Abstract:

To enhance the economic efficiency of engineering geological investigations for construction projects and accurately reveal the correlations between the physical and mechanical parameters of cohesive soils, in this study taking the cohesive soils in central urban area of Changchun as the research object, based on the data from 1 092 physical tests, compression tests and direct shear tests, two multiple regression models were constructed to emphatically analyze the rationality of the model’s premise assumptions and the interaction effects among explanatory variables. The results show that: 1) The residual probability density plot and residual normal Q-Q （quantile-quantile） plot confirm that the model error terms follow a normal distribution, with homogeneous and mutually independent variances, satisfying the basic assumptions of the multiple regression model. 2) The two-factor interaction effect plots and Johnson-Neyman plots indicate significant interaction effects among water content, density, void ratio and liquidity index. A regression model considering interaction effects can more accurately predict the compression coefficient. It is feasible to use multiple linear regression to analyze the correlations between physical and mechanical parameters of cohesive soils.

Key words: cohesive soil')">cohesive soil, multiple linear regression, model diagnostics, model evaluation, interaction effect