摘要
基于随机度和分治思想,提出并实现了前向反馈神经网络中的循环自构形算法,该算法能有效解决自构形算法的收敛不一致问题,将网络收敛到一致的最精简的结构。通过MATLAB实验表明,循环自构形算法可将网络隐层节点数从15和20一致精简到6。将该算法应用于风机的故障诊断中,实践表明该方法具有更有效的故障知识表达性能,能更好地解决故障诊断问题。
Based on the randomness and divide algorithm,the circular self-configuring algorithm of BP neural network is introduced in this paper.The algorithm is realized in the experiment.It shows this algorithm can converge uniformly and reduce the neural network to most simplify.The experiment of application in the centrifugal fan fault diagnosis underlines the theoretical resuits.
出处
《计算机工程与应用》
CSCD
北大核心
2007年第24期220-222,229,共4页
Computer Engineering and Applications
基金
江西省科技公关计划(the Key Technologies R&D Program of Jiangxi Province
China under Grant No.20041B100100)
关键词
循环自构形神经网络故障诊断风机
circular self-configuring
Neural Network
fault diagnosis
centrifugal fan
