System integration of neural network analizers by state diagnosis of engineering net

Abstract

A set of the multivariable balance and diagnostic models and software tools are proposed for analyzing the flow balance of engineering net of an oil-mines site. The models are realized in neural networks basis in the form of local neural networks analyzers of realtime data with piecewise linear and quadratic approximation of separable functions. The state diagnosis of the oil-mines site engineering net is fulfilled in the process of multilevel system integration (metaintegration) of local neural networks analyzers by use of checking results of hypotheses about the causes of flow disbalance rise in the nodes of engineering net. A neural network models are generated by use of engineering net graph description stored in the industrial data base of software complex of the oilmines site.