Dry Dual Clutch Transmission Systems: a Model Predictive Approach to Clutch Slipping Control = Dry Dual Clutch Transmisson system: clutch slipping control using Model Predictive approach

In this thesis project, developed in collaboration with Centro Ricerche Fiat (CRF), a Model Predictive Control approach is proposed to control the Dry Dual Clutch of an Automated Manual Transmission (AMT) system during its slipping phase. In particular, the analysis is focused on the Dry Dual Clutch Transmission (DDCT) developed by Fiat Power-train Technologies. During gear shifting operations, the clutch disks that are about to close do not immediately take on the same speed: the controller’s goal is to ensure that the speed difference between the disks (slipping speed) follows a given profile and reaches the null value in a smooth way. The results taken as a reference (and therefore to be improved) are those obtained with a PID controller designed by CRF. This controller was the one that guaranteed best performances among other controllers developed using different techniques (like loop shaping or H-infinity control). Firstly, the Model Predictive Controller is simulated using the same configuration used with the PID: only the clutch torque is an input computed by the controller, with the engine torque considered as a known, external non-modifiable input, and only the slipping speed is controlled as output. Then, with the aim of improving the obtained results, some changes are made: the engine torque is considered as another manipulated input rather than a disturbance, and other two outputs are controlled: the integral error of the slipping speed and the drive shaft torsion speed. A successive tuning of the controller is also performed in order to compensate for a disturbance acting on the engine torque: this disturbance in the real word can be identified with the driver’s action. The simulated results show that in the configuration with only one manipulated variable the slipping speed reaches the null value in a faster way with respect to what was possible to achieve using a PID controller, while the second type of configurations ensure a significant reduction of the oscillations. In conclusion, the choice to use the Model Predictive approach for the clutch slipping control turns out to be advantageous in terms of performances, but also for the flexibility proper of the MPC, that can be easily extended to MIMO systems, allowing to control multiple outputs and to compensate for external disturbances.