oa Full Waveform Inversion Based on Strain Signals from Distributed Fiber Optic Sensing

Recently, the Distributed Acoustic Sensing (DAS) system, utilizing optical fiber technology, has gained attention for seismic data acquisition. It allows for permanent underground deployment, enabling large-area, high-density observations with strong anti-interference capabilities. Compared to conventional seismic geophone, the stability and repeatability of DAS technology render it better suited for urban activity monitoring and oil and gas exploration. To better analyze underground structures using DAS data, accurate velocity analysis is necessary. As a state-of-the-art velocity analysis tool, full waveform inversion (FWI) can reconstruct underground model parameters by minimizing the differences between observed seismic waveforms and simulated waveforms. Hence, FWI imposes more stringent demands on the precision of simulated data. However, DAS acquisition technology typically employs strain or strain rate as the seismic signals, which are used to generate DAS data. This presents more significant challenges for FWI based on data matching. The conventional FWI based on DAS data is achieved by converting DAS data into geophone signals. This conversion-based approach can result in the amplification of low wavenumber noise, particularly in cases characterized by a low signal-to-noise ratio. The signal-to-noise ratio of data in DAS technology tends to be low because it exclusively records axial strain rate responses from optical fibers. The utilization of this conversion-based method can substantially impact the accuracy of the model inversion.

To address this problem, we develop an FWI method for DAS data, utilizing strain parameters as the foundation for the inversion process. We first implement a strain-based forward modeling method for DAS data based on the elastic wave equation. This approach can simulate and analyze strain response, facilitating the accurate seismic modeling of DAS data. Then we define the objective function by directly measuring the discrepancy between the observed strain signal and the simulated strain signal. Through this, we enable FWI directly based on DAS data by deriving the gradient and adjoint equations using a gradient-based approach. Compared to conventional FWI with DAS data, the primary difference between the two methods lies in the adjoint source. In addition, conventional FWI with geophone has the flexibility to employ either a single component or simultaneously multiple components in the inversion. In contrast, DAS focuses on the residual of the axial component in its backpropagation process, as it exclusively receives axial strain signals along the fiber. By avoiding the requirement to convert DAS data into geophone data, our method can make the direct updation of model parameters using strain signals from DAS data. This strategy results in effective parameter optimization and enhanced inversion outcomes. Finally, we apply the proposed method to synthetic data to verify its feasibility and effectiveness. A Numerical experiment has shown that this method can significantly improve the accuracy of the parameters of the inversion model.