Robust Control and Fidelity Assessment of Real-Time Hybrid Substructuring of Contact Problems
Final Report Abstract
Real-Time Hybrid Substructuring (RTHS) is a method for the efficient and precise investigation of complex dynamic systems. A part of the system is investigated experimentally by coupling it with a co-simulation of the rest of the system in real time using actuators and sensors. Transferring this test principle to the investigation of foot prostheses enables a precise investigation of the interaction between humans and prostheses, which is essential for the development of new and optimized prostheses. Since RTHS tests for foot prostheses pose a particular challenge due to the alternating ground contact, this project developed methods for the accurate and stable performance of RTHS tests with alternating ground contact and for interpreting the test accuracy. A combination of Iterative Learning Control (ILC) and Normalized Passivity Control (NPC) proved to be particularly promising. While the ILC iteratively improves the actuator’s tracking performance and thus the test accuracy, the NPC ensures test stability even for lightly damped systems. This control strategy was successfully investigated and validated both simulatively and experimentally using a one-dimensional RTHS experiment with alternating contact. Other methods such as adaptive feedforward filters and hybrid position and force control were also investigated. As no reference solutions are usually available for RTHS tests, interpreting the accuracy of the test results is a particular challenge. As part of this project, a method was developed that enables the test accuracy to be predicted on the basis of measurable error variables, such as the control error of the actuator. The error values are actively varied in order to be able to extrapolate the sensitivity of the variables to be measured with the error values in such a way that the measured variable can be predicted in the absence of an error, which is not possible in the experiment. On this basis, the test accuracy with a minimum possible error can be set in relation to the ideal test result in order to interpret the accuracy of the test. This methodology was also tested on different one-dimensional RTHS tests with promising results. Implementation details of the developed methods were made publicly available together with tutorials on GitLab. In the last phase of the project, a first RTHS test with a real foot prosthesis was carried out. The numerical system used was an abstract gait model based on the Virtual-Pivot-Point. The RTHS test was able to reproduce the gait characteristics qualitatively, but not yet quantitatively due to the limited motion possibilities of the actuator.
Publications
-
Fidelity Assessment of Real-Time Hybrid Substructure Testing: a Review and the Application of Artificial Neural Networks. Experimental Techniques, 46(1), 137-152.
Insam, C. & Rixen, D. J.
-
Hardware-in-the-Loop Test of a Prosthetic Foot. Applied Sciences, 11(20), 9492.
Insam, Christina; Ballat, Lisa-Marie; Lorenz, Felix & Rixen, Daniel Jean
-
Normalized passivity control for hardware-in-the-loop with contact. International Journal of Dynamics and Control, 9(4), 1471-1477.
Insam, Christina; Peiris, L. D. Hashan & Rixen, Daniel J.
-
Robust and high fidelity real-time hybrid substructuring. Mechanical Systems and Signal Processing, 157, 107720.
Insam, Christina; Kist, Arian; Schwalm, Henri & Rixen, Daniel J.
-
Fidelity assessment of Real-Time Hybrid Substructuring based on convergence and extrapolation. Mechanical Systems and Signal Processing, 175, 109135.
Insam, Christina & Rixen, Daniel J.
