base flanges covered from both sides, and MT where surface cracking is suspected. Each weld report names the acceptance document and edition used. EMA is based in Southern California and mobilizes nationwide. For thermite weld UT or a review of a weld inspection procedure, see ultrasonic examination services or contact EMA Industries.

Related reading: Rail weld inspection pillar · Types of rail welding · PAUT vs TOFD for rail welds · Detecting rail defects · Flash-butt weld inspection · Continuous welded rail

Frequently asked questions

What probe angles are used to inspect a thermite weld?

Typically a 0° probe and 70° probes on the head, 45° (sometimes with 60°) from the head into the web and web-base fillet, and 45° to 70° on the top of the base flanges. Each zone is scanned from both sides of the weld. TTCI's modeling recommended a fixed 45° for the fillets and 45° to 70° with a full skip for the base. Your written procedure sets the final angles.

Does AWS D1.1 apply to thermite rail welds?

No. AWS D1.1 is the structural steel welding code. Thermite rail welds are accepted to the track owner's specification, which in the US often adopts AWS D15.2/D15.2M:2022 and AREMA MRE Chapter 4, and in Europe to EN 14730-2.

When is a thermite weld defective under FRA rules?

Under 49 CFR 213.113(d)(6), a weld in service is defective when transverse discontinuities exceed 5% of the rail head area individually or 10% in aggregate. Lack of fusion, slag or sand, shrinkage cracking and fatigue cracking are all named causes. The owner then verifies the indication and applies the remedial action for its size.

Why is the base of a thermite weld hard to inspect?

Detector cars scan the head and web, and only the part of the base under the web. Most thermite breaks start in the web, the fillets and the base. Hand UT from the top of each base flange, using a full skip off parent rail, is the usual way to cover it.

Sources

• 49 CFR 213.113, Defective rails, eCFR (text read on Cornell LII, October 2026)
• 49 CFR 213.119, Continuous welded rail (CWR); plan contents, eCFR (text read on Cornell LII, October 2026)
• 49 CFR 213.237, Inspection of rail, eCFR
• Poudel, Phased Array Ultrasonic Beam Modeling for Rail Thermite Welds, TD21-027 (Sept 2021), TTCI/MxV Rail
• Kristan, Thermite Weld Failure Survey, TD-00-013 (July 2000), TTCI/MxV Rail
• Kristan, Laboratory Evaluation of Robotic Slot Weld for Railhead Repair, TD-04-011 (July 2004), TTCI/MxV Rail
• Brice, Full Section, High Hardness Thermite Welds, TD19-010 (July 2019), TTCI/MxV Rail
• AWS D15.2/D15.2M:2022 public preview, American Welding Society
• AREMA MRE Chapter 4 table of contents (2025), AREMA
• EN 14730-2, qualification of aluminothermic welders, approval of contractors and acceptance of welds, BSI Knowledge listing
• UFGS 34 11 19.00 20, Welding Crane and Railroad Rail – Thermite Method (08/18), Whole Building Design Guide (US DoD)
• Chen, Lawrence, Barkan and Dantzig, "Weld defect formation in rail thermite welds", Proc. IMechE Part F 220 (2006)