Estimating Fault Friction from Seismic Signals in the Laboratory (1710.04172v4)

Abstract: Nearly all aspects of earthquake rupture are controlled by the friction along the fault that progressively increases with tectonic forcing, but in general cannot be directly measured. We show that fault friction can be determined at any time, from the continuous seismic signal. In a classic laboratory experiment of repeating earthquakes, we find that the seismic signal follows a specific pattern with respect to fault friction, allowing us to determine the fault's position within its failure cycle. Using machine learning, we show that instantaneous statistical characteristics of the seismic signal are a fingerprint of the fault zone shear stress and frictional state. Further analysis of this fingerprint leads to a simple equation of state quantitatively relating the seismic signal power and the friction on the fault. These results show that fault zone frictional characteristics and the state of stress in the surroundings of the fault can be inferred from seismic waves, at least in the laboratory.

Estimating Fault Friction from Seismic Signals Using Machine Learning

This paper, presented by Bertrand Rouet-Leduc et al., details an investigation into the frictional state of faults through a laboratory setting using machine learning to analyze seismic signals. The paper demonstrates a robust relationship between fault friction and seismic signal characteristics, providing a novel methodology to infer fault frictional states that are otherwise challenging to directly measure due to the lack of continuous observational data.

Overview and Methodology

The research is anchored in classic laboratory experiments simulating repeating earthquake events. The experimental setup consists of a biaxial shear apparatus that applies controlled shear stress analogous to tectonic forces on a granular fault gouge. The continuous seismic signal emitted is collected via piezoceramics. Machine learning algorithms, specifically gradient boosted trees, are employed to extract statistical features from the seismic signals during these controlled experiments. These features are used to construct models that predict the instantaneous frictional state of the fault.

Key Findings

1. Discovery of a Predictive Relation: Through the application of machine learning, a direct relationship between the seismic signal's statistical characteristics and the frictional state of the fault is established. The paper identifies seismic signal variance, proportional to power, as a critical feature that correlates with fault friction.
2. Equation of State: The research uncovers an equation of state linking friction to the seismic signal's power. This relationship remains consistent across different stress conditions, suggesting a generalizable model that can predict fault behavior across varied experimental environments.
3. Hysteretic Behavior: Analysis reveals hysteretic cycles within the fault's seismic cycle. The paper draws parallels between observed hysteretic behavior and the discrete memory phenomenon in dilatated rocks, providing deeper insights into fault dynamics.

Implications and Future Directions

The implications of this paper extend beyond laboratory settings. The findings indicate a potential for real-world applications where machine learning techniques could be utilized to infer frictional states from seismic data without direct observation. This can significantly advance our understanding of earthquake mechanics and prediction models, providing a more comprehensive approach to fault monitoring.

Furthermore, the transfer of the equation of state across different load levels paves the way for further research aiming to replicate these findings in more complex and variable natural conditions. Future work could focus on applying these methodologies to large-scale simulations and field data, enhancing our ability to interpret and predict seismic activities with higher accuracy.

Conclusion

This paper presents a significant advancement in the integration of machine learning with geophysical analyses of fault mechanics. By revealing a predictable link between seismic signals and fault friction, the research sets a foundation for new methodologies in earthquake monitoring. The findings encourage further exploration into the application of continuous signal analysis in natural settings, offering promising directions for future seismic research and practical earthquake forecasting tools.