基于ART1神经网络的设计检索  

Design retrieval based on ART1 Neural networks

作　　者：王玉[1] 邢渊[1] 阮雪榆[1] 

机构地区：[1]上海交通大学国家模具CAD工程研究中心,上海200030

出　　处：《机械设计与研究》2000年第4期27-28,共2页Machine Design And Research

摘　　要：深入研究了基于ART1神经网络的设计检索 ,并对ART1网络的算法进行了改进。改进后的算法可大大减少不必要的零运算 ,提高网络运行效率。与基于BP、Hopfield设计检索系统相比 ,ART1网络具有较大的优越性。主要体现在设计师可根据自己的意图 ,通过调节警戒参数得到所希望的检索结果。文中给出了齿轮类锻件的设计检索实例。This paper studied the design retrieval system based on ART1 Neural networks in detail,and made some improvements for the ART1 algorithm.The modified algorithm can reduce unnecessary computations and enhace the networks operation efficiency.Compared with the BP and Hopfield based design retrieval system,the ART1 networks possess a greater advantage,that is,by adjusting the vigilance parameter,the designer can,according to his/her design intent,get the desired retrieval results.A gear like forging design retrieval case was presented in this paper to demonstrate the systemfeasibility and efficiency.

关 键 词：设计检索 ART1神经网络 齿轮类锻件 

分 类 号：TH132.41[机械工程—机械制造及自动化] TP183[自动化与计算机技术—控制理论与控制工程]