Degree Type

Dissertation

Date of Award

1999

Degree Name

Doctor of Philosophy

Department

Mechanical Engineering

First Advisor

Ron M. Nelson

Abstract

The objective of this research is to develop a model using multivariate statistical methods to identify system faults in an air-handling unit (AHU). The effects of using a reduced amount of available information are investigated. The faults are simulated and are applied in an actual flow loop facility;The process model fault scheme as presented in Iserman (1987) was adopted for this research. Unlike the parameter estimation from Iserman study, the procedures for detecting and diagnosing AHU faults in this study use multivariate Statistics and Probability; To aid the research, a factorial experimental design was developed and experimental normal and fault data were collected for the AHU operation. The investigation included three types of faults: fan belt slippage, coil blockage, and sticking control valve. The data included temperatures, flow rates, pressure rise of the fan, fan speed, and power measurement. To extract the dimensionality of the problem, a principal component analysis (PCA) was performed and cross-examined by the method of discriminant analysis. An additional energy analysis was performed to validate the model and provide the grounds for the explanation of the physical fault detection with the minimum set of variables for the AHU;The logistic regression analysis was applied to distinguish between normal and faulty operation. In determining the minimal set of variables, stepwise discriminant analysis provided four variables: inlet air temperature, outlet water temperature, inlet water temperature, and fan pressure rise. Standardized discriminant coefficients allowed freedom from scale and correctly reflected the joint contribution of the variables to the discriminant function as it maximally separated the groups;Discriminant and classification analysis showed valuable information for detecting and diagnosing the three faults in an AHU. The experimental data gathered using a factorial experiment design gave good evaluation of the significant variables involved to segregate different faults. The results of this research demonstrated an effective fault detection and diagnostic mechanism for an air-handling unit, leading to improved system performance and decreased energy use and demand.

DOI

https://doi.org/10.31274/rtd-180813-13839

Publisher

Digital Repository @ Iowa State University, http://lib.dr.iastate.edu/

Copyright Owner

Kyung-Jin Jang

Language

en

Proquest ID

AAI9924726

File Format

application/pdf

File Size

189 pages

Share

COinS