A joint effort: Probabilistic methods for hybrid joint strength prediction

Abstract

This study investigates the strength prediction of hybrid joints that combine adhesive bonding with mechanical fasteners using probabilistic methods, focusing on four cold-curing adhesives: DP460, SW7240, SF479, and S370, which exhibit varying mechanical properties and performance outcomes; experimental results reveal that DP460 has the highest average joint strength at 329.4 ± 19.1 kN, while SF479 shows the lowest at 286.2 ± 42.7 kN, highlighting the significant influence of adhesive properties on load-bearing capacity. Previous studies had shown that the strength of hybrid joints was governed mainly by the adhesive's E-modulus, whereas bulk tensile strength dominated in single-lap bonded joints, indicating different failure mechanisms. A statistical analysis on the strength expressed in terms of hydrostatic pressure and maximum principal stress using Weibull distributions quantifies the variability in adhesive strength and predicts failure probabilities under non-uniform stress distributions, indicating that the bilinear failure criterion enhances prediction accuracy, particularly for adhesives like SW7240 and SF479, which demonstrate complex failure behaviours. Predictions versus experimental load capacities show that a linear failure criterion combined with a two-parameter Weibull distribution generally provides reasonable predictions with deviations of +1 % to +15 %, while the bilinear criterion can underestimate strengths for epoxies by up to 24 % and overestimate for elastoplastic materials like SF479. The study also addresses scatter in strength predictions, finding that the two-parameter Weibull distribution underestimates the lower bound (AVG – 1 × SD) by about 10 % and overestimates the upper bound (AVG + 1 × SD) by 18 %, with scatter analysis revealing that the probabilistic approach effectively captures variability in failure behaviour, despite challenges in maintaining consistent sample geometry and stress distributions, particularly in off-axis specimens. The study highlighted the need for accurate measurement of adhesive thickness and bolt pretension, showing that the probabilistic dimensioning method reliably predicted joint performance without empirical adjustments. It underscored the decisive role of adhesive properties and failure criteria, demonstrating that probabilistic methods improve load and failure predictions while supporting more reliable hybrid joint designs in engineering applications.