Utilização de redes neurais artificiais para determinar o tempo de resposta de sensores de temperatura do  tipo RTD / Time response of temperature sensors using neural networks

AUTOR(ES)

Roberto Carlos dos Santos

FONTE

IBICT - Instituto Brasileiro de Informação em Ciência e Tecnologia

DATA DE PUBLICAÇÃO

16/09/2010

RESUMO

In a PWR nuclear power plant, the primary coolant temperature and feedwater temperature are measured using RTDs (Resistance Temperature Detectors). These RTDs typically feed the plants control and safety systems and must, therefore, be very accurate and have good dynamic performance. The response time of RTDs is characterized by a single parameter called the Plunge Time Constant defined as the time it takes the sensor output to achieve 63.2 percent of its final value after a step change in temperature. Nuclear reactor service conditions are difficult to reproduce in the laboratory, and an in-situ test method called LCSR (Loop Current Step Response) test was developed to measure remotely the response time of RTDs. From this test, the time constant of the sensor is identified by means of the LCSR transformation that involves the dynamic response modal time constants determination using a nodal heat-transfer model. This calculation is not simple and requires specialized personnel. For this reason an Artificial Neural Network has been developed to predict the time constant of RTD from LCSR test transient. It eliminates the transformations involved in the LCSR application. A series of LCSR tests on RTDs generates the response transients of the sensors, the input data of the networks. Plunge tests are used to determine the time constants of the RTDs, the desired output of the ANN, trained using these sets of input/output data. This methodology was firstly applied to theoretical data simulating 10 RTDs with different time constant values, resulting in an average error of about 0.74 %. Experimental data from three different RTDs was used to predict time constant resulting in a maximum error of 3,34 %. The time constants values predicted from ANN were compared with those obtained from traditional way resulting in an average error of about 18 % and that shows the network is able to predict accurately the sensor time constant.

ASSUNTO(S)

neural networks redes neurais rtd rtd sensores de temperatura temperature sensors tempo de resposta time response

Documentos Relacionados

• Detecção de adulteração de combustíveis com sensores poliméricos eletrodepositados e redes neurais artificiais.
• ARTIFICIAL NEURAL NETWORKS IN TIME SERIES FORECASTING
• Utilização de redes neurais na análise e previsão de séries temporais
• REDES NEURAIS TEMPORAIS PARA O TRATAMENTO DE SISTEMAS VARIANTES NO TEMPO
• Modelagem constitutiva de areias usando redes neurais artificiais