TL;DR
This paper introduces a novel, fast, and automated inversion algorithm for analyzing air-coupled Spectral Analysis of Surface Waves (SASW) data. By directly relating the dispersion curve to shear velocity profiles using particle displacement characteristics, the method bypasses slow, iterative forward simulations, enabling near real-time subsurface profiling for geological and pavement assessments.
What Problem Does This Paper Address?
Traditional Spectral Analysis of Surface Waves (SASW) is a powerful non-destructive technique for determining subsurface soil and pavement properties. However, its widespread adoption, especially for rapid pavement maintenance, is hindered by the slow and labor-intensive inversion process. This process typically relies on iterative "guess-and-check" methods that require manual adjustments and forward simulations, making it inefficient for quick, on-site evaluations. The paper aims to overcome this limitation by developing a significantly faster and automated inversion method.
Methods
The core of this paper is a new fast inversion algorithm that leverages the fundamental physics of surface waves. Instead of relying on traditional forward simulations, the algorithm establishes a direct, weighted averaging relationship between the dispersion curve (phase velocity vs. wavelength) and the shear velocity profile of layered subsurface materials. This relationship is derived by considering the distribution of particle displacement along the depth of penetration for a given surface wave wavelength. The vertical component of particle displacement is used as a weighting factor, reflecting its contribution to the overall wave motion.
The inversion process is structured in two stages:
- Initialization: The measured dispersion curve is converted into a discrete function of phase velocity versus approximate layer depth. Each depth is treated as a thin layer, and initial shear velocities are estimated assuming a half-space model for each layer sequentially.
- Iterative Adjusting: This stage refines the shear velocity estimates. For each layer, the shear velocity is adjusted based on the difference between a calculated and an assumed value, with a convergence criterion to ensure stability. This iterative process is purely algebraic, avoiding complex differential equations.
The method is designed to be fully automated, requiring no manual input or adjustment of trial profiles.
Datasets
The paper uses simulated datasets derived from established methods for validation. Specifically:
- Ascending Stiffness Profile: A layered structure with increasing shear velocities with depth was simulated. The dispersion curve was generated using the stiffness matrix method and compared with published results.
- Descending Stiffness Profile: A layered structure with decreasing shear velocities with depth was simulated.
- Published Examples: Two examples from existing literature (Nazarian et al. [7] and Joh [8]) were used as benchmarks, employing their reported dispersion curves as input for the proposed inversion algorithm.
- Field Test Data: Data was collected from a mobile testing system at an asphalt parking lot.
Key Insights and Ideas
- Direct Relationship: The paper demonstrates that the dispersion curve can be directly related to the shear velocity profile through a weighted averaging of layer velocities, with particle displacement distribution serving as the weighting factor. This bypasses the need for computationally expensive forward modeling.
- Automation Potential: The purely algebraic nature of the inversion algorithm makes it highly amenable to automation, removing the need for expert judgment in setting initial profiles.
- Speed Advantage: The proposed method is significantly faster than traditional iterative approaches, achieving inversion in less than one second, which opens possibilities for real-time applications.
- Fundamental Mode Dominance: The algorithm is primarily based on the fundamental mode of surface wave propagation. This works well for geological sites with ascending stiffness but requires careful handling of higher modes in scenarios like pavements with stiffer top layers.
Key Results
- The proposed fast inversion algorithm successfully estimated shear velocity profiles for both ascending and descending stiffness layered systems, showing good agreement with original profiles and published results.
- Validation against the stiffness matrix method confirmed the accuracy of the inversion.
- Comparison with two examples from previous studies indicated that the proposed method yields comparable results to established, albeit slower, inversion techniques.
- The field test on an asphalt parking lot demonstrated the algorithm's capability to estimate elastic modulus profiles, though accuracy was noted to be lower for thin layers due to discretization limitations.
- The inversion calculation for simulated data was completed in less than one second, highlighting its speed advantage.
Limitations and Future Work
- Fundamental Mode Assumption: The algorithm's reliance on the fundamental mode of surface wave propagation is a limitation. For scenarios with dominant higher modes (e.g., pavements with stiff top layers), advanced pre-processing techniques (like Gabor filters) or alternative methods (like Lamb wave inversion) are needed to extract the fundamental mode or handle multi-mode vibrations.
- Layer Discretization: In the field test, the accuracy for thin layers was affected by the fine discretization required by the algorithm.
- Future Work: The authors suggest embedding pre-processing techniques for extracting the fundamental mode or handling higher modes directly into the real-time workflow. Further research into handling multi-mode vibrations and improving accuracy in complex layered structures is also implied.
Summary updated