Computer Engineering and Applications ›› 2013, Vol. 49 ›› Issue (2): 9-12.

Previous Articles     Next Articles

Application of counterpropagation proeess neural network in well fault diagnosis

ZHANG Qiang, XU Shaohua, LI Panchi   

  1. School of Computer and Information Technology, Northeast Petroleum University, Daqing, Heilongjiang 163318, China
  • Online:2013-01-15 Published:2013-01-16


张  强,许少华,李盼池   

  1. 东北石油大学 计算机与信息技术学院,黑龙江 大庆 163318

Abstract: To better solve the pattern diagnosis problem of the indicator diagram for pumping wells, this paper identifies the indicator diagram as pattern recognition problem of dynamic system continuous curve(displacement-time curve and load-time curve) according to drawing principles of indicator diagram. Process neural can deal with two-dimensional information of time and space simultaneously and can extract the process pattern characteristics of the set for time-varying functions automatically. The method on mechanism has good adaptability to solve classification problems of time-varying signals. Based on above, a counterpropagation process neural network diagnostic model and learning algorithm are proposed. The model can be trained and be used to identify fault using well measured data, and achieves good application results.

Key words: counterpropagation process neural network, learning algorithm, indicator diagram, fault diagnosis

摘要: 为更好解决抽油机井示功图模式诊断问题,依据示功图绘制原理,将示功图识别看作动态系统连续曲线(位移-时间曲线和载荷-时间曲线)的模式识别问题。利用过程神经元能同时处理时、空二维信息,可自动抽取时变函数样本的过程模式特征,在机制上对时变信号的分类问题具有较好的适应性,提出一种基于对传过程神经元网络诊断模型及其学习算法。以油井实测数据对模型进行训练和故障识别,取得了较好的应用效果。

关键词: 对传过程神经网络, 学习算法, 示功图, 故障诊断