September 10, 2025Open Access
View Full PaperFault prediction of pumping well based on CNN-SVM
Key Points
- The hybrid cnn-svm model significantly improves fault prediction accuracy in oil pumping wells.
- Performance metrics showed the model achieved higher accuracy between 85% and 89% and improved recall values from 74% to 83%.
- A total of 130 iterations were completed with the model reaching a loss of 0.4, demonstrating efficiency in convergence.
- The findings may enable more reliable automation and fault diagnosis methods in the oilfield industry.
Abstract
Abstract Fault prediction in oil pumping wells is crucial for oilfield production, directly impacting well stability and economic efficiency. Pumping well equipment is prone to various failures during long-term operation. This study developed a hybrid CNN-SVM (Convolutional Neural Networkand Support Vector Machine)model for oil well fault prediction, addressing the limitations of standalone CNN and SVM models. By integrating CNN’s powerful feature extraction capabilities with SVM’s classification strengths, the proposed model enhances prediction accuracy and efficiency. Compared to standalone CNN and SVM models, the hybrid model demonstrates superior performance including faster convergence (130 iterations, loss of 0.4), higher accuracy (85%–89%), improved recall (74%–83%), and better F1 scores (87%–91%). It provides a reliable and efficient solution for intelligent fault diagnosis in oilfield automation.