Cable Winch Gear Dynamics

The Gears

The gears simulated here have a specific provenance. They once drove a cable winch in the Beringen coal mines of Belgium, and were brought to KU Leuven's campus in Heverlee after the mines closed in the late 1980s. They are large, industrial, and equipped with a dog clutch that selects between two gear speeds — a mechanism familiar to anyone who has studied race-car gearbox design, albeit at a considerably larger scale.

Predicting their dynamic transmission error and stress distributions required more than a straightforward application of standard gear dynamics tools. These gears have lightweight cut-outs in the gear body that disturb the clean cyclic symmetry of the tooth crown. That clean symmetry is exactly what the MUTANT toolbox developed at KU Leuven LMSD exploits, through model reduction, to keep the computational cost of gear dynamics manageable. The cut-outs break it, and with it much of the solver's efficiency.

Model and Assumptions

The solution draws on ideas from dynamic substructuring. Rather than treating the entire gear as one monolithic flexible body, the analysis separates the gear body from the contact analysis of the gear teeth. The tooth contact region retains its fine mesh and full nonlinear contact treatment, while the body beneath it is handled with a reduced-order model that can accommodate the non-periodic geometry without sacrificing accuracy.

This split allowed the gear dynamics solver to regain much of its computational efficiency even for components that violate the assumptions underlying the standard cyclic symmetry approach. The resulting model predicts transmission error, dynamic contact forces and stress distributions across the tooth flanks at a fraction of the cost of a brute-force finite element simulation.

Reflections

This project is computationally a trip back to methods developed during earlier research, applied to a machine that predates the simulation tools used to analyse it by several decades. There is something instructive about that gap: the gears are robust, reliable objects that have outlasted the industrial context that required them. The simulation methods applied here are designed with a similar philosophy — to remain predictive without becoming brittle, and to exploit structure in the problem rather than brute-forcing through it.

Whether the gears still belong on a modern engineering campus is debated; from a modelling perspective, they present a genuinely non-trivial challenge — and a reminder that the right reduction, rather than raw computing power, is what makes such a problem tractable.