Inverse Method Of Material Characteristic Parameters Based On Digital Image Correlation And Manifold Learning

With the rapid development of science and technology,modern engineering puts forward higher and more stringent requirements for material properties.At present,the research material properties mainly by experiment means,however,traditional measurement technology such as strain electrical measuring method is less number of measurement point,affected by environmental factors such as insufficient,can not accurately obtain full displacement change,on the surface of the test specimen makes traditional testing with single point or limited point is easy to appear when measurement information for material properties identification "mysterious material" phenomenon.Digital image correlation technique,as a common non-contact measurement method of surface deformation,has been widely used in the field of experimental mechanics.Based on the speckle deformation information of the gauge surface obtained by the digital image acquisition system,the full-field displacement was obtained by the correlation calculation,and the full-field strain was obtained by the assumption of the continuity of the deformed body.Digital image correlation can provide sufficient measurement information for material characterization,but for parameter inversion,full field measurement information as inverse input is too redundant,which will lead to inverse problem and "dimension curse" and other phenomena.How to make full use of the whole field measurement information while obtaining the lowest input information dimension is a difficult point in the current research.On the other hand,under the loading conditions of different strain rates,the mechanical properties of the same material have obvious differences.Therefore,it is of great theoretical value and engineering significance to study how to reduce dimension of digital image correlation measurement information as parameter inverse input,predict and evaluate material parameters under different loading conditions,and determine accurate material model.This paper carries out research on the above difficulties,and the main work is as follows:(1)Optimal points layout method based on digital image correlation and orthogonal matching greedy algorithm.Selected specimen gage section of the center point as the first point,all the displacement response of orthogonalization for its characteristic vector,and with a characteristic vector direction for the new coordinate system,will be full of each measuring point response to the characteristics of the projection vector coordinate system,based on the principle of greedy algorithm iterative optimal combination of measuring points,determine the materials used for parameter reverse surface of the specimen gage section of the best position and number of points.The process of eliminating redundant sample information and normalizing relevant information is completed,and the dimensionality reduction of the input end of the inverse parameter of digital image is realized,so as to improve the solving efficiency of the inverse problem of material parameter identification and avoid the discomfort of the inverse process.(2)An inverse method of material characteristic parameters based on load-response curve manifold learning is proposed.Carried out under different strain rate loading conditions of material tensile test,based on digital image correlation material for full static/dynamic load-response information,can assume that all the material characteristic parameters combination by simulating the load-response curves are smooth manifold of projection in a lower dimensional space,and is used to identify the material parameters in the low dimensional space.Based on principal component analysis and manifold learning method,the finite element model to output response build manifold space,get response experiment in manifold projection coordinates in space,by using the polynomial agent model building parameter space and load-the mapping relationship between the response curve of the manifold space,using a genetic algorithm for optimization and optimizing objective function iteration inverse problem parameters interval,Under the conditions of convergence and loading at different strain rates,the material characteristic parameters obtained by reverse calculation are optimal.In this paper,the local linearization of the nonlinear inverse problem in the process of performance identification solves the difficulty in establishing the spatial mapping relationship between the multi-dimensional response of the test measurement and the parameters to be identified.The high nonlinearity of the inverse problem is overcome,and the problems such as the unsuitable nature of the inverse problem are avoided,thus improving the accuracy and stability of the identification of the dynamic/static properties of materials.(3)An inverse method of material characteristic parameters based on characteristic morphology manifold learning in tensile tests is proposed.The change history of the characteristic morphology on the surface of the gauge section of the specimen in the tensile test was obtained by the image acquisition system,and the time history range of the characteristic morphology was determined by the method of finite element simulation and experimental comparison.The optimal time and quantity of the characteristic morphology were obtained by using the full-field optimal layout method,and the node coordinate information corresponding to the experiment was extracted to form the sample space of the morphology.The morphology sample space was transformed into a manifold space by principal component analysis and manifold learning,and the experimental projection coordinates of the characteristic morphology in the manifold space were obtained.The difference between the experimental projection coordinates and the morphologic manifold coordinates of the simulation model was calculated as the inverse objective function,and the mapping relationship between the parameter space and the morphologic manifold space was constructed by using the proxy model.The genetic algorithm was optimized to realize the accurate inversion of material parameters.It solves the problem that it is difficult to establish the spatial correspondence between the multi-dimensional morphological characteristics measured in the experiment and the parameters to be identified,and further improves the accuracy and stability of material property identification.