Practical Virtual Sensor Deployment for Indirect Torque Estimation in a Range Rover Drivetrain

This chapter presents a straightforward framework for input characterization when challenging modeling conditions hinder the use of parametric estimators for accurate source reconstruction. The study case of a Range Rover Evoque drivetrain is utilized as a testing platform to address such problems. The aim of this work is to indirectly estimate the torque generated in the prop-shaft of the Evoque while considering the full assembly and complex interactions among all the drivetrain components. To achieve this, a least squares alternative approach leveraging training data is proposed to practically deploy a virtual sensor solution in the physical asset. The study provides an overview of the technical specifications of the Evoque drivetrain and the sensor layout designed to capture the system’s dynamic response. Operational datasets obtained under various conditions are described, serving as the basis for the analysis. The performance of the data-driven estimator is assessed in both the time and frequency domains, with a particular focus on phase differences between the training data and operational data. The results highlight the effectiveness of the proposed framework in addressing the challenges posed by complex multibody systems. The data-driven approach demonstrates promising performance with relatively small errors in the order of 5% of the range of the measured torque signals, enabling reliable source reconstruction.