An intelligent diagnostic method for lost circulation based on transient pressure waves

Accurate diagnosis of leakage location can reduce the loss caused by lost circulation, and this study incorporates the use of Improved Partly Ensemble Local Characteristic scale Decomposition (IPELCD), Phase Randomization Technique (PRT), and Pearson Correlation Coefficient and Gray Wolf Optimization (GWO) to obtain improved Support Vector Machines (PGSVM). The IPELCD-PRT-PGSVM intelligent diagnostic method is proposed for accurately locating leaky layer and reducing losses caused by lost circulation. The method innovatively applies the IPELCD algorithm and PGSVM to enhance end point detection, which overcomes modal aliasing and end point effects during transient pressure wave processing. In low-frequency component processing, IPELCD outperforms other algorithms, such as conventional Local Characteristic scale Decomposition (LCD) and Improved adaptive noise Complete Ensemble Empirical Modal Decomposition with Adapted Noise, with the Signal-to-Noise Ratio and Peak Signal-to-Noise Ratio of the experimentally reconstructed signals improved by more than 3 and 2 times, respectively, and the structural similarity index close to 1. In terms of intelligent diagnosis, the PGSVM model achieves 97.5% accuracy in leakage localization experiments, maintaining high accuracy and stability at different leakage levels and borehole sizes. Compared with other methods, the IPELCD-PRT-PGSVM method improves the stability by at least 11.48%, reduces the misclassification rate by at least 27.2%, and has the optimal performance in Precision, Recall, and F-Score. The results show that the method performs well in terms of leak location accuracy and application effect and can meet the inspection needs of different drilling sites.