Application of SBAS Technique Combined with BP Neural Network in the Settlement of the Yinxi Industrial Park in Baiyin

摘要

In recent years, due to the obvious ground settlement and other phenomena of the Yinxi Industrial Park in Baiyin, it has brought many hidden dangers to the local development, it is of great practical significance to monitor the deformation of the area for a long time series. The ground deformation field of Yinxi Industrial Park from June 2018 to April 2021 was obtained by processing Sentinel-1A data using SBAS technology, and the high coherence point D1 was predicted and analyzed by BP neural network. The results show that subsidence occurs in several places in the Yinxi Industrial Park, and the average annual subsidence rate ranges from -19.28 mm to 5.08 mm, the areas of severe settlement have a clear geographical distribution, mainly concentrated in road and building areas, other areas have a more stable ground base; the mean square error in the BP neural network prediction result is 2.56 mm, and the average relative error is 6.06%, which is a high prediction accuracy. The predicted cumulative settlement value at point D1 in 2023 is 45 mm, and there is a tendency for the settlement to intensify. The prediction results are of great significance for the early identification and prevention of ground settlement in the study area.