Random analysis of deposit landslide deformation under uniformly increasing rainfall using machine learning

Abstract

Landslides are one natural disaster in the mountains, causing damage to properties, destruction of infrastructure and loss of life. While landslide deformation and displacement are commonly observed through experimental investigation of the prototype models, this time-consuming approach might involve human errors and inaccurate constitutive models. This study aims to propose a numerical approach based on machine learning for substituting experimental investigation. In pursuit of this objective, the probability-weighted moments (PWMs) for generalised extreme value distribution (GEVD) were implemented to predict the deposit landslide deformation under uniformly increasing rainfall. The deposit landslide of the Dahua was selected as the case study. The accuracy and reliability of the proposed PWM are validated through the maximum likelihood method and the Jenkinson method. The PWM with GEVD is used to derive the progressive failure of the deposit landslide stability under the return period, the pore water pressure variation and the return period’s effect. The results show that the proposed solution is suitable for predicting the responses of landslides. Moreover, the results demonstrate the importance of soil density; 30% of the deposited soil was moved when the soil density was 2 g/cm³ besides, and more than 90% were moved at the top of the slope for the soil density equal to 3 g/cm³.

Keywords:

• Landslide deformation
• displacement
• constitutive model
• probability-weighted moments
• generalized extreme value distribution
• machine learning

Disclosure statement

The authors declare that they have no known competing interests.

Data availability statement

The data used to support the findings of this study are included in the article.

Additional information

Funding

This research was supported by the National Key R&D Program of China (2017YFC1501100).