Weld Failure on a Pressure Vessel Nozzle
A cylindrical air receiver, 900 mm internal diameter and 6 mm wall, operates at 1.1 MPa. A 50 mm nominal bore nozzle is fillet-welded to the shell. During hydrostatic testing a dye-penetrant check reveals a 12 mm long toe crack at the weld on the shell side, oriented circumferentially. The vessel is made from mild steel with yield 250 MPa and ultimate 430 MPa. The inspector must decide whether to reject, repair, or accept with monitoring, knowing the crack is small but sits in the highest-stress region. Production wants the vessel released because 'the crack is only 12 mm and the wall is 6 mm'.
1. Estimate the hoop stress in the shell and assess whether the crack is in a benign or critical stress orientation.
Read the worked answer
Thin-cylinder hoop stress: sigma = pD/(2t) = (1.1 x 900)/(2 x 6) = 990/12 = 82.5 MPa. Longitudinal stress is half that, about 41 MPa.
The crack is circumferential, so its plane is normal to the axial direction. A crack grows when the stress acts normal to its plane, i.e. the axial stress of about 41 MPa. So the crack is actually loaded by the lower principal stress, which is favourable. Had the crack been longitudinal, it would open under the 82.5 MPa hoop stress and be far more dangerous.
Judgement: the orientation is benign relative to hoop stress, but the weld toe is a stress concentration, so local stress is well above nominal. This is not a simple accept.
2. Production argues the crack length is trivial compared with the wall. Evaluate this argument and recommend an action.
Read the worked answer
The comparison of crack length to wall thickness is the wrong ratio. What matters is crack depth through the wall and the fracture toughness relative to applied stress. A 12 mm surface crack in a 6 mm wall, if it has grown through most of the thickness, leaves very little ligament; a through-thickness crack in a thin wall can leak or tear rapidly.
I would require measurement of crack depth by ultrasonic testing, not just surface length. If depth exceeds about half the wall (3 mm), reject for repair. If shallow, grind out and re-weld, then re-test. Release without depth data is unjustified because the dye test only shows surface extent. The recommendation is repair, not accept with monitoring, since a pressure vessel has no safe failure mode for a weld-toe crack.