Unbalance identification method based on SINDy applied to an SFD rotordynamic system

Abstract

In recent years, there has been an increasing interest in Data Science and Machine Learning in different topics like financial and health, this have led to start using these methods on engineer applications. This paper is focus on identify the equivalent unbalance on Squeeze Film Damper – SFD bearing using a recent machine learning technique “Sparse Identification of Nonlinear Dynamics – SINDy”. Four different cases will be examined from Bonello’s work, all of which we introduce 4 different conditions of noise to the acceleration of the system. The data for this work was obtained via a simulation of the SFD system reported on Bonello’s thesis. From the simulation only the last 20 cycles were used to feed the SINDy. This study uses a combinatorial polynomial search space over preselected functions with the purpose to identify the equivalent imbalances. Both hyperparameters: the degree of the combinatory k and the threshold value λ remaining static during all the study. There was no error between the original equations and the identified system.