| Plant fiber-reinforced resin-based composite materials have high specific strength and modulus,are environmentally friendly,and renewable.They have great application prospects in fields such as automobiles,construction,and packaging.However,the discreteness of the structural parameters of plant fiber-reinforced resin-based composite material components,the complexity of the mesoscopic geometric structure,and the vast structural parameter space have led to an unclear relationship between the composite material structure and mechanical properties.This has resulted in a lack of guidance for the preparation and optimization of composite materials,limiting material optimization and applications.This paper studies bamboo fiber-reinforced palm oil-based resin composite materials using a combination of numerical simulation,deep learning,and integrated learning to achieve rapid prediction of mesoscopic mechanical mechanisms,stress distribution,and mechanical property prediction under multi-factor coupling.To address the unclear relationship between the complex mesoscopic structure and mechanical properties of bamboo fiber-reinforced palm oilbased resin composite materials,a secondary development module based on PYTHON’s ABAQUS is constructed.This module achieves parameterized automatic modeling of the representative volume cell model of the composite material and finite element simulation of its stress and strain distribution.By comparing experimental results and finite element simulation results,the reliability of the established composite material mesoscopic mechanical finite element model is verified.Furthermore,the finite element simulation system is used to study the effects of bamboo fiber length,diameter,proportion,and parameter coupling relationships on material mechanical properties.The results show that the mechanical properties of the resin matrix,interface bonding type,and volume fraction of bamboo fibers have a significant impact on the mechanical properties of the composite material.In response to the significant time and economic costs associated with analyzing material properties due to the vast combination of structural parameters in the composite material,a deep learning method based on the mesoscopic geometric images of the composite material is constructed to quickly predict its stress distribution.The deep learning model is trained using a dataset of 832 two-dimensional stress distribution finite element simulation results of composite materials.The results show that the constructed model can extract and recognize geometric features of different constituent phases,and can predict the nonlinear changes in stress distribution of materials under different loading conditions through the composite material strain rate.The correlation coefficient(model fitting effect)on the validation set is approximately 0.92.Compared to the finite element analysis method,this model can obtain the stress distribution results of composite materials in two-dimensional structures along the fiber direction,two-dimensional structures perpendicular to the fiber direction,and three-dimensional structures in just a few seconds,offering higher efficiency,practicality,and accuracy.In response to the difficulty of predicting and analyzing the mechanical properties of composite materials,which are influenced by multiple factors and have a large parameter space,an integrated machine learning model is constructed based on composite material structural parameters to quickly predict tensile strength(random forests,gradient boosted decision trees,extreme gradient boosting,and category enhancement).The integrated machine learning model is trained using a dataset of 82 finite element simulation results of representative unit cells’ tensile strength and optimized using cross-validation and the Hyperopt tool.The results show that the gradient boosted decision tree model performs the best,with a correlation coefficient of 0.791 on the validation set.Feature importance analysis indicates that the resin type has the most significant impact on the mechanical properties of composite materials,followed by the volume fraction of bamboo fibers.The integrated learning method can quickly predict and analyze the main factors of composite material mechanical properties based on a relatively small amount of numerical simulation data.This paper explores the finite element analysis of the mechanical properties of bamboo fiber-reinforced palm oil-based composite materials,deep learning based on geometric morphology for stress distribution prediction,and integrated machine learning based on composite material structural parameters for mechanical property prediction.This research reveals the relationship between the mesoscopic structure and mechanical properties of the material,providing theoretical value for the study of the mechanical mechanisms of such composite materials and application value for the development and optimization of the materials. |
