Parallel implicit sparse identification of nonlinear dynamics for a 2DOF system

Yan, Minrui; Khanesar, Mojtaba A.; Armstrong, Harry; Milella, Ferdinando; Herschmann, Sam; Piano, Samanta; Branson, David

Abstract

Data-driven methods have emerged as a pivotal area of system identification and among these the SINDy-PI (parallel implicit sparse identification of nonlinear dynamics) algorithm stands out for its ability to identify underlying parameters in equations for dynamic systems from measurement data. This method benefits from the capacity to deal with sparse nonlinear dynamical systems and robustness to noisy data. In this work, a dynamic model for a 2 Degree of Freedom, 5 bar linkage, system with two actuated inputs based on equations is developed in Matlab; by using SINDy-PI, unknown parameters of this model are estimated from simulated data with low RMSE(root mean square error) of 0.0501 to simulated results; The case demonstrates the feasibility of applying SINDy-PI on a 2DoF mechanical system and results indicate substantial improvements in noice robustness.

Citation

Yan, M., Khanesar, M. A., Armstrong, H., Milella, F., Herschmann, S., Piano, S., & Branson, D. (2025, June). Parallel implicit sparse identification of nonlinear dynamics for a 2DOF system. Paper presented at Euspen’s 25th International Conference & Exhibition, Zaragoza, Spain

• Presentation Conference Type: Conference Paper (unpublished)
• Conference Name: Euspen’s 25th International Conference & Exhibition
• Start Date: Jun 9, 2025
• End Date: Aug 13, 2025
• Acceptance Date: Mar 26, 2025
• Online Publication Date: Jun 9, 2025
• Publication Date: Jun 9, 2025
• Deposit Date: Aug 15, 2025
• Peer Reviewed: Peer Reviewed
• Keywords: Nonlinear system identification, SINDy-PI, robotics, simulation
• Public URL: https://nottingham-repository.worktribe.com/output/52838903
• External URL: https://www.euspen.eu/events/25th-international-conference-exhibition-9th-13th-june-2025/?subid=25th-international-conference-exhibition-9th-13th-june-2025