An 85′ heavyweight passenger car derailed reliably on one 18″ curve, always the same way: the leading outer wheel climbed the railhead and the car stepped off to the outside. Shorter cars were fine on the same curve; the passenger car was fine on 24″ and broader.
Climb-to-the-outside at the lead wheel points at the flange fighting the rail, so the suspects are lateral-force (how hard the flange is pushed into the rail) and the Nadal limit (how steep the flange face is). Back-to-back and gauge checked in-spec, so it wasn’t a gauge-narrowing bind. The car’s long truck centers meant large overhang on an 18″ curve — the body was forcing the trucks to a sharp swivel, and the leading truck was running with its wheels skewed hard against the outer rail.
Two geometry facts combine. First, a long car on a tight radius develops heavy overhang and a large truck swivel angle, which drives the leading flange hard against the outer railhead — a big lateral force L. Second, whether that wheel climbs is set by Nadal’s limit: the wheel climbs when L/V exceeds about tan(flange-face angle). The RP-25 flange on this car, plus the high L from the overhang-driven angle of attack, pushed L/V past the limit. On 24″ the overhang (and so L) drops enough to bring L/V back under the line, which is why broader curves cured it.
Lower L, or raise the L/V the wheel can take:
- Respect the car’s minimum radius. An 85′ car wants broad curves; the overhang math says 18″ is simply below its practical floor. Re-route it to 24″+ if at all possible.
- Reduce angle of attack — free up truck swivel, confirm the truck isn’t binding on the bolster or a wire, and make sure both axles are rolling freely so the lead axle isn’t being dragged into the rail.
- Check the flange profile. A deeper, steeper RP-25 flange tolerates a higher L/V before climbing than a shallow fine-scale one.
- Weight it properly so V isn’t low — a light car has a worse L/V for the same lateral push.
A wheel climbing to the outside is Nadal’s limit being exceeded: too much lateral force for the flange angle and wheel load. On a long car, the lateral force is manufactured by overhang on a too-tight curve — so the real fix is almost always radius, not the wheel.