Preliminary results on a near-real-time rock slope damage monitoring system based on relative velocity changes following the September 5,2022 M_(S) 6.8 Luding,China earthquake

Relative seismic velocity change(dv/v)is important for monitoring changes in subsurface material properties and evaluating earthquake-induced rock slope damage in a geological disaster-prone region.In this paper,we present a rapid damage assessment on three slow-moving rock slopes by measuring dv/v decrease caused by the 2022 M_(S) 6.8 Luding earthquake in Southwest China.By applying the stretching method to the cross-correlated seismic wavefields between sensors installed on each slope,we obtain earthquake-induced dv/v decreases of~2.1%,~0.5%,and~0.2%on three slopes at distances ranging from~86 to~370 km to the epicenter,respectively.Moreover,based on seismic data recorded by 16 sensors deployed on the rock slope at a distance of~370 km away from the epicenter,a localized dv/v decease region was observed at the crest of the slope by calculating the spatial dv/v images before and after the earthquake.We also derive an empirical in situ stress sensitivity of -7.29×10^(-8)/Pa by relating the dv/v change to the measured peak dynamic stresses.Our results indicate that a rapid dv/v assessment not only can help facilitate on-site emergency response to earthquakeinduced secondary geological disasters but also can provide a better understanding of the subsurface geological risks under diverse seismic loadings.

Optimal CNN-based semantic segmentation model of cutting slope images

This paper utilizes three popular semantic segmentation networks,specifically DeepLab v3+,fully convolutional network(FCN),and U-Net to qualitively analyze and identify the key components of cutting slope images in complex scenes and achieve rapid image-based slope detection.The elements of cutting slope images are divided into 7 categories.In order to determine the best algorithm for pixel level classification of cutting slope images,the networks are compared from three aspects:a)different neural networks,b)different feature extractors,and c)2 different optimization algorithms.It is found that DeepLab v3+with Resnet18 and Sgdm performs best,FCN 32s with Sgdm takes the second,and U-Net with Adam ranks third.This paper also analyzes the segmentation strategies of the three networks in terms of feature map visualization.Results show that the contour generated by DeepLab v3+(combined with Resnet18 and Sgdm)is closest to the ground truth,while the resulting contour of U-Net(combined with Adam)is closest to the input images.