Fault identification in hydrodynamic bearings using machine learning

• Grant number: 22/12565-1
• Support Opportunities: Scholarships in Brazil - Doctorate (Direct)
• Start date: March 01, 2023
• End date: February 28, 2027
• Field of knowledge: Engineering - Mechanical Engineering - Mechanics of Solids
• Principal Investigator: Gregory Bregion Daniel
• Grantee: Matheus Victor Inacio
• Host Institution: Faculdade de Engenharia Mecânica (FEM). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil

Abstract

Hydrodynamic bearings occupy a prominent place in industry due to their significant influence on the dynamic behavior of the rotating system. Among the types of hydrodynamic bearings, the most widespread is the radial cylindrical bearing, which has low manufacturing cost, high load capacity, low frictional energy losses, and long operating life. Despite the advantages already highlighted, this component can present flaws that damage its performance, such as wear and ovalization. In general, these failures alter the circular profile of the bearing, impacting the vibrational response of the rotor and impairing the overall behavior of the rotating system. Among the undesired changes, a local reduction in oil film thickness may occur, which in turn reduces load capacity and favors metal-to-metal contact between shaft and bearing.Within this context, this Ph.D. research project aims to develop a new approach to identifying hydrodynamic bearings operating with non-circular profiles. The core of the project consists in creating a database with a large amount of information in different formats of hydrodynamic bearing profiles, which are associated with the parameters of their respective dynamic responses. From this, a model can be trained using machine learning techniques to identify the shape of the hydrodynamic bearing profile through specific parameters of its dynamic response.In this proposal, the hydrodynamic bearing profile will be represented by a vector that contains the radius of the discretized profile as a function of the bearing's angular position, and the interpolation of this vector will result in the continuous hydrodynamic bearing profile. The main advantage of this approach is in directly evaluating the profile, thus avoiding the use of previously existing fault models to represent non-circular profiles because these models are built from simplifications, limiting non-circular hydrodynamic bearing profiles to specific shapes. Finally, it is important to highlight that the proposed approach is innovative, since it allows the physical integrity (profile) of hydrodynamic bearings to be assessed, in a generalist way, from the dynamic response of the rotor, corroborating the development and innovation in the area of intelligent monitoring of machines and equipment. (AU)

Scientific publications

(References retrieved automatically from Web of Science and SciELO through information on FAPESP grants and their corresponding numbers as mentioned in the publications by the authors)

INACIO, MATHEUS VICTOR; CAVALCA, KATIA LUCCHESI; DANIEL, GREGORY BREGION. Identification of non-circular profiles in hydrodynamic journal bearings. MECHANISM AND MACHINE THEORY, v. 203, p. 23-pg., 2024-10-19. (22/12565-1)