一种基于半监督流形学习的抽油机故障诊断方法

Pumping Unit Fault Diagnosis Method Based on Semi-supervised Manifold Learning

针对油田实际生产中示功图已知识别样本匮乏的情况,提出一种基于半监督流形学习的识别方法,利用标识和未标识的样本数据通过流形学习算法对采集的样本数据进行特征提取获得低维特征空间,并在低维特征空间进行故障模式分类以达到故障诊断的目的,取得了很好的实际应用效果。

According to indicator diagram of known samples in actual production of oil is deficient,A semi supervised recogni?tion method based on manifold learning is proposed.Low-dimensional feature space can be obtained using manifold learning algo?rithm by labeled and unlabeled sample data to the data collected for feature extraction.It can achieve the purpose of fault diagnosisin the low-dimensional feature space for fault pattern classification,and achieve good effect in practical application.