Aluminum structures are everywhere—from ships to bridges to lightweight vehicles—and their welded joints often operate under fluctuating loads that make fatigue a critical design constraint. Yet despite aluminum’s growing industrial relevance, fatigue design approaches for welded aluminum joints have historically lagged behind those for steel.

A new 2025 study published in the International Journal of Fatigue fills this gap by evaluating two modern, local fatigue-assessment methods:

• Effective Notch Stress (ENS)
• The 4R Method, including full elastic–plastic mean stress correction

The study also includes a global sensitivity analysis using SimDec to quantify how model parameters influence results.

Why this study matters

Designers increasingly rely on local fatigue approaches to predict fatigue strength accurately while enabling lighter, more efficient structures. Aluminum’s lower residual stresses, different hardening behavior, and strong mean-stress sensitivity make steel-based design rules unreliable without adjustment.

This study systematically re-examines ENS for aluminum and applies the 4R method to aluminum for the first time, using:

• 1031 experimental datapoints (various joint types, alloys 5xxx & 6xxx)
• Finite Element Analysis for notch stress factors (Fig. 4, p. 5)
• Monte Carlo + SimDec sensitivity analysis for the 4R method (Section 4)

What the researchers did

1. Built a massive fatigue database

Across butt welds, T-joints, attachments, cruciforms, and lap joints (see Fig. 3, p. 4), the study collected over 1000 fatigue results, all under constant amplitude loading. Residual stresses were found to be very low in most small coupon specimens (p. 6), simplifying some analyses.

2. Evaluated ENS for aluminum

• ENS uses a 1 mm fictitious notch radius (Fig. 1, p. 2).
• The IIW recommends a reference fatigue curve with Δσc = 71 MPa at 2 million cycles.

But the study found:

The IIW curve is nonconservative by a factor of ~1.25, especially in low-cycle fatigue (see Fig. 6, p. 7).
The recalibrated characteristic value was 57 MPa, not 71 MPa.

3. Applied the 4R method to aluminum for the first time

The 4R method uses:

• local elastic–plastic behavior
• real notch geometry
• mean-stress correction
• residual stress
• structural notch stress

This study adapted the 4R model to aluminum alloys by collecting or estimating Ramberg–Osgood parameters (Table 2, p. 5).

Results:

• 4R significantly reduced scatter (see Fig. 8, p. 8)
• It unified weld toe and root failures into a single S–N curve
• Performed especially well for high-cycle fatigue

4. Ran a full SimDec global sensitivity analysis

This is one of the first applications of SimDec to local fatigue modeling.

The sensitivity analysis (Fig. 10, p. 10) found:

Most influential parameters:

1. Nominal stress range Δσ
2. Mean stress (R-ratio)
3. Residual stress σres
4. Notch stress factor Kf (geometry & weld quality)

Almost no influence from:

• Monotonic & cyclic Ramberg–Osgood parameters (<2%)
• Material uncertainty has much lower impact than loading conditions

Why this matters:

The fatigue strength of aluminum welds is controlled far more by loading and geometry than by uncertainties in material properties.

Key Insights From the Study

1. ENS is usable but requires correction

• Current IIW curve is too optimistic for aluminum
• A revised ENS characteristic curve at 57 MPa is safer
• Considering mean stress via SWT improves accuracy (Fig. 7, p. 8)

2. 4R outperforms ENS in data consistency

4R captured:

• mean stress effects
• geometry effects
• residual stress variation

It produced tighter scatter, especially in high-cycle fatigue.

3. Mean stress matters more for aluminum than steel

The study confirms aluminum’s strong mean-stress sensitivity, which ENS (without correction) does not fully capture.

4. Residual stresses are low in small aluminum specimens

Most coupons had near-zero residual stress (p. 6), unlike steel.

5. For weld roots, geometry dominates

High Kf reduces importance of residual stress—plasticity overwhelms its influence.

Overall conclusions

The authors show:

• ENS alone is not fully reliable for aluminum unless adjusted
• 4R is accurate, adaptable, and unifies different failure locations
• SimDec sensitivity analysis quantifies uncertainty and validates assumptions
• Residual stress & mean stress are critical parameters
• Data analysis methods (aggregated vs. disaggregated) significantly influence results

This study sets the stage for modern fatigue design rules for aluminum structures—and demonstrates how simulation-based methods can support engineering decisions with greater confidence.

Reference

Havia, J., Ahola, A., Kozlova, M., Baumgartner, J., & Björk, T. (2025). Fatigue strength assessment of arc-welded aluminum joints by local approaches. International Journal of Fatigue, 193, 108803. https://doi.org/10.1016/j.ijfatigue.2024.108803

‍