Masters Theses

Abstract

"Reciprocating compressors are commonly used machinery for industrial applications. Unscheduled downtime and maintenance activity on the compressors causes considerable loss in throughput and efficiency of a plant. Of all the failures that cause unscheduled downtime in reciprocating compressors, valve related causes are predominant. Most of the failures associated with the valves are tracked to the failure of moving elements within the valve. Achieving higher reliability of critical reciprocating systems requires continuously monitoring the system and performing dynamic analysis of the sensory data for valve fault diagnosis. Continuous monitoring will improve the time and cost to repair through keeping a constant vigil for failure events. Though there has been a good amount of work done for condition monitoring of compressors, there has been very little work on detecting and predicting valve failures. The objective of this thesis is to research detection and prediction of valve failures by wavelet analysis, logistic regression and neural network analysis of pressure and temperature signals, which are the most common measurements on a reciprocating compressor system. Valve failures are seeded on a reciprocating compressor testbed that is instrumented with only temperature and pressure sensor order emulate the reciprocating compressor systems used in the industry. The parameters are measured on a continuous basis and baselines are established for normal (or acceptable) behavior and failure (or fault) condition. Deviation of the system from the normal condition and the time for the system to reach the fault mode is quantified with the help of the above mentioned tools."--Abstract, page iii.

Advisor(s)

Leu, M. C. (Ming-Chuan)

Committee Member(s)

Chandrashekhara, K.
Sarangapani, Jagannathan, 1965-

Department(s)

Mechanical and Aerospace Engineering

Degree Name

M.S. in Mechanical Engineering

Sponsor(s)

Boeing Company
Chevron Corporation

Publisher

University of Missouri--Rolla

Publication Date

Fall 2007

Pagination

viii, 56 pages

Note about bibliography

Includes bibliographical references (pages 83-85).

Rights

© 2007 Raghuram Puthali Ramesh, All rights reserved.

Document Type

Thesis - Open Access

File Type

text

Language

English

Library of Congress Subject Headings

Failure Analysis System (Computer system)
Logistic regression analysis
Neural networks (Computer science)
Valves -- Performance
Valves -- Reliability

Thesis Number

T 9306

Print OCLC #

793771075

Electronic OCLC #

793756908

Share

 
COinS