| Popis: |
Assembly of steel structures, modules, and pipe spools requires cycles of fitting and alignment in fabrication facilities and on construction sites. To minimize this work, good discrepancy feedback for automated refitting and realignment is required. A framework for such feedback is presented here that overcomes current limitations. It commences with a constrained registration step to overcome the incapabilities of the current discrepancy analysis approaches. By borrowing concepts from robotic kinematics and 3D image alignment theories, forward kinematics models are generated link by link, and thus provide the means for a local discrepancy analysis for quantifying the deviations autonomously. Experiments show that the proposed approach is suitably accurate and sufficiently fast to be employed for real-time feedback in order to systematically and automatically develop the realignment plans required for refitting and realigning assemblies, which is the key contribution of the work presented in this paper. |
Full Text from ScienceDirect