Everyone says “clean track pulls better,” but by how much? If tractive effort is capped by μ × weight on the drivers, then raising the wheel–rail adhesion coefficient μ by cleaning should raise the maximum pull in direct proportion. We expected a measurable jump from dirty to clean, and a smaller further gain from a light abrasive polish.
One four-axle diesel (no traction tires), weighed at 13.4 oz on the drivers. We measured drawbar pull by coupling the loco to a thread over a pulley to a cup, adding water until the drivers slipped at a steady crawl, then weighing the cup on a kitchen scale (±1 g). Five slip trials per rail condition, same 18″ tangent of nickel-silver track, same loco, same session. Conditions: (1) two weeks’ dust, untouched; (2) wiped with isopropyl alcohol; (3) alcohol-wiped then burnished with a very fine abrasive pad. Pull in grams-force, converted to the implied μ = pull / driver-weight.
| Rail condition | Mean pull at slip | Spread (min–max) | Implied adhesion μ |
|---|---|---|---|
| Dirty (2 weeks dust) | 68 gf | 61–74 gf | 0.18 |
| Alcohol-cleaned | 103 gf | 98–109 gf | 0.27 |
| Cleaned + fine burnish | 114 gf | 108–120 gf | 0.30 |
Cleaning lifted mean pull from 68 to 103 gf — a 51% increase — moving implied μ from 0.18 to 0.27, right into the clean-nickel-silver band the Lab assumes. The burnish added another ~11 gf (to μ ≈ 0.30) but with overlapping spread, so that last step is real but marginal. The trial-to-trial scatter (~±5 gf) is small next to the dirty-vs-clean gap, so the effect is not measurement noise.
In practical terms: at a 4 oz average car on the level, 35 gf of extra pull is on the order of several more cars before the drivers slip — the difference between a train that climbs the grade and one that stalls.
Clean track measurably and substantially increases pulling power — here, about half again more pull just from removing dirt. The fine burnish helped a little more but is in diminishing-returns territory. Caveats: one loco, one railhead, one humidity/day; a water-cup rig is crude; “dirty” is not standardized. Treat the magnitude as indicative, not a spec — but the direction and rough size are robust. Feed the measured μ into RailGrade instead of guessing.