Frank Naets

Design and Experimental Validation of a Stable Two-Stage Estimator for Automotive Sideslip Angle and Tire Parameters

This paper proposes and experimentally validates a two-stage approach for coupled lateral vehicle state and tire model estimation. In a first stage, an extended Kalman filter is employed which provides vehicle slip angles and lateral tire forces from commercial low-cost vehicle sensors. The obtained estimates are exploited in the second stage, where a (quasi-static) tire model is fitted to this data. A major issue in this estimation process is the typical instability of these estimators for situations with (prolonged) straight driving. This issue is traced back to a lack of local observability. The use of a variable model covariance is introduced as a practical method to obtain a stable estimator, irrespective of the unobservability. The developed methodology has a low computational load and the Kalman estimator is able to run in real time, whereas the tire model parameter fitting is cheap enough to run online. The proposed methodology is validated experimentally and provides reliable results in variable driving conditions.

Improved human-computer interaction for mechanical systems design through augmented strain/stress visualisation

Strain/stress evaluation is a crucial operation performed during severa l stages in the typical mechanical product development/life cycle. Howeve r it is often difficult to obtain an appropriate estimation of the strain and stress distribution due to the difficult to model operational conditions, unknown inp ut forces and/or parameters. This makes it particularly difficult for a des ign r to evaluate whether the final product under testing meets all the operationa l design specifications. This work presents an extended Kalman filtering ap proach to obtain accurate strain and stress estimates of a structure under operatio nal loading. This information is exploited in an augmented reality application to visualize strains and corresponding stresses on a real component. Th is provides a very efficient human-computer interface to evaluate strains and stres ses on a physical prototype. This approach is therefore very suitable to improv e the design of components due to the good overview of the performance. In order to obtain an efficient formulation, the developed approach is based on the exploitation of reduced order mechanical models based on high fidelity fin eelement design models. By including a parameterization in the reduced mod el, the approach can be made robust with respect to unknown operational pa rameters, like boundary conditions. The obtained paradigm is validated on a flexible beam with unknown input forces and length. The proposed approach permits a m ore natural visualization and interpretation of operational conditions. Our results enco urage the adoption of the proposed approach not only for design validation but also on-line monitoring of structural components, opening new possibility in the fi eld of Augmented Reality for Maintenance. Copyright

A Novel Approach to Real-Time Flexible Multibody Simulation: Sub-System Global Modal Parameterization

This paper introduces a novel model reduction technique, namely Sub-System Global Modal Parameterization (SS-GMP), for real-time simulation of flexible multibody systems. In the past, other system-level model reduction techniques have been proposed for this purpose, but these were limited in applicability due to the large storage requirements for systems with many rigid degrees-of-freedom (DOFs). However, in the SS-GMP approach, the motion of a mechanism is split up into a global motion and a relative motion of the (sub-)system. The relative motion is then reduced according to the Global Modal Parameterization, which is a model reduction procedure suitable for closed chain flexible multibody systems. In combination with suitable explicit solvers, the SS-GMP approach enables (hard) real-time simulations due to the strong reduction in the number of DOFs and the conversion of a system of differential-algebraic equations into a system of ordinary differential equations. The proposed approach is validated numerically with a quarter-car model. This fully flexible mechanism is simulated faster than real-time on a regular PC with the SS-GMP approach while providing accurate results.Copyright

Broadband Load Torque Estimation in Mechatronic Powertrains Using Nonlinear Kalman Filtering

An important bottleneck in the design, operation, and exploitation of mechatronic powertrains is the lack of accurate knowledge of broadband external loading. This is caused by the intrusive nature of regular torque measurements. This paper proposes a novel nonintrusive approach to obtain torsional load information on mechatronic powertrains. Online coupled state/input estimation is performed through an augmented nonlinear Kalman filter. This estimation approach exploits general lumped-parameter physics-based models in order to create a widely applicable framework. This paper considers both extended (EKF) and unscented Kalman filtering approaches. Contrary to previous works, no considerable difference in accuracy is obtained from experiments, with a considerably lower computational load for the EKF. This paper reveals the benefits of including rotational acceleration measurements from a theoretical perspective, which is demonstrated through the experimental validation. This drastically increases the broadband accuracy. The result of this paper is an accurate and noninvasive virtual torque sensor with a sufficiently broad bandwidth for use in condition monitoring, control, and future design optimization.

State Estimation: A Model-Based Approach to Extend Test Data Exploitation

Design models can drastically improve the applicability of testing and allow measuring previously unmeasurable quantities and designing reduced test configurations. A common workflow is followed: a multiphysics system model provides a prediction of the system states which is corrected by the estimation algorithms using the measurement data. The model can then generate data of the non-measurable quantities (e.g. virtual sensors). A wide range of models can be used, including analytical, 1D lumped parameter and 3D distributed parameter models. Key is that they are easy to evaluate and have a small number of states, while capturing the dominant physics. Novel model order reduction techniques enable the use of more complex models. A wide range of state estimation approaches has been developed such as the (linear, extended, unscented, …) Kalman Filter and the Moving Horizon Estimator. All approaches require a trade-off between accuracy and computational load so that conventional estimators must be tailored to deal with high-fidelity nonlinear models of industrial complexity. The approach is illustrated with two cases: the estimation of hard-to-measure vehicle body forces using the extended Kalman filter and the application to an electro-mechanical drivetrain subject to unknown input forces. Methodological aspects are evaluated and different estimators are compared.

Use of Flexible Models in Extended Kalman Filtering Applied to Vehicle Body Force Estimation

Accurate knowledge of wheel loads is of great value in vehicle design and control. However, a direct measurement of these forces is generally not feasible. This motivates the use of model-based estimation techniques, such as the Kalman filter to obtain operational wheel forces. The general approach in literature is to use simple ad-hoc models (like the bicycle model) in the Kalman filter. In many applications however, including vehicle dynamics, this results in a system that is not observable for all the variables of interest, e.g. the individual tyre forces. In this light, this work proposes the use of general flexible multibody models for Kalman filtering. The introduction of flexible deformations in the model enables the observation of variables which cannot be obtained from a rigid model. This allows the filter to differentiate between the contributions of different input forces. This approach is demonstrated by employing an augmented extended Kalman filter to perform a combined estimation of the current vehicle state and wheel forces of a 2D vehicle model. The system is modeled in a floating-frame-of-reference (FFR) approach and the vehicle body is described by a reduced order finite element model. An observability analysis is performed and the observability conditions for the unknown input forces are derived. The proposed approach is validated numerically and compared to an estimator with a rigid assumption.

An extended Kalman filter approach for augmented strain/stress visualization in mechanical systems

This work presents an extended Kalman filtering approach to obtain accurate strain and stress estimates of a structure under operational loading. This information is exploited in an augmented reality application to visualize strains and corresponding stresses on a real component. A parametrized reduced physical model allows an efficient computation of the stresses in the Kalman filter. The model is parametrized in order to give good robustness to uncertain parameters, by estimating the parameters concurrently with the states. In order to allow unknown loading conditions, also the unknown input forces are estimated. This approach offers a very efficient and robust estimation approach. On the other side, using augmented reality as the visualization paradigm, offers two major benefits: visualizing operational strains and stresses field instead of discrete quantities; collocating the results on top of the real component under investigation. The obtained paradigm, validated with a demonstration case through an experimental validation on a beam, permits a more natural visualization and interpretation of operational conditions. Our results encourage the adoption of the proposed approach for on-line monitoring of structural components, opening new possibility in the field of Augmented Reality for Maintenance.