A Study on Machine Learning Based Signal Prediction in Flow Control Systems of Nuclear Power Plants

This study proposes a machine learning-based approach for predicting lost signals specifically, flow rate(Q) and pressure(P) in flow control system by utilizing the inherent characteristics of control valves, particularly the flow coefficient(C_v). Experimental data were obtained by measuring system flow rate(Q) and pressure(P) corresponding to various valve opening rates(LVDT). Using the acquired dataset, correlations among variables were analyzed, and signal recovery models developed using Random Forest(RF) and Multi-Layer Perceptron(MLP) algorithms. The model evaluation results showed high prediction performance, with the coefficient of determination(R²) reaching up to 0.96. The proposed method demonstrates potential for enhancing the operational safety and reliability of flow control systems.