Automatic calibration method for discrete element linear stiffness parameters during geotechnical material simulation

Abstract

The invention discloses an automatic calibration method for discrete element linear stiffness parameters during geotechnical material simulation. According to the method, an analytic formula is used as an initial estimated value of a linear elastic contact parameter, a particle rigidity parameter is used as an independent variable, a numerical simulation actual result is used as a dependent variable, and the contact parameter in a discrete element model is automatically trained. According to the invention, the target parameters can be obtained through automatic training, and the time requiredby a traditional manual checking method is greatly reduced; the macroscopic deformation analytical solution based on the isotropic discrete sample is used as an initial value of parameter training, sothat the problems of local optimal traps, overlong training time and the like in the parameter training process are effectively avoided; elastic parameters are calibrated through response of the sample under the small strain condition, and the parameter calibration efficiency is high. The implementation process is simple, and the method is suitable for rapidly calibrating the mesoscopic deformation parameters of various discrete materials such as sandy soil and rockfill materials and bonded discrete materials such as sedimentary rock and ceramics during simulation research by adopting discrete elements.

Description

technical field [0001] The invention belongs to the field of geotechnical engineering. Specifically, the invention relates to an automatic calibration method of discrete element linear stiffness parameters when simulating rock and soil materials. Background technique [0002] Due to the natural discrete nature of geotechnical materials such as sand and rockfill materials, the discrete element method with particles as the basic unit is widely used in the analysis and research of geotechnical materials and geotechnical engineering problems. However, since the mesoscale parameters adopted by the discrete element algorithm are not easy to measure directly through physical tests, most of the current discrete element simulations for geotechnical engineering must first simulate certain types of conventional geotechnical tests (such as triaxial tests). By manually adjusting the discrete element mesoscopic parameters until the simulated object is basically consistent with the physica...

AI Technical Summary

Problems solved by technology

The disadvantages of this method are: (1) It is necessary to simulate a sufficient number of numerical samples to provide a basis for sensitivity analysis and statistical fitting, and the calculation is relatively large; (2) after the numerical output, a single macro The main mesoscopic influencing parameters behind the indicators, and then Latin hypercube sampling and ant colony algorithm to optimize parameter values, rely too much on statistical methods and ignore the physical connection behind the macroscopic and mesoscopic parameters of bulk materials, and the implementation process is relatively complicated

However, there are also the following shortcomings: (1) All mesoscopic parameters are trained at the same time, and each parameter interferes with each other in the process, which affects the calibration efficiency; (2) The update strategy comes from empirical assumptions, lacks strict mathematical foundation, and cannot guarantee sufficient convergence efficiency (3) The input mesoscopic initial value comes from the deformation relationship of particles with a single particle size and regular arrangement, which is quite different from the reality that the particle structure and particle size are randomly distributed in the actual geotechnical materials

The initial estimate of the macroscopic response that deviates too much from the actual bulk not only makes the training time too long, but also increases the risk of the material facing the local optimal trap during the training process; (4) The parameter update process ignores the macroscopic and microscopic Parameter relationship, for example: normal stiffness and tangential stiffness have an impact on the elastic modulus and Poisson's ratio in the macro parameters at the same time, according to the parameter update strategy used in this method, only the elastic modulus or Poisson's ratio is used to update the normal direction Stiffness or tangential stiffness are actually inappropriate; (5) do not distinguish between two-dimensional models and three-dimensional models in terms of parameter training strategies and processes

For example: using the same initial value estimation for 2D and 3D models, the model cannot achieve the best convergence efficiency

Images