Types of Rail Welding: How Each Rail Weld Is Made and When It's Used

The main types of rail welding are thermite (aluminothermic), flash-butt in a plant or in track, gas pressure and enclosed-arc welding, plus arc welding for repair. Most North American field joints are thermite or mobile flash-butt welds. A conventional flash-butt weld consumes up to 1.5 in. of rail; a standard thermite weld needs a 1 in. gap.

The processes differ in heat source, whether the steel melts, how much rail each weld uses up and which defects form. This guide explains each one, compares them in one table and works a thermite-or-flash-butt decision for a continuous welded rail (CWR) rail change. Inspection is covered in our rail weld inspection guide and our page on ultrasonic inspection of aluminothermic welds.

Contents: What are the main types of rail welding? · Thermite · Plant flash-butt · Mobile flash-butt · Gas pressure · Enclosed-arc · Arc repair · Comparison table · Worked example · Rules and standards · Limits · FAQ

What are the main types of rail welding?

Rail is joined by fusion welding (thermite, enclosed-arc) or by solid-state forge welding (flash-butt, gas pressure). A 2013 Nippon Steel technical report names the same four as "the four welding methods used for rails presently in the world" (Saita et al., Nippon Steel Technical Report No. 395). Arc welding is used mostly to repair rail, not to join it.

Japan Railways (JR) made about 604,000 rail welds from fiscal 1988 to 2007: thermite 40%, gas pressure 29%, flash-butt 23% and enclosed-arc 8% (the same Nippon Steel report). No comparable North American split is published.

Lists of "the five basic weld types" often give butt, fillet, lap, corner and edge. Four of those are joint types (AWS A3.0 lists five joints: butt, corner, T, lap and edge), and a fillet is a weld type that joins T, lap and corner joints. Every rail weld in this guide is a butt joint.

How is a thermite rail weld made?

A thermite weld is a small steel casting poured into a mold around a gap between two rail ends. Aluminum powder reduces iron oxide in a crucible above the mold, giving molten steel at about 2,000 °C and an aluminum oxide slag (Nippon Steel 2013). Track owners use thermite for most field joints because the kit is portable and needs no power supply.

The sequence varies by kit supplier, but the main steps are these:

1. Cut and prepare the rail ends. Torch-cut ends aren't allowed in Class 3 to 5 track except as a temporary emergency repair, at no more than Class 2 speed (49 CFR 213.122). Rail to be welded is normally saw-cut.
2. Set the gap and alignment. A standard thermite gap is 1 in., and a wide-gap weld uses 2.75 in. (TTCI TD-09-012, April 2009). The rail ends are usually peaked slightly, so that the weld pulls straight as it cools.
3. Fit and seal the mold. Refractory mold halves clamp around the gap and are sealed. TTCI names "improper mating between the weld mold and the rail surface" as one cause of defects.
4. Preheat. A propane burner heats the rail ends through the mold, for a time set by the process. A 1976 patent describes conventional preheat as about 1,000 °C over 5 to 25 minutes, and claims a short preheat of up to 2 minutes at 300–700 °C (US 3,942,579). Follow the kit procedure, not a number from an article.
5. Ignite and tap. The portion reacts in the crucible. Many kits use a pre-loaded single-use crucible that taps automatically; reusable crucibles also exist.
6. Strip and trim. After a set cooling time, the crew removes the mold and shears the excess head metal while the weld is still hot.
7. Grind and finish. The head is ground to profile and checked with a straightedge before the weld goes into service.

What does "portion type" mean?

A thermite portion is the pre-weighed charge of aluminum, iron oxide and alloying additions for one weld. Suppliers make portions for a specific rail section, steel grade (standard or head-hardened) and gap. A portion for the wrong grade gives weld metal that is too soft or too hard against the rail. EN 14730-1:2017 approves aluminothermic processes for the same profile and grade only, and says approval "does not of itself ensure" suitability for specific track and traffic.

What goes wrong in thermite welds

Thermite welds fail more often than flash-butt welds. In TTCI's survey of 265 revenue-service thermite weld failures, slag and porosity caused 40% and fatigue from the weld runout 32% (Kristan, TD-00-013, July 2000). TTCI also judged "approximately 91 percent" of those failures preventable through better installation practice. Moisture in the portion, mold or sealing material causes porosity. Thermite uses no shielding gas, and field thermite welds get no post-weld heat treatment.

Key takeaway: most thermite weld failures start with installation: gap, preheat, mold fit, dry materials and finishing.

How is a plant flash-butt weld made?

A flash-butt weld is a forge weld made by passing high current through two rail ends, flashing them until they are hot and plastic, then forcing them together. The upset pushes molten metal and oxides out of the joint, and a shear removes the upset collar. Plants use flash-butt welding to join rolled rail into long CWR strings, which are then joined in the field.

The plant cycle has four stages:

1. Clamp and align. The welder clamps both rails through electrodes on the rail web or base.
2. Preheat and flash. The ends touch and separate repeatedly, and then flash continuously as the machine feeds one rail toward the other. This burns off the end faces and builds a heat zone.
3. Upset (forge). The machine drives the ends together under high force. Most of the molten layer and its oxides squeeze out.
4. Shear and finish. The upset collar is trimmed hot, and the weld is ground to profile and straightened.

One logged weld from a stationary K1000 machine shows the scale: 26.1 mm of flashing allowance, 15.5 mm of upset and a 67 s weld time (Rudenko et al., Paton Welding Journal, 2017, No. 5–6). Each plant's settings come from its qualified procedure.

Flash-butt welds aren't defect-free. Their typical defects are planar: "cold welds" from too little heat before upset, which leave a flat fracture face, and lack-of-fusion areas with oxide or silicate films, sometimes called "matt spots" or "flat spots". These sit in the bond plane, which is why upset control and ultrasonic testing (UT) of the weld matter.

TTCI reports that flash-butt welds "fail at a lower rate than thermite welds" (TD-06-008, April 2006). Flash-butt welding is often called the "strongest" method, but no published data supports a blanket ranking, and TTCI notes that neither thermite nor flash-butt welds match the mechanical performance of parent rail (TD-04-011).

Key takeaway: flash-butt welds fail less often than thermite welds, but their planar defects still need testing.

How does mobile (in-track) flash-butt welding work?

Mobile flash-butt welding uses the same process, carried on a truck or rail vehicle with a weld head lowered onto rail in track. It has been routine in North America for decades for field and closure welds in CWR. In Europe, EN 14587-2 covers flash-butt welding by mobile machines in track, and EN 14587-1 covers welding in a fixed plant.

The weld head clamps the rails, flashes, upsets and shears in one cycle, so a mobile weld has no mold and no portion. The head's clamps can also pull the rail, which lets a crew close a gap and adjust rail tension in the same operation. That matters in CWR, where the rail's neutral temperature has to be restored after a rail change.

Rail consumption is the main limit. TTCI put it plainly:

"The conventional EFB process … can consume as much as 1.5 inches of rail."

That is TTCI's TD-06-008 (April 2006) on electric flash-butt (EFB) welds; the reduced-consumption process it tested cut burn-off by "nearly 50 percent". The crew must add or pull rail to make up the loss. Access is the other limit: a mobile welder needs a road or rail approach, while a thermite kit goes almost anywhere.

Key takeaway: flash-butt welding is a standard field process too; its main constraint is the rail it consumes.

How does gas pressure welding join rail?

Gas pressure welding heats the rail ends with an oxy-fuel burner to a plastic state, about 1,200–1,300 °C, and presses them together without melting. The bond forms by diffusion and recrystallization across the clean end faces, and the bulge is trimmed afterward (INNOTRACK D4.6.5, 2009). It is a standard process in Japan. RTRI writes that "Gas pressure welding is commonly used for rail welding in Japan" (Yamamoto et al., Quarterly Report of RTRI 58(3), 2017).

RTRI also notes that the method "requires specialist skills" for end-face grinding and heating.

Gas pressure welds are not defect-free. Their typical defect is a lack of bond at the interface, which shows up on a fracture face as a flat, dull area called a "white spot" or "flat spot".

• A 2021 Southwest Jiaotong University study calls white spots "the typical defect of rail gas pressure welded joints", caused by contamination or under-heating (Zheng and Dai, Materials Reports 35(14)).
• INNOTRACK found earlier test welds prone "to form planar lack of bond defects in the foot tip". Its 2009 series showed none but questioned wear resistance of the wider heat-affected zone (INNOTRACK D4.6.5).

In North America, gas pressure welding has been tested rather than adopted. Ten test welds at FAST from 2006 outlived the thermite welds around them but showed batter, shells and a web fracture from a brand stamp (TD-08-010, February 2008).

Key takeaway: gas pressure welding makes a good solid-state weld when the end faces are clean, and its signature defect is a planar lack of bond that a visual check can't see.

Is enclosed-arc welding a real rail welding process?

Yes. Enclosed-arc welding is a recognized rail butt-welding process, but it is not submerged arc welding. The welder deposits weld metal in layers from coated electrodes into the gap, between copper or high-copper shoes that form the rail profile. Nippon Steel lists it as one of the four rail welding methods in world use, and it made 8% of JR's welds from 1988 to 2007.

Claims that enclosed-arc welding isn't a recognized rail-joining process are wrong. AWS D15.2 covers arc welding of rail (clause 5.9) and its procedure qualification (clause 6.4), as our WPS vs PQR guide explains. Equally wrong is tying it to submerged arc welding (SAW) under granular flux in a plant. It is a manual field process with coated electrodes. Patent literature describes it as skill-dependent, high-preheat and low-productivity, and Nippon Steel gives about 60 minutes of welding time. Typical defects are lack of fusion and slag between layers, porosity, and HAZ cracking if the preheat is wrong.

No figures were found for current North American use, so treat it as uncommon in US track and confirm the owner's specification allows it.

Key takeaway: enclosed-arc welding is a slow, skill-dependent field joining process with an established role in Japan and a small one elsewhere.

When is arc welding used on rail?

Arc welding on rail is used mainly to repair and build up, not to join running rail. The common jobs are:

• rebuilding battered rail ends, wheel burns and engine burns
• building up worn frogs, crossings and switch points
• slot welds that replace a shallow head defect

AWS D15.2 covers SMAW, GMAW and FCAW on rail, and it leaves the decision to recondition by welding to the track owner.

TTCI tested a robotic GMAW slot weld that repaired head defects up to 1.375 in. deep. The crew mapped the defect by UT, milled it out, checked the cavity by magnetic particle testing (MT) and preheated by induction before welding (TD-04-011, July 2004). The FAST prototypes show why arc repairs need inspection: 7 of the first 12 had ultrasonic anomalies when installed, and their hardness varied by more than 100 HB (TD-06-008).

Carbon rail steel needs controlled preheat and slow cooling to avoid hard, crack-prone heat-affected zones, while cast manganese frogs follow their own supplier procedures. The owner's qualified procedure sets the limits.

How do the rail welding processes compare?

Thermite is slow but portable; flash-butt is fast and consistent but consumes rail and needs heavy equipment. The table compares speed, location, typical defects and routine quality control (QC).

Welding times are for the weld step only (Nippon Steel 2013), without cutting, setup, cooling, grinding or testing.

Production QC for rail welds is UT, visual testing (VT), MT or PT, hardness and profile or alignment checks. Radiography (RT) and residual-stress measurement are research and qualification tools, not routine production QC. A 1970 AAR study of thermite weld radiographs concluded that "more study is needed" on the technique's effectiveness, and field RT needs a radiation safety crew.

Key takeaway: choose the process for the site and the track window, then specify the QC that matches that process's defects.

Worked example: thermite or mobile flash-butt for a CWR rail change?

For a rail change in CWR, the rail each weld uses up and the neutral temperature decide how long the plug rail must be. The inputs are illustrative; your CWR plan and weld procedures set the real ones.

Situation: a detail fracture is found in 136 lb CWR on Class 4 track. The owner will cut out the defect, install a plug rail and weld both ends.

Inputs (stated assumptions):

• Opening between the saw-cut rail ends, held by rail tensors after cutting: O = 240.0 in. (20 ft)
• Rail temperature during the work: 60 °F
• Target rail neutral temperature (RNT) under the owner's CWR plan: 95 °F
• Rail length that will be adjusted (unanchored on both sides, plus the plug): L = 400 ft = 4,800 in.
• Thermal expansion coefficient of rail steel: α = 6.5 × 10⁻⁶ per °F
• Thermite gap: 1.0 in. per weld (TD-09-012)
• Mobile flash-butt consumption: 1.5 in. per weld, the conventional maximum (TD-06-008)

Step 1: find the length to remove for the target RNT. At 60 °F, rail installed stress-free would have an RNT of 60 °F, which is 35 °F below target. To restore 95 °F, the adjusted rail must be shortened by the amount it would contract over 35 °F:

ΔL = α × L × ΔT = 6.5 × 10⁻⁶ × 4,800 in. × 35 °F = 1.09 in.

Step 2: thermite plug length. The plug has to leave two 1 in. gaps after the tensors pull out the 1.09 in.:

P(thermite) = O − 2 × 1.0 − ΔL = 240.0 − 2.0 − 1.09 = 236.91 in. (about 19 ft 8.9 in.)

Step 3: mobile flash-butt plug length. Each weld burns off up to 1.5 in., so the plug must be longer than the opening:

P(flash-butt) = O + 2 × 1.5 − ΔL = 240.0 + 3.0 − 1.09 = 241.91 in. (about 20 ft 1.9 in.)

Step 4: check the result. The two plugs differ by 2.0 + 3.0 = 5.0 in., and 241.91 − 236.91 = 5.00 in. Both plugs include the same 1.09 in. adjustment, so the RNT outcome is the same with either process if the crew pulls the full amount.

Step 5: decide. Mobile flash-butt welding wins on weld time and on TTCI's failure-rate data. It needs a welder truck on site that can pull the rail through that temperature difference. Thermite wins where a truck can't reach the site and the track window can absorb about an hour of welding. If the crew can't pull to the target, 213.119(d)(2) applies: rail must not be added below the plan's installation temperature "without provisions for later adjustment".

What the example leaves out: saw-cut tolerances, the plug's own temperature if it was stored in the sun, the owner's minimum plug length, anchoring patterns and the time for UT after welding. A cut-length error of 0.25 in. shifts the RNT by about 0.25 ÷ (6.5 × 10⁻⁶ × 4,800) ≈ 8 °F over this 400 ft length.

Which rules and standards apply to rail welds?

FRA rules define a defective weld and require CWR plans and records. They don't qualify welding procedures, qualify welders or set acceptance criteria for new welds. Those come from the track owner's specification, which commonly adopts AWS D15.2 and AREMA practice.

Under 213.113(d)(6), a defective weld is "a field or plant weld containing any discontinuities or pockets" over 5% of the head area individually or 10% in aggregate, in or near the transverse plane.

Section 213.119(j) requires "Rail temperature, location, and date of CWR installations", kept at least one year, plus a record of nonconforming CWR work. Per-weld logs come from the owner's specification or D15.2, not the rule.

Three sections sometimes cited as weld rules aren't. Section 213.109 covers crossties, 213.110 gage restraint measurement systems and 213.115 rail end mismatch at joints. None of the three mentions welds (Cornell LII text, checked October 2026). Section 213.121 covers joint bars, not welds. AWS D15.1 covers railcars and locomotives, not rail in track.

Key takeaway: cite 213.113(d)(6) for welds in service and 213.119 for CWR work, and take new-weld acceptance from D15.2 or the owner's specification.

How are rail welds inspected?

New rail welds are checked visually and for alignment, then tested by hand UT. Welds in service are covered by the owner's internal rail testing under 49 CFR 213.237. Test cars don't scan most of a weld's base. TTCI found that vehicle UT covers the head and web and "not the majority of the base" (TD-00-013), so thermite welds need hand UT of the base and the web-base fillets.

See our rail weld inspection guide, the aluminothermic weld UT page and our flash-butt weld inspection page for probe angles, base coverage, reference levels and disposition.

What this guide can't settle

• The welding times are for the weld step only, from one 2013 manufacturer comparison.
• The 1.5 in. flash-butt consumption is a maximum for the conventional process (TTCI, 2006). Programs and reduced-consumption methods use less. Use the machine's qualified setting.
• Gas pressure and enclosed-arc welding have no documented current production use in North America in the sources found.
• This page doesn't give acceptance criteria. D15.2 Tables 8.1 and 8.2 and the owner's specification do. Check the edition named in your contract.
• Every number in the worked example is illustrative. The CWR plan, the measured rail temperature and the weld procedure govern.

How EMA Industries applies this

EMA Industries inspects rail welds for transit agencies, short lines and industrial track owners. We're based in Southern California and mobilize nationwide. After a rail change, a weld report should record the process, location, rail temperature, visual and alignment results, and UT of head, web and base. To scope weld inspection after a rail change or a welding program, see our ultrasonic rail and weld testing services or contact us.

Related reading: Continuous welded rail (CWR) · Lack of fusion in welds · FRA defect codes explained · PAUT vs TOFD for rail weld inspection

Frequently asked questions

What is the most common type of field rail weld in North America?

Thermite welds are the most common field weld. TTCI wrote that "several hundred thousand thermite welds" are installed in North American track each year (TD-06-008, 2006). Mobile flash-butt welding is also routine for field and closure welds where a welder truck can reach the site.

Can flash-butt welding be done in the field?

Yes. Mobile flash-butt welders weld rail in track from a truck or rail vehicle, and EN 14587-2 covers the process in Europe. The limits are access and rail consumption, up to about 1.5 in. per weld.

What is CWR, and is it a welding process?

Continuous welded rail is the product, not the process. Under 49 CFR 213.119, CWR is rail "welded together into lengths exceeding 400 feet". It is made by plant flash-butt welding and joined in the field by thermite or mobile flash-butt welds.

What are WPS and PQR for rail welding?

A welding procedure specification (WPS) sets how a weld is made, and a procedure qualification record (PQR) documents the tests that qualified it. For rail, AWS D15.2 Clause 6 sets procedure qualification for thermite, flash and arc welds, and Clause 7 qualifies welders and operators. AWS D1.1 doesn't apply to rail in track.

Which rail weld is strongest?

No process is strongest in every case, and none is defect-free. TTCI data show flash-butt welds failing at a lower rate than thermite welds, but TTCI also notes that neither matches parent rail.

Sources

• Saita, Karimine, Ueda, Iwano, Yamamoto, Hiroguchi, Trends in Rail Welding Technologies and Our Future Approach, Nippon Steel Technical Report No. 395, 2013 (four processes, welding times, JR FY1988–2007 counts)
• 49 CFR 213.113, Defective rails, eCFR (text read on Cornell LII, October 2026)
• 49 CFR 213.119, CWR plan contents, eCFR (read on Cornell LII)
• 49 CFR 213.122, Torch cut rail, eCFR (read on Cornell LII)
• 49 CFR 213.109, 213.110 and 213.115, Cornell LII (checked: no weld content)
• AWS D15.2/D15.2M:2022, Specification for Joining Railroad Rail and Related Rail Components, public preview, American Welding Society
• AREMA Manual for Railway Engineering, Chapter 4 table of contents (2025), AREMA
• Kristan, Thermite Weld Failure Survey, TD-00-013, TTCI, July 2000
• Kristan, TD-04-011 (rail weld volume, slot-weld repair), TTCI, July 2004
• Kristan, LoPresti, Garcia, Initial Results of Robotic Slot Weld and Reduced Consumption EFB Weld Testing at FAST, TD-06-008, TTCI, April 2006
• Gutscher, Gas Pressure Welds and Hollink Slot Welds Testing at FAST, TD-08-010, TTCI, February 2008
• Gutscher, Li, McDaniel, Preliminary Performance of Wide-Gap Welds at Eastern Mega Site, TD-09-012, TTCI, April 2009
• INNOTRACK D4.6.5, Gas Pressure Welding: Quality of Test Welds, EU Sixth Framework Programme, November 2009
• Yamamoto, Tatsumi, Kakizaki, Rail Gas Pressure Welding Process without Using Expert Skills, Quarterly Report of RTRI 58(3), 2017
• Zheng and Dai, Research on White Spot Defect of Rail Gas Pressure Welded Joint, Materials Reports 35(14), 2021
• Rudenko et al., Influence of Flash Butt Welding Process Parameters on Strength Characteristics of Railway Rail Butts, The Paton Welding Journal, 2017
• Guntermann, US Patent 3,942,579, Process for the aluminothermic welding of rails, 1976
• AREA Bulletin 71 (1970), AAR Studies of Butt Welded Rail Joints, TRID abstract
• US Patent 5,877,468 (background on enclosed-arc and thermite rail welding)