Monitoring of Temperature Measurements for Different Flow Regimes in Water and Galinstan with Long Short-Term Memory Networks and Transfer Learning of Sensors

• DOI: 10.3390/computation10070108
• Publication Date: 2022-06-30T00:47:56Z
• Authors: Stella Pantopoulou; Victoria Ankel; Matthew T. Weathered; Darius D. Lisowski; Anthonie Cilliers; Lefteri H. Tsoukalas; Alexander Heifetz
• Keywords: Thermocouple

Temperature sensing is one of the most common measurements a nuclear reactor monitoring system. The coolant fluid flow in core depends on power state. We investigated and estimation thermocouple time series using machine learning for range regimes. Measurement data were obtained, two separate experiments, loop filled with water liquid metal Galinstan. developed long short-term memory (LSTM) recurrent neural networks (RNNs) sensor predictions by training sensor’s own prior history, transfer LSTM (TL-LSTM) correlated history. Sensor cross-correlations identified calculating Pearson correlation coefficient series. accuracy TL-LSTM temperature was studied as function Reynolds number (Re). root-mean-square error (RMSE) test segment each shown to linearly increase Re both Galinstan fluids. Using linear correlations, we estimated values which RMSE smaller than measurement uncertainty. For fluids, showed that provide reliable estimations typical regimes reactor. runtime be substantially acquisition rate, allows performing validation real time.