A centrifugal clutch that Edvin Börthas and I designed for a go-kart’s drivetrain in KTH’s Product Development course: sized by hand, modelled in CAD and checked in an acceleration model. A design, not a built part.
A go-kart’s engine has to idle while the kart stands still, then pass its torque to the wheels as the revs rise, with nothing for the driver to operate. In KTH’s Product Development course the kart was split between subteams, and Edvin Börthas and I designed its drivetrain: a centrifugal clutch, the engine mount and a chain drive for a Honda GC190 engine. I developed the clutch concept, did most of its final CAD and made the KeyShot renders on this page; we did the sizing calculations and wrote the report together. Nothing was built. Every number on this page is calculated or simulated.
The carrier, two shoes and return springs inside the drum. Original KeyShot render.
01 / How it engages
Speed moves the shoes. Friction takes the load.
The carrier turns with the engine shaft. As the revs rise, centrifugal force pushes the two shoes outward against their springs until the linings, with a friction coefficient of 0.35, touch the drum; from then on the torque passes through friction to the chain sprocket.
The design problem is to choose the shoes and the springs together. The torque the clutch can carry grows with the square of the shaft speed, less what the springs hold back, so two conditions settle the two unknowns: no torque at the speed where the shoes should first touch, and the full engine torque of 11.2 N·m at the speed where they should stop slipping. Those two equations give the shoe mass and the spring preload.
Exploded clutch assembly: the shaft interface, carrier, shoes, drum and fasteners.
INPUT
Shaft interface and carrier
CONTACT
Two shoes, linings and return springs
OUTPUT
Drum, bearing and chain sprocket
The sizing approach+
Alongside the hand calculation, a Python sweep searched shoe angle, drum radius, lining thickness, air gap and spring rate, and rejected any combination above 1.2 MPa of lining pressure or 0.8 MPa·m/s of pressure times sliding speed. It found acceptable combinations, not the best one. It assumed simple ring-shaped shoes, which the final shoes are not, and it ignored the cost of the lining and how well it sheds heat. The report draws the same conclusion: the idea is sound, but it should optimise for low mass, heat dissipation and cost.
Source: Reviewed drivetrain report, PDF pp. 8–11, 17; Python sweep appendix pp. 47–51.
02 / Around the clutch
From the shaft to the frame.
The clutch also has to sit on the engine shaft, stay supported, drive a chain and stay serviceable in the frame. It is keyed to the shaft, held with a circlip and runs on two SKF 61908 deep-groove ball bearings. Those details decide whether the calculated torque has a real path to the wheels, which is why the CAD around the clutch mattered as much as the clutch itself.
The chain drive uses 13- and 39-tooth sprockets for a 3:1 ratio. A ratio of 3.5 first looked necessary to reach the acceleration target, but a higher ratio costs top speed, and 3:1 met both requirements in the model. We sized the engine mount, the chain and the joints alongside the clutch.
Bearing and shaft detail in the designed assembly.
Designed and purchased parts+
We identified the shaft transition, drum, shoes, linings and carrier as custom components; bearings, springs, fasteners and sprockets are selected parts. The rendered vehicle includes work by the other subteams; our credit here is the drivetrain.
Source: Report PDF pp. 12–24; custom-component table p. 17.
03 / System behaviour
From engagement to acceleration.
The course’s acceleration model connects engine torque, clutch capacity, reflected inertia, and rolling and aerodynamic load. With our parameters, the shoes begin to slip at about 1,200 rpm, the clutch stops slipping at about 2,500 rpm, and the kart reaches 50 km/h in 9.692 seconds, inside the 10-second requirement. Those are model results, not a measured run.
The model is only as good as its engine data, which stops at 4,000 rpm. The report’s top speed lies beyond that range, so I leave it out. Within the documented range, the model gives about 68 km/h, above the 65 km/h the course required.
The chain drive in the team’s go-kart assembly. Our part is the drivetrain.
Model provenance and limits+
The acceleration framework is the course’s, written by Anders Söderberg. We configured it with our parameters and with our own fitted curve for the clutch torque. The milestones above come from that simulation; none of them is a test of a built clutch.
Source: Report PDF p. 27; PRO35_MATLAB.zip, acceleration and torque functions.
04 / Design review
The details decide what is ready.
We sized the parts against a safety factor of 2.1, chosen with Pugsley’s method. Two checks came out below it: the bearing pressure in the chain at 1.355, and the sprocket joint at 1.513. Other parts went the other way: the sprocket clamps are over-dimensioned and could be made smaller.
We proposed a third fastener for the sprocket joint. Before anyone builds it, the design should be revised and both margins recalculated.
Two safety factors below the 2.1 target
Check
Safety factor
Target
Review
Chain bearing pressure
1.355
2.1
Below target
Sprocket joint
1.513
2.1
Below target
What I took from the project+
In my reflections at the time I listed four contributions: breaking the system into functions, developing the clutch concept, doing most of its final CAD, and bringing in KeyShot for the presentation renders. The lesson I keep is that splitting a machine into functions is only half the job. The other half is putting it back together, with an explicit check at every interface.
Source: Report PDF pp. 14, 23–24; my course reflections, October and December 2025.
Guided scroll · scroll or press Escape to take over
