Date of Award

2024

Document Type

Campus Access Master's Thesis

Degree Name

Master of Science in Mechanical Engineering (MS)

Administrative Home Department

Department of Mechanical Engineering-Engineering Mechanics

Advisor 1

Jim DeClerck

Advisor 2

Jason Blough

Committee Member 1

Sriram Malladi

Abstract

Analyzing harmonic content in a signal can be challenging when there are multiple noise contributions. The periodic spacing of the harmonics can become hidden among content caused by other noise in the system, disabling the possibility of root causing the source. There are many systems that introduce harmonic series to a system, but one example to analyze is a ball screw assembly.

Currently, high load electric power steering systems utilize ball screw assemblies to translate rotational movement to linear movement. Noise is produced by these systems due to inconsistencies on the ball screw assembly components’ surfaces, presenting a problem for the vehicle driver. Due to the nature of the assembly, this noise manifests itself as harmonic content, with periodic peaks in the frequency spectrum. A consistent method to analyze this content and define defect type is necessary to discriminate between acceptable and unacceptable systems for consumer use.

Cepstral analysis, a type of homomorphic filtering, is one method that can be utilized to perform this task. By taking the logarithm of power spectrum data and then taking the inverse fast Fourier transform (IFFT) of those results, the data is transformed to the cepstral domain. Within this domain, if any harmonic signals are found in the original frequency data, they will be identified and displayed as the reciprocal of the spacing between peaks of the signal. While frequency data is useful to evaluate resonant frequencies and modal response of a rotating system, it can mask harmonic content when energy is smeared across the spectrum due to inconsistent speed. Smearing can cause the harmonics to become spread as a broadband signal across the spectrum, with no clear indication of periodic peaks.

In a rotating system, the data can be analyzed not just in the frequency domain, but in the order domain. The order domain processes the data in terms of position rather than time. Cepstral analysis can then be performed on order domain data rather than frequency domain data. Just as frequency data can be processed in the quefrency domain, order data can be processed in the roder domain. Cepstral analysis of order domain data is a technique that can be utilized to identify harmonic content in rotating equipment signals, independent of speed.

This technique can be utilized as a diagnostic tool to determine the cause of noise, monitor wear of assemblies, and identify acceptable and unacceptable assemblies. For these reasons, cepstral analysis of order domain data can be used as both a development and production tool to capture and evaluate harmonic signals.

Available for download on Saturday, April 26, 2025

Share

COinS