When performing statistical analyses on PCP run-life data, one can easily reach incorrect conclusions if subtle but important issues are not fully understood. This paper illustrates some of these issues using examples from analysis work conducted recently with a relatively large and complete set of data collected from several large operators. Issues examined include misinterpretation of single-variable correlation results and run-life measures. It is impractical to design and conduct lab or field experiments to investigate the effect of individual variables on PCP run-life. On the other hand, with normal field data, it can be difficult to isolate the effects of individual variables. The assumption of "all else being equal?? rarely applies to field data. Additionally, different run-life measures can give different results at different times in the life of a field, even when the underlying system reliability has not changed. Awareness of these issues is essential to all companies that collect and analyze PCP failure and run-life information for the purpose of evaluating and improving PCP system performance.
PCP failure tracking systems are an important part of identifying and correcting issues that negatively affect PCP system run-life and increase operating costs [Robles 2001; Kollatschny and Rodriguez 2004; Barrios and Fernández 2004]. They allow operators and vendors to record not only failure information, such as the failed component and the failure cause, but also equipment characteristics and well operating conditions. This data can then be analyzed to identify possible actions that may result in increased run-life.
This type of analysis, when conducted properly, can provide insight into problems arising from improper system design, equipment manufacturing, equipment installation, or system operation. Furthermore, lessons learned through analysis of data from one field can transfer over to new fields with similar operating conditions. For example, mitigating actions taken to reduce the effect of sand on PCP run-life, such as selection of sand screens or specialized elastomers, can be applied to new fields with similar issues.
However, it is very important to understand the limitations of the analysis techniques and metrics used [Alhanati 2008]. In this paper, dangers inherent in four commonly used analysis techniques and system run-life measures are highlighted with examples from analysis work conducted recently with a relatively large and complete set of data collected from several large operators.
The first danger occurs when correlations between important parameters and run-life measures are misinterpreted as causal relationships. Often, in this type of data, a correlation between two parameters is a result of other underlying factors and not necessarily due to the influence of one parameter on the other. A second danger occurs when the reliability of PCP system components is evaluated using their average runtimes or their failure frequencies separately. These run-life measures must be evaluated together to provide a good understanding of the PCP system's reliability. A third danger occurs when evaluating the run-life performance in a field using a single run-life measure. Several measures must be used to properly evaluate the effects of recent changes in operating practices or selected equipment on the overall run-life of systems. The fourth danger discussed here occurs when assuming that failure mechanism statistics provides sufficient information for identifying the main system reliability issues present. Understanding failure causes (or root causes of failures) should be the goal of any failure analysis investigation, as only looking at them can lead to improvement in system design, equipment manufacturing or system operation.
Author: Sheldon, J., Alhanati, F. J. S., & Skoczylas, P.
Publisher: SPE Progressing Cavity Pumps Conference, 12-14 September, Alberta, Canada
Purchase URL: Purchase this paper