Consistence-Enhanced Dual-Weighted Broad Learning System for Effective Cementing Pump Fault Diagnosis

Abstract Cementing pumps are the key equipment in oilfield well cementing operations. How to diagnosis their running status is an important and challenging task. To handle this issue, this paper presents a consistence-enhanced dual-weighted broad learning system (CE-DWBLS) for effective cementing pump fault diagnosis. Firstly, the vibration signals of cementing pump are analyzed to obtain the features of time domain and frequency domain. For making full use of the time series property of signals, the phase space reconstruction technology is introduced to generate the augmented features. Secondly, from two distinct perspectives of variables and samples, a consistence-enhanced dual weighting strategy is designed to improve the basic broad learning system (BLS). The variable-dimension weights are constructed by the partial F test method to highlight the features contributing largely to the classification performance, while the sample-dimension weights are developed by the sample divergence analysis for emphasizing the key samples. For further suppressing the noise influence and enhance the model robustness, a consistence constraint is introduced to BLS optimization objective. Thirdly, in order to provide a reasonable explanation about the decision result, an error gradient-based input attribution method is proposed to indicate which features are closely related to the classification results. Lastly, one real cementing pump condition monitoring experimental platform is constructed and application on the experimental dataset is used to validate the method performance. The application results show that the proposed method has a better fault classification accuracy than the basic BLS method, and provide a reasonable explanation for the decision procedures.