Abstract: In the paper global modeling of complex systems with regard to quality of local models of simple plants is discussed. Complex systems consists of several sub-systems. As a global model multilayer feedforward neural networks are used. It is desirable to obtain an optimal global model, as well as optimal local models. A synthetic quality criterion as a sum of the global quality criterion and local quality criteria is defined. By optimization of the synthetic quality criterion can be obtained the global model with regard to the quality of local models of simple plants. The quality criterion of the global model contains coefficients which define the participation of the local quality criteria in the synthetic quality criterion. The investigation of influence of these coefficients on the quality of the global model of the complex static system is discussed. The investigation is examined by a complex system which is composed from two nonlinear simple plants. In this paper complex system means real chemical object (i.e. a part of the line production of ammonium nitrite).

Keywords: complex system, neural network, global modeling

ACM Classification Keywords: I.2.6 ARTIFICIAL INTELLIGENCE, Learning - Connectionism and neural nets

Link:

STUDY THE QUALITY OF GLOBAL NEURAL MODEL WITH REGARD TO LOCAL MODELS OF CHEMICAL COMPLEX SYSTEM

Grzegorz Drałus

http://www.foibg.com/ibs_isc/ibs-26/ibs-26-p11.pdf