A Hybrid F–K Domain Feature Extraction and Enhancement Framework for Low-Frequency DAS Production-Logging Data: A Single-Well Field Case Study

Distributed optical fiber acoustic sensing (DAS) has become an important technology for production logging because it can record dense strain or strain-rate responses along an optical fiber under high-temperature, high-pressure, and corrosive downhole conditions. This single-well field case study investigated a hybrid low-frequency DAS processing framework for distributed optical fiber production logging. First, a finite impulse response (FIR)-based preprocessing step was used for low-pass smoothing before an F-K domain analysis. The DAS records were then transformed into the frequency-wavenumber (F-K) domain, where particle swarm optimization (PSO) was used to tune the nu parameter of a one-class support vector machine (OCSVM) for automatic feature extraction. Rule-based feature enhancement and a small-sample support vector classifier (SVC) were then applied to suppress residual F-K domain noise and retain the V-shaped features associated with upgoing and downgoing waves. Finally, linear regression was applied to the enhanced F-K domain branches to estimate the apparent propagation velocities and derive the flow velocity through the field interpretation relationship. The workflow was demonstrated using 15 s field DAS segments from an oil-water two-phase production well, and the six-window validation showed errors below 3.13% relative to the field-reference values. These results demonstrate the feasibility of the proposed workflow for the investigated well, but do not constitute general validation across different wells or acquisition conditions.