Abstract: A method of detecting the gross error of slope monitoring data is presented based on singular spectrum analysis (SSA) and density-based spatial clustering of applications with noise (DBSCAN). The method integrates the advantages of SSA in signal extraction and DBSCAN in distinguishing gross errors and outliers. Firstly, SSA is used to decompose and reconstruct the monitoring series to accurately extract the main signal and obtain the residual components. Secondly, DBSCAN is employed to analyze the residual components. The two methods are combined to determine and eliminate the gross errors. Examples of slope monitoring series affected by multiple factors are introduced for verification. Moreover, the present method is compared with the median absolute deviation method (MAD) and Grubbs criterion method (Grubbs), and results suggest that the present SSA-DBSCAN method is of excellent performance and low misjudgment rate compared with the abovementioned methods.

Key words: slope engineering, singular spectrum analysis, time series, safety monitoring data, gross error detection, DBSCAN