Ultrasonic thermometry is a promising technique for temperature sensing in harsh environments; however, the practical applicability of this method is hindered by waveguide oxidation, acoustic attenuation, and unreliable echo localization at extreme temperatures. To overcome these challenges, this report proposes a novel segmented ultrasonic temperature sensor comprising a W-Re transmission section and an Ir-Rh sensing section. This heterogeneous design balances long-distance, low-attenuation acoustic transmission with high oxidation resistance at the hot end, thus increasing the durability of the sensor compared to traditional W-Re or ceramic waveguides. A resistance-welded joint is integrated to preserve acoustic integrity by creating a stable heterogeneous interface that undergoes minimal thermal damage. Furthermore, a learning-assisted extraction scheme is applied, whereby a You Only Look Once v11 model automatically identifies echo features to define robust regions of interest, followed by an energy-weighted cross correlation to eliminate cycle-skipping errors that are common in traditional algorithms. Furnace calibration from 20 to 1600 °C reveals a monotonic non-linear temperature-velocity relationship fitting a second-order polynomial model (R2 = 99.94%). The developed system demonstrates a repeatability variation of <0.3% and a unit sensitivity of 24-67 ns·(°C m)-1. Comparative benchmarking analysis confirms that the proposed segmented sensor provides excellent robustness and stability under high-temperature oxidizing conditions where conventional sensors typically degrade or fail.
Wei et al. (Mon,) studied this question.