Abstract: A novel method of data processing for three-dimensional landslide model test is proposed. First, “3σ criteria” is adopted to detect singular data. In subsequence, self-adaptive filtering method of time series is employed to predict the moment of singular data, and the singular values are replaced by the predicted values. Finally the cubical smoothing algorithm with five-point approximation is used to process the fluctuation data subjected to noise interference. The method is applied to the data processing for three-dimensional landslide model test, and the results show that this method could remove singular data and glitches effectively without altering the normal trend of the original data.

Key words: 3σ criteria, time series, self-adaptive filtering method, cubical smoothing algorithm with five-point approximation