Torque Reconstruction for a Full-Scale Maritime Powertrain Using Trend Estimation

The aim of this article is to present a trend-filtering approach for the simultaneous estimation of unknown input torque and initial states, which can be used to reconstruct the torsional response in the drivetrain of an azimuthing thruster. Accurate information of shaft torques is essential for condition monitoring and improving system design practices. Torque reconstruction with virtual sensors allows more freedom in choosing sensor locations compared to using only physical sensors, and enables the estimation of unknown external disturbances which are difficult to measure directly. Empirical analysis of the torque estimation method is carried out with simulations. Verification is done on experiments on a laboratory testbench and full-scale operational measurements of an azimuthing thruster. The results show that regularized least-squares estimation can be used to accurately estimate propeller shaft torque in a full-scale azimuthing thruster with measurements from the driving motor shaft, by applying regularization constraints based on physical characteristics of the external forces and torsional response of the thruster.