(上海交通大學(xué)國家模具工程研究中心，上海 200030)
摘 要: 開發(fā)了一個基于神經(jīng)網(wǎng)絡(luò)的Ti-17合金的本構(gòu)關(guān)系模型。首先利用Thermecmastor-Z型熱模擬機等溫壓縮Ti-17合金，研究在不同變形溫度、變形程度和應(yīng)變速率等工藝參數(shù)條件下流動應(yīng)力的變化情況。然后用實驗所得的熱變形工藝參數(shù)與性能間的數(shù)據(jù)訓(xùn)練人工神經(jīng)網(wǎng)絡(luò)。訓(xùn)練結(jié)束后的神經(jīng)網(wǎng)絡(luò)變成為一個知識基的本構(gòu)關(guān)系模型。利用該模型預(yù)測的流動應(yīng)力的值與實驗結(jié)果間的誤差較小。
關(guān)鍵字: 人工神經(jīng)網(wǎng)絡(luò)；Ti-17合金；本構(gòu)關(guān)系；BP算法
（National Die and Mold CAD Engineering Research Center,
Shanghai Jiao Tong University, Shanghai 200030, P. R. China）
Abstract: Artificial neural networks have been applied to acquire the constitutive relationships of a Ti-17 alloy at elevated temperature, using data obtained from homogeneous compression experiments carried out on a Thermecmastor-Z hot simulator. During building up the neural network model of the constitutive relationship for the alloy, deformation temperature, equivalent strain rate and equivalent strain were taken as the inputs and flow stress was taken as the output. At the same time, four layers were constructed, twelve neurons were used in the first hidden layer and eight neurons were used in the second hidden layer. The activation function in the output layer of the model obeyed a linear function, while the activation function in the hidden layer was a sigmoid function. The neural network became stable after 31530 repetitions in training . Comparison of the predicted and experimental results shows that the neural network model used to predict the constitutive relationship of the Ti-17 alloy has good learning precision and good generalization. Meanwhile, the neural network methods are found to show much better agreement than the statistical regression methods in dealing with the experimental data.
Key words: artificial neural network; Ti-17 alloy; constitutive relationship; BP algorithm
