摘要
本文提出基于神经网络可修复非冗余数字系统可靠性分析新的算法。算法利用前馈递归神经网络实现系统的离散Markov模型。网络权重的能量函数及校正等式的建立使网络能够收敛于设计系统所需的可靠性,并且满足系统所需可靠性的故障率及修复率,可从收敛处的神经元权重提取。文中已经给出实验结果并用马尔可夫连续解加以验证。
This paper proposes a new algorithm which can implement the reliability analysis of nonredundant digital system with repair based on neural network. The algorithm can determine the continuous time solutions of the Markow models using a feedforward, recursive neural network. Energy functions and updated equations of the weights make neural networks converge to the desired reliability of systems, The failure and repair rates of the systems can meet the desired reliability and are obtained as functions of the neural weights at convergence. The experimental results are presented and verified with the continuous time solutions of the Markov models.
出处
《系统工程学报》
CSCD
1993年第2期61-66,共6页
Journal of Systems Engineering
