NC simulation emulates the CNC machine tool and the cutter move along its axis. It depicts the material removal to better visualize the machining process. Until now, its main goals are still confined to check an unproved NC program to avoid potential collisions, and to analyze undercuts and overcuts. In this work, by using the simulated output-the in-process model (IPM), and rebuilding a machining feature based model, applications of NC simulation are extended to establish the feedback link from CNC to CAM, which can automatically pass modifications of manufacturing engineers back through the digital chain. In cases when old CAD archives of parts are incompatible with new systems and machines, or are damaged or lost, the original CAD model is neither available nor usable, and the part program is the only data available, rebuilding the feature based model from the NC simulation is necessary. This paper proposes an approach for machining feature recognition from IPM. As per the IPM characteristics, the approach adopts a novel approach of curvature based region segmentation and valuated adjacency graph based feature recognition.