Investigation of Size Effects for Fatigue Life Assessment in Heavy Machinery

Large-scale industrial components are subject to cyclic loading throughout their operational lifetime, requiring accurate fatigue life estimation for maintenance planning and failure prevention. Fatigue properties of laboratory specimens cannot be reliably transferred to these components due to size effects that cause systematic reduction in fatigue strength with increasing dimensions. This thesis establishes an application-oriented methodology for quantifying and modeling size effects in two industrially relevant alloys: the cast brass CuZn35Mn2Al1Fe1-C-GS and the quenched and tempered steel 27NiCrMoV15-6. To investigate the cast brass material, specimens were extracted from a component exhibiting cooling-rate-driven microstructural variations with grain sizes ranging from 1 to 9 mm. Fracture mechanical characterization establishes cyclic R-curves and NASGRO crack propagation parameters at multiple load ratios, revealing higher crack growth resistance for fine-grained microstructures and extending the fracture mechanical database for cast brass alloys. High cycle fatigue testing investigates technological size effects arising from grain size differences and statistical size effects through systematic variation of highly stressed volume. The interaction between technological and statistical size effects is modeled by integrating the Hall-Petch grain size relationship into the volumetric size effect model, enabling the Weibull exponent ¿ to vary as a function of grain size and highly stressed volume. Statistical size effects are quantified experimentally across highly stressed volumes spanning 1800 to 55000 mm³ for cast brass and 0.60 to 42000 mm³ for quenched and tempered steel. Testing encompasses notched and unnotched specimens under tension-compression and rotating-bending loading conditions, establishing fatigue strength reference values for large-scale component design. For the steel material, an extension to the volumetric size effect model is proposed featuring a volume-dependent Weibull exponent ¿(V) that transitions asymptotically between limiting values, extending the applicability across multiple scales of highly stressed volume. Quantification of non-metallic inclusion and fish-eye size distributions using generalized extreme value statistics reveals relationships between highly stressed volume and critical defect dimensions, validated through Murakami's defect-based model and Kitagawa-Takahashi analysis with good agreement to experimental fatigue strengths. This methodology enables fatigue life estimation for large-scale industrial components, supporting maintenance planning and failure prevention in critical applications.