Preload Decay and Stress Relaxation: Why Long-Service Joints Come Loose
Executive Summary
Torque passed at assembly, yet on re-tightening the bolt feels loose while the nut never turned—this is preload decay. Five paths—embedding, stress relaxation, creep, thermal-expansion mismatch and fretting wear—cause preload loss over long service. This article quantifies each path's contribution, gives the residual-preload calculation method and anti-decay design points.
Table of Contents
- A Troubleshooting Case: "Bolt Didn't Loosen, but the Joint Did"
- Path 1: Embedding Loss (Short-Term, Dominant)
- Path 2: Stress Relaxation and Creep (Long-Term, Dominant)
- Path 3: Thermal-Expansion Mismatch and Temperature Cycling
- Path 4: Fretting Wear
- Path 5: "False Decay" from Insufficient Design Margin
- Design and Process Points Against Decay
- Conclusion
A Troubleshooting Case: "Bolt Didn't Loosen, but the Joint Did"
After 8000 hours of operation, a high-temperature pipeline flange developed leakage. Inspection showed the nut's torque witness mark had not moved at all, meaning the nut did not rotate; but after removal, measuring bolt elongation and back-calculating residual preload showed only 62% of the assembly value.
This is preload decay: the bolt did not turn and the joint looks intact, yet nearly 40% of clamp force has been lost in service. It is different from "bolt loosening"—the latter is relative rotation of the thread pair; the former is loss of elastic elongation. Confusing the two leads to entirely the wrong countermeasure.
Path 1: Embedding Loss (Short-Term, Dominant)
Micro-asperities on the thread flanks and bearing surface are flattened under pressure, reducing the bolt's elastic elongation and lowering preload. This process typically completes within hours to tens of hours after assembly, making it short-term rapid decay.
| Contact-Surface Condition | Typical Embedding | Preload Loss | Typical Completion Time |
|---|---|---|---|
| Precision-machined surface, single joint face | 5~10 μm | 3%~8% | A few hours |
| Common machined surface | 10~20 μm | 8%~15% | Hours to 1 day |
| Multiple joint faces or rough surfaces | 20~40 μm | 15%~25% | 1~3 days |
| Coating layer or soft shim at the interface | 40 μm+ | Up to 30% | Days to weeks |
Key insight: embedding is approximately proportional to the number of contact surfaces. A joint made of 3 parts with 2 joint faces loses significantly more by embedding than a single joint face. Reducing the number of joint faces is the most effective means of controlling embedding loss, and it costs almost nothing.
Path 2: Stress Relaxation and Creep (Long-Term, Dominant)
Under sustained load and temperature, the material creeps and elastic deformation gradually turns into plastic deformation, slowly lowering preload. Temperature is the dominant accelerating factor: the higher the temperature, the faster the creep rate, often exponentially.
| Service Temperature (relative to material creep threshold) | Relaxation Behavior | Design Countermeasure |
|---|---|---|
| Far below threshold (e.g. steel < 200 °C) | Slow relaxation; < 10% loss over thousands of hours | Conventional design suffices |
| Near threshold (e.g. steel 250~350 °C) | Obvious relaxation; 10%~25% loss over thousands of hours | Reserve decay margin; use creep-resistant materials |
| Above threshold (e.g. steel > 400 °C) | Marked relaxation; becomes the dominant mechanism | Must calculate residual preload over service life |
The clamped parts also creep. Aluminum and magnesium alloys already creep appreciably at room temperature, and more so above 100 °C. This is one reason light-alloy joints show excessive preload loss after long service.
Path 3: Thermal-Expansion Mismatch and Temperature Cycling
When bolt and clamped parts use different materials, the difference in thermal-expansion coefficients changes preload as temperature changes. Take a steel bolt with an aluminum part: aluminum's linear expansion coefficient is about twice that of steel; when temperature rises, the aluminum expands more and the joint is actually clamped tighter; when temperature drops, the aluminum contracts more and preload falls.
| Material | Coefficient of Linear Expansion (×10−6/K) | Relative to Steel |
|---|---|---|
| Carbon / alloy steel | 11~13 | Reference |
| Austenitic stainless steel | 16~17 | ~1.4× |
| Aluminum alloy | 22~24 | ~2× |
| Magnesium alloy | 25~27 | ~2.2× |
| Titanium alloy | 8~9 | ~0.7× |
This raises two engineering issues:
- Temperature cycling causes preload fluctuation: even without creep, preload oscillates as temperature cycles between high and low; its amplitude is estimated from the material expansion difference and clamp length.
- Temperature cycling superimposed on creep causes cumulative loss: creep loss at high temperature is not recovered at low temperature. Therefore, for temperature-cycling service, loss must be accumulated by equivalent duration at the high-temperature stage.
Path 4: Fretting Wear
Joint faces undergo micrometer-scale relative motion under alternating load; the contact material is repeatedly plowed, oxidized and spalled. There are two consequences:
- The contact layer thins, equivalent to added embedding and lowering preload;
- Microcracks form, becoming fatigue sources and eventually causing bolt fracture.
Diagnostic features: reddish-brown iron-oxide debris, annular wear marks, local bright spots (cold-weld traces) on the bearing surface or joint faces. Once started, fretting wear self-accelerates—the debris becomes abrasive particles that further intensify wear.
Control measures: raise bearing-surface pressure and hardness (hardened washers), use anti-fretting coatings, and eliminate or reduce relative slip (add anti-shear elements).
Path 5: "False Decay" from Insufficient Design Margin
This is not true material decay, but it looks the same in the field: preload is set too low at design; after deducting normal embedding and relaxation losses, residual preload is inherently insufficient to resist working load, so the joint quickly shows "loosened" in service.
The calculation is simple:
FM,res = FM × (1 − ηembedding) × (1 − ηrelaxation)
FM,res must still satisfy anti-slip and anti-separation criteria, with a 1.2~1.5× margin
Experience suggests the short-term residual ratio after assembly should be no less than 80%, and after long service no less than 60%. If the calculated result is below these values, initial preload must be raised or the joint faces improved.
Design and Process Points Against Decay
| Decay Path | Control Measure | Expected Effect |
|---|---|---|
| Embedding loss | Reduce joint-face count, improve surface quality, add hardened washers | Loss drops from 20% to 5%~8% |
| Stress relaxation | Use creep-resistant materials, raise initial preload, calculate over life | Residual ratio in high-temperature service rises by 10%~20% |
| Thermal-expansion mismatch | Choose materials with close expansion coefficients; increase clamp length to reduce stiffness sensitivity | Lower fluctuation amplitude from temperature cycling |
| Fretting wear | Raise bearing-surface hardness and pressure, add spigot or dowel, anti-fretting coating | Eliminate debris and microcrack sources |
| Design margin | Check anti-slip and anti-separation criteria using residual preload | Avoid "false decay" |
Conclusion
Preload decay and bolt loosening are two different failure paths: the former is loss of elastic elongation with the nut not turning; the latter is relative rotation of the thread pair. Only after judging which one applies will the countermeasure be correctly placed.
Of the five decay paths, embedding loss occurs within hours after assembly; stress relaxation and creep occur over the whole service life; thermal-expansion mismatch oscillates with temperature cycling; fretting wear self-accelerates under alternating load; and insufficient design margin amplifies the effect of all paths.
The core response is two things: build the decay margin into the design, so that residual preload still meets anti-slip and anti-separation criteria; and recover embedding loss in process by reducing joint faces, improving surface quality, delayed re-tightening and using Belleville washers. Do both, and long-service joints truly will not come loose.
