Rochester Institute of Technology
Data Driven Approach to Non-stationary EMA Fault Detection and Investigation Into Remaining Useful Life
Abstract
dc:description.abstractGrowing interest in using Electromechanical Actuators (EMAs) to replace current hydraulic actuation methods on aircraft control surfaces has driven significant research in the area of prognostics and health management. Non- stationary speeds and loads in the course of controlling an aircraft surface make fault identification in EMAs difficult. This work presents a time- frequency analysis of EMA thrust bearing vibration signals using wavelet transforms. A relatively small EMA system is designed and built to allow for simple, quick, and repeatable component replacement. A simulated signal is developed to test four potential faults in the system. Classification is performed using an artificial neural network (ANN), which yields over 99% accuracy. Indentation faults from moderate and heavy loads are seeded in thrust bearings, which are then tested to generate data. The ANN achieves 95% classification accuracy in a two class scenario using healthy and moderately indented bearings. A three class test is executed using thrust bearings at each level of damage to perform preliminary remaining useful life (RUL) testing, where an ANN is able to identify the fault severity with an accuracy of 88%.
Degree
thesis:*- Name thesis:degree_name
- Mechanical Engineering (MS)
- Level thesis:degree_level
- Thesis
- Discipline thesis:degree_discipline
- Mechanical Engineering (KGCOE)
Author and committee
dc:creator, dc:contributor.*- Author dc:creator
-
- Craig, William S
- Contributors dc:contributor
-
- Jason Kolodziej
Subjects
dc:subject × 3Identifiers
dc:identifier.*- Repository record dc:identifier
- https://repository.rit.edu/theses/8586