Research On Prediction Of Sheet Metal Extrusion Forming Force Based On Deep Learning

Sheet metal extrusion is a typical process in sheet bulk metal forming,having a wide range of applications.Forming force,as an important parameter in the forming process of sheet metal extrusion,is an basis for corresponding die design and press selection,its prediction research has great engineering significance.Many force calculation models are based on mechanical analysis and forming mechanism to express the complicated non-linear relationship between the forming force and many parameters.But most material model and process model are appropriately simplified,resultingly,these calculation models’ applicable scope and calculation accuracy are limited.With the rapid development of artificial intelligence,deep learning can directly mine the relationship between input and output from a large number of data,and has definite advantages in building complex nonlinear models.Therefore,based on deep learning technology,this thesis has carried out research on the prediction model of sheet metal extrusion forming force.The main research contents and results are as follows:(1)Research for batch data acquisition method based on sheet extrusion experiment and numerical simulation was carried out.The reliability of the finite element model was verified by comparing the forming force results of sheet metal extrusion simulations with corresponding experiments.Based on this,the finite element model template update technology was used to obtain the data of different process conditions in batches,thereby laying the foundation for the establishment of prediction model.(2)Combined with different material harden models,auto-encoder was used to establish material parameter compression model,which compressed the flow stress-strain curves of different materials into 4 material characteristic values.In this way,the input of material parameters in the prediction model was simplified,and applicable scope of the model was enlarged.Furthermore,compared with the prediction model without material parameter compression,the training samples can be effectively reduced and the prediction accuracy can be improved.(3)Based on deep neural network,the prediction model,which input variables were radius of die,reduction in area,stroke,sheet thickness,friction coefficient,and material characteristics,was established.The prediction model was optimized by adjusting the number of neurons and the number of layers,the results showed that the prediction accuracy increase rapidly with more neurons and layers.Under the optimal parameters,the prediction model had a good generalization ability.(4)Aiming at the influence of forming section shape on forming force,research on forming force prediction model for different shapes was carried out based on convolutional neural network.Through constructing forming force dataset of basic shapes,and using convolutional neural network to identify the features of shape,prediction model for different forming section shape was established.Through verification and evaluation,the prediction model had high prediction accuracy for forming force of basic shapes and combined shapes.