Development of Design System of Numerical Interface Between Physical Elements of Agricultural System Based in Fuzzy Theory and Heuristic Self Organization Method

构成工厂生产系统的主要要素是厂房、机械和环境。在这些主元素上可以观察到的数据或信息的类型通常是不兼容的。本研究项目的目标是开发一个数值界面设计系统，使我们能够从一个元素到另一个元素传输不同类型的数据。对农业系统各物理要素存在的问题进行了大量的研究，积累了许多有益的经验。从农业系统的不同要素获得的研究成果之间的交叉关系并不总是确定的。这是因为植物-机器-环境系统是复杂的宏观结构。在不同元素数据之间进行数据传输接口的重要性已经得到了认识。大多数这类问题已经被系统科学方法解决了。本课题将工厂-机械-环境系统视为一个非线性多输入多输出系统，并尝试建立一种确定数值界面的通用方法，使系统中不同类型的基本外设相互握手。数据处理成组方法（GMDH）是一种有效的非线性系统辨识方法。FTS的显著优势在于不需要先验信息来了解产出和投入之间的关系。GMDH预言家选择模型的结构。然而，传统的GMDH讨厌包含多输入多输出（MIMO）系统的问题。这是因为传统的GMDH算法采用二阶多项式函数作为基本函数，自变量作为单对象变量。本文的目的是开发一种改进的GMDH算法，既能处理MIMO系统，又能保留GMDH在非线性系统辨识中的所有有用特征。三层神经网络的应用已经被认为是通用的，因为它们可以做任何更多层的网络可以做的事情，这使得在GMDH算法中避免使用多项式函数成为可能。本文提出的neurogmdh是神经计算和GMDH的结合算法。少

Major elements that compose the plant production system are plant, machinery and environment. The types of data or information that can be observed on those mplor elements are incompatible in general. The objective of this research project is to develop a design system of numerical interface that enable us to transfer different types of data from one element of another. Many studies on problems existing in each physical element in agricultural system have been done and accumulated many useful resofts. Cross relationships between research resofts obtained from different elements of agricultural systems are not always determinable. This is because that the plant-machine-environment system is complex and macrostructured. The importance of interfacing data transfer between different elemental data has been recognized. Most of this type of problems have been handied by system science approach.In this project, by considering the plant-machinery- environment system as a non linear MIMO system … More , An establishment of general method for determining numerical interface that can make all different types of elemental peripherals of the system shake hands was attempted. The group method of data handling(GMDH)is known as a useful technique of nonlinear system identification. fts molt remarkable advantage is not to require a priori information on relationship between output and inputs. GMDH seer-selects the structure of the model. The conventional GMDH, however, abhors problems containing Mufti Input Mufti Output(MIMO)systems. This is because the conventional GMDH algorithm employs a second order polynomial function as a basic function of which independent variable is treated as a single object variable.It is the purpose of this paper to develop an Improved algorithm for GMDH that can handle MIMO systems wdh retaining all useful features of GMDH in non-linear system identification. An application of three-layered neural networks, which have been known to be universal in the sense that they can do anything that a network with more layers can, makes ft possible to avoid the use of polynomial function In GMDH algorithm. The Neuro-GMDH presented here is a combined algorithm of the neuro computing and GMDH. Less