Application of pool‐based active learning in physics‐based earthquake ground‐motion simulation

Abstract

We present the results of the application of the active learning method in developing surrogates as physics‐based earthquake ground‐motion simulators. The surrogates, which map input parameters into output values without demanding intensive computations, are an essential part of any parameter optimization, sensitivity, and uncertainty analysis. Artificial neural networks (ANNs), as an example of surrogates, are very effective in estimating any complicated model. ANNs use a set of training data to learn the mapping process. Training data are a set of input parameters and their corresponding output values.

Generating training data requires conducting numerous regional scale ground‐motion simulations. These numerical simulations are computationally challenging. Therefore, a step‐by‐step learning method should be employed to reduce the need for generating unnecessary training data. We demonstrate how active learning methods can significantly reduce the computational burden while maintaining the accuracy of physics-based ground motion prediction models.

Key Contributions

• Computational Efficiency: Significant reduction in required physics-based earthquake simulations
• Active Learning Framework: Novel application of pool-based active learning in seismological modeling
• Surrogate Development: Efficient methodology for creating accurate ground motion simulators
• Parameter Optimization: Enhanced framework for sensitivity and uncertainty analysis in earthquake engineering
• Neural Network Integration: Systematic approach for training ANNs with minimal but informative data

Applications in Earthquake Engineering

This work addresses critical computational challenges in:

• Regional Seismic Hazard Assessment: Efficient evaluation of ground motion across large geographical areas
• Earthquake Scenario Modeling: Rapid assessment of potential earthquake impacts
• Uncertainty Quantification: Systematic analysis of model uncertainties with reduced computational cost
• Real-time Applications: Development of fast surrogate models for emergency response systems

Methodology and Impact

The research demonstrates how machine learning techniques can be effectively integrated with physics-based earthquake simulation to:

• Reduce computational requirements for large-scale seismic studies
• Maintain accuracy while dramatically decreasing simulation time
• Enable more comprehensive uncertainty and sensitivity analyses
• Support real-time earthquake impact assessment applications

This interdisciplinary approach bridges computational seismology, machine learning, and earthquake engineering, contributing to more efficient and practical earthquake hazard assessment methodologies.

Recommended citation: Khoshnevis, N., & Taborda, R. (2019). Application of pool‐based active learning in physics‐based earthquake ground‐motion simulation. Seismological Research Letters, 90(2A), 614-622. https://doi.org/10.1785/0220180144
Download Paper