Machine learning-assited optical thermometer for continuous temperature analysis inside molten metal

This paper demonstrates a robust optical fiber thermometer (OFT) for temperature measurement under extreme environments. To date, the development of sensors for continuous temperature measurement in environments with temperatures over 1000 °C, severe electromagnetic interferences, and strong oxidizing agents has been very challenging. The proposed nano-OFT system consists of a ceramic tube, a nanorod coated sapphire fiber, and a near-infrared (NIR) spectrum analyzer for continuous measurement of molten steel temperature in furnaces. The nanorod layer functions as an effective cladding material for the sapphire fiber to sustain a reliable transmission of NIR thermal emissions. The thermal radiation from the ceramic tube's tip was coupled out of the nano-OFT probe via the sapphire fiber and measured using the NIR spectrometer. The NIR emissions were analyzed using a convolution neural network to determine the probe temperature. Our results show that the nano-OFT probe can measure furnace temperature in the temperature range from 1,000–1,650 °C, with the error percentage as low as 0.5 %. The nano-OFT system can be employed by the steel industry to monitor steel temperature continuously, and thus enhance steel production efficiency and reduce energy consumption.