Investigation of motion characteristics of catastrophic landslide using material point method and gene expression programming

This study presents two approaches (i.e., the material point method (MPM) and gene expression programming (GEP)) for prediction of landslide runout. The MPM was modified based on a strain-softening constitutive model. A collapse test and physical model test were conducted to verify the suitability of the modified MPM. The sensitivity analysis was performed to investigate the importance of the softening parameters on the motion characteristics and final accumulation state. In addition, a predictive equation for prediction of the maximum horizontal distance was proposed based on GEP, with the coefficient of determination of 0.8825. The effects of the maximum vertical distance, landslide volume and landslide posture on the maximum horizontal distance were quantitatively analyzed. Moreover, the GEP model was verified by comparison with other machine learning approaches. Finally, taking the Jiweishan landslide as an example, the simulation and prediction results obtained from the MPM and GEP approaches were compared. The results show that the two are both within a reasonable range, with the errors of 14.1% and 10.6%, respectively.

Automated summary

This study presents two approaches (i.e., the material point method (MPM) and gene expression programming (GEP)) for prediction of landslide runout. The modified MPM was verified through collapse and physical model tests. Sensitivity analysis was performed to investigate the importance of the softening parameters. A prediction equation based on GEP achieved a coefficient of determination of 0.8825. The effects of various factors on the maximum horizontal distance were quantitatively analyzed. The simulation and prediction results of MPM and GEP were compared and found to be within reasonable ranges with errors of 14.1% and 10.6%, respectively.