Fault Diagnosis Method Based on Dynamic Axis Nucleation KPLS for Pumping Unit

Fault diagnosis of pumping unit system has long been a challenging issue owing to the system that exhibits nonlinearity, coupled parameters and time-varying in the production process. In this paper, a novel fault diagnosis method based on dynamic axis nucleation kernel partial least squares (DANKPLS) is proposed for pumping unit. First, the multiple dichotomous regression (MDR) model is established to reflect the hidden dynamic relationship between variables efficiently. Then, a novel axis nucleation kernel partial least squares method is proposed to map the principal axis into a high-dimensional space. In particular, the correlation of the data can be further and clearly represented. Finally, the proposed method is applied to the pumping unit system. Experimental results show the effectiveness and favorable diagnosis rate in false alarm and missing alarm.