Calibration and Estimation of Dog Teeth Positions in Synchronizers for Minimizing Noise and Wear during Gear Shifting

Sammanfattning: Electric motors are used more widely in automotive to reducing emissions in vehicles. Due to the decreased usage of internal combustion engines which used to be the main noise source, impacts from synchronizers cannot be ignored during gear shifting, not only causing noise and wear but also delaying gear shifting completion.  To minimize the impacts during gear shifting, a dog teeth position sensor is required but the high calculation frequency leads to a high cost, due to the high velocity of synchronizer portions and the dog teeth number.  In this thesis, the gear shifting transmission is being modelled, in order to study the process of gear shifting and engagement. The transmission model, which is expressed with electrics and dynamics formulations. In order to avoid the impact without the dog teeth position sensor, this thesis proposes an estimation algorithm based on the transmission model to approve the gear engagement if the first and second portions of synchronizers are engaged in the mating position without impacts.  Two different learning algorithms: direct comparison and particle swarm optimization application, are presented in the thesis as well, which are used to calibrate a parameter in the off-time test as part of the end of the calibration line, the so-called relevant initial phase being used in the real-time estimation.  The transmission model is simulated in Simulink and different algorithms are running in MATLAB. All these results are plotted and analyzed for further evaluation in different aspects in the result chapter. The direct comparison algorithm has a simpler structure of computation but the quantity of required actuation is uncertain in this algorithm with a probability of failure to find the solution. The application of particle swarm optimization in this case succeeds in calibrating the objective parameter with a small error than the other algorithm. Actuation quantity affects the accuracy of the solutions and errors but not the failure rate.