基于BP神经网络的复合材料性能预测  

Prediction of Composite Material Properties Based on BP Algorithm of Artificial Neutral Network

作　　者：尹海莲[1] 胡自力[1] 

机构地区：[1]南京航空航天大学无人机研究院,南京210016

出　　处：《南京航空航天大学学报》2006年第2期234-238,共5页Journal of Nanjing University of Aeronautics & Astronautics

基　　金：航空科学基金(05G52054)资助项目

摘　　要：针对复合材料性能表征十分复杂、困难的情况,利用人工神经网络的BP算法,建立了复合材料性能预测模型。模型由3层神经元组成,分别为输入层、隐含层和输出层。以碳/陶瓷复合材料性能与成分的关系为研究对象,选取了38组实验数据作为学习样本,模型总误差为0.18,用建立的网络预测未知,并给出预报曲线。和试验值相比较表明,所建立的网络能反映碳/陶瓷复合材料组分与其材料性能之间的关系,为实验设计提供了新的思路,节省了时间和劳力。A model is established to predict the relation between components and properties for the composite material based on the back propagation （BP） algorithm of the artificial neural network （ANN）. The model is composed of three neuron layers： input layer, hidden layer and output layer to simulate the real structure of the human brain. The relation of properties and components for carbon-ceramics composite is studied. 38 samples are used as study data and total error of the model is 0. 18. Then the unknown data can be predicted by the model and the predicting curves are drawn. Compared with the experiment, the model can reflect the relation between components and properties for carbon-ceramics composite. It provides a new route for test-design, thus saving the time and labor.

关 键 词：BP神经网络 学习样本 材料性能 复合材料 

分 类 号：TB302[一般工业技术—材料科学与工程]