论文部分内容阅读
本文采用的单体模糊神经网络(MFNNs)在智能控制中兼有模糊控制与人工神经网络控制的特点,不仅具有良好的控制特性,而且能够进行模型自学习和在线自调整.结合本身结构、人工神经网络理论和模糊理论,可得到MFNNs自学习的一系列算法.本文对基于ANNs梯度算法的MFNNs进行了研究,结果表明,MFNNs结构的自学习、自调整是可行、简便和有效的,有比一般ANNs快得多的收敛性,能够成功地应用于智能控制系统,包括复杂对象的控制.此外所提出的学习方法很容易以VLSI硬件实现.
The single fuzzy neural networks (MFNNs) used in this paper have the characteristics of fuzzy control and artificial neural network control in the intelligent control. They not only have good control characteristics, but also can carry out model self-learning and on-line self-tuning. Combined with its own structure, artificial neural network theory and fuzzy theory, a series of algorithms for MFNNs self-learning can be obtained. This paper studies the MFNNs based on the gradient algorithm of ANNs. The results show that the self-learning and self-tuning of the MFNNs structure is feasible, simple and effective, and has much faster convergence than the ANNs. It can be successfully applied to the intelligent control system , Including the control of complex objects. In addition, the proposed learning method is easily implemented in VLSI hardware.