A Research on System Error Correction for a High Temperature Hydrogen Detector Based on Neural Network Technique

Abstract

A mathematical model is established for the High Temperature Hydrogen Detector (HTHD) used in severe accident conditions of nuclear power plants. The system error caused by the temperature difference of the internal wall between the working thermal conductivity cells and the reference conductivity cells is analyzed. Then the back propagation neural network algorithm is introduced to correct the system error. The test results show that BP neural network can effectively suppress this system error, and it has well generalization performance. At the same time, this method can be extended to correct measurement errors caused by other disruptive factors, such as supply voltage fluctuation, velocity variation due to pressure change, and interfering components (e.g. steam).