Data-Driven Characteristic Frequency Extraction and Operational Parameter Mapping in DAS Monitoring of Hydraulic Fracturing

Real-time distributed acoustic sensing (DAS) monitoring during hydraulic fracturing is indispensable for optimizing fracture design and ensuring operational safety. However, current field practices for interpreting fiber-optic monitoring data still rely predominantly on manual analysis, resulting in low efficiency, poor real-time responsiveness, and only coarse full-frequency-range spectral characterization. Moreover, existing methodologies have not thoroughly investigated phase-specific differences in spectral responses nor provided a quantitative framework to analyse the correlations between acoustic signal features and treatment parameters. To overcome these limitations, we propose a data-driven characteristic frequency extraction and operational parameter mapping method. This method integrates Fourier transformation with morphological quantification and genetic-algorithm parameter optimization, enabling efficient, automated identification of characteristic frequency bands in complex, noise-contaminated environments. Next, we develop a multidimensional feature-engineering workflow: DAS time-domain signals are transformed via fourier analysis to compute their energy spectra, from which multiple features are extracted to quantitatively characterize the energy-signal properties. Finally, based on these features we establish a quantitative correlation-analysis framework linking Flow Rate, proppant concentration, and DAS features within each frequency band. The presented method was applied to the real-time DAS monitoring data from a shale reservoir in China to show the workflow. The one-dimensional energy spectral data from a specific perforation cluster was investigated to reveal dynamic coupling mechanisms between proppant transport processes and fiber-optic acoustic frequency responses. Results demonstrated that significant high-frequency responses (>2000 Hz) are triggered during the initial proppant injection phase, with characteristic frequency energy values increasing by 1-2 orders of magnitude compared to the base fluid injection stage. With the pumping going on, the responsive frequency band progressively expanded to higher ranges. During the later injection stage, the amplitude of dominant frequency energy stabilizes while exhibiting enhanced uniformity in spectral energy distribution across the frequency band. The proposed algorithm consistently extracts proppant-dynamics-correlated characteristic frequency bands throughout all fracturing phases. These findings provide a multidimensional data-driven decision framework for real-time optimization and effectiveness evaluation in shale gas fracturing operations. This proposed method provides a multidimensional, data-driven analytical framework for real-time optimization and performance evaluation of fiber-optic monitoring in hydraulic fracturing, establishing a more automated acoustic diagnostic paradigm. It significantly enhances feature-recognition efficiency and robustness, offering a practical tool to meet industry demands for intelligent fracturing process monitoring and control.