基于PNN网络和Freeman链码的抽油机井工况诊断

Working Condition Diagnosis of Pumping Unit Based on PNN and Freeman Chain Code

利用概率神经网络(PNN)对抽油机井工况进行诊断,建立了抽油机井工况诊断的概率神经网络模型;对示功图提取特征值的质量好坏直接影响识别效率和可靠性,提出了用Freeman链码对等效的电流示功图提取特征参数,进行预处理,建立抽油机典型工况的链码特征样本库;将Freeman链码作为特征向量,利用MATLAB对网络进行训练;结果表明,Freeman链码能够有效地识别各种典型工况示功图,并且该概率神经网络学习速度快、诊断准确率高,可用于抽油机井工况的实时监测和诊断。

A probabilistic neural network was used to diagnose the working conditions of pumping unit. This paper attempts to feature extraction parameters through the equivalent current indictor diagram by the Freeman chain code, establish typical operating conditions pumping characteristic parameters of sample chain code libraries, to solve the effects of the accuracy of indicator diagram＇s eigenvalue. Use Freeman chain code as a feature vector, the model of probabilistic neural network was set up and trained by Matlab. The results show that Freeman chain--code can effectively identify a variety of typical working condition indicator diagrams, and this network features fast learning, high diagnostic accuracy for real time detection and diagnostics of pumping unit operating conditions.