Prediction Of Mechanical Properties Of Carbon Fiber Composites Based On Deep Material Network

The excessive consumption of energy and the continuous destruction of the environment are huge challenges that human development must face.Therefore,the new energy industry and lightweight technology are the mainstream trends in the current development of the automotive industry.Lightweight technology plays a vital role in reducing the quality of vehicle maintenance,reducing fuel consumption and reducing exhaust emissions.As a leader in the field of emerging materials,composite materials are one of the best ways to achieve lightweight technology,and carbon fibers make carbon fiber reinforced composites a leader among lightweight materials due to their light weight,high strength and hardness,and large specific modulus.It is more applied to the lightweight design of automobiles.During the design of the automobile structure,different batches of carbon fiber composite materials will be made into laminates with different layers and different layup sequences.A large number of composite material mechanical performance experiments must be performed on all combinations to obtain their basic properties.Parameters,which bring high time and material costs to the design and performance analysis of the entire vehicle.With the development of artificial neural network and deep learning technology,it becomes possible to apply deep learning technology to the prediction of mechanical properties of composite materials to save the cost of material development and application.Convolutional neural networks(CNN)can process complex images and extract feature information from the graphs.Long-term and short-term memory networks(LSTMs)are good at processing data with sequence features.Based on the deep material network,a CNN-LSTM neural network model is proposed to predict the mechanical properties of carbon fiber composites.Training of deep learning models requires a large amount of data,and there is currently no open source dataset of mechanical properties of carbon fiber composites.Therefore,the Digimat platform was first used to obtain the homogenized flexible matrix of carbon fiber composites,and it was converted into picture information.It was processed into 1200 groupsof data samples with sequence characteristics and corresponding label files were made.The CNN-LSTM neural network model is trained and the accuracy of the network model is tested using the produced data set.The results show that the accuracy of the CNN-LSTM mechanical property prediction of carbon fiber composite materials reaches 0.976.It is proved that deep learning is feasible for predicting the mechanical properties of carbon fiber composites.At the same time,the tensile mechanical properties experiments of carbon fiber composite laminates were carried out in this paper to obtain the mechanical parameters of the samples,and the corresponding input samples were constructed,and the CNN-LSTM model was used for prediction.Then compare the experimental results with the model prediction results.The results show that the prediction results of the mechanical properties of carbon fiber composites by CNN-LSTM are within 5% of the corresponding experimental results of tensile mechanics.Accuracy of material mechanical property prediction.