The application of thermoelastic stress analysis in compound structure is particularly complicated because of the different material components, which determines the different thermoelastic effect to be depended on the different material property and mechanical performance. This paper describes a theoretical and experimental analysis on full-filed stress distribution from thermoelastic measurements and its application to determination of stress concentration for compound Steel-Foam-GFRP structure. A finite element modeling is proposed to calculate the sum of the principal stress under the condition of dynamic cyclic load. The sum of the principal stress can be measured by means of thermal stress analysis (TSA). Lock-in thermography has been applied to measure the sum of principal stress distribution of component by its high thermal resolution. In this study, Experiments were carried out with Steel-Foam-GFRP compound structure under dynamic periodic load. The thermoelastic constant is calibrated for different component of compound structure, respectively. An artificial neural network (ANN) is proposed to identify the different component stress distribution on whole compound structure. The experimental result shows that the stress distribution of compound structure can be measured and analyzed using lock-in thermography. It is found that the stress distribution of compound structure can be evaluated with good accuracies by lock-in thermography.