In this work a Monitoring System was developed using GMDH (Group Method of Data Handling) algorithm and Artificial Neural Networks (ANN) which was applied in the IEA-R1 research reactor at IPEN. GMDH is used in two different ways to perform an input data preprocessing to the ANN. The system perform the monitoring by comparing the estimative calculated values with the measured ones. Each Monitoring System was developed and tested for five different sets of input data, ancluding data from a Reactor Teoretical Model and for four different sets of reactor variables. The results for the most of cases show an improvement when GMDH is combined with ANN algorithm in the Monitoring Systems. The good results obtained in the present work show the viability of using GMDH algorithm in the study of the best input variables to the ANN, thus making possible the use of these methods in the implementation of a new Monitoring Methodology applied in sensors of a nuclear power plant.
Abstract-In this work, two different computational intelligence techniques -Neural Networks and Neuro-fuzzywere used in the development of a Monitoring and Fault Detection system applied to the IEA-R1 experimental nuclear reactor. The monitoring is made by comparing an estimative of the variable generated by the computational intelligence techniques with the actual value. Both techniques have been successfully applied to the Monitoring System and were effective in estimate a monitored variable. The Neuro-Fuzzy technique showed strongly better performance compared with Neural Network.
Resumo -Este trabalho tem como objetivo apresentar um estudo comparativo entre as metodologias de Redes Neurais Artificiais (RNAs) e GMDH(Group Method of Data Handling) aplicados na Monitoração de sensores de temperatura de um reator experimental. Foram desenvolvidos dois modelos de Monitoração: um utilizando Redes Neurais e o outro o algoritmo GMDH. Para tanto foi utilizada uma base de dados gerada por um modelo teórico do reator. Os resultados obtidos utilizando o algoritmo GMDH foram melhores do que os obtidos com as Redes Neurais.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.