Application of an artificial neural network to pump card diagnosis

Journal Article · Thu Dec 01 00:00:00 EST 1994 · SPE Computer Applications (Society of Petroleum Engineers); (United States)

OSTI ID:6917567

Ashenayi, K ^[1]; Lea, J F ^[2]; Kemp, F ^[3]; Nazi, G A

---

1. Univ. of Tulsa, OK (United States)
2. Amoco Production Research, Tulsa, OK (United States)
3. Kemp (Franklin), Dallas, TX (United States)

Beam pumping is the most frequently used artificial-lift technique for oil production. Downhole pump cards are used to evaluate performance of the pumping unit. Pump cards can be generated from surface dynamometer cards using a 1D wave equation with viscous damping, as suggested by Gibbs and Neely. Pump cards contain significant information describing the behavior of the pump. However, interpretation of these cards is tedious and time-consuming; hence, an automated system capable of interpreting these cards could speed interpretation and warn of pump failures. This work presents the results of a DOS-based computer program capable of correctly classifying pump cards. The program uses a hybrid artificial neural network (ANN) to identify significant features of the pump card. The hybrid ANN uses classical and sinusoidal perceptrons. The network is trained using an error-back-propagation technique. The program correctly identified pump problems for more than 180 different training and test pump cards. The ANN takes a total of 80 data points as input. Sixty data points are collected from the pump card perimeter, and the remaining 20 data points represent the slope at selected points on the pump card perimeter. Pump problem conditions are grouped into 11 distinct classes. The network is capable of identifying one or more of these problem conditions for each pump card. Eight examples are presented and discussed.

• OSTI ID: 6917567
• Journal Information: SPE Computer Applications (Society of Petroleum Engineers); (United States), Journal Name: SPE Computer Applications (Society of Petroleum Engineers); (United States); ISSN 1064-9778
• Country of Publication: United States
• Language: English

Related Subjects

02 PETROLEUM
03 NATURAL GAS
99 GENERAL AND MISCELLANEOUS//MATHEMATICS, COMPUTING, AND INFORMATION SCIENCE
ARTIFICIAL LIFTS
PERFORMANCE
NATURAL GAS INDUSTRY
NEURAL NETWORKS
NATURAL GAS WELLS
OIL WELLS
PETROLEUM INDUSTRY
COMPUTER CALCULATIONS
COMPUTER CODES
INDUSTRY
WELLS
020300* - Petroleum- Drilling & Production
030300 - Natural Gas- Drilling
Production
& Processing
990200 - Mathematics & Computers