Metal additive manufacturing is associated with thermal cycles of high rates of heating, melting, cooling, and solidification. Some areas within the build experience thermal cycles depending on the paths of the energy source. In addition, geometrical features, such as thin walls and overhangs, can lead to heat accumulation which affects the microstructure, fatigue life, and induced residual stresses that may lead to dimensional distortion and cracking. Identification of significant heat accumulation can be used for part quality monitoring, to inform the design process to enhance the quality of printed parts, and to optimize the process parameters. In the present work, we use in-situ thermal monitoring of builds by IR imaging. A computational framework, employing unsupervised machine learning, is developed to detect zones of heat accumulation in the CAD geometry. The effectiveness of this framework is demonstrated by implementation on builds with different geometrical features.