Estimation of torque in heavy duty vehicles with focus on sensor hysteresis
Doctoral thesis, 2016

For diagnosis and feedback control of vehicle powertrains, it is highly desired to monitor the torques at different locations online. The most obvious way to do this is to measure cylinder pressures, from which cylinder individual delivered torques can be calculated, or to use torque sensors. Unfortunately, both the pressure and the torque sensors are expensive as parts and to install. Also, pressure sensors have durability issues due to the harsh environment and, for heavy duty applications, it appears as if it is only realistic to place torque sensors on the gearbox side of the flywheel. This motivates the still active research area on torque estimation from flywheel oscillations, which involves hardly no hardware costs. In this work, two applications of a torque sensor on the gearbox input shaft in a heavy duty vehicle are explored. For the first application, which is for control of AMT gear boxes, the physical location of the sensor is ideal. However, the sensor has an hysteresis which comes into play for this application. A general method for sensor hysteresis compensation is to apply an inverse hysteresis model in series with the hysteresis, and hysteresis model identification methods are here derived and applied to the sensor. The sensor is subject to an ageing process in the vehicle and a method for adapting the inverse hysteresis model has therefore been developed. Experimental data indicates that the method can essentially eliminate ageing effects for the current application. General results for hysteresis models, such as identifiability conditions for the generalized Prandtl-Ishlinskii model, are also reported. The other application is estimation of the cylinder individual pressures, for detection of periodic combustion faults. A new estimation method based on Fourier series and a cylinder pressure parameterization is described. It is motivated that if a parameterization with sufficiently few parameters is used, then an overdetermined system of equations for the cylinder pressures can be formulated. A method for identifying minimal cylinder pressure bases from data is therefore derived. By replacing the torque sensor by a model for the driveline, the method is directly applicable for estimation using only the flywheel sensor. Data for the vehicle indicates that clutch ageing can result in significant changes of driveline dynamics, and therefore driveline model errors, but it is motivated that estimators robust to model errors or even online identification of driveline models may be possible by using the derived estimation method.

hysteresis modelling

adaptive hysteresis compensation

hysteresis compensation

magnetoelastic torque sensor

Preisach model

generalized Prandtl-Ishlinskii model

cylinder pressure estimation

torque estimation

EC, Hörsalsvägen 11, Campus Johanneberg
Opponent: Professor Ciro Visone, University of Sannio, Italy


Marcus Hedegärd

Chalmers, Signals and Systems, Systems and control

Convex identification of models for asymmetric hysteresis

American Control Conference,; (2014)p. 4753-4758

Paper in proceeding

Adaptive hysteresis compensation using reduced memory sequences, M. Hedegärd, T. Wik, and C. Wallin

Identifiability of generalized Prandtl-Ishlinskii models, M. Hedegärd, T. Wik, and B. Kulmar

Convex identification of minimal bases for cylinder pressure, M. Hedegärd, T. Wik, K. Fredriksson, and J. Engbom

Subject Categories

Control Engineering

Signal Processing



Doktorsavhandlingar vid Chalmers tekniska högskola. Ny serie: 4152



EC, Hörsalsvägen 11, Campus Johanneberg

Opponent: Professor Ciro Visone, University of Sannio, Italy

More information

Latest update

4/2/2019 9