石油钻井过程故障检测的多模核主元分析方法

摘要

A kernel principal component analysis ( KPCA ) method applicable to the drilling process fault detection was put forward. Firstly, process data were classified by using threshold classification algo-rithm, and the data of the steady state condition were obtained. Secondly, according to the classification data the corresponding KPCA model was established, and these corresponding KPCA models were com-bined together to realize fault detection. After multiple tests, the method was proved to be suitable for fault detection of drilling process, the detection sensitivity was improved and the error was reduced.%提出了一种适用于石油钻井过程故障检测的多模核主元分析方法。首先，利用门限值分类算法对过程数据进行分类，可以得到钻井过程各个稳态工况下的数据；其次，取不同工况的数据分别建立相对应的核主元模型，将这些核主元模型组合到一起构成一个核主元模型组进行故障检测。经实验数据分析，该检测方法适用于石油钻井过程，提高了检测灵敏度并减少了误差。