Acquisition of kHz-frequency two-dimensional surface temperature field using phosphor thermometry and proper orthogonal decomposition assisted long short-term memory neural networks

Abstract

A reconstruction technique of kHz time-resolved two-dimensional (2D) surface temperature field was achieved with the discrete point measurements and low sampling rate 2D thermographic phosphor (TP) thermometry measurements using a long short-term memory (LSTM) based artificial intelligence framework. The 2D surface temperature field of a 350 °C plate with a 2.5 Hz swing cooling jet was measured using TP thermometry at a sampling rate of 20 Hz. At the same time, high-frequency thermocouples with a sampling rate of 1 kHz were recorded for the construction of LSTM neural networks training and for validation. The 20 Hz 2D surface temperature field was analyzed with proper orthogonal decomposition to acquire the energy modes and model coefficients. The mode coefficients are then trained with the discrete but high-frequency time-resolved temperature information from the thermocouple by LSTM to acquire the time-resolved mode coefficients. Finally, the high-frequency time-resolved 2D surface temperature field is obtained by reconstructing modes and the time-resolved coefficients. The reconstructed result shows that the current technique can obtain high time-resolved and spatially resolved 2D surface temperature fields very well.