Abstract:
Accurately detecting bridge plug position and leakage rate is crucial for mastering plug working conditions and achieving real-time quantitative monitoring of multi-cluster staged hydraulic fracturing operations. It holds significant importance for evaluating fracturing effectiveness and optimizing processes. This paper focuses on the core objectives of precise bridge plug position detection and effective leakage rate identification, employing a combination of numerical simulation and digital signal processing methods to conduct the following research.Firstly, by analyzing the operational conditions of bridge plugs, an acoustic monitoring approach is proposed. Subsequently, based on the k-ω turbulence model and the convective wave theory, a real-time monitoring numerical model for bridge plug working conditions is established. Secondly, leveraging equipment working principles and numerical signal processing theory, Empirical Mode Decomposition (EMD) is utilized to extract signal feature parameters. Finally, a digital signal processing methodology is implemented, employing Principal Component Analysis (PCA) for dimensionality reduction and Random Forest (RF) for classification and regression tasks.Supported by the equipment scheme and digital signal processing methods introduced in this paper, the detection accuracy for bridge plug position achieves over 96%. Under conditions of a limited dataset, using the Random Forest algorithm to identify the bridge plug leakage rate yields a trained model with an R-squared value exceeding 0.9 and a mean absolute error of only 0.0105 m
3/min.The findings of this research possess substantial theoretical significance and application value for enhancing fracturing operation efficiency, optimizing fracturing design, and reducing operational risks. Furthermore, it provides a novel approach and methodology for realizing real-time monitoring of the fracturing construction process. This study demonstrates the feasibility and effectiveness of integrating physical modeling with advanced data-driven techniques for addressing key challenges in downhole monitoring, paving the way for more intelligent and data-informed fracturing operations. The proposed framework, combining EMD, PCA, and Random Forest, shows strong potential for handling complex, non-linear acoustic signals in noisy downhole environments to deliver accurate and reliable diagnostics.