Automated bolting and head handling for multi-flange pressure-vessel closures: a simulation-based design evaluation with a delayed-coking case study

A simulation-based design evaluation. No system has been built or tested; every performance figure is a model prediction from stated assumptions. Submitted to The International Journal of Advanced Manufacturing Technology. An automated architecture for opening and closing a 92-bolt, three-flange delayed-coking drum (orbital pneumatic torque runners, a bolt carousel, a counterbalanced head manipulator and an interlocking sequence controller) is tested against the targets originally claimed for it. A Monte Carlo model built on the actual star-pattern tool travel shows that one torque head per flange needs a median 7.96 h to unbolt and re-bolt the drum and, within the investigated parameter bounds, cannot meet a 2.5 h target. Removing bottlenecks in turn gives 2.28 h with four synchronised heads and 1.82 h with a bolt-transfer channel per head, the configuration the analysis supports for experimental development (98 % of samples within 2.5 h); one fewer star pass would give 1.60 h but requires qualification. The ranking of architectures held across all 20,000 simulated samples under uniform, triangular and correlated input models. An exactly optimised tool path (Held-Karp, load-spreading constraint) adds little, and the complete operation remains about 3.4 h, governed by head handling and leak testing; a servicing-time envelope is derived. Specifying ±2 % torque accuracy barely affects preload scatter, which nut-factor variation dominates; under an assumed 3 % angle-based preload-estimate uncertainty, torque-angle re-torque cuts bolts outside ±10 % of target from 32 % to 5 %. Explicit-state verification of the interlock logic found no violation of seven safety invariants but exposed deadlocks after interrupted sequences (1,232 of 2,105 reachable states for a 12-bolt flange), which a resume rule removed; mutation tests confirm each invariant can fail. Pneumatic head handling needs a counterbalance carrying most of the load. An experimental validation and falsification plan with acceptance criteria is given.