The Silent Guardian: How Wheel Flanges Keep Trains on Track

At 14 minutes past midnight on 20 September 1967, a rake of 32 chalk hopper wagons became detached from its locomotive at Cliffe in Kent and ran away down the 1 in 100 grade toward the Thames estuary for approximately 1.9 kilometres before derailing at a curve. Post-incident investigation by the Ministry of Transport’s Inspectorate of Railways found multiple failures of the train’s brake system, but the derailment sequence at the curve provided an unusually clean set of evidence about how flange-to-rail contact forces develop under high lateral load. The curved track showed a distinctive gouge pattern on the gauge face of the outer rail — a continuous groove 8–12 mm wide and 2–3 mm deep, running for 43 metres before the point where the leading wagon’s wheel finally climbed the rail. The groove was made by the wagon flange pressing laterally against the gauge face with sufficient force to plastically deform the rail steel: the flange was not guiding the wheelset, it was cutting into the rail. The metallurgical analysis of the gouge matched the force history reconstructed from wheelbase geometry and rolling radius calculations: the lateral force on the outer flange had reached approximately 120 kN — well above the 40 kN or so that would represent normal flange-rail contact in a curve of that radius, and well into the range that Nadal’s criterion, derived more than half a century earlier, had identified as the threshold for potential wheel climb. What the Cliffe investigation contributed to the literature was a rare piece of physical evidence — the rail itself as a force-measuring instrument — confirming that the theoretical Y/Q criterion was not merely a conservative design parameter but a boundary that, once crossed, was measurable in the steel. Understanding why that threshold exists, what controls it, and how flange geometry determines whether a wheel will climb or be guided, is the subject of this article.

What Is the Wheel Flange?

The wheel flange is the projecting circumferential ridge on the inner face of a railway wheel rim. It runs parallel to the wheel’s rotation axis, projects inward toward the centre of the track, and its inner vertical face engages the gauge face of the rail head when the wheelset is displaced sufficiently far toward that rail to bring the flange into contact. The flange is not a separate component — it is an integral part of the monobloc wheel forging, machined to precise dimensional tolerances defined in EN 13715 and the associated profile standards.

Three parameters define the flange geometry as specified in EN 13715: flange height Sh (the vertical distance from the tread running surface to the top of the flange — typically 28–32 mm new, minimum 26 mm worn, maximum 36 mm); flange thickness Sd (measured at a defined depth of 10 mm from the flange tip — typically 29–33 mm new, minimum 22 mm worn); and flange steepness qR (the vertical height difference between the flange profile at 2 mm and 10 mm lateral distance from the outer edge of the flange face — a measure of how steeply the flange face tapers, with operational limit ≤ 6.5 mm per EN 15313).

Nadal’s Derailment Criterion: The Physics of Wheel Climb

The mathematical relationship between the forces acting on a flanging wheel and the onset of wheel climb was derived by M.J. Nadal of the French railways in 1908 and has remained the fundamental criterion for derailment safety assessment in every subsequent railway engineering standard worldwide.

Force Geometry at the Flange Contact

When a wheel flange presses against the gauge face of a rail, the contact point is at the upper portion of the flange face, where the flange transitions from its nearly vertical face toward the flange top. The normal force at this contact point acts perpendicular to the local flange surface — which is inclined at angle δ (the flange angle) from the vertical, typically 60–75° for modern S1002 profiles. Two forces act at this contact: the lateral force Y (the horizontal force pushing the wheelset toward the rail) and the vertical force Q (the wheel load, acting downward). The question Nadal asked was: under what combination of Y and Q will the frictional force at the flange contact point be insufficient to prevent the wheel from riding upward along the inclined flange surface — climbing over the rail head?

The Nadal calculation reveals two important and sometimes counterintuitive results. First, lubricated flanges are geometrically safer (higher Y/Q limit) than dry flanges because the lower friction coefficient makes the climbing mechanism less efficient — the wheel tends to slide laterally rather than climb. This explains why flange lubrication reduces derailment risk rather than increasing it. Second, worn flanges with shallower face angles (reduced δ) are significantly more prone to climbing than new flanges with steep face angles — which is why qR (flange steepness) is a safety parameter and not merely an aesthetic one.

Flange Wear: Mechanisms, Rates, and Measurement

Flange wear — the progressive reduction of flange thickness Sd — occurs by adhesive and abrasive wear at the flange-gauge face contact interface. Unlike tread wear, which is distributed around the full circumference of the wheel, flange wear is concentrated at the specific circumferential zone where the flange face contacts the rail gauge face — typically a 5–15 mm wide band on the flange face at the contact angle. The wear rate depends on four primary factors: the lateral force magnitude, the creep (slip ratio) between flange and gauge face, the friction coefficient, and the material hardness of both surfaces.

High Rail vs. Low Rail: Asymmetric Flange Wear in Curves

On a curve, the outer (high) rail receives the majority of the lateral guidance loading. The high-rail wheel’s flange presses against the outer gauge face with a force proportional to the curving cant deficiency (the degree to which the train is travelling faster than the curve’s equilibrium cant speed). The inner (low) rail wheel’s flange typically does not contact the inner gauge face at all on properly superelevated track. This asymmetry means that, in service, wheel flanges on one side of a vehicle tend to wear faster than the other: on routes with predominant curve direction (more left-hand or more right-hand curves), the wheels on the high-rail side show faster flange wear. Railways with balanced curve topology — similar numbers and radii of left and right curves — show more symmetric flange wear across the fleet.

The qR Parameter: Flange Steepness and Switch Safety

The qR parameter — the flange steepness — is defined as the height difference between the flange profile at lateral positions 2 mm and 10 mm from the outer edge of the flange face. A new S1002 wheel has qR ≈ 4–5 mm — a moderately sloping transition. As the wheel wears in service and the flange face contact zone shifts inward, the outer portion of the flange face can become nearly vertical (steep), increasing qR toward the 6.5 mm EN 15313 limit. A steep flange face is dangerous at switch toes for two reasons: first, the steep geometry provides less of a “ramp” effect that would naturally guide the wheel around a misaligned switch blade rather than over it; second, a steep flange can lodge in the small gap between the switch blade and the stock rail (the “switch check”) at certain blade tip positions, wedging the blade open rather than riding past it. This switch-picking mechanism has been identified in several freight wagon derailments at switches and is the primary motivation for the qR parameter existing as a separate measurement from Sd.

The Mechanics of Wheel Climb: How a Derailment Actually Happens

Wheel climb derailment is not instantaneous. It is a process that develops over a finite distance — typically 0.5–5.0 m of travel — during which the flange rides progressively higher up the rail gauge face, with the tread simultaneously rising above the rail head level, until the flange tip clears the top of the rail and the wheel falls outward off the running surface. Understanding this process explains why derailment can sometimes be prevented by track geometry correction or speed restriction even after the Y/Q limit is exceeded transiently.

The Three Stages of Wheel Climb

Stage 1 — Flange contact initiation: The wheelset is displaced laterally by a cant deficiency force, hunting oscillation, or track geometry defect until the flange face contacts the gauge face of the rail. Initial contact is at moderate force (Y/Q 0.3–0.7); the flange is in rolling contact with the gauge face, with modest creep. The tread remains fully on the rail head.

Stage 2 — Flange climbing: If the lateral force continues to increase — either because the curve tightens, the speed is excessive for the cant, or the wheelset is dynamically loaded by track irregularity — the Y/Q ratio rises toward and beyond the Nadal limit. The frictional resistance at the flange contact zone is insufficient to prevent the wheel from riding upward along the inclined flange face. The tread begins to lift off the rail head on the flanging side. The wheel is now in partial support on the flange. Rail head contact shifts to the outer tread edge.

Stage 3 — Wheel over-rail: The flange tip reaches the level of the rail head top surface. At this point, the wheel has effectively climbed the rail and the contact transfers to the outer surface of the tread above the rail head. If the vehicle continues in this state, the axle is no longer constrained laterally by the rail, and the vehicle derails.

The False Flange: How Hollow Tread Wear Mimics Flange Failure

The “false flange” is one of the less intuitive hazards in wheel-rail contact, yet it is implicated in a disproportionate number of switch-area derailments on fleets where tread wear monitoring focuses primarily on dimensional limits rather than profile shape.

When a wheel tread wears preferentially in the contact band (the central 20–40 mm of tread width where most normal contact occurs), the tread develops a concave (hollow) cross-section. The outer edge of the tread — the rim shoulder between the tread and the flange root — may remain relatively unworn, projecting downward as a ridge. This ridge is the “false flange.” It occupies a lateral position approximately 60–70 mm from the flange face — well outside the normal flange-rail clearance envelope — and its tip may be at a lower elevation than the true running surface if the hollow is deep enough.

When a wheel with a significant false flange negotiates a switch, the false flange can engage the switch blade tip or the check rail approach geometry in locations and orientations not designed for lateral contact. The check rail and switch blade geometries are designed for engagement by the true flange at its specified lateral position and angle; a false flange arriving at the switch from an unexpected position can apply lateral forces to the switch blade tip that deflect the blade or, in extreme cases, lift the wheel over it. EN 15313 (Railway applications — In-service monitoring of wheelsets — Requirements for in-service and off-vehicle wheelset monitoring) addresses false flange through the parameter of maximum hollow wear depth — typically 3–5 mm depending on operator — as a complementary measurement to the flange parameters Sh, Sd, and qR.

Flange Lubrication Systems: Reducing Wear Without Contaminating the Rail

Flange lubrication is the primary maintenance strategy for extending flange life on routes with significant curving. The objective is to apply a thin, controlled film of lubricant to the wheel flange face and/or the rail gauge face, reducing the friction coefficient at the flange-gauge face contact interface from the dry value (~0.35) to a lubricated value (~0.08–0.15). The Nadal calculation shows that this actually increases the theoretical Y/Q safety margin while dramatically reducing flange and gauge face wear rates — typically by 70–80% on tight-curve sections.

Types of Flange Lubrication Systems

The Tread Contamination Risk

The critical engineering requirement for any flange lubrication system is zero lubricant on the rail crown — the top surface that the wheel tread rolls on. Grease or oil on the rail crown reduces wheel-rail adhesion from approximately 0.25–0.35 (dry) to potentially 0.03–0.08 (grossly contaminated), creating wheel slide conditions that defeat WSP systems, extend stopping distances, and in the worst cases prevent trains from restarting on grades. The lubrication zones for gauge face and crown are separated by only 30–40 mm horizontally, and lubricant migration — by centrifugal spreading from wheel rotation, by splash under rain, or by wheel flange sliding the lubricant laterally — is the primary failure mode of poorly designed lubrication systems.

Modern GPS-triggered onboard systems address this by applying lubricant only at precisely mapped curve entry points, using micro-dosing pumps that deliver 0.1–0.5 ml per application — far less than older spring-contact systems which could apply 2–5 ml continuously. Network Rail’s Technical Specification for Flange Lubricators (RT/LS/P/010) requires that any onboard or trackside lubrication system demonstrate, during commissioning testing, that the adhesion coefficient at 30 m past the lubrication application point remains above 0.15 on the rail crown under the worst-case temperature and application rate conditions.

Check Rails and Guard Rails: Infrastructure-Side Flange Guidance

Where the consequences of a wheel climbing the rail or taking the wrong route at a junction are particularly severe — on bridges, in tunnels, at level crossings over main lines — the track infrastructure itself incorporates a secondary guidance system that reinforces the flange’s derailment prevention role: the check rail (also called a guard rail in some applications).

A check rail is a short section of rail (or a specially profiled steel section) mounted inside the running rail, positioned so that its face is approximately 41–44 mm from the gauge face of the running rail (for standard gauge 1,435 mm track). When a wheel is displaced far enough laterally to bring the back face of the flange into contact with the check rail, further outward displacement is physically prevented. The check rail thus limits the maximum lateral travel of the wheelset to a defined value — preventing the outer wheel’s flange from climbing the outer rail even if the Y/Q forces at the outer contact exceed the Nadal limit.

Check rails are mandatory at specific locations under Network Rail’s NR/SP/TRK/3060 standard: on all movable bridges and swing bridges; through tunnels with restricted evacuation access; at complex junctions with limited flangeway clearance; and at level crossings over main lines. The standard specifies a maximum check rail clearance of 44 mm from the outer rail gauge face (measured at the narrowest point) and a minimum flangeway width of 38 mm (to allow flange passage without wedging). The check rail section must withstand a 50 kN lateral load applied by the back of the flange at any point along its length without deflection exceeding 1 mm — a structural requirement that distinguishes the load-bearing check rail from the cosmetic “guard iron” used solely to prevent ballast intrusion into the flangeway.

Flange Contact States: Normal Operation vs. Warning vs. Critical

Flange-Related Derailments and Engineering Lessons

Frequently Asked Questions

1. Why does lubrication of the flange increase the Y/Q safety margin rather than reducing it — is this counterintuitive result actually correct?

It is counterintuitive but correct, and the Nadal formula shows exactly why. The Y/Q limit for wheel climb is: (tan δ − μ) / (1 + μ × tan δ). As the friction coefficient μ decreases (more lubrication), the numerator (tan δ − μ) increases and the denominator (1 + μ × tan δ) decreases — both changes push the Y/Q limit upward, meaning more lateral force is required to cause wheel climb. Physically, this makes sense: a highly lubricated flange-rail interface allows the wheel to slide laterally along the gauge face rather than climbing it. The climbing mechanism requires friction — the flange face must “grip” the rail surface and push the wheel upward. With low friction, the wheel simply slides sideways along the gauge face without climbing. This means that a well-lubricated curve is intrinsically safer against wheel climb than a dry curve, all else being equal. The practical concern is not lubrication reducing safety — it is lubricant migration onto the tread running surface reducing traction and braking adhesion, which is an entirely separate and real hazard. The solution is precise application geometry (lubricate only the gauge face, not the crown) and controlled dosing, not avoidance of lubrication. A perfectly lubricated gauge face with zero crown contamination is the safest possible state from both wear and derailment perspectives simultaneously.

2. What is the difference between a check rail and a guard rail — are these the same thing, and where is each used?

The terms “check rail” and “guard rail” are often used interchangeably but technically describe slightly different applications of the same concept. A check rail is mounted inside the running rail, close to the gauge face of the opposite rail, with a clearance gap of approximately 41–44 mm. It prevents the back of the far-side wheel’s flange from moving outward — limiting the maximum lateral displacement of the wheelset. Check rails are used where derailment must be absolutely prevented: on bridges, in tunnels, at complex junctions, and at any location where a derailed vehicle could fall (bridge decks) or block evacuation routes. A guard rail is used in a slightly different way: it is typically mounted alongside the running rail, parallel to it and just outside it, specifically at level crossings and other locations where the risk is not lateral displacement of the wheelset but longitudinal intrusion of the wheel flange into a gap between track components — preventing the flange from dropping into the gap and derailing the vehicle. At level crossings, the guard rail runs on both sides of each running rail to prevent wheel flanges from catching the edge of the road surfacing or expansion joint. In current Network Rail standards, “guard rail” and “check rail” are both described in NR/SP/TRK/3060 but with different dimensional requirements reflecting their different loading and function: check rails are designed for 50 kN lateral load; guard rails at level crossings are designed for 30 kN vertical impact load. The common thread is that both are passive mechanical stops that exploit the flange’s geometry to redirect or absorb lateral forces that the tread-steering mechanism has failed to manage.

3. How does the track gauge directly relate to flange safety — and what happens if the gauge widens beyond its maintenance limit?

Track gauge — the distance between the inner faces of the two running rails, nominally 1,435 mm for standard gauge — defines the lateral space available for wheelset movement. The nominal back-to-back dimension of a standard gauge wheelset (AR, the distance between the inner flange faces) is 1,360 mm, leaving a nominal lateral clearance of (1,435 − 1,360) / 2 = 37.5 mm between each flange face and its adjacent rail gauge face. This clearance is by design: it allows slight lateral misalignment between track and wheelset without generating flange contact forces, and it accommodates the wheelset yaw needed for curve negotiation without jamming. If gauge widens progressively — through rail fastening loosening, spike withdrawal, or sleeper deterioration — the clearance increases. At gauge widths above approximately 1,470 mm (35 mm wider than nominal), the flange back-to-back clearance increases to 55 mm on one side — sufficient for the wheel to drop through between the rails if the vehicle is on a section of wide gauge at a switch heel or other lateral support discontinuity. This is the mechanism behind the Potters Bar derailment (2002): the switch geometry had been altered by a loosened clamp bolt to the point where, at a specific switch position, the effective gauge was wide enough for the wheel to move laterally past the gauge face of the stock rail, losing lateral support. Network Rail’s Plain Line Pattern Recognition (PLPR) system on its New Measurement Train monitors track gauge continuously at 100 mph with ±0.5 mm accuracy, specifically to detect progressive gauge widening before it reaches the point where flange guidance fails.

4. Why does the Santiago de Compostela derailment (2013) illustrate the Y/Q limit failure mode — and what made the curve geometry particularly dangerous?

The Santiago de Compostela derailment is the best-documented recent example of speed-induced Y/Q exceedance causing a wheel-climb derailment on a main-line passenger train. The Alvia train was travelling at 179 km/h on a curve with a design speed of 80 km/h and a radius of approximately 400 m. At 179 km/h on a 400 m radius curve, the centrifugal acceleration experienced by the vehicle was approximately 179²/(3.6² × 400) = 6.2 m/s² — more than six times the acceleration the track cant (typically 150 mm on a 400 m curve) was designed to compensate. The resulting lateral force on the outer bogies, distributed over the wheelsets, produced Y/Q ratios well above the Nadal limit of approximately 1.2 for the bogie geometry involved. The specific geometry of the Angrois curve also included a descending grade at the curve entry, meaning that the train’s suspension was loaded asymmetrically (weight transfer toward the outer side of the curve due to both centrifugal force and longitudinal deceleration from the grade), increasing the vertical load on the outer rail wheelsets and the lateral force simultaneously. The derailment sequence showed the leading bogie’s outer wheelset climbing the outer rail approximately 30–40 m after the curve entry — consistent with the wheel climb distance calculation for Y/Q values of 1.5–2.0 at the measured speed excess. The ATP system that should have prevented the speed excess had been deactivated on that section of track due to a compatibility issue between ETCS and the older national ATP system, leaving the driver’s attention as the sole speed enforcement mechanism — which, on that curve, failed.

5. What is flange oiling versus flange greasing — and does the lubricant type matter for effectiveness and rail contamination risk?

Flange lubricant type significantly affects both effectiveness and contamination risk, and the industry has moved toward more sophisticated lubricant formulations over the past two decades. Traditional rail gauge face lubrication used petroleum-based grease (NLGI Grade 1 or 2, typically lithium soap thickener base) applied by brush-contact applicators. Grease has the advantage of high viscosity — it stays where applied and does not run, reducing crown contamination risk — but it attracts ballast dust and rail abrasive particles, forming a grinding compound that can actually accelerate wear if the lubricant layer is thick. Modern systems favour two alternatives. The first is solid lubricant sticks — waxy, talc-based or PTFE-compound sticks that are applied to the wheel flange by spring contact and transfer to the rail gauge face as the wheel rolls. These produce a very thin, dry-film lubricant layer with low migration onto the rail crown and excellent environmental acceptability (no petroleum spillage risk, particularly relevant in water catchment areas). The second is water-soluble oil emulsions applied by micro-dosing spray — these provide good lubrication immediately after application and evaporate or wash off within 500–1,000 m, limiting the carry distance of the lubricant along the rail and reducing crown contamination accumulation over a route. For tram systems in pedestrian zones, biodegradable ester-based lubricants are mandatory in most European cities under contaminated land regulations. The EN 15427 standard (Railway applications — Wheel/rail friction management — Performance criteria) specifies minimum friction reduction requirements (at least 50% reduction in friction coefficient versus dry baseline) and maximum carry-distance limits (lubricant must not reduce friction below 0.15 on the rail crown at any point more than 50 m beyond the application point) for all new flange lubrication systems installed on EN-compliant railway infrastructure.