In this project the potential use of placing the ABB torque sensor (Torductor) between the flywheel and the gearbox is explored. In particular the research is focused on two applications; zero torque prediction and combustion diagnosis.
Placing the sensor on the engine shaft, in front of the flywheel, has been studied in several previous CERC projects with focus on combustion phasing. In heavy duty vehicles this location is not feasible and hence the new location in the gearbox. This location, on the other hand, means that the sensor can be used for combustion analysis and to better predict the zero torque point; thereby improving the timing of gear shifts. In turn, improved gearshift timing
may allow for implementation of Automatic Manual Transmissions (AMTs) in heavy duty vehicles (e.g. garbage trucks, construction vehicles and city busses) where high-loss hydraulic transmissions are used today.
The presence of a flywheel implies that the measured torque signal can be regarded as a lowpassed signal of the engine shaft torque, though with the complication of clutch dynamics. Having that torque signal, prior CERC results can be used to separate the signal into torque contributions from specific cylinders, which we aim to use primarily for diagnosis of the injectors. Depending on the quality of the separated torque signals there is also a possibility that feed-forward injection control can be complemented with phasing feedback.
In these applications it is important to have very accurate torque measurements. The sensor has high accuracy in absolute terms relative to the maximum torque. However, a large part of the existing error is caused by complicated magnetostrictive hysteresis that in relative terms can be very high at low torques. In order to reduce these effects, the hysteresis is modelled by an extended generalized Prandtl-Ishlinski model. The effects of hysteresis can be significantly reduced (up to 80-90%) by applying the inverse model to the measurement signal. To carry out individual calibration of mass produced sensors it is important to find a rapid calibration cycle and optimization of model parameters. For that purpose an efficient method for convex identification of extended generalized Prandtl-Ishlinski models has been developed and evaluated on different types of hysteretic sensors. During the lifetime of a truck, the sensor and its hysteresis change significantly due to aging and therefore an adaptive scheme based on extended Kalman filters and known zero torque measurements (open clutch) has been developed.