An Automated Framework for Probabilistic Back-Analysis of Rockfall Catalogs Using Bayesian Optimization and Radar Tracking.
Saved in:
| Title: | An Automated Framework for Probabilistic Back-Analysis of Rockfall Catalogs Using Bayesian Optimization and Radar Tracking. |
|---|---|
| Authors: | Xie, Arnold Yuxuan1 (AUTHOR) w450198403@gmail.com, Huang, Zhanyu2 (AUTHOR), Li, Bing Q.1 (AUTHOR) bing.li@uwo.ca |
| Source: | Rock Mechanics & Rock Engineering. Aug2025, Vol. 58 Issue 8, p8701-8720. 20p. |
| Subjects: | Rockfall, Tracking radar, Engineering models, Parameter estimation, Surrogate-based optimization, Climate change, Kriging, Risk assessment |
| Abstract: | Increasing temperatures and extreme weather events driven by climate change have heightened the risk of rockfall hazards. Reliable estimates of rockfall hazards have a significant impact on the design of roadways, slopes, open pits, and underground excavations. The reliability of these estimates relies on numerical models calibrated using rockfall catalogs. Recent advancements in radar technology enable high-resolution monitoring of rockfall events, providing valuable data on their propagation. However, there are few methods to back-analyze numerical model parameters from radar data. Hence, we propose a Bayesian Optimization Framework to address this gap and demonstrate the framework through a Doppler-radar-tracked rockfall catalog comprising 21 events and 19,356 rock positions. Specifically, we use Gaussian Process Regression (GPR) as a surrogate model to minimize the number of simulations required to determine the best-fitting coefficients of restitution (CORs). The framework supports event-based and site-wise analysis. The former can readily identify unmatchable low-quality events, and the latter excels in fitting CORs generalizable across multiple events. Our framework achieves comparable resolution using five times fewer simulations compared to a benchmark grid-search method. Trajectories simulated using the mean and covariance of the best-fit CORs exhibit reasonable agreement with the measured data. This framework accelerates data inspection and parameter searching when back-analyzing numerical model parameters using a catalog of multiple rockfall events. Highlights: We present a method for automated back-analysis of multiple rockfall events and demonstrate its performance on a catalog in an open-pit mine. Quantify the best-fit parameters and their uncertainties with a mean and covariance matrix, which generates trajectories matching the measured data. The method can also identify when events are monitored with poor data quality, such as due to an inaccurate digital terrain model. [ABSTRACT FROM AUTHOR] |
| Copyright of Rock Mechanics & Rock Engineering is the property of Springer Nature and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Engineering Source |
|
Full text is not displayed to guests.
Login for full access.
|
|
| Abstract: | Increasing temperatures and extreme weather events driven by climate change have heightened the risk of rockfall hazards. Reliable estimates of rockfall hazards have a significant impact on the design of roadways, slopes, open pits, and underground excavations. The reliability of these estimates relies on numerical models calibrated using rockfall catalogs. Recent advancements in radar technology enable high-resolution monitoring of rockfall events, providing valuable data on their propagation. However, there are few methods to back-analyze numerical model parameters from radar data. Hence, we propose a Bayesian Optimization Framework to address this gap and demonstrate the framework through a Doppler-radar-tracked rockfall catalog comprising 21 events and 19,356 rock positions. Specifically, we use Gaussian Process Regression (GPR) as a surrogate model to minimize the number of simulations required to determine the best-fitting coefficients of restitution (CORs). The framework supports event-based and site-wise analysis. The former can readily identify unmatchable low-quality events, and the latter excels in fitting CORs generalizable across multiple events. Our framework achieves comparable resolution using five times fewer simulations compared to a benchmark grid-search method. Trajectories simulated using the mean and covariance of the best-fit CORs exhibit reasonable agreement with the measured data. This framework accelerates data inspection and parameter searching when back-analyzing numerical model parameters using a catalog of multiple rockfall events. Highlights: We present a method for automated back-analysis of multiple rockfall events and demonstrate its performance on a catalog in an open-pit mine. Quantify the best-fit parameters and their uncertainties with a mean and covariance matrix, which generates trajectories matching the measured data. The method can also identify when events are monitored with poor data quality, such as due to an inaccurate digital terrain model. [ABSTRACT FROM AUTHOR] |
|---|---|
| ISSN: | 07232632 |
| DOI: | 10.1007/s00603-025-04608-3 |
