Multiaxial Fatigue Evaluation of Structural Steels in Heavy Duty Vehicles : Comparison of Uniaxial and Multiaxial Fatigue Approaches for Non-Proportional Stress States

Sammanfattning

Structural components in heavy-duty vehicles are subjected to complex loading conditions during service, where stresses act simultaneously in multiple directions and change orientation over time. Conventional fatigue assessment methods are typically based on simplified assumptions that do not fully capture this complexity, which can lead to either overly conservative or potentially unsafe life predictions. This thesis evaluates and compares uniaxial and multiaxial fatigue criteria applied to stress histories representative of bus structural components, with the aim of identifying which methods are appropriate under which loading conditions and whether the added complexity of multiaxial approaches is justified in practice. Two categories of methods are examined. The first concerns endurance-limit criteria, which assess whether a component is expected to sustain a given loading indefinitely. Three multiaxial criteria Crossland, Dang Van, and Findley are implemented and compared against conventional scalar stress measures. Comparisons are performed on synthetic load cases, literature-based experimental data, and finite element stress histories from a transient vehicle simulation provided by Volvo Buses. The second category concerns damage accumulation models, which estimate the total fatigue damage produced by a variable-amplitude loading history. A simplified uniaxial approach based on the maximum principal stress is evaluated alongside three multiaxial critical-plane methods, all implemented in Python. The endurance-limit results show that conventional scalar methods fail to represent fatigue severity for shear-dominated, non-proportional, and equibiaxial loading conditions. Among the multiaxial criteria, Crossland produces the most conservative estimates, while Dang Van and Findley demonstrate stronger sensitivity to the combined influence of shear and normal stress on physically relevant material planes. For damage accumulation, the simplified maximum principal stress method overestimates damage by a factor of 10–30 under tension-dominated loading and completely fails to detect damage under pure shear and rotating load conditions both of which are representative of real vehicle service histories. The multiaxial critical-plane methods are shown to produce physically consistent results across all evaluated load cases, with negligible additional computational cost relative to the finite element analysis that generates the input.