Non-destructive temperature measurement of manufactured parts by inverse calculation method

For manufactured parts with complex structures and variable service environments, insufficient accuracy remains the primary constraint limiting the application of multi-wavelength pyrometry, and accurate initial emissivity values or emissivity models are vital for improving its measurement accuracy. This study introduces the least squares covariance method, which estimates initial emissivity more accurately than existing methods. A hybrid algorithm combining the least squares covariance and sequential quadratic programming, which is based on the smooth newton method, is proposed to achieve accurate temperature measurement of manufactured parts. The reliability of the proposed algorithm is first validated by inverting temperatures and emissivity from standard radiance data, which includes six types of classic models. Simulation results demonstrate an average relative error of 0.33% at 1800 K. Then, the temperatures and the emissivity of 304 stainless steel tubes, rectangular GH3536 plates, and GH4169 nozzles are measured using a self-built experimental setup, and an average temperature relative error of 0.47% between the measured values and those inverted by the proposed algorithm further proves its practicability. These findings indicate that the proposed method has promising application potential in engineering fields, including heat transfer analysis and non-destructive in-situ thermal characterization of manufactured parts.