Study On Compression Mechanical Properties Of Mesoscopic Metal Foam Based On Deep Learning And Finite Element Methods

As a type of porous material,metal foam possesses exceptional mechanical properties,including high specific strength and specific stiffness,surpassing those of traditional materials.Its application extends across diverse engineering domains such as aerospace,transportation,and construction.Metal foam exhibits a prolonged plateau stage when subjected to compression.Owing to its lightweight composition and remarkable energy-absorbing cushioning traits,metal foam holds tremendous promise in applications related to automotive collision safety and lightweight design.Metal foam exhibits a coupled material structure at the macro-meso level,featuring intricate microstructural characteristics at the mesoscopic scale.The diverse meso-structures of metal foams contribute significantly to variations in their macro-mechanical properties.While metal foams offer excellent design flexibility,this complexity cell poses challenges in accurately predicting their macro-mechanical behavior.Investigating the correlation between the meso-structure and macro-mechanical properties of metal foam is essential for a profound understanding of the meso-structural impact on macro-mechanical properties.This understanding can enhance the design efficiency and engineering applications of metal foam structures.Against this backdrop,this study conducts mechanism research on the mechanical properties and structural deformation of metal foam based on the finite element method at the meso scale,and accurately establishes the correlation between the microstructure and mechanical properties of metal foam based on the deep learning method.This study provides new technologies and methods for the rapid analysis of the performance and the design on demand of the foam metal energy absorbing structure in auto body,thus improving the lightweight design level of the auto body structure.The primary research objectives are as follows:Based on the Voronoi model and image recognition algorithm,multi-dimensional mesoscopic models for metal foam were established.A random cell wall thickness solid model generation algorithm was proposed on the basis of the Voronoi model,allowing for the rapid establishment of a three-dimensional metal foam meso-model driven by meso-structure parameters.The surface cell pore structures of the actual metal foam were scanned,and the cell structure boundaries of metal foam were extracted using image recognition technology,enabling the swift creation of a two-dimensional metal foam model.Quasi-static compression tests and simulations were conducted on metal foam.The structural deformation and compressive stress-strain response of the metal foam in the test and simulation results were compared,demonstrating the effectiveness of the established metal foam model.The established mesoscopic metal foam model now serves as the foundation for subsequent research on mechanical properties.The mesoscopic dynamic compression mechanical properties and strain rate effect of metal foam were investigated.Dynamic compression finite element simulations were conducted within a strain rate range of 0 s?1 to 4000 s?1,considering the strain rate effect of both the metal foam material matrix and the gas within cell pores.This aimed to explore the dynamic compression behavior,mechanical properties,and energy absorption characteristics of mesoscopic metal foam.The study delved into the flow pattern of gas within the cells,the gas pressure distribution in the cell pores during the compression process,and the impact of gas within the cells on the deformation of the metal foam structure based on the simulation results.The investigation also sought to understand the generation mechanism of the strain rate effect in metal foam.The findings revealed a noticeable strain rate sensitivity in metal foam under compressive load.In the late plateau and densification stage of dynamic compression,the gas stress enhancement effect was significant.Under high strain rates,the gas within the cell pores contributed to the densification of the metal foam.The interaction between gas and cell structure amplified the deformation and stress of the local cell structure,ultimately enhancing the overall load-bearing capacity of the metal foam structure,which was identified as the primary reason for the significant gas stress enhancement effect during the densification stage of metal foam under high strain rates.To account for the strain rate effect,modifications were made to the dynamic compression mechanical properties of metal foam.A dynamic stress prediction equation for metal foam,considering its microstructure,was established.This equation incorporated high-order strain terms and the reinforcing effect of gas on the cell structure.Additionally,a gas stress enhancement prediction equation,including a structural stress enhancement term,was formulated.Simultaneously considering the stress enhancement in both the matrix material and the gas within the cell pores,a dynamic stress enhancement prediction equation for metal foam was developed to effectively predict its dynamic stress-strain response.The influence of microstructural parameters on the mechanical properties of metal foam was studied.The macro/meso-structural parameters of metal foam were correlated,and a parametric modeling method for a specific porosity metal foam model was proposed.Mesoscopic metal foam models with cell pore diameters between 3 mm and 4 mm under three porosity ratios of 90%,80% and 70% were established,and quasi-static and dynamic compression simulation were conducted.The study found that under the same porosity,as the pore size of the metal foam decreases,the stress in each stage of the stress-strain curve of the metal foam under quasi-static and dynamic compression increases,but the maximum energy absorption efficiency decreases and the energy absorption capacity weakens.A prediction equation for the compression mechanical properties of metal foam taking into account the microstructural parameters was established.The prediction equation can better represent the uniaxial compressive stress-strain relationship of metal foam under different porosity and cell pore diameters,and more accurately predict the quasi-compression mechanical behavior of metal foam.By introducing data-driven artificial intelligence algorithms,a deep learning prediction framework for the mechanical properties of metal foams was established.The point cloud and voxelization methods were proposed,a three-dimensional metal foam deep learning dataset was established,and a three-dimensional convolutional neural network(3D-CNN)was used to identify and extract the unique three-dimensional meso-structure of metal foam.In addition,a deep learning framework for metal foam mechanical property prediction and reverse structural design has been established to achieve two-way high-precision mapping of metal foam’s "meso-structure and mechanical properties".A two-dimensional metal foam deep learning dataset was established,and the unique meso-structural characteristics of metal foam were identified and extracted through a two-dimensional convolutional neural network(2DCNN)to achieve efficient and accurate prediction of the compressive mechanical properties of metal foam.Conditional generative adversarial network(CGAN)realizes the inverse structural design process from mechanical properties to microstructure.The results show that the established deep learning models have good performance in identifying the microstructure characteristics of metal foam,predicting mechanical properties,and designing structures.