Measurement and assembly methods for bolted flange joints using three-dimensional digital image correlation

Bolted flange joints are extensively used in pressure vessels and piping equipment. The assembly coaxiality accuracy and bolt pre-tightening force are particularly important in causing the risk of seal and leakage failure. This paper proposed a novel non-contact measurement and assembly adjustment method to reduce the coaxiality angle deviation efficiently. The flange joint is first bolted according to standard flange joint assembly guidelines and then adjusted using the proposed method. In the proposed method, the surface deformation and the coaxiality of the flange are evaluated using a 3D-DIC (three-dimensional digital image correlation) technique; RMS (root mean square) of the surface deformation is used to drive the correction of the bolts to further reduce the coaxiality angle deviation. Compared with existing bolting patterns, the established measurement and assembly adjustment method, which combines 3D-DIC measurement and adjustment based on RMS, can effectively reduce the coaxiality angle deviation. The experiments using a correction strategy demonstrated that the coaxiality angle deviation is reduced by 48.9 % and 60.1 % on average compared with the well-known star bolting pattern and modified star bolting pattern provided by ASME PCC-1, respectively, and even by 30.6 % and 51.9 % compared with their rotational pass patterns. Further experiments with an iterative strategy showed that the coaxiality angle deviation could be reduced by around 70 % and 75 %, tending to be stable around 0.01°–0.035°, after only 5 iterative adjustments. •An assembly method assisted by 3D-DIC for flanges with grouped bolts were proposed.•The surface deformation and coaxiality of flanges are evaluated using a 3D-DIC technique.•Tightening is determined by the RMS of displacement field on the top surface of the flange.•The proposed correction strategy can significantly improve coaxiality angle accuracy.•The coaxiality angle error is reduced by 80%–90 % after only 4–5 iterative adjustments.