Background radiation compensation calibration method for film cooling infrared temperature measurement based on BP neural network

Abstract

The precise non-contact temperature measurement method is essential for studying gas turbine component heat transfer. While infrared thermal imaging provides high-resolution two-dimensional temperature fields, errors arise from the measurement environment's complexity, emissivity variations, and reflection radiation effects. This study utilizes BP neural networks to correct infrared radiation on gas film cooled plates. By modeling background radiation and analyzing infrared emission processes, factors influencing temperature distribution are identified. Experimental data from uncooled plate are used for network training. Experimental measurements were conducted on uncooled plates to obtain training data. A BP neural network was established based on Planck's law and the radiation transmission process to relate the mainstream temperature to the background radiation. Calibration analysis was performed based on experimental data from the cooled plate, achieving the restoration of the two-dimensional temperature field under multiple operating conditions and significantly reducing measurement errors. This method comprehensively considers the entire process of radiation transmission, thereby eliminating background radiation embedded in the infrared images. The obtained measurement results vividly depict the real temperature distribution, offered significant support for gas turbine engineering applications.