Ivan Scansani Gregori

A data-driven learning system that reads the torque signal of a dry clutch, tells six friction materials apart, and predicts judder in the vehicle from a laboratory bench test.

Multivariate statistics applied to real bench-test torque signals separates clutch facings by judder performance and manufacturing characteristics, and establishes a reference torque behaviour for material development.

Finite element analysis is used to predict when a clutch facing bursts, with numerical results showing clear agreement with experimental burst tests.

A methodology for classifying clutch facing judder sensitivity on a sub-scale bench test instead of a full vehicle, with preliminary results showing good correlation to the car.

The original proposal for moving clutch facing judder classification off the vehicle and onto a sub-scale bench machine, motivated by test time and material development speed.

A two-part methodology for characterising facing judder sensitivity: numerical simulation of the judder bench test to save time, validated against an experimental judder test in a passenger car.