Thermal Equivalence Networks for Analysis of Transient Thermographic Data

The network identification by deconvolution method is applied to transient thermal data obtained from infrared thermography. The use of thermographic images is challenging with respect to signal-to-noise ratio, sampling frequency, and recorded data volume. Here, locally weighted scatterplot smoothing (LOWESS) in combination with adaptive window length based on Stein's unbiased risk estimate (SURE) is used to overcome these issues. The network identification by deconvolution method is applied to each pixel of the thermographic image sequence separately. The resulting spatially resolved time constant spectra are displayed as a pseudo-color image sequence which allows an intuitive interpretation of the cooling behavior. As proof of concept, measurements on two different types of LED modules are conducted.