Abstract: This paper details the relationship between the content of clay particles and the density of debris flow by projection pursuit regression analysis, and the kinetic mechanism of debris flow and the mechanism of particle suspension were analyzed. The results show that the average relative error of PPR model and polynomial model is 5.3% and 6.5%, respectively when the PPR model is simple-modeled, when the relative error is less than 20%,15%,10% and 5%, the rate of the PPR model increases by 39%,10.6%,23% and 24%.When using PPR reservation test analysis, compared with the polynomial model, the qualification rate of the modeling analysis group and the reservation test group was improved. In addition, there is a non-linear correlation between the content of clay particles and the density of debris flow. When the content of clay particles is less than 6.78%, the density of debris flow increases with the increase of the content of clay particles, and when the content of clay is more than 6.78%, the density of debris flow decreases with the increase of the content of clay particles. When the clay content is between 6%~10%, the average bulk density of debris flow is the largest and fluctuates between 2.05~2.20 g/cm3. This paper proves that the PPR model has obvious advantages over the polynomial model, its relative error is smaller, the regression fitting is stable, the analysis result accords with the kinetic energy mechanism of debris flow and the conclusion of particle suspension mechanism, and further illustrates that PPR model has more advantages in compatibility and quantitative information utilization.

Key words: Clay content, Debris flow density, Projection pursuit regression analysis method, Kinetic energy mechanism of debris flow, Mechanism of particle suspension