Masters Theses

Date of Award

12-1999

Degree Type

Thesis

Degree Name

Master of Science

Major

Mechanical Engineering

Major Professor

Masood Parang

Committee Members

Stan Johnson, Wes Hines

Abstract

Mechanical seals on centrifugal pumps provide the junction between the rotating shaft and stationary housing on a pump assembly which prevents unnecessary fluid leakage. When these seals fail, the pumps no longer run at peak performance, and sometimes will not run at all. Since mechanical seals fail approximately 85% of the time as opposed to wear out, the early detection of their failure could assist in preventing system down time and possibly preventing the seal failure. The seals in question were instrumented to determine various operating conditions during different phases of the seal's performance spectrum. By training artificial intelligence computer codes to classify the relative performance of specific seals based on the information obtained by instrumentation, a prediction technique was modeled. These artificial intelligence codes were programmed using the general judgment criteria found from a literature survey as well as discussions with plant engineers directly associated with the failing seals. These codes are dependent on a variety of pump performance characteristics. These models for predictive analysis along with defined goals for increased data collection practices not only prove the feasibility of using artificial intelligence codes to help predict mechanical seal failure, but also lay the foundation for an expanded and robust failure analysis study.

Download
Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS