Local parameter identification with neural ordinary differential equations

Abstract

The data-driven methods extract the feature information from data to build system models, which enable estimation and identification of the systems and can be utilized for prognosis and health management (PHM). However, most data-driven models are still black-box models that cannot be interpreted. In this study, we use the neural ordinary differential equations (ODEs), especially the inherent computational relationships of a system added to the loss function calculation, to approximate the governing equations. In addition, a new strategy for identifying the local parameters of the system is investigated, which can be utilized for system parameter identification and damage detection. The numerical and experimental examples presented in the paper demonstrate that the strategy has high accuracy and good local parameter identification. Moreover, the proposed method has the advantage of being interpretable. It can directly approximate the underlying governing dynamics and be a worthwhile strategy for system identification and PHM.

A mechanical-free designing method for tailoring nonlinearity in bi-stable piezoelectric energy harvesters

“…These experiments include: (1) measuring the free-vibration displacement of the linear system for determining the damping ratio, open-circuit and shortcircuit natural frequency, (2) measuring the displacement and open-circuit voltage of the linear system under weak sinusoidal excitation to determine the force factor, (3) measuring the output voltage frequency response at given acceleration for parameter optimization procedure, and (4) measuring the magnetic induction field of the magnets to determine the magnetization and local demagnetizing coefficients. For further parameter identification methods, see for examples [62,63].…”