A Fault Diagnosis Approach for Triplex Pump Based on Digital Twin and Ensemble Learning

Plunger pumps are the most commonly used type of pumps in the power plant. Condition monitoring of such machine can significantly reduce repair costs and unexpected break downs. In this paper, we present a fault diagnosis method for triplex pump based on digital twin (DT) modeling and Adaboost algorithm. First, the digital twin model of a triplex pump is constructed from three dimensions of the pump's appearance, behavior, and process using the Simscape tool of MATLAB software. Second, different types and severity of faults are simulated by modifying the settings of the DT model, and the pressure signals of the output fluid is collected as a basis for classification. Third, the features with high differentiability are extracted from the acquired pressure signals by Fourier transform. Fourth, the fault classifier model based on Adaboost is proposed. Finally, the performance of our proposed algorithm is compared with the traditional decision tree algorithm is verified through a case study of a triplex pump for which the accuracy of the proposed algorithm reaches 86.7% and the AUC is 95%.