Triply periodic minimal surfaces (TPMS) are a type of mathematical surface that is periodic in three directions and has a mean curvature of zero at every point. TPMS structures can be used to develop architected cellular materials that promise to revolutionize the market of lightweight components, due to their high strength-to-weight ratio, excellent energy absorption, and ability to be precisely manufactured using Additive Manufacturing (AM) processes [1]. This contribution introduces a purely conduction-based model for PBF-LB/M process simulations, enabling an effective representation of the primary heat loss mechanisms that occur at the part scale. This model is implemented using an adaptive immersed boundary method, namely the Finite Cell Method (FCM), for thermal analysis of PBF-LB/M process. Such a numerical solution is particularly well-suited for simulations of complex geometric parts, such as TPMS structures [2]. The proposed physical model and the immersed boundary framework are validated using experimental online data acquired on an in-house developed PBF-LB/M system during the manufacturing process of five gyroid cells made of stainless steel 316L. This talk will discuss the challenges related to accurately capturing the part-scale temperature evolution with this model. In particular, the effective material properties of the continuum powder model significantly influence the buildup of residual heat on the part-scale level. Moreover, accurately reproducing top-surface temperatures measured by infrared imaging requires modeling the actual powder layer thickness, which can be several times larger than the programmed powder layer thickness.