Valve Leakage Detection System Analysis and Development for Quituplex Pumps

Abstract

This thesis undertakes the study of digital signal processing in the purpose of an optimized and dynamic approach to the valve leakage detection problem. Several field tests were analyzed to evaluate the system properties in search for valuable assets for the development process. Numerous analyses were conducted both in the time-domain and the frequency-domain. The importance of the findings varied, but led to a better understanding of the system characteristics and breakthroughs in regards to possible approaches to the problem. The shaft execution sequence was identified with the commissioning video. This sequence had a vital role together with knowledge about phase behavior when determining the pump state, which is crucial when identifying the valve that causes the leak. The program language platform was predefined by National Oilwell Varco as LabVIEW, with belonging National Instruments devices for the developed system. The leakage detection system developed in this thesis consists of two compact reconfigurable I/O modules (cRIO) and a desktop computer. A simulator was created, because the Quintuplex Pumps are located in Huston, Texas. In order to perform a pilot test that was close to real, the simulator reproduces measurements from the previous gathered field tests. One cRIO has the role of a simulator and the other performs system and leak evaluation from the measurements provided, before sending the results with serial communication. The desktop leakage program reads data from the serial port, displays the values as number indicators and trends over time and store the results as text logs. The result is a stable and dynamic leakage detection system that uses envelope detection and root mean square, in combination with the phase offset table, which with minor modification can be applied to a pump system with a different set of valves.