Microseismic source imaging using physics-informed neural networks with hard constraints

Xinquan Huang,Tariq Alkhalifah

DOI: https://doi.org/10.1109/tgrs.2024.3366449

2024-02-15

Abstract:Microseismic source imaging plays a significant role in passive seismic monitoring. However, such a process is prone to failure due to aliasing when dealing with sparsely measured data. Thus, we propose a direct microseismic imaging framework based on physics-informed neural networks (PINNs), which can generate focused source images, even with very sparse recordings. We use the PINNs to represent a multi-frequency wavefield and then apply inverse Fourier transform to extract the source image. To be more specific, we modify the representation of the frequency-domain wavefield to inherently satisfy the boundary conditions (the measured data on the surface) by means of a hard constraint, which helps to avoid the difficulty in balancing the data and PDE losses in PINNs. Furthermore, we propose the causality loss implementation with respect to depth to enhance the convergence of PINNs. The numerical experiments on the Overthrust model show that the method can admit reliable and accurate source imaging for single- or multiple- sources and even in passive monitoring settings. Compared with the time-reversal method, the results of the proposed method are consistent with numerical methods but less noisy. Then, we further apply our method to hydraulic fracturing monitoring field data, and demonstrate that our method can correctly image the source with fewer artifacts.

Geophysics,Machine Learning,Computational Physics

What problem does this paper attempt to address?

The problem that this paper attempts to solve is that in micro - seismic source imaging, due to sparse and irregular data, traditional methods are difficult to image accurately. Specifically, traditional micro - seismic source location methods rely on travel - time inversion, which requires time - consuming first - arrival time picking, and in the case of low signal - to - noise ratio data and multi - event situations, picking the arrival time becomes very challenging. In addition, the resolution and accuracy of these methods may be relatively low.

To overcome these problems, the author proposes a direct micro - seismic source imaging framework based on Physics - Informed Neural Networks (PINNs). This framework represents the frequency - domain wave field by introducing hard constraints and uses the inverse Fourier transform to extract the source image. The hard constraints avoid the difficulty of balancing data loss and PDE loss in PINNs by imposing constraints on the boundary conditions (i.e., the data measured on the surface). In addition, the author also proposes a causality loss function to improve the convergence speed and accuracy of PINNs.

The main contributions of this method include:
1. Developing a direct source imaging framework based on PINNs and hard constraints.
2. By introducing the reference frequency loss function and the causality loss function in the depth direction, ensuring that the training process is stable and reliable.
3. Verifying the effectiveness of this method on synthetic data and actual data, and the results show that this method can provide reliable and less - noisy source images in single - source or multi - source events and passive monitoring settings.

Through these improvements, this method performs well in dealing with sparse and irregular observation data, showing great potential in practical applications.