Infrared thermography is an effective method for diagnosing thermal bridges in the field. However, it is less applied in a quantitative manner particularly under dynamic conditions. This study proposed digital twin-based research, wherein a structure is described as a numerical model to identify insulation defects and thermal bridges at the window-wall joint. The window-wall joint construction level was divided into cases with or without insulation for the window-wall joints using a chamber describing dynamic external temperature changes. Thermal images for the indoor wall surface temperature around the window-wall joint were obtained and described using a numerical model capable of high-speed calculation. Regarding the case of absent insulation of the joint, the obtained thermal image results exhibited a large difference in temperature pattern compared to the reference model, which can be reduced to a smaller difference through multiple numerical models obtained by changing the thermal conductivity of the joint. The reduced model proposed for practical optimization simulation exhibited superior performance in terms of accuracy and calculation time compared to a commercial simulation model.