LSTM Autoencoder 기반의 사면 붕괴 예측 연구

초록

Numerous researchers have evaluated the stability of the slope, as its failure can cause severe damage to life and property. To this end, real-time data were collected and analyzed to evaluate the real-time risk in various slopes. However, most of the previous studies have either derived an empirical regression equation or relied on statistical methodologies, including statistical process control (SPC). In this study, a machine learning model, long short-term memory network (LSTM) Autoencoder, based on slope failure prediction was developed and validated by applying it to data on actual slope failure. As a result, it was possible to detect signs of collapse earlier than the SPC method used in previous studies and is expected to ensure lead time to cope with such collapse.

키워드