A transient temperature measurement method for extreme environments based on the combination of MF-LIBS and machine...

In response to the limitations of traditional temperature measurement methods in rapidly changing extreme environments, this study introduces an advanced approach integrating magnetically confined laser-induced breakdown spectroscopy (MF-LIBS) with a whale optimization-back propagation neural network (WOA-BP). The research explores how magnetic field confinement influences the emission spectrum of laser-induced plasma. It was observed that an external magnetic field significantly boosts both the intensity and SNR of the spectral signals. Specifically, under the condition of a sample temperature of 300 °C, magnetic field confinement increased the intensity of the Fe II 578.39 nm spectral line by 1.67 times, with the SNR improving by 25.41 %. Further analysis of the spectrum and interferometric images revealed that magnetic field confinement impacts plasma properties and evolution, primarily by inhibiting plasma diffusion through the Lorentz force. Principal Component Analysis (PCA) was employed to streamline the spectral data, facilitating the development of a temperature prediction model using WOA-BP. Compared with traditional methods, this model, when applied in magnetic field conditions, achieved superior performance, evidenced by a coefficient of determination (R²) of 0.9982 and a root mean square error (RMSE) of 3.57, representing a 29.31 % reduction in RMSE compared to conditions without magnetic field confinement. Thus significantly improving temperature prediction accuracy. The findings affirm the potential of combining magnetic field constraints with machine learning to enhance transient temperature measurement amid rapid thermal fluctuations.