The Numerical Substructure Simulation Based On Data-driven

The 21^st century is considered as the era of big-data,and data science has become a new approach to scientific research,which is profoundly affecting the progress of science and technology in an unprecedented way.In classical mechanical systems,constitutive laws are often derived from the fitting of experimental data and then analytical functions are constructed using empirical parameters.Errors and uncertainties will be introduced during this process and it also requires a high cost of intelligence and time.With the development of computer technology,data science is shining brightly.The use of measured data to replace constitutive relations in analytical form has great potential and important practical value.Based on this,this paper proposes a mechanics-based data-driven stress updating algorithm which can be used for the analysis of one dimensional elastoplastic problems,and then builds a bridge between the proposed algorithm and the classical elastoplastic algorithm through the Numerical Substructure Method and applies it in engineering problems.The main contents are as follows:（1）Based on the concept of data-driven,a set of stress-strain data obtained from uniaxial tensile tests is used to calculate the mechanical response of the 1D elastoplastic problems.Without constructing explicit constitutive expressions of materials,the boundary problem involving materials satisfying isotropic hardening,kinematic hardening or combined hardening are solved.Compared with the results of classical models,this method can accurately presict the elasyoplastic behavior of materials under uniaxial cyclic load.（2）During numerical analysis,Numerical Substructure Method could take into account both computational efficiency and accuracy.The paralell computing mode of master and sub-structure also provides convenience for the combination of different algorithms.Combining the proposed data-driven stress updating algorithm with the Numerical Substructure Method,a numerical substructure simulation method based on data-driven stress updating algorithm was proposed to simulate the mechanical responses of key components in engineering structures,and the accuracy of the algorithm was preliminarily verified by two ideal shear frame-support structures.（3）The bucking restraint brace is a kind of energy dissipating damping member with stable perfomence and full hysteretic curve.With the data-driven stress updating algorithm proposed in this paper,the measured data could be directly used to simulate and analyze the response of BRB members,which avoids the analytical solution of the traditional mechanical model and greatly reduces the difficulty of data processing.Combined with Numerical Substructure Method,the simulation experiment of a steel frame-braced structure under static cyclic load is carried out,and the results show that the proposed algorithm can accurately fit the mechanical response of braced structure under cyclic static test,which proves that the proposed algorithm can be used in practical engineering.