Application Research Of SVM Algorithm In Adaptive Random Experiment Design And PBX Stress-Strain Relation Law Mining

Polymer bonded explosive(PBX)is a high-energy material widely used in the field of weapons and ammunition,with relatively high detonation performance and low sensitivity.In the process of production,transportation and storage,accidental deformation or destruction of structural parts occurs from time to time,by the external stimuli.These cases will affect the service safety and reliability of the weapon system.Therefore,it is of great significance to study the deformation and destruction laws of PBX.In this research,in the research of PBX deformation and failure law,the experimental design and experimental data law mining two existing problems in two levels,carried out related research.First,the traditional experimental design method can not determine the number of tests to be carried out under the condition,that the experimental law is unknown,and the output model can not obey a certain quantitative standards.Second,because PBX has obvious tensile and compressive asymmetry and temperature correlation,the tension-compression stress-strain relationship is difficult to describe.The article proposes an adaptive random experiment design(ARED)algorithm.The algorithm takes real-time model error as an adaptive condition,and based on the automatic selection of test cases,outputs a model that conforms to the error quantization standard.According to the actual experimental scenario,the similar number of test cases were selected between the ARED method and the comparative experimental design method under the bimodal Gaussian function,the bimodal surface function and the peaks function,respectively.simultaneously,the support vector machine(SVM)algorithm is used to construct the model for the selected test cases,and the verification surface(or curve)is predicted.The qualitative and quantitative analysis is carried out at two-slice of applicability and precision.The results show that the ARED method can be applied to the experiment of multi-factor,and has better precision and applicability than the comparative experimental methods.The tensile-compression stress-strain relationship of PBX under uniaxial quasi-static is the basic work established by the constitutive model,and it is of great significance for the corresponding prediction of structural mechanics.This paper based on the uniaxial tension-compression stress-strain test data at 17 temperature points of TATB-based PBX explosives,a Support Vector Machine algorithm is proposed to construct a constitutive model of the material.A SVM constitutive model that can simultaneously describe the mechanical behavior of all temperature points is established by analyzing the influencing factors of the constitutive behavior of PBX materials,taking temperature and strain as independent variables and stress as the dependent variable,the optimal parameters are searched using grid search method and cross validation method.Comparing with the experimental data,the maximum mean absolute percentage error of the SVM constitutive model is 7.768%,the maximum value of the average absolute error is 0.8476 MPa,and the correlation coefficient of the prediction is all above 0.9987,which provides new ideas for the study of the mechanical properties of similar materials.To construct a tensile and compressive constitutive equation for the tensile and compressive asymmetry materials of polymer-bonded explosives,and to study its long-term service performance,the applicability of support vector machine algorithm in the construction of constitutive model is explored.Uniaxial tensile and compressive creep tests are carried out on a TATB-based PBX material with three different stress levels at 20 ~oC,35~oC,and 50 ~oC.Then constructed the train data,selected the kernel function,and the SVM constitutive model is established.Furthermore,the model is analyzed by the correlation coefficient,and a new model description method is proposed.The results show that under a set of parameters,the minimum correlation coefficient is 0.99903.The model curves and the error band are in good agreement with the experimental results,which has engineering applicability.