基于人工神经网络的BaTiO_3陶瓷配方研究  

INVESTIGATION OF BaTiO_3 FORMULATION THROUGH ARTIFICIAL NEURAL NETWORK TECHNIQUE

作　　者：郭栋[1] 王永力[1] 夏军涛[2] 李龙土[1] 桂治轮[1] 

机构地区：[1]清华大学材料科学与工程系新型陶瓷与精细工艺国家重点实验室,北京100084 [2]北京理工大学化工与材料学院,北京100081

出　　处：《硅酸盐学报》2002年第3期329-334,共6页Journal of The Chinese Ceramic Society

基　　金：国家自然科学基金 (No .5 9995 5 2 3 ) .

摘　　要：人工神经网络具有巨量并行、结构可变、高度非线性等特点 ,其建立数学模型并不需要预先知道太多有关问题背景的知识 ,这尤其适用于陶瓷配方研究中某些机理尚未完全清楚、传统数学方法无法分析的情况 .本工作将人工神经网络技术用于介电陶瓷的配方性能分析 ,以BaTiO3为研究对象选取了几种掺杂剂 ,在均匀实验设计的基础上 ,用BP人工神经网络对所得实验结果进行了分析 ,并且用图形化方式直观地表达了出来 .根据实验结果 ,并与多重非线形回归模型相比发现 ,人工神经网络模型比多重非线形回归模型更加准确且能给出配方组成与性能更丰富的信息 。Artificial neural network (ANN) technique is endowed with certain unique attributes such as the capability of universal approximation, the ability to learn from and to adapt to its environment and the ability to invoke weak assumptions about the underlying physical phenomenon responsible for generation of the input data. ANN seems to be of most interest, in particular, when a solid theoretical basis or mathematical relationship is not available in advance. In this study ANN technique is used to model the dielectric properties of BaTiO 3 based system. Based on the homogenous experimental design, the experimental results of 21 samples were analyzed by a three_layer BP network modeling. The results were also expressed by intuitive graphics. Comparison of the results from ANN model and multi_nonlinear regression (MNLR) model indicates that the BP network is a very useful tool in dealing with problems with serious non_linearity encountered in the formulation design of dielectric ceramics.

关 键 词：BATIO3陶瓷 配方 钛酸钡 介电性能 人工神经网络 多层反问误差传播算法 

分 类 号：TQ174.758[化学工程—陶瓷工业] TP183[化学工程—硅酸盐工业]