External Load That Separates a Preloaded Joint

Every advantage of a preloaded joint depends on one condition: the members must stay in compression. Separation is the load at which that condition fails, and it is a cliff rather than a slope.

The arithmetic is short. An external load P relieves the members by $(1-C)P$. The members started with the preload $F_i$ of compression in them. They run out when $(1-C)P = F_i$, which gives $P_0 = F_i/(1-C)$. Below that load the joint behaves as the stiffness-ratio page describes, and the bolt's tension creeps up by only $CP$. At and above it, the members have nothing left to give and the bolt takes EVERY additional newton on its own.

That transition is why the calculation matters more than its simplicity suggests. Consider a joint with C = 0.2 and a 25 kN preload: it separates at 31.25 kN. At a 30 kN working load the bolt tension is 25 + 0.2(30) = 31 kN, barely above preload. At 40 kN — a third more load — the bolt is at 40 kN, up thirty per cent. The bolt's stress curve has a knee in it at P₀, and a joint cycling through that knee fatigues at a rate the closed-joint calculation gives no hint of. Fretting starts, the faces work against each other, and preload is lost, which lowers P₀ further. It is a one-way process.

Design the working load well below P₀. A factor of about 1.5 on separation is a common target, more where the load is uncertain, where the joint has to seal, or where opening it even briefly would be unacceptable. Then check the preload from the other side: the bolt has to survive $F_i + CP$ without yielding, so preload plus the bolt's share of the external load, divided by the thread stress area, must stay under the proof strength. A widely used target for a reusable joint is seventy-five per cent of proof load, which leaves room in both directions at once.

Finally, the preload you calculate is not the preload the joint keeps. Embedment — the surface asperities flattening under the clamp force — costs real clamp force in the first hours and can be several per cent on a rough or multi-interface stack. Gasket creep, thermal cycling with dissimilar materials, and paint or plating under the joint faces all take more. This is why critical joints get re-torqued after a heat cycle, and why torque, with its notoriously scattered nut factor, is the least accurate of the preload methods — bolt elongation, turn-of-nut and ultrasonic measurement all beat it, and NASA RP-1228 tabulates the scatter honestly.