Prediction Of Nuclear Fuel Rod Damage Based On Artificial Neural Network

With the development of nuclear power generation,the safety and efficiency of nuclear power plants have received more and more attention.As the core component of nuclear reactors,the timely monitoring of damage of the nuclear fuel rod is critical to the safe and efficient production of nuclear power.The fuel assembly of a heavy water reactor is composed of a nuclear fuel rod bundle,which is loaded in a core basket of a nuclear reactor.The flow space of the coolant in the reactor is interconnected,and the reactor does not have a separate process channel for each fuel assembly.Therefore,it is very difficult to diagnose and locate broken fuel assemblies and fuel rods.Traditional nuclear fuel rod damage detection uses a chemical sampling analysis method,that is,the cooling water in the primary circuit is regularly extracted,and the radioactivity of the radioactive products in the cooling water is measured to determine whether the nuclear fuel rod casing is damaged.Due to the generally high cost of detection,the sampling period is generally long,and the damage to the reactor and fuel components and fuel rods cannot be detected in a timely manner,and the damage status cannot be continuously monitored.This subject combines artificial neural network technology,based on the monitoring parameters of nuclear power plant feedback from the implementation monitoring system of nuclear power plant,selects certain characteristics,builds a suitable neural network model,and completes the prediction of nuclear fuel damage.Based on traditional detection,it can complete the prediction of nuclear fuel rod damage while having the characteristics of low destructiveness,low cost and good real-time performance for the reactor.Using these advantages of neural network,through the design and optimization of the network model,a damage prediction method based on artificial neural network is proposed.Firstly,data preprocessing is performed on the data obtained from the nuclear reactor online monitoring system.There are many data that can be used to analyze the damage of nuclear fuel rods,but the correlations are different.One type of characteristics,that is,the content of radioactive material in the coolant in the primary circuit,can directly reflect the damage of the fuel rod,and it has a high correlation with the damage of nuclear fuel rods;another important type of important parameters of nuclear power plants,including channel operation The status parameters and core operating status parameters are the status parameters of the nuclear power plant that are fed back in real time by sensors installed in the nuclear power plant.This paper uses the physical and chemical principles based on previous studies and the feature extraction scheme based on machine learning feature engineering.The nuclear reaction process involves a large number of non-linear and uncertain issues,and it is difficult to describe and model by definite mathematical or physical expressions.Artificial neural networks can autonomously learn and store complex input-output mapping relationships.In this paper,by constructing BP neural network,convolutional neural network,and recurrent neural network models,the mapping between input and output data is established from different data processing perspectives.By adjusting the network structure and related parameters,continuous optimization is performed to finally complete the nuclear fuel rod Broken forecast.The simulation results show that the BP neural network model,convolutional neural network model and recurrent neural network model based on logistic regression have good prediction effect and stability for nuclear fuel rod damage,and reduce the destructive detection of nuclear reactors.Based on the detection,it can get a good prediction effect.Based on the traditional method,it has a great performance improvement and prediction accuracy rate.The prediction accuracy rate can basically reach more than 95%,and it has a relatively good performance for both damaged and unbroken nuclear fuel rods.