Liliana Borcea,Josselin Garnier,Alexander V. Mamonov,Jörn Zimmerling
Abstract:Waveform inversion is concerned with estimating a heterogeneous medium, modeled by variable coefficients of wave equations, using sources that emit probing signals and receivers that record the generated waves. It is an old and intensively studied inverse problem with a wide range of applications, but the existing inversion methodologies are still far from satisfactory. The typical mathematical formulation is a nonlinear least squares data fit optimization and the difficulty stems from the non-convexity of the objective function that displays numerous local minima at which local optimization approaches stagnate. This pathological behavior has at least three unavoidable causes: (1) The mapping from the unknown coefficients to the wave field is nonlinear and complicated. (2) The sources and receivers typically lie on a single side of the medium, so only backscattered waves are measured. (3) The probing signals are band limited and with high frequency content. There is a lot of activity in the computational science and engineering communities that seeks to mitigate the difficulty of estimating the medium by data fitting. In this paper we present a different point of view, based on reduced order models (ROMs) of two operators that control the wave propagation. The ROMs are called data driven because they are computed directly from the measurements, without any knowledge of the wave field inside the inaccessible medium. This computation is non-iterative and uses standard numerical linear algebra methods. The resulting ROMs capture features of the physics of wave propagation in a complementary way and have surprisingly good approximation properties that facilitate waveform inversion.
In this arxiv version two important typos are corrected when compared to the published version. The typo was in the second equation in Theorem 3 and carried over into Corollary 1. The proofs are correct.
What problem does this paper attempt to address?
### Problems the paper attempts to solve
The paper aims to address the challenges in waveform inversion. Waveform inversion is a technique for estimating heterogeneous media by transmitting probing signals in the media and recording the generated waves. This problem is a classic inverse problem with a wide range of applications, such as medical diagnosis, non - destructive testing of bridges and buildings, aircraft fuselage testing, radar imaging, underwater acoustics, seismology and geophysical exploration, etc.
However, the existing waveform inversion methods have many deficiencies. The main difficulty lies in the non - convexity of the objective function, which easily leads to getting trapped in local minima during the optimization process. Specifically, the reasons for these problems are as follows:
1. **The mapping from unknown coefficients to the wave field is nonlinear and complex**: The propagation of the wave field is affected by the media parameters, and this influence is nonlinear.
2. **The sources and receivers are usually located on one side of the media**: This means that only the back - scattered waves can be measured, not the waves passing through the media.
3. **The probing signals are band - limited and contain high - frequency components**: This makes waveform inversion more difficult because high - frequency signals are easily affected by noise and multipath effects.
To solve these problems, the author proposes an approach based on data - driven reduced - order models (ROMs). These reduced - order models are calculated directly from the measured data without the need to know the wave field information inside the media. Through this method, the physical characteristics of wave propagation can be captured and the effect of waveform inversion can be improved.
### Main contributions
1. **Data - driven reduced - order models**: Two reduced - order models of operators related to wave propagation are proposed. These models can be calculated directly from the measured data without iteration.
2. **Improved waveform inversion method**: By using these reduced - order models, the physical characteristics of wave propagation can be better approximated, thereby improving the accuracy of waveform inversion.
3. **Theoretical analysis**: A detailed theoretical analysis of the properties of these reduced - order models is carried out, proving their superiority in waveform inversion.
### Conclusion
By introducing data - driven reduced - order models, the paper provides a new perspective and method for waveform inversion, which is expected to overcome the non - convexity and local minima problems in the existing methods, thereby improving the accuracy and robustness of waveform inversion.