(吉林大學(xué) 南嶺校區(qū) 材料科學(xué)與工程學(xué)院，長春 130025)
摘 要: 采用人工神經(jīng)網(wǎng)絡(luò)研究了在不同型溫、澆溫和轉(zhuǎn)速條件下以離心法制備Al16%Si FGM時(shí)初晶硅的分布規(guī)律，并通過實(shí)驗(yàn)進(jìn)行了驗(yàn)證。在建立神經(jīng)網(wǎng)絡(luò)模型時(shí)，以型溫、澆溫、轉(zhuǎn)速等工藝參數(shù)作為人工神經(jīng)網(wǎng)絡(luò)的輸入，以內(nèi)生初晶硅分布的相對厚度作為輸出。實(shí)驗(yàn)表明，預(yù)測結(jié)果與實(shí)際測定結(jié)果比較吻合，說明采用神經(jīng)網(wǎng)絡(luò)預(yù)測離心法制備梯度功能材料中內(nèi)生顆粒的分布是可行的。
關(guān)鍵字: 離心法； 梯度功能材料； 人工神經(jīng)網(wǎng)絡(luò)；內(nèi)生顆粒；顆粒分布
（College of Materials Science and Engineering,
Nanling Campus, Jilin University, Changchun 130025， P.R.China）
Abstract: Artificial neural network has been applied to acquire the constitutive relationships of endogenetic particle distribution in FGM prepared by centrifugal casting at different mould temperature, pouring temperature and rotating speed. Building up the neural network model of the constitutive relationship for the alloy, mould temperature, pouring temperature and rotating speed are taken as the inputs and relative thickness of endogenetic particle distribution in FGM is taken as the output. At the same time, four layers are constructed, six neurons are used in the first hidden layer and four neurons are used in the second hidden layer. The activation function in the output layer of the model obeys a linear function, while the activation function in the hidden layer is a sigmoid function. Comparison of the predicted and experimental results shows that the neural network model used to predict the constitutive relationship of the endogenetic particle distribution in FGM has good learning precision and good generalization. It's available to forecast endogenetic particle distribution in FGM prepared by centrifugal casting based on artificial neural network.
Key words: centrifugal casting； functionally gradient material； artificial neural network； endogenetic particle； particle distribution
