Almost every bolted joint in service was installed by someone turning a wrench to a torque figure, and almost every torque figure is a proxy for something else. The joint is designed around clamp force, the tension in the fastener that holds the members together and keeps the bolt from seeing the working load directly. Torque is the crude and highly variable means by which that tension is produced. When a joint fails and the assembly records show the specified torque was applied, the record establishes considerably less than it appears to establish.
The joint is designed around preload
In a sound bolted joint the fastener is tensioned so that the clamped members remain in compression across the full range of service load. The bolt then absorbs only a fraction of any external load, and a cyclic external load produces only a small cyclic stress in the fastener. This is why correctly preloaded bolts rarely fail in fatigue and inadequately preloaded ones frequently do.
Lose the preload and the joint changes character entirely. Members separate and reseat, the bolt takes the load range directly, and relative motion between mating threads begins the self-loosening process characterised by Junker and reproduced in the transverse vibration testing described in ISO 16130 and DIN 65151.
Where the torque actually goes
Only a small fraction of applied torque becomes bolt tension. The large majority is consumed by friction under the turning bearing face and in the threads. How that splits depends on surface finish, plating, lubrication, prior installation, and corrosion or debris in the threads.
Because the useful fraction is small and the frictional fraction large, a modest change in friction produces a large change in achieved tension at the same torque. A lubricated bolt torqued to a dry specification can be pulled past yield; a corroded one torqued to the same figure may reach only a portion of the intended clamp.
Scatter is a property of the method
Torque control produces substantial scatter in achieved preload even under good shop conditions, which is why designs that rely on it carry a margin to absorb that variation. The scatter is inherent to the method rather than evidence of poor workmanship, and treating a torque wrench as though it delivered a specific tension misunderstands the tool.
It follows that an assembly record showing the correct torque is consistent with a wide band of actual clamp forces, including values that were never adequate for the joint.
Methods that control tension more directly
Turn-of-nut control uses rotation past snug to produce a controlled elongation and is much less sensitive to friction. Direct tension indicators, covered by ASTM F959, compress at a known load. Hydraulic tensioners and ultrasonic or strain-gauged measurement act on elongation itself. The RCSC Specification for Structural Joints Using High-Strength Bolts recognises several pretensioning methods.
Which method the design called for, and which was actually used, has both a documentary answer and a physical one; where they diverge the hardware usually governs.
What the grade marking promises
SAE J429 grades and ISO 898-1 property classes define material and strength requirements. The ASME B18 series covers dimensions and head markings. ASTM F606/F606M gives the methods by which proof load, wedge tensile strength and hardness are verified. ASTM F3125 governs the high-strength structural bolts used in steel construction, and the NASM and NAS specifications cover much of the aerospace population.
None of these standards specify a torque. They specify what the fastener can carry, leaving the designer to set a target preload and the procedure to achieve it. That division is where responsibility questions arise.
Assembly is itself a specified activity
For pressure boundary flanges, ASME PCC-1 sets out assembly practice in detail: tightening patterns, target stress determination, lubrication, tool calibration, and qualification of the people performing the work. Structural steelwork has the RCSC specification. Both treat assembly as an engineered activity generating its own records.
Where such a document governed and was not followed, the deviation is a finding in its own right, independent of the fracture mechanism.
Reconstructing the preload after the fact
Preload is inferred after a failure, not measured. Thread deformation and burnishing on bearing faces indicate how hard the joint was pulled up. Residual breakaway torque on surviving fasteners bounds the clamp still present. Permanent elongation, measured against an unused bolt from the same lot, is direct evidence of yielding during installation.
Friction coefficients measured on the actual hardware in a torque-tension fixture convert a recorded torque into a credible tension range. That range, rather than a single calculated number, is what the evidence supports.
Design, installation, maintenance or environment
Attribution usually resolves along the preload question. A fastener undersized or too few in number for the load is a design issue. A sound design assembled dry, out of sequence, or with an uncalibrated tool is installation. A joint that relaxed through gasket creep, embedment or thermal cycling and was never retightened is maintenance. A joint that lost clamp to corrosion product or coating breakdown is environmental.
These categories are not exclusive, and most contested fastener failures involve more than one. Stating which contributions the evidence supports, and how strongly, is more useful and more durable than selecting a single cause.
Where these opinions are challenged
Commonly: that the torque-to-tension conversion used a handbook nut factor rather than one measured on the actual parts; that a torque record was treated as a preload record; that method scatter was never quantified; and that relaxation over years of service was left out of the analysis entirely.
This article is general technical orientation, not a failure analysis, an engineering opinion, or advice on any specific matter. Determining the cause of a particular incident requires hands-on examination by a credentialed expert.