Abstract

Unexpected main bearing failure on a wind turbine causes unscheduled maintenance and increased operation costs (mainly due to crane, parts, labor, and production loss). Unfortunately, historical data indicates that failure can happen far earlier than the component designed lives (due to manufacturing problems, for example). For the legacy fleet, which composes the majority of the installed basis, fatigue has become a major issue. Although bearing fatigue can be expressed with physics-informed models, they often inherent large uncertainties due to operation and unknown lubricant degradation mechanism. Apart from the unknown physics of failure, additional uncertainties associated with the grease that surrounds the bearing can be listed as the lack of fidelity in the observations due to visual inspections, and quality variation from one batch to the other. As opposed to detailed laboratory analysis, grease visual inspection can lead to large uncertainties in characterization of grease condition (although visual inspection can be cost and time effective). Eventually, a main bearing fatigue model that can quantify the model-form uncertainty (unknown grease degradation mechanism), observation uncertainty (visual inspections), and input uncertainty (grease quality variation), becomes a necessity for managing and optimizing maintenance of aging wind turbines. In this research, we investigate the effect of lubricant state on main bearing fatigue. After we demonstrate the importance of modeling grease, we propose a novel modeling approach that is hybrid and designed to merge physics-informed and data-driven layers within deep neural networks. The result is a cumulative damage model where the physics-informed layers are used to model the relatively well-understood physics (bearing fatigue damage accumulation) and the data-driven layers account for the hard to model components (i.e., grease degradation). In addition, we introduce a trainable classifier tailored for our application, to map continuous grease damage into discrete visual inspection rankings. Finally, we improve our model to estimate the variation due to lubricant quality, and provide probabilistic life estimations for the component.

Graduation Date

2021

Semester

Spring

Advisor

Viana, Felipe

Degree

Doctor of Philosophy (Ph.D.)

College

College of Engineering and Computer Science

Department

Mechanical and Aerospace Engineering

Degree Program

Mechanical Engineering

Format

application/pdf

Identifier

CFE0008561; DP0024237

URL

https://purls.library.ucf.edu/go/DP0024237

Language

English

Release Date

5-15-2021

Length of Campus-only Access

None

Access Status

Doctoral Dissertation (Open Access)

Download

Share

COinS