基于PCA-GA神经网络模式识别的炭纤维复合材料导电综合性能优化及预测的研究  

Optimization and prediction research on carbon fiber composite integrated conductive performance based on PCA-GANN pattern recognition

作　　者：杨榛[1,2] 顾幸生[2] 梁晓怿[1] 张睿[1] 凌立成[1] 

机构地区：[1]华东理工大学化工学院,上海200237 [2]华东理工大学信息科学与工程学院,上海200237

出　　处：《计算机与应用化学》2008年第12期1543-1548,共6页Computers and Applied Chemistry

基　　金：国家自然科学基金(50672025);上海市科技登山计划(065258033;06JC1401)

摘　　要：在炭纤维/ABS树脂基复合材料导电性和拉伸强度实验数据的基础上,利用主成分分析-遗传算法后向传播(Principal Component Analysis-Genetic Algorithm Back Propagation,PCA-GABP)神经网络模式识别法,对炭纤维复合材料导电综合性参数进行智能分析识别。其中主成分分析法作为前处理过程优化样本集的选择,GABP神经网络较好地克服了BP网络非线性映射易陷入局部极小值问题。给出性能优良的目标参数优化区,并用实验予以验证。该模式识别方法可减少实验工作量,提高工作效率,在复合材料设计领域具有理论应用前景。On the basis of experimental data about carbon fiber/ABS resin matrix composites conductivity and tensile strength, PCA- GABP （Principal Component Analysis-Genetic Algorithm Back Propagation Principal） neural network method for pattern recognition was used for material design. To optimize the impact factors of sample sets, principal component analysis was used as a pre-treatment process. GABPNN could effectively overcome shortcoming of BPNN such as easily leading into local minimum. The method was used to recognize and analysis integrated conductive parameters of carbon fiber composites. Then, the optimized district of performance parameters was given. The predicted results agreed well with the experimental ones. Therefore, PCA-GABPNN pattern recognition prospect theoretical design method for composites design.

关 键 词：主成分分析 GABP神经网络 模式识别 炭纤维复合材料导电综合性能 

分 类 号：TB332[一般工业技术—材料科学与工程]