基于修剪策略的D-FNN直接逆控制算法研究

神经网络是模拟人脑结构，它具有大规模并行及分布式信息处理能力，但是不能处理和描述模糊信息.模糊系统具有推理过程容易理解，但它很难实现自适应学习的功能.如果结合神经网络与模糊系统，可以取长补短.基于此，本文提出了一种新型动态模糊神经网络（Dynamic fuzzy neural network，D-FNN）学习算法.因为它具有结构和参数同时调整且学习速度快等优点，所以既可以在模糊逻辑系统中包含低级的神经网络学习和计算功能，也可以为神经网络提供高级的类似人的思维和推理的模糊逻辑系统.此外，本文还开发了生物医学工程应用算法程序，针对药物注射系统的直接逆控制案例进行了仿真，结果表明:D-FNN具有实时学习和控制能力强、参数估计和结构辨识同时进行等优点.

Research on D-FNN Direct Inverse Control Algorithm Based on Pruning Strategy

The neural network simulates the human brain structure with the capabilities processing large-scale parallel and distributed information, which cannot process and describe fuzzy information. The inference processing of the fuzzy system is easy to understand, but it is difficult to realize the adaptive learning. If combining neural networks with fuzzy systems, they can learn from each other. This paper proposes a novel dynamic fuzzy neural network (D-FNN) learning algorithm. Because it has the advantages of simultaneous adjustment of structure and parameters and fast learning speed, it can not only include low-level neural network learning and calculation functions in fuzzy logic systems, but also provide the neural network with the high-level fuzzy logic system which is similar to human thinking and reasoning. In addition, this paper also develops a biomedical engineering application algorithm program, which simulates the direct inverse control case of the drug injection system. The simulating results show that D-FNN has many advantages such as real-time learning, robust control ability, simultaneous parameter estimation and structure identification.