The Application Of Extension Theory In RBF Neural Network

Neural Networks use black-box model to simulate the complex relationship between variables, the modeling results meet the needs of industrial applications, so they are applicated widely, but they lack of effective description of the information. Extension theory characterizes the information based on the basic-element theory and takes extension methodology as the guide, in order to work out the conflicts, which is a thransverse cross disciplinary of philosophy, mathematics and engineering. This paper integrates extension theory and neural network and starts the extension neural network research. The main contens are as follows:First, starting from the basic-element theory of extenics, the extension distance and extension transformation are integrated to RBF neural network research, the algorithm process description is standardized, an extension theory-based Radial Basis Function (ERBF) neural network is proposed. Datasets from UCI database are used to show the feasibility and effectiveness of the algorithm in solving the classification and clustering issues.Second, selective neural network ensemble technology is studied, then ERBF is taked as the subnet, the definition of diversity based on the error curve is modified, and Manhattan distance is used to instead of Euclidean distance, which is commonly used to correct the diversity formula based on the subnet output, a new ensemble neural network is built. UCI datasets are taken as test datas to validate the experiment results, which show that the new method is more practical.Third, basing on.Net Framework, OO, COM, design patterns and other related technologies, an extension theory-based ensemble neural network platform is developed to experimental research. Meanwile, PTA solvent system in one subplant of petrochemical is used as practical modeling, from which the technical supports of production forecasting, operation optimization and energy reduction are provided. Modeling results confirm that the algorithm has practical value in process industrial field. The system is user friendly, features clearly and high portability.The research results show that:The Extenics is introduced to chemical engineering, process automation and intelligent engineering field, which is conducive to the improvement and development of the process systems engineering methods, and improves the analysis and application capabilities of the industrial automation systems.