<p>The mechanical performance and reliability of additive manufacturing (AM) parts—particularly those produced via laser powder bed fusion (L-PBF)—are strongly influenced by dross formation and surface roughness. Accurate prediction of these defects is essential for minimizing material waste and ensuring performance in mission-critical applications. However, the inherent complexity of multiphysics interactions, combined with the stochastic nature of L-PBF, has made such predictions extremely challenging—particularly at the part scale and within reasonable computation times. To date, no simulation framework has been presented that enables rapid and high-fidelity prediction of both dross formation and surface roughness across an entire part. This study is the first to introduce such a capability. This paper presents a novel stochastic model, termed the cellular automata-driven probability propagation (CA-PP) model, developed as a stochastic model along with its parameter calibration process. The model employs high-density voxel representation to predict dross formation and surface roughness with high fidelity. Validation was conducted using two separate specimens: AlSi10Mg for dross formation and Haynes^® 230 alloy^™ for surface roughness. The CA-PP model successfully predicted the magnitude and shape of dross as well as surface roughness trends across varying overhang angles. Computational analysis revealed that the model runs approximately 207 times faster than the actual L-PBF process at a voxel resolution of 0.06&#xa0;mm. These findings highlight the CA-PP model’s potential as a high-speed simulation engine for digital twin systems, enabling real-time defect prediction and proactive process control to enhance the reliability and efficiency of L-PBF fabrication.</p>