摘要
Petri网是对具有产生式规则的故障诊断系统建模的有力工具,但其学习能力不强。以Petri网的基本定义为基础,结合模糊逻辑和Petri网模型,并在此基础上引入人工神经网络技术,定义了模糊学习Petri网模型。模型中的隐含库所和变迁将人工神经网络中隐含神经元内部信息处理过程明确化,然后对该模型提出一种逐层调整变迁阈值的训练算法,该算法通过逐层调整的方式来获取变迁的阈值,改善了网络的学习效率,并对算法的收敛性进行了证明。最后,以故障推理实例验证了算法的有效性与实用性。
Petri net is a useful modeling tool for fault diagnosis system based on production rules.But Petri net lacks a stronger learning ability.Based on the basic definition of Petri nets and combined with the fuzzy logic,a fuzzy learning Petri Nets model was presented by adopting the artificial neural networks technology.In this model,semantic relations of the hidden layers with their predecessor and successor layers can be justified.Then an advanced training algorithm for adjusting the thresholds of transitions step by step was proposed and its convergence was proved,which improved the training efficiency greatly.Finally,a fault example was given to show the effectiveness and practicability of the proposed model.
出处
《海军工程大学学报》
CAS
北大核心
2010年第2期11-15,共5页
Journal of Naval University of Engineering
基金
国家自然科学基金资助项目(60774029)
