Prediction Of Failure Behavior Of Composite Stiffened Panels Using Artificial Neural Network

Composite hat-stiffened panels are typical composite structures that embody the concept of high strength and light weight.These panels have several design parameters and multiple buckling failure modes.Consequently,the accuracy of predicting their failure behavior is low,particularly when the traditional simplified method of engineering calculation is employed.Although the finite element method can achieve high-precision predictions,the process is timeconsuming for design engineers.In this study,the mechanical behavior of the composite hatstiffened panel is examined by combining finite element(FE)simulation and experimental verification,and a fast method for the prediction of structural bearing capacity by using artificial neural network(ANN)is provided.The main works of this paper are as follows.(1)Two kinds of ANNs are established to predict the compression buckling behavior of composite hat-stiffened panels.First,the compression buckling behavior of the composite hatstiffened panel is studied using the combination of FE simulation and experimental verification.Then,considering the variations of four mechanical properties of the stiffened panels,the FE model set is used to generate the training dataset and testing dataset of ANN in batches.Two kinds of ANNs are comparatively selected to predict the buckling load and buckling failure mode.Based on the testing dataset and the new dataset,the performance and generalization ability of the artificial neural networks are examined.The results show that the compression buckling behavior of the composite hat-stiffened panel can be effectively revealed by numerical and experimental investigation,and the trained ANN can accurately and efficiently predict the buckling behavior of composite hat-stiffened panels under axial compression.(2)An ANN is established to predict the failure behavior of composite hat-stiffened panels under in-plane shear and this proposed ANN successfully predicts the buckling and ultimate loads.First,the failure behavior of a composite hat-stiffened panel under in-plane shear are examined by combining FE simulation and experimental verification.Then,based on the parametric modeling technology afforded by the secondary development of ABAQUS,the FE model set is generated via batch calculation;thereafter,it is randomly divided into training,validation,and testing datasets of the ANN.Finally,an autoencoder is employed to compress the original characteristics,and a back propagation neural network is established to predict the buckling and ultimate loads.The performance and generalization ability of the ANN are examined based on the testing set.The results show that the buckling evolution process and failure mechanism of the composite hat-stiffened panel under in-plane shear can be effectively revealed by numerical and experimental investigation,and the trained ANN can accurately and efficiently predict the buckling and ultimate loads of composite hat-stiffened panels under inplane shear.