Ineffective tolerance analysis of prefabricated building components has a significant impact on the quality of accelerated construction projects. Emerging accelerated construction methods have adopted prefabrication techniques to improve construction productivity and balance workflows. The complex three-dimensional relationships among curved and deformed pipes and ductworks bring difficulties of analyzing how fabrication and installation errors of densely located pipes influence each other, and how such errors propagate throughout connected curvilinear components. This paper presents a computational framework that integrates 3D imagery data with Building Information Modeling to detect and analyze fit-up problems of curvilinear components. Using the detected deviations, an automated deviation classification algorithm derives the tolerance information of each individual component. This process generates a tolerance network describing how geometric variations of individual components influence each other. This tolerance network provides a quality control framework that enables adaptive redistribution of prefabrication and installation errors to resolve fit-up problems during accelerated construction.