Deep mining of big data of tests and constitutive relation of dilative soils

Due to the restriction of the traditional constitutive theory and the lack of in-depth studies on the common change laws of the basic mechanical characteristics of soils, most of the constitutive models established at present cannot reflect the actual deformation mechanism of soils well. A big data processing platform of Hadoop and Spark is built. By using the functional network and the AIC criteria, a distributed adaptive auto-regressive algorithm is proposed for deep mining of big data of tests on dilative soils. Based on the big data characteristic relationship of each plastic coefficient, combined with its significant and secondary influence factors, the constitutive model for dilative soils is established based on the theory of generalized plastic mechanics. Through the model verification experiments, the results show that the proposed model is better than the modified Cambridge model and the similar Cambridge model considering the dilatancy, and has strong adaptability to the expression of the mechanical properties of the dilative soils under different stress paths. The big data technology and generalized plastic mechanics are applied to the studies on the constitutive relationship of soils, which effectively breaks through the shackles of the traditional constitutive theory, and is of more extensive theoretical significance. At the same time, it also provides a new idea for the studies on the constitutive relationship of soils.