Railway Thermit Welding: Process, Safety & Weld Quality

It is 2:40 a.m. on a single-track subdivision, and a rail has pulled apart in a 40-foot gap that will stop the first coal train of the morning. The track crew has no welding plant, no power, and about ninety minutes of track-and-time. What they do have is a cardboard box of iron oxide, aluminium granules and a ceramic crucible — and by dawn the gap will be a solid, cast-steel joint. That box is railway thermit welding, and understanding how it works explains a great deal about how modern track is built, inspected and kept safe.

Why a Handful of Rust and Aluminium Can Hold a 200-Ton Train

Picture a planned closure instead of an emergency. A four-hour possession has just begun on a subdivision that carries coal, intermodal freight and a busy commuter schedule. Two rail ends sit a few millimetres apart inside a sand mold, a burner hisses under a preheat cap, and a crew of four waits for a temperature reading before anyone strikes an igniter. Behind them, a dispatcher is watching the clock. The entire job, from saw cut to finished profile, has to fit inside that window, and the joint that leaves with the first train must survive years of 200-ton cars at track speed.

That joint is railway thermit welding, also called aluminothermic welding. It joins two rail ends by pouring superheated liquid steel, produced by a thermite reaction, into a single-use mold clamped around a prepared gap. The steel floods the gap, fuses into both rail faces and solidifies into one continuous casting. There is no power pack, no hydraulic clamping machine and no welding plant involved. A cardboard box of iron oxide, aluminium granules and a ceramic crucible does the work that a small factory would otherwise do.

The chemistry behind that is beautifully simple. Aluminium has a stronger affinity for oxygen than iron does, so it strips the oxygen away from iron oxide: Fe₂O₃ + 2Al → 2Fe + Al₂O₃. The reaction is strongly exothermic, releasing roughly 4 MJ per kilogram of mixture, and it leaves behind two useful products, liquid iron and an aluminium oxide slag that floats clear. That same oxygen hunger is why aluminium is such a demanding metal in conventional fabrication, where surface oxide must be broken down before a sound joint forms; our aluminum MIG welding guide covers that side of the metal in detail.

The numbers explain the appeal. The reaction zone reaches on the order of 2,500 °C and produces molten steel around 2,000 °C, hot enough to fuse rail ends without any external power supply. The pour itself lasts only seconds, yet a complete weld cycle in track commonly runs 30 to 60 minutes once cutting, gap preparation, alignment, preheat, cooling, mold removal and grinding are counted. Rail steel is a high-carbon pearlitic steel, typically about 0.7–0.8% carbon with roughly 1% manganese, which is exactly why this is a casting process rather than an arc welding job.

What this guide covers

Over the sections ahead we will work through the chemistry and history of the process, the field procedure step by step, the metallurgy that explains why a cast weld behaves differently from a forged one, and the standards and inspection regime that prove a weld is sound, including the FRA Track Safety Standards in 49 CFR Part 213. We will also look at defects and troubleshooting, safety and roadway protection, how thermit compares with flash butt and other methods, the mistakes that recur in the field, cost and training, and where the technology is heading.

Why it survives

Flash butt welding produces a stronger, more consistent joint, and it dominates new continuous welded rail construction and welding plant output. Thermit welding survives anyway, and the reasons are practical rather than metallurgical. The kit fits in a truck. It needs no electrical power. It works with a small crew, in tight locations, on isolated territory and in the middle of a repair that cannot wait for a mobile flash butt unit to travel hundreds of miles. When a rail pulls apart at 2 a.m., portability beats perfection.

Who this is for

Track maintenance staff, welding students, engineers, inspectors and curious rail enthusiasts will all find something here. The material assumes no prior metallurgy and no welding ticket, but it does not talk down to people who already own a track gauge.

The three questions readers actually have

Most people arrive with one of three questions. How does a box of rust and aluminium turn into a structural weld? Why do these welds sometimes break, often years after they were made? And how does a crew prove that a finished weld is good enough to carry traffic? This article answers all three, in that order and in plain language.

What Railway Thermit Welding Actually Is: Chemistry, History and Where It Fits

Railway thermit welding, also called aluminothermic welding, is a field casting process. Two prepared rail ends are clamped inside a single-use mold, and superheated liquid steel is poured into the gap. The steel is not supplied from a furnace or a welding machine. It is created on the spot by a chemical reaction inside a ceramic crucible. When the metal solidifies, the rail ends are locked into one continuous casting, and the mold is broken away.

That distinction matters. Thermit welding is not arc welding, and it is not forge welding. There is no electric arc, no shielding gas, and no mechanical upsetting. The joint is created by molten metal filling a cavity and freezing in place. Modern kits are alloyed to approximate the chemistry and hardness of the parent rail, with controlled additions of carbon, manganese, silicon, and sometimes chromium or vanadium, plus grain-refining elements. The result is a cast-steel joint that behaves differently from the rolled rail on either side of it.

Chemistry: Rust, Aluminium and a Very Hot Exchange

The reaction is elegant in principle. Aluminium has a powerful affinity for oxygen, stronger than iron does at these temperatures. When aluminium granules and iron oxide, often mill scale, are ignited together, the aluminium pulls oxygen away from the iron. Iron is liberated as liquid steel, while aluminium oxide forms a slag that floats on top.

In simplified form: Fe₂O₃ + 2Al → 2Fe + Al₂O₃. The reaction is strongly exothermic and self-sustaining once started. It produces molten steel at roughly 2,000 °C and a reaction zone closer to 2,500 °C. No external power is needed after ignition, which is exactly why the process works in remote territory. Anyone who has battled the stubborn oxide layer in aluminum MIG welding will recognise the same chemistry from the other direction: aluminium loves oxygen, and thermit welding turns that appetite into a controlled industrial tool.

Inside a Thermit Kit

A field kit is a compact foundry. The main components are:

• Iron oxide and aluminium granules in a pre-measured charge, often with mill scale as the oxide source.
• Alloying additions and grain refiners to match rail chemistry, hardness, and toughness.
• A single-use ceramic crucible, commonly alumina-graphite, fitted with a self-tapping pin or thimble that opens the pour hole at the right moment.
• A single-use sand or resin-bonded mold, clamped around the rail and sealed with luting sand.
• An igniter and a preheat burner for drying and heating the mold and rail ends.

The crucible is both reactor and pouring vessel. The charge reacts inside it, and the same vessel then delivers the steel into the mold. Most modern practice uses one-shot crucibles rather than reusable graphite types because a fresh crucible removes a major variable: contamination and cracking from previous pours. A single-use mold and crucible also simplify quality control, since every weld starts with known, traceable consumables.

A Short History

The process was commercialised by Hans Goldschmidt in the 1890s, building on the aluminothermic reaction that still carries his name. Early railway and tramway applications followed around the turn of the twentieth century in Germany, where the ability to join track without a power supply or a heavy plant was immediately useful. The kit format has changed enormously since then, but the underlying reaction has not.

Where Thermit Fits Today

Thermit welding occupies a specific niche rather than competing directly with factory welding. It is used for field closures and cut-ins on continuous welded rail (CWR), pull-apart and defect repairs, welds in confined or remote locations, joining rail to special trackwork, and renewal work where a mobile flash butt unit cannot reach. It is also the fallback when a track window is short, access is poor, or the nearest welding plant is hundreds of miles away.

Rail welding practice is addressed by standards such as AWS D15.2, recommended practices for welding rails, and EN 14730-1 and EN 14730-2 for aluminothermic process approval and welder qualification, alongside the AREMA Manual for Railway Engineering, Chapter 4 on rail. Those documents matter because they separate a legitimate structural weld from a hopeful field repair.

One nearby technology is worth separating out. Exothermic signal bonding uses a related reaction to attach copper cables to rail for track circuits. It is a small, localised connection, not a structural rail weld. The chemistry looks familiar, but the purpose, mold, and quality requirements are entirely different.

The Field Process, Step by Step: From Saw Cut to Finished Profile

Welding a rail in the field is less like striking an arc and more like running a very small foundry inside the four-foot. Every step protects one of two things: the cleanliness of the surfaces that must fuse, or the geometry that has to carry a wheel load at track speed. Skip one and the weld can still look fine from the ballast shoulder, right up until it doesn’t.

Plan the Weld Before You Touch the Rail

The job starts with paperwork, not fire. The briefing sets the track window, foul time limits, lookout and watchman assignments, and the escape route. Under 49 CFR Part 214 roadway worker protection, nobody lifts a tool until that briefing is complete and understood. Then the crew measures and marks cut points, the specified gap, and a reference mark on the field side of both rail ends so alignment can be rechecked once the mold is in place.

A planned closure weld runs on a comfortable rhythm. An emergency repair under slow order does not: the crew is working against a train, often in the dark, and the temptation is to compress the steps that matter most, preheat, alignment and cooling. That is exactly when discipline pays.

The Sequence, in Order

1. Protect the track and confirm the window.
2. Measure, mark and record rail temperature.
3. Cut out the old weld or defect with an abrasive rail saw.
4. Square and deburr both ends, head and foot.
5. Clean to bright metal: no rust, scale, paint, oil or moisture.
6. Set the gap with the kit gauge and verify it.
7. Align running surface, gauge face and vertical bend with a straightedge and wedges.
8. Fit, seal and lute the mold.
9. Preheat to the specified temperature and measure it.
10. Insert crucible and igniter; pour and let the reaction complete.
11. Allow the soak, then cool.
12. Remove the mold, riser and gate.
13. Grind running surface and gauge face in the direction of travel.
14. Check with straightedge and profile gauge; document the weld.

Why Each Step Exists

The cut. Rail ends are cut square with an abrasive saw. Flame cutting is avoided because it leaves a heat-affected, carbon-depleted and irregular edge, plus oxide and slag, and that poor edge undermines fusion at the very interface the weld depends on. If you have watched a torch wash over high-carbon steel, you know the result is not a machined face. The same logic applies to any thermal cutting method used near a joint that must later be welded.

The gap. Set with a gauge matched to the kit and rail section, typically around 20–30 mm, and verified before the mold goes on. Gap controls how much weld metal sits in the joint and how shrinkage develops.

Cleanliness and alignment. Ends are cleaned to bright metal, and moisture is an enemy rather than an inconvenience. The same prep discipline that decides quality in other welding processes applies here with far less margin. Alignment covers running surface, gauge face and vertical bend, checked with a straightedge and wedges. A few tenths of a millimetre of vertical misalignment becomes a dynamic impact load with every wheel pass.

Rail temperature. For continuous welded rail the weld is planned around the rail neutral temperature, commonly about 90–110 °F (roughly 32–43 °C) in North American practice, so the joint is not locked in at an extreme rail temperature and left to fight the rest of the string.

Mold and preheat. The mold is fitted, sealed and luted so no steel escapes and no air is drawn in. Preheat generally falls around 600–950 °C depending on rail section, kit design and ambient conditions, verified with a contact pyrometer or temperature-indicating crayon, never judged by eye, because a dull red in the dark and a dull red at noon are not the same temperature.

Pour, soak and cool. The reaction runs hot and fast, and the soak that follows is part of the process, not dead time. Rushing the cool cycle is how crews manufacture cracks they find later.

Finishing. The riser and gate come off, then the weld is ground on running surface and gauge face in the direction of travel. Geometry is held to tolerances measured in fractions of a millimetre, confirmed with a straightedge and profile gauge, and written down against the weld record.

Weather Changes the Method, Not the Standard

Note: Wind pulls heat out of the mold cavity and disturbs the preheat flame, so crews build screens. Rain and standing water demand covers over the rail ends and extra care with luting, because water and molten steel are a violent combination. Cold weather lengthens preheat times. In every case the measured temperature, not the clock, decides when the pour happens.

Metallurgy: Why a Cast Weld Behaves Differently From a Forged One

Almost every practical rule in thermit welding traces back to one uncomfortable fact: you are not forging a joint, you are casting one. A rail is a carefully rolled product with a fine, uniform microstructure. A thermit weld is a small steel casting poured into a sand mold in a ditch at two in the morning. That difference explains why preheat matters, why the collar exists, why cooling rate is specified, and why inspectors stare so hard at geometry. Filler chemistry drives joint behaviour in other steels too, as this alloy steel electrode selection guide shows.

Three Microstructures in One Joint

Rolled rail is wrought steel. Hot rolling breaks up the cast structure and leaves fine pearlite with grains elongated in the rolling direction: tough, consistent and forgiving of load. A flash butt weld is forged, with the ends heated and upset under pressure, so the joint keeps a fine-grained wrought character and needs no filler metal.

Thermit weld metal is neither. It arrives as liquid steel at roughly 2,000 °C and freezes inside a mold, so grains nucleate at the cooler mold faces and grow inward as long columnar dendrites. The result is a coarse cast structure with lower toughness and lower fatigue resistance than the parent rail. In plain terms, the casting is the weak link metallurgically, and the rest of the process exists to stop that weakness from turning into a crack.

Shrinkage, the Riser and That Awkward Head-Web Junction

Steel shrinks as it solidifies, and a casting has to be fed with liquid metal while it does. That is the job of the riser above the weld and the collar around it: a reservoir of hot metal that keeps feeding the shrinking joint. The last regions to freeze are the centreline and the area around the head-web junction, which is exactly where porosity and centreline shrinkage tend to cluster. These are inherent casting risks, not signs of a careless crew, which is why mold design and collar geometry are engineered rather than incidental. A well-fed weld pushes shrinkage up into the riser, where it can be cut off and discarded.

The Heat-Affected Zone and the Martensite Problem

Outside the cast nugget sits the heat-affected zone, parent rail that got hot but did not melt. Rail steel is high-carbon and hardens readily. If that zone cools too quickly, it can transform into untempered martensite, a brittle constituent prone to cracking under wheel loads. Head-hardened and premium rails are especially sensitive because they arrive at the site already heat-treated.

This is why controlled cooling is written into good practice: forced air where the specification calls for a defined cooling rate, insulating covers where the aim is slow, even cooling in cold weather. European practice under EN 14730 addresses hardness and cooling requirements for aluminothermic welds, including acceptance testing of production welds, and industry bodies such as AREMA and the UIC publish related guidance. The takeaway for the field is simple: the cooling curve is part of the weld, not an afterthought.

Why Geometry Decides Fatigue Life

Thermit welds generally show lower fatigue performance than flash butt welds, and the cast grain structure is only half the reason. Fatigue in rail welds typically initiates at a geometric discontinuity: vertical or horizontal misalignment, a poorly blended collar, a grinding gouge, a rough transition at the running surface. A sharp notch concentrates stress, a small internal defect sits beneath it, and millions of wheel passes do the rest. That is why alignment tolerances, collar blending and grinding direction receive so much attention in inspection standards. A weld can have perfect chemistry and still fail early because its profile was rushed.

Why the Collar Stays Put

New crews often ask why the collar is not simply ground away for a cleaner-looking joint. In most practice it is left largely intact and only blended, because cutting it back removes the mass that feeds solidification and supports the head-web region, and because a full removal risks gouging the parent metal. The collar does create a modest stress concentration at its blend, which is precisely why the blend radius and finish are specified rather than left to the eye.

The practical summary is that a thermit weld is a casting with a forged neighbour on each side, and its life is governed by how well the transition between them is managed. Moisture control, preheat, mold sealing, cooling rate and finishing are all attempts to make a cast structure behave as close to wrought steel as physics allows. Where aluminium is welded instead, the heat-flow rules differ sharply, as this aluminum MIG welding guide explains, but the underlying lesson is the same: microstructure follows the thermal history you give it.

Standards, Inspection and Quality Assurance: Proving the Weld Is Good

A thermit weld can look flawless from the ballast and still be worthless. The collar can be beautifully blended, the running surface mirror-smooth, the gauge face clean — and somewhere under the head, a pocket of shrinkage is quietly waiting for the next few million wheel loads. That gap between appearance and reality is exactly why rail welding is wrapped in a quality system. The chemistry is reliable; the proof is not automatic. Somebody has to define what “good” means, measure it, and write it down.

The Rulebook: Who Decides a Weld Is Acceptable

In the United States, the governing document is 49 CFR Part 213, the Track Safety Standards. It is a performance rule, not a welding manual. In plain terms it requires that rail joints be sound, that finished welds present correct geometry to the wheels, and that defective rail or joints be removed, repaired or protected by the appropriate speed restriction until they are. Enforcement has teeth: FRA inspectors can require the removal of welds that do not meet the standards, and a railroad’s own specifications are usually stricter than the federal floor. Class of track, traffic density and curving all shape how much deviation a given property will tolerate.

The Consensus Standards Behind the Rulebook

Beneath the regulation sit the technical documents that describe how to actually do the work. The AREMA Manual for Railway Engineering, Chapter 4, is the standard North American rail reference for track structure, including welding practice and acceptance guidance. AWS D15.2 provides recommended practices for welding rails and related rail components, covering procedure qualification and inspection. In Europe, EN 14730-1 covers approval of the aluminothermic welding process itself, while EN 14730-2 covers approval and certification of welders and welding operators, including periodic re-verification. That two-part split is worth noting: the process is approved as a system, and the people running it are approved separately. Both can lapse.

Six Layers of Quality Control

1. Consumables. Batch traceability, dry storage, shelf-life control. A kit that sat in a damp trailer for a season is a defect waiting to happen.
2. Procedural. An approved welding procedure, matched to rail section and metallurgy, executed by a qualified operator.
3. In-process. Gap measurement, alignment, preheat temperature, mold sealing and luting — checked, not assumed.
4. Post-weld visual and geometry. Straightedge, profile gauge, surface finish, collar blend, gauge face alignment.
5. Non-destructive testing. Ultrasonic and, where specified, other volumetric methods.
6. Destructive and statistical verification. Test welds cut up on a schedule to prove the process, not just the individual joint.

The Honest Limit of Ultrasonic Inspection

Here is where thermit welds are genuinely harder to qualify than flash butt welds. The cast, coarse-grained structure of aluminothermic weld metal scatters and attenuates ultrasound far more than the fine forged structure of a flash butt joint. Small defects hide behind grain noise. That pushes inspectors toward specialised probes, tuned procedures and, increasingly, phased array techniques that can steer and focus the beam rather than relying on a single fixed angle. It also means inspection alone cannot carry the quality burden, which is why disciplined workmanship and documentation matter more here than in almost any other weld type — a principle that holds across standards-driven welding work in other safety-critical industries.

Production Testing: Cutting Up Good Welds on Purpose

Every credible thermit program includes periodic test welds — either cut from production or made alongside it under the same conditions — subjected to bend tests, hardness surveys and macro or microstructural examination. A bend test opens the weld to the eye; a hardness traverse shows whether the heat-affected zone has gone too hard; a macro section reveals shrinkage, porosity and fusion at the rail/web interface. This is statistical proof that the process, the kits and the crew are still capable. It is also how problems get caught before they become derailments. Programs built around documented procedure and repeatable records — the same discipline seen in marine and structural welding standards — tend to be the ones that survive audit and incident review.

Traceability is the backbone of all of it. When a weld fails two years later, the record is what tells investigators whether the kit was in date, whether the preheat was short, or whether the rail was the wrong grade for the procedure. Quality assurance is not paperwork bolted onto welding. It is the mechanism that turns a portable casting process into something a railroad can stake a 200-ton train on.

Defects, Failure Modes and Field Troubleshooting

Almost every thermit weld defect traces back to one of four things: moisture, heat, alignment or finishing. Learn the signature and you can usually name the mistake before the cut-out torch leaves the truck.

Fusion-Related Defects

Lack of fusion and cold lap happen when poured steel meets rail ends too cold to melt into. Under-preheating is the classic culprit: the liquid metal washes over the rail end, solidifies against it, and never actually fuses. The surface can look perfectly sound, which is exactly what makes it dangerous. Incomplete penetration at the rail/web interface is a close relative — awkward geometry, thin mold walls, and any chill from a damp mold or a mis-set gap shows up as a cold shut. Run-out, where molten steel escapes the mold entirely, usually means incomplete luting, a damaged or mismatched mold, or a mold that shifted during the pour.

Gas-Related Defects

Moisture is the most common cause of porosity in thermit welds. Wet molds, damp luting sand, condensation on a cold rail and rain during the pour all generate gas that gets trapped as the steel solidifies. Blowholes and scattered porosity are the visible signature; finer gas damage may only appear in a macro etch or radiograph. Oil, paint, grease and rust on the rail ends do the same thing on a smaller scale, which is why ends must be cleaned to bright metal. Prevention is boring and effective: dry consumables, pre-warmed molds in cold weather, shelter from rain, and a moisture check before the crucible goes in.

Shrinkage-Related Defects

Cast steel shrinks as it freezes, so thermit kits use a riser and collar to feed liquid metal into the last regions to solidify. When feeding fails you get centreline shrinkage, piping and internal voids, clustering near the head-web junction and the collar — exactly where bending stress is highest and inspection is hardest. A short pour, a cold mold or a blocked riser all raise the risk.

Geometry-Related Defects

Vertical or horizontal misalignment creates a step at the running surface that hammers every wheel passing over it. A dipped or battered weld sits low; a poor collar blend leaves a notch at the transition; grinding gouges cut transverse scratches that act as stress raisers. These are finishing errors, and they hurt fatigue life more than most internal defects of similar size.

Base-Metal-Related Defects

Foot burn comes from a preheat flame held too close or too long on the rail foot. Arc strikes from nearby work introduce untempered martensite in high-carbon rail steel, and martensite cracks. On head-hardened and premium rails, fast cooling of the heat-affected zone produces the same brittle structure, which is why cooling rates are controlled rather than left to the weather.

The Failure Timeline: Why Small Defects Become Broken Rails

Most thermit weld failures are fatigue-driven, not instantaneous. A small internal defect or a geometric notch initiates a crack, and repeated wheel loads grow it slowly — thousands of cycles, often invisible — until the remaining section can no longer carry the load and the rail breaks transversely. The NTSB has investigated numerous track failures that followed precisely this pattern. Inspection intervals, not just inspection quality, determine risk: a defect found early is a planned closure, a defect found late is a service disruption.

Troubleshooting Table

Symptom	Likely cause	Corrective action
Blowholes on collar or head	Moisture — wet mold, damp sand, rain	Cut out; dry consumables and shelter the mold
Steel escaped the mold	Poor luting, damaged or shifted mold	Cut out; re-lute and pin the mold
Suspected lack of fusion	Under-preheat or cold rail ends	Cut out; check preheat with a calibrated gauge
Visible step at running surface	Misalignment before pour	Re-weld if out of tolerance; verify with straightedge
Dipped weld after traffic	Soft weld metal or over-grinding	Assess hardness and profile; cut out if life is compromised
Crack from a gouge or notch	Grinding across the rail or poor collar blend	Cut out; grind along the rail and blend the collar

The Only Real Remedy: Cut It Out

The standard corrective action for a defective thermit weld is complete removal — the weld plus a short length of rail on each side — followed by a fresh weld. Surface repair is not an option: you cannot gouge out porosity, chase a crack with a burr, or patch cast steel and call it fixed. Attempting it with carbon arc or any arc process simply adds brittle martensite to an already compromised joint; the difference between air arc gouging for removal and arc welding for repair is the difference between demolition and a false repair.

Emergency Response: Pull-Aparts, Slow Orders and Re-Railing

When a weld has already broken, the priority order is protection, then movement, then repair. Set up roadway protection and notify the dispatcher before anyone touches the rail. If a thermit kit, mold and preheat set are on hand and the gap is within kit limits, a field weld restores the track in a couple of hours. If not, slow orders, joint bars or a temporary closure keep trains safe until a crew and materials arrive. Re-railing — replacing the affected length — is sometimes the better engineering answer when the break is part of a wider pattern of poor welds. Document every event; a cluster of failures in one territory is a process problem, not bad luck.

Safety, Fumes, Fire and Roadway Protection

Thermit welding is not a process you can improvise safely. A crew is handling a reaction hot enough to melt steel, working beside live track, often at night and often in bad weather. Most incidents are mundane: a damp mold, slag that rolls into a tie crib, a lookout whose attention slips. The controls are mundane too. They work only when they are treated as part of the job rather than paperwork.

Thermal Hazards and the Water Rule

Molten thermit steel is around 2,000 °C, and the reaction zone reaches roughly 2,500 °C. Spatter can travel several metres, slag stays hot long after it loses its glow, and the mold vents steam and flame during the pour. Clear the area of anyone not needed, keep people out of line with the crucible, and treat every mold, crucible and slag lump as hot until it has been checked.

The critical rule concerns water. Molten steel and water produce an explosive steam expansion, so a wet mold, a puddle in the crib or a damp crucible can turn a pour into a blowout. Water must never be applied to a burning thermite reaction. Dry sand or a suitable Class D extinguishing agent is the correct medium. The same discipline underpins the hot-work guidance in OSHA’s welding and cutting resources.

Eye, Face and Body Protection

The reaction is intensely bright. A welding helmet with an appropriate filter shade, plus face and neck protection, is the minimum, and it should be worn before the igniter is fitted, not raised a second later. Add leathers or a welder’s jacket, gauntlets, spats over boots and hearing protection. Trousers go outside boots, never tucked in, because a single spatter ball inside a boot causes a serious burn. Avoid synthetic clothing that can melt onto skin. For a wider look at optical hazards and filter selection, our Alpha Laser Welding safety guide is a useful companion.

Fumes and Dust

The plume contains iron oxide and aluminium oxide particulate, manganese fume from both the alloy addition and the rail steel, and respirable crystalline silica from mold and luting sand. Manganese matters: OSHA’s Table Z-1 sets a ceiling limit of 5 mg/m³ for manganese fume. Silica is regulated separately under OSHA’s silica standards, including 29 CFR 1910.1053 for general industry. Work upwind where possible, ventilate trenches and confined locations, and match the respirator to the exposure. A half-face respirator is not a substitute for ventilation in a deep, still trench.

Fire Prevention

Hot work rules are not optional. OSHA 29 CFR 1910.252 requires a hot work permit programme, protection of combustibles within 35 feet, fire watches and suitable extinguishing equipment. NFPA 51B describes the same principles as good practice. On track, the obvious fuels are wooden ties, dry vegetation, signal cable and debris in the cribs. Dampen surrounding timber before work where permitted, contain the slag, and post a fire watch that stays through the required post-work period.

Roadway Worker Protection

Thermit crews are fully subject to FRA’s Roadway Worker Protection rule at 49 CFR Part 214. That means a job briefing before work, defined working limits, and lookouts or watchmen with the authority to stop work. Foul time and adjacent-track provisions must be agreed with the dispatcher, not assumed. The hardest discipline is personal: never turn your back on a live track, and agree an escape route and place of safety before the pour begins. A crew watching a crucible is a crew not watching for a train.

Short Note: Tool Trailer Checklist. Hot work permit and fire extinguisher; Class D agent and dry sand; helmet with spare filter lenses; leathers, gauntlets, spats and hearing protection; respirators and correct cartridges; gas monitor for confined locations; first-aid kit and burn dressing; lookout flags, radios and briefing form; and a signed weld record.

Thermit vs Flash Butt vs Other Rail Welding Methods

Choosing a rail welding method is not a beauty contest. It is a trade between joint quality, logistics, and how much track time you can afford. The same railway may run a stationary flash butt plant for new continuous welded rail (CWR), a mobile flash butt machine for a planned possession, and a thermit kit at 3 a.m. for an emergency repair. Guidance from bodies such as AREMA reflects that reality: no single process wins every job.

Flash Butt Welding: Forging Under Pressure

Flash butt welding clamps both rail ends, brings them together under high current, and lets a controlled flashing action burn away impurities while the metal reaches a plastic state. A powerful axial force then upsets the joint, forging the ends together. The result is a fine-grained wrought structure with fatigue performance generally superior to a cast thermit weld, which is why flash butt dominates welding plant output and new CWR construction.

The catch is infrastructure. A stationary plant needs heavy equipment and a large electrical supply, and even a mobile welder demands serious power, a skilled crew, and enough possession to set up and cycle the machine. Right tool for planned work, not for an isolated pull-apart in remote territory.

Thermit Welding: A Foundry in a Box

Thermit takes the opposite approach. A single-use ceramic crucible holds iron oxide, aluminium, and alloying additions. Once ignited, the aluminothermic reaction produces superheated liquid steel that is poured into a sand mould clamped around the prepared gap, fusing the rail ends into one continuous casting. No mains power is needed, the kit is portable, and a small crew can work where no welding machine could reach, which is why thermit still dominates field closures, remote locations, and emergency repairs.

The trade-off is metallurgical. The weld metal solidifies as a casting rather than forged steel, so fatigue performance is generally lower and quality depends heavily on moisture control, preheat, alignment, and finishing. Kits carry alloying additions and grain refiners to match the rail section, much like choosing the right alloy steel welding electrodes for a structural job.

Gas Pressure and Enclosed Arc Welding

Gas pressure welding heats the rail ends with an oxy-fuel flame while applying axial pressure, forging the joint without an arc or a casting. It is used extensively in Japan and parts of Europe and can produce excellent joints. It has not displaced thermit in North America largely because it needs specialised equipment, precise end preparation, and a trained operator cadre, while thermit already has a mature supply chain and repair workflow across remote territory.

Enclosed arc welding traps an arc inside a mould and fills the gap with deposited steel. It has seen some rail use but is not mainstream for modern main line CWR. Ordinary arc welding of rail ends is generally unacceptable on main line track: rail steel is a high-carbon pearlitic grade that is highly crack-sensitive, and controlling dilution and hydrogen in a field joint is impractical. A process that works beautifully on mild steel, including stick welding, is not a safe substitute here.

Side-by-Side Comparison

Criterion	Thermit	Flash butt	Gas pressure	Arc / enclosed arc
Joint quality and fatigue	Good, but cast structure; lower fatigue resistance	Best and most consistent; forged, fine-grained	Very good; proven field process	Not acceptable for main line joints
Equipment and capital cost	Low; hand tools and single-use kits	High; plant or specialised mobile machine	Moderate to high; specialised rig	Low, but unsuitable for the application
Power and site needs	No mains power; small generator for tools	High electrical demand; planned site	Oxy-fuel and pressure gear; no high current	Electrical supply; poor field control
Crew and cycle time	Two to four people; under an hour, preheat and cooling dominate	Larger crew; minutes per weld, hours of mobilisation	Small specialised crew; minutes after setup	Small crew, but the method is not approved
Mobility and access	Excellent; remote and confined sites	Limited; needs machine access and possession	Moderate; equipment-dependent	Portable but technically unacceptable
Weather sensitivity	High; wind, rain, and moisture demand protection	Moderate; worksite and machine dependent	Moderate to high	High
CWR production vs repair	Repair, closures, cut-ins, special trackwork	Production welding and new CWR construction	Field production welding in some regions	Not a production rail joining method
Inspection difficulty	Higher; coarse cast grain scatters ultrasound	Lower; forged structure inspects more predictably	Generally favourable	Not relevant; joint quality cannot be assured

Match the method to the job: flash butt for planned production and the best fatigue life, thermit for portability and repair, gas pressure where the equipment and expertise exist, and arc processes nowhere near a main line rail joint.

Common Mistakes, Best Practices, Cost and Training

Almost every failed thermit weld in an investigation report failed for a boring reason: someone skipped a measurement, rushed the preheat, or ground the head too hard at 4 a.m. The reaction is forgiving; the procedure around it is not. Here are the errors that repeat, the habits that prevent them, what a weld really costs and how operators get qualified.

The Mistakes That Keep Repeating

• Welding on wet or frozen rail. Moisture flashes to steam the instant molten steel arrives, producing blowholes, cold laps and occasionally a violent ejection of metal. Frozen rail also drags the preheat down and distorts the cooling curve.
• Skipping the preheat measurement. Flame colour is not a temperature. Under-preheated ends do not fuse cleanly; over-preheated ones burn the mold and coarsen the heat-affected zone. Both return later as a transverse crack.
• Using the wrong kit for the section. Wrong steel volume, wrong gap fill, wrong thermal mass: short pouring, shrinkage voids, or a weld that never reaches full head profile.
• Reusing a single-use mold. A second pour leaks, misaligns the collar and traps sand in the fusion zone.
• Over-grinding, or grinding across the rail. Grinding is a finishing operation with metallurgical consequences: too deep thins the head and adds tempering damage to a surface that must survive millions of wheel passes, while transverse scratches become stress raisers aimed straight at wheel load. Finish strokes should run with traffic.
• Misreading rail temperature. On continuous welded rail, temperature relative to neutral temperature decides whether the weld is stress-free or locked in tension. A reading taken in shade, on a shiny surface or with an uncalibrated device is worse than no reading.
• Ignoring wind and slag. Wind strips preheat, chills the slag and cold-shuts the gauge face. Slag dropped on a creosoted tie or dry vegetation starts a fire that burns quietly.
• Skipping documentation. No record means no traceability, no trend data and no way to link a later defect to a suspect batch of consumables.

The Best-Practice Playbook

• Control consumables like perishables. Kits have a shelf life and are moisture-sensitive, so dry storage, stock rotation and batch traceability are quality controls, not housekeeping.
• Brief the job and check the kit before fouling track. A pre-job briefing under 49 CFR Part 214 is the natural moment to confirm section, kit number, gap gauge and burner.
• Work to a written procedure, and verify gap and alignment independently. The person who set the gap should not be the only person who checks it.
• Measure temperature with a calibrated device and record the number, not the word “hot”. Control cooling to the rail grade rather than letting head-hardened rail quench itself in cold wind.
• Finish to a standard, not to “looks alright” — straightedge, profile gauge and a documented finish direction — then complete the record the same shift, while the details are still true.

What a Weld Really Costs

Consumable kits are the visible cost and the least important one. Track time usually dominates: possession, slow orders, crews standing around, trains waiting on the other side of the closure. Cycle-time discipline and first-time-right quality therefore carry more economic weight than shopping for a cheaper box. Flash butt alternatives invert the structure, with heavy capital, plant logistics and mobilisation set against a low per-weld consumable cost.

The most expensive weld is the one that fails early. A premature failure triggers emergency possession, slow orders, an investigation and, in the worst case, derailment liability, as the NTSB accident record shows far too regularly. Those costs dwarf any consumable line item — the whole argument for ten extra minutes on preheat.

Training and Qualification

Nobody becomes a rail welder from a video. Manufacturer and railway-run courses teach the chemistry, kit, failure modes and paperwork, then put candidates through supervised welds before they touch live track. Qualification welds are usually cut out and destructively tested — bend, hardness, macro — because a weld that looks perfect can be hollow inside. In European practice, operator approval follows EN 14730-2, which requires approval testing and periodic re-verification rather than a one-time certificate (see UIC for the wider standards framework). Many people entering the trade start with a broader aluminium welding course to build fundamental technique before specialising.

Cross-training in inspection is the quiet multiplier. An operator who has looked at failed welds under a macro etch, or watched a bend test open a lack-of-fusion plane, preheats differently afterwards.

The Future of Rail Welding and What It Means for Thermit

If this guide has one takeaway, it is that thermit welding is a portable casting process, not a magic trick with rust and aluminium. The chemistry is settled and thoroughly documented: aluminium reduces iron oxide, liquid steel fills a mold, and a joint is born in about a minute of reaction time. What separates a weld that lasts thirty years from one that cracks in eighteen months is rarely the recipe in the box. It is moisture control, measured preheat, honest alignment, disciplined finishing and inspection that actually gets recorded. Everything ahead for this process builds on that same truth.

Where the Process Is Heading

Consumable design keeps improving. Mold and crucible geometry is being refined to feed shrinkage more reliably, reduce slag entrapment and make seating and sealing less dependent on operator feel. That matters because the mold is where so many field variables are either absorbed or exposed.

Rail metallurgy is the bigger pressure point. Head-hardened and bainitic grades are spreading on premium track, and both demand closer alloy matching and tighter cooling control than standard carbon pearlite rail ever did. A weld that is metallurgically acceptable in ordinary rail can produce untempered martensite or a soft, battering zone in these grades. Alloy matching is becoming the same kind of discipline that governs aluminum MIG welding, where filler selection and heat input decide the outcome before the arc ever starts.

Two other shifts are quietly reshaping quality assurance:

• Digital weld records. Capturing operator identity, kit batch, rail temperature, preheat temperature, ambient conditions and grinding results is becoming routine, and it transforms failure investigation from guesswork into evidence.
• Better non-destructive testing. Phased array ultrasonic testing and improved probe designs are being applied specifically to overcome the scattering caused by coarse cast grain structure, the long-standing weak point of thermit inspection.

Competitive Pressure, and Why Thermit Stays

Mobile flash butt welding keeps getting better, and it should. It produces forged, fine-grained joints with superior fatigue performance, and it will continue to dominate new continuous welded rail construction and high-output welding. Research into alternative joining methods will continue too. None of that eliminates the case for thermit.

Thermit remains the practical choice for emergency pull-apart repairs and isolated field closures where mobilising a flash butt welder is impossible within the available track window. No power supply, a small crew, a kit that fits in a truck, and a process that works in a trench, on a bridge or at the end of a branch line. As long as track breaks at inconvenient times and in inconvenient places, the cardboard box stays on the shelf.

A Closing Framework: Plan, Control, Prove

1. Plan the weld with track protection, the correct kit for the rail section, and a written procedure.
2. Control moisture in consumables, molds, luting sand and the rail ends themselves.
3. Measure preheat rather than estimating it, on both rail ends.
4. Align and finish to tolerance for running surface, gauge face and profile.
5. Inspect visually, dimensionally and with NDT where required.
6. Record every variable while it is still fresh.

The Three Things That Matter Most

First, moisture is the enemy. Nearly every gas-related defect traces back to it. Second, alignment and finish geometry dominate fatigue life, because most failures begin at a notch rather than in the middle of sound metal. Third, documentation is not bureaucracy; it is how a railroad proves a weld was made correctly and learns from the ones that were not.

The governing standards will keep evolving, and readers should follow the current editions of AREMA Chapter 4, AWS D15.2 and the relevant EN documents. In the United States, 49 CFR Part 213 sets the baseline for rail joints and defective rail. If you want to go deeper into related territory, our guides on welding safety, rail and structural steel, and inspection are a good next stop.

Frequently Asked Questions

Is thermite welding still used on railways today?

Yes. It remains the standard method for field closures on continuous welded rail, emergency pull-apart repairs, and welds in remote or confined locations where a flash butt welder cannot be mobilised. Flash butt welding dominates new rail production and depot work, but thermit welding is still embedded in day-to-day track maintenance.

How hot does a thermite rail weld get?

The aluminothermic reaction reaches roughly 2,500 °C in the crucible, and the molten steel that enters the mold is around 2,000 °C. That is hot enough to fuse the rail ends without any external power, and it is why moisture control and PPE are non-negotiable.

Why do thermite welds fail more often than flash butt welds?

Thermit weld metal is cast rather than forged, so it has a coarser grain structure, lower toughness and a greater likelihood of shrinkage porosity. Failures usually start at an internal defect or a geometric notch such as misalignment or a poorly blended collar, then grow by fatigue under repeated wheel loads.

What is the difference between thermite welding and aluminothermic welding?

They describe the same process. Aluminothermic is the technical and European standard term for a reaction in which aluminium reduces iron oxide; thermite is the generic name for the mixture, and Thermit is the historic Goldschmidt trade name.

Can you join rail ends with an ordinary arc welder?

Not for main line track. Rail steel is a high-carbon pearlitic steel that is highly sensitive to cracking, and field arc welding cannot control dilution, hydrogen or cooling adequately. Arc processes are confined to specific repair and cladding applications under strict procedures, not to structural rail joints.

How long does a thermite weld take from start to finish?

The pour itself takes only seconds, but the full cycle — cutting out the old joint, squaring and cleaning the ends, setting the gap, aligning, fitting and luting the mold, preheating, pouring, cooling, stripping the mold and grinding the profile — commonly takes 30 to 60 minutes depending on section and conditions.

Can thermite welds be inspected with ultrasonic testing?

They can, but it is harder than for flash butt welds because the coarse cast grain structure scatters and attenuates ultrasound. Specialised probes, phased array techniques and qualified procedures are used, and periodic destructive testing of production welds remains an important part of quality assurance.

Do you always need to preheat the rail before a thermite weld?

Preheat is a standard part of the procedure for most rail sections and kit designs, with temperatures typically in the several-hundred-degree range depending on the kit and rail section. The specific value comes from the welding procedure and manufacturer instructions, and it must be measured rather than estimated.

Is water ever used on a thermite reaction or a hot weld?

No. Water contacting molten thermite or molten steel can flash to steam explosively and throw molten material. Dry sand or a suitable Class D extinguishing agent is the correct response, and molds, rail ends and luting sand must be dry before the pour.