【作者】 刘卫红；

【导师】 黄席樾；

【作者基本信息】 重庆大学 ， 控制理论与控制工程， 2002， 硕士

【摘要】 论文以国家863计划306主题的关键项目“智能化农业信息技术应用示范工程——重庆示范区”（项目编号863-306-ZD-05-03-F）中的一个子项目“农作物栽培智能决策系统”为研究对象，对基于神经网络与专家系统集成的智能决策系统进行了深入研究。本文对人工智能技术的两个分支神经网络与专家系统的发展应用及各自的技术特点进行了阐述，分析了基于神经网络和专家系统集成的混合系统功能框架、并以番茄栽培智能决策为例，提出了基于神经网络和专家系统集成的农作物栽培智能决策系统的结构。本文分别从知识获取、知识表示、知识的检测和求精等四个方面对农作物栽培智能决策系统的知识库的构造进行了详尽的分析。提出了采用神经网络和传统符号知识表示相结合的领域知识混合表示模式，论文还对知识的一致性和完整性进行了分析，提出了相应的检测算法。并且提出了一种对知识库中知识进行求精的算法。本文对农作物栽培智能决策系统的推理机提出了“元推理+目标推理”这一新型推理模式，应用元知识推理协调定性推理、定量推理和神经网络推理。在确定性理论的基础上，提出了一种新的不确定性推理模型，以番茄营养诊断为例对这种新的不确定性推理模型进行了验证，同时给出了系统推理过程的实现程序步骤。本文提出了病虫害诊断神经网络块的设计。通过对病虫害症状的整理和收集，利用神经网络的理论，建立并实现了多个子神经网的结构模型，并用MATLAB程序语言实现了病虫害神经网络的训练和网络应用仿真。更多还原

【Abstract】 By taking“the application demonstration project of the intelligent agricultural information technology — demonstration area of Chongqing City”Which is financed by intelligent computer topic of national 863 (project index:863-306-ZD-05-03-F) as background, this paper has a deep analysis of an intelligent decision-making application system based on Expert System (ES) and Artificial Neural Network (ANN) .The paper analyzes the feature of ES and ANN which are the two active branches of intelligence technology. The function structure of intelligent decision-making application system of using the tomato planting as an example to propose basing on combining ES and ANN .In the paper about knowledge base, this article discussed the knowledge base of the neural-expert hybrid system in four parts: knowledge acquirement, knowledge representation, knowledge examining and knowledge correcting. And put forward a new method which integrated with ES technology and ANN technology to represent the knowledge At last, the article analyzed the consistency and integrality of knowledge base and introduces the examining algorithm and correcting algorithm.In inference engine of intelligent decision-making application system, meta-knowledge inference engine is applied to coordinate with quantitative inference, qualitative inference and ANN inference. Then basing the theory of certainty, a new uncertain reasoning model was introduced and this model was validated in the example of tomato. Then, the step of realization of the neural-expert hybrid expert system’s inference engine was introduced. A diagnose method of diseases and insect pests base on module of neural network was presented and realized in this paper. Through the arrangement and collection of the symptoms of diseases and insect pest, a complex multi-layers structure model of the neural network, which is made up of lots of sub-network, was built taking the advantage of neural network theory. And put the network training and network simulating into reality by using MATLAB language更多还原