< Back to previous page


MultiBody Model-based Virtual Sensing. A Generic Indirect Sensing Approach for Mechatronic Systems with Nonlinear Bushing Connections

We are in the era of digitisation where accurate and reliable data are repeatedly exchanged fueling innovation in emerging high-growth technologies. 

In mechanical engineering applications, digitization innovates among others the on-board sensing and model-based design.
Despite the advancements in sensing, some quantities are troublesome to obtain in a cost-effective, non-intrusive manner. For instance, forces and moments are difficult or even impossible to be directly measured without affecting the system layout and behaviour. 
Model-based solutions are continuously growing and increasingly being considered in the conception, design, analysis, and operational phases of the system. Such approaches bring clear benefits in mechatronic solutions enabling improved prototyping efficiency with rapid innovation, maintenance actions, and reduced investment risks. Importantly, numerical models rely on accurate knowledge of system inputs and properties. This information might be in practice undefined or bounded to specific operating conditions limiting the general applicability of purely model-based solutions for real-life problems.

The combination of both shapes the digital twin paradigm: physical products come along with their digital companion in which experimental and simulated data are exchanged to one another. In such a way, measurements can be leveraged to update the model states and parameters to accurately represent physical reality. The empirically-informed model can be exploited to evaluate physical information in an inexpensive and non-invasive fashion. An emerging application of the digital twin concept is virtual sensing technology, where ready-to-use measurements and models are combined by means of estimation algorithms to derive difficult-to-measure quantities indirectly. 

In mechatronic applications, a multitude of modelling strategies exist. However, not all models reveal the same level of information with an equal degree of confidence. Multibody models are regularly employed in the engineering community to accurately represent complex mechanisms with several rigid or flexible bodies which are linked through kinematic joints and force elements. A practical example of a connecting element is the rubber bushing component which is widely used in mechanical systems to reduce the transmitted vibration and pretension of interconnected bodies.

While bushings provide necessary functionality, the component behaves in a nonlinear fashion which makes modelling and inclusion in the multibody environment challenging. Moreover, the multibody dynamics and kinematics are generally described by differential algebraic equations in redundant coordinates which cannot be straightforwardly employed in common estimation schemes.

In this work, a versatile component-level modelling strategy is developed to combine state-of-the-art and new lumped-parameter bushing models. Given the large parameter space of such components, an off-line optimisation problem is solved to identify the model parameters from experimental data.
Subsequently, the identified bushing is included in a system-level multi-body model, which is transformed in a state-space form through an efficient linearisation approach. The state-space allows the model to be embedded in a Kalman filter to jointly estimate system states and unknown inputs. The latter is achieved via accurate sensor models and optimal sensor selection. Finally, the indirect virtual sensing and bushing modelling approaches are applied and experimentally validated.

The outcome of the thesis is a generic indirect sensing framework combining physics-inspired models, including nonlinear connections, in tandem with easy-to-use low-cost sensors. 

Date:21 Oct 2016 →  31 Jan 2022
Keywords:Virtual sensing, Multibody modeling, Kalman filtering, Rubber bushing, Inverse load estimation
Disciplines:Control systems, robotics and automation, Design theories and methods, Mechatronics and robotics, Computer theory
Project type:PhD project