Exploring The Strength Behavior Of Large-Sized Coarse-Grained Soils Through Sequence Modeling Networks

In order to achieve green and low-carbon transformation,China continuously carries out reforms and developments in the energy sector,increases investment and development in clean energy,and emphasizes the development of hydroelectric resources.Coarse-grained soil is an important building material for constructing earth-rock dams in hydroelectric engineering.The mechanical characteristics of coarse-grained soil have important practical value in actual engineering construction,and its stress-strain relationship is the most important feature.We need to study the relationship between this relationship and the soil strength of coarse-grained soil.The mechanical characteristics of large-grained coarse-grained soil cannot be obtained through traditional experimental instruments,and soil mechanics experiments require a large amount of manpower and material resources.Therefore,the large-grained material can only be processed by the reduced-scale method so that it can be tested by laboratory instruments to obtain physical experimental data,and a mathematical model is established based on experience to deduce the mechanical properties of large-grained coarse-grained soil on-site.However,these models have a large number of physical parameters that have no practical significance,and their predictive effects cannot fully meet the needs.This paper collects a large amount of real triaxial compression test data and proposes a prediction model for the mechanical properties of coarse-grained soil.Through the idea of dealing with time series problems,the mechanical properties of large-grained coarse-grained soil are further explored.The main research content of this paper is as follows:(1)We collected,organized,and preprocessed a large amount of real triaxial experimental data and built an expandable database,laying a good foundation for future research on the scale effect of coarse-grained soil.After basic data cleaning and format conversion,we analyzed and processed the data based on the sequence features of coarse-grained soil using the idea of time series modeling.(2)In order to fully utilize the temporal dependencies among the coarse-grained soil datasets,this paper utilized the bidirectional long short-term memory network and Self-Attention mechanism to construct the main model for predicting the mechanical properties of small-grained coarse-grained soil with a maximum particle size of dmax=60mm.The accuracy of the model was validated through experiments.(3)Considering the systematic deviation that may occur in the prediction process of the main model for coarse-grained soils with large particle size,this paper added an auxiliary model on the basis of the existing learning ability of the main model,to learn the differences in mechanical behavior of coarse-grained soils with large particle size.The proposed method was experimentally verified to achieve high accuracy and reliability in predicting and analyzing the mechanical properties of coarse-grained soils with large particle size.