To realize high quality, additively manufactured parts, realtime process monitoring and advanced predictive modeling tools are crucial for accelerating quality assurance and quality control in additive manufacturing. While previous research has demonstrated the effectiveness of physics-and model-based diagnosis and prognosis for additive manufacturing, very little research has been reported on real-Time monitoring and prediction of surface roughness in fused deposition modeling (FDM). This paper presents a new data-driven approach to surface roughness prediction in FDM. A real-Time monitoring system is developed to monitor the health condition of a 3D printer and FDM processes using multiple sensors. A predictive model is built by random forests (RFs). Experimental results have shown that the predictive model is capable of predicting the surface roughness of a printed part with very high accuracy.