A bolted flange that loosened after 6 months. The flange was 8 bolts, M12 8.8 grade, torqued to 85 N·m. The joint was a hydraulic manifold at 140 bar. The customer re-torqued the bolts. They loosened again after 2 months. The issue wasn’t under-torque — it was preload relaxation. The bolt preload decayed over time due to gasket creep and bolt relaxation. This is about bolt preload retention and why re-torque is sometimes necessary.

The preload loss mechanism

When a bolt is torqued, it stretches. The stretch creates clamp force. Over time, the clamp force decays. Three mechanisms cause this:

Gasket creep: a compressed gasket (spiral wound, rubber) flows under constant compression. The gasket thickness decreases by 0.1-0.5 mm over months. The bolt stretch reduces. The clamp force drops. For a 12 mm bolt with 1 mm stretch: a 0.1 mm gasket relaxation drops the preload by 10%.

Bolt relaxation: the bolt itself stretches under load. At stress above 60% of yield, the bolt undergoes creep (slow plastic deformation). The bolt gets slightly longer. The preload drops. For a properly torqued 8.8 bolt (60% of yield), this is minimal. For an over-torqued bolt (80% of yield), relaxation is significant.

Surface settlement: the high spots on the flange and nut bearing surfaces compress. The embedment takes 0.02-0.05 mm. The preload drops by 5-10% in the first 24 hours.

What was changed

1. Re-torqued after 24 hours. The bolts were torqued to 85 N·m during assembly. After 24 hours, they were re-torqued to 85 N·m. The embedment was complete. The preload stayed. The flange didn’t leak for 2 years.

2. Used a spiral-wound gasket with inner ring. The original was a compressed non-asbestos gasket that crept. The spiral-wound gasket (stainless windings, graphite filler) creeps 80% less. With an inner ring to prevent gasket extrusion, the relaxation dropped to under 5% over a year. No re-torque needed.

3. Specified bolt lubrication. The bolts were dry (unlubricated). The torque-tension relationship is uncertain: dry bolts have a friction coefficient of 0.15-0.25. Lubricated bolts (with moly paste) have 0.10-0.12. A dry bolt torqued to 85 N·m might only reach 60% of the intended preload. With moly paste, the same torque gives 90% of intended preload. The customer applied anti-seize to all bolt threads. The preload became predictable. The flange didn’t leak.

The torque-tension relationship

The torque applied doesn’t directly equal clamp force. Most of the torque overcomes friction (under the head and in the threads). Only 10-15% creates clamp force:

F = T / (0.16 × d)

For dry bolts, the factor is 0.20 (more friction). For lubricated bolts, 0.13. For M12 (d=12 mm) at 85 N·m: F = 85,000 / (0.16 × 12) = 44,270 N. That’s about 44 kN. The bolt yield is 64,000 N (for 8.8 grade, stress area 84.3 mm² × 640 MPa). The preload is 69% of yield. That’s at the upper limit. With lubrication (factor 0.13): F = 54,487 N = 85% of yield. Too high. The correct torque for lubricated M12 8.8 is 65 N·m, not 85. The customer was over-torquing dry bolts. The over-torque caused bolt relaxation.

The preload rule: lubricate threads, torque to 60-70% of yield, re-torque after 24 hours. The loosened flange wasn’t under-torqued — dry bolts at 85 N·m had uncertain preload and over-yield relaxation. Use moly paste, torque correctly, re-torque once. For critical flanges, use a spiral-wound gasket with inner ring. Predictable preload beats guessed torque.