Research On The Prediction Of Fluidity And Shrinkage Of Aluminum Alloy Based On BP Neural Network

In this study,BP neural network is employed to construct models to predict fluidity and shrinkage of casting aluminium alloys.To make it easier to use,softwares have been developed based on the prediction models.Then,the prediction software has been used to predict fluidity and macroscopic shrinkage of binary and multicomponent casting aluminum alloys separately.In the process of constructing and optimizing of the models,the average MSE,average R and average AARE are all used to evaluate the performance of BP neural network models.And a method of permutation and combination is applied to study the effects of training algorithms and transfer functions.Results show that: for both the prediction model of fluidity and prediction model of shrinkage of casting aluminum alloys,(1)only considering the effect of training algorithms,the performance of BP neural network model using training algorithm of “trainbr” or “trainlm” are superior to that using training algorithm of “traingd”,"traincgf" or "trainoss";(2)only considering the effect of transfer functions,the performance of BP neural network model with transfer function combination of “TF1 = 'tansig' + TF2 = 'purelin'” or “'TF1 = 'logsig' + TF2 = 'purelin'” are better than that with transfer function combination of “TF1 = 'tansig' + TF2 = 'logsig'” or “'TF1 = 'logsig' + TF2 = 'logsig'”.Based on BP neural network and training dataset of fluidity of casting aluminum alloys,a prediction model with a structure of 8-9-1 has been constructed to predict the fluidity of casting aluminum alloys.The inputs of the model are contents of Al,Si,Fe,Cu,Mn,Mg,Zn and pouring temperature,and the output is fluidity of casting aluminum alloys.The transfer function of input layer to hidden layer of the model is “tansig”,the transfer function of hidden layer to output layer is “purelin”,and the training algorithm is “trainbr”.The test dataset of fluidity of casting aluminum alloys was used to check the accuracy of the model.Results show that this model can well predict the fluidity of casting aluminum alloys,with a maximum error of 11.81% and an average error of 6.56%.Also,based on the prediction model of fluidity of casting aluminum alloys,the fluidity prediction software has been developed and has been successfully used to predict the fluidity of binary and multicomponent casting aluminum alloys.A BP neural network model with a structure of 7-46-1 was constructed to predict the shrinkage of casting aluminum alloys,based on the training dataset of shrinkage of casting aluminum alloys.This model establishs the relationship between contents of Al,Si,Fe,Cu,Mn,Mg,Zn and macroscopic shrinkage of casting aluminum alloys.In this model,the transfer function of input layer to hidden layer is “logsig”,the transfer function of hidden layer to output layer is “purelin”,and the training algorithm is “trainlm”.Compared the prediction data using this model with the test dataset of shrinkage of casting aluminum alloys,it is thought that this model can well predict the macroscopic shrinkage of casting aluminum alloys,with a maximum prediction error of 7.82% and an average prediction error of 3.67%.The develpoed shrinkage prediction software,based on the prediction model of shrinkage of casting aluminum alloys,is efficient and accurate in predicting the macroscopic shrinkage of binary and multicomponent casting aluminum alloys.