应用模式识别技术实现抽油机井下故障的自动诊断

The Application of Pattern Recognition Technique in Automatic Diagnosis of the Downhole Faults of the Sucker-Rod Pumping Units

本文将模式识别技术应用于抽油井井下故障诊断,以达到对泵示功图定量分析的目的。作者提取和选择出反映示功图形状的四个特征:圆形度、细长度、散射度和凹性度,这四个特征满足特征不变量要求。在几种典型故障示功图样本训练集的基础上,基于距离这一判别函数对井下故障进行分类决策,实例表明,特征向量是稳定和有效的。

In this paper, the pattern recogniation technique is introduced briefly and then is used to diagnose the downhole faults of the sucker rod pumping units.Four specific variables describing the characteristics of the dynagraph shape are proposed and selected, namely, circularity,ellipticity,scattering and concavity,which meet the demands of the feature invariables. On the basis of several kinds of typical fault dynagraphs and according to the descriminating function-the distance,the downhole faults are classified and decisions are made to solve them.Practical examples show that the features of the shape are stable and effective.