High cycle fatigue characterisation and modelling of 316L stainless steel processed by laser powder bed fusion

• Main Author: Zhang, Meng
• Other Authors: Li Hua
• Format: Thesis-Doctor of Philosophy
• Language: English
• Published: Nanyang Technological University 2020
• Subjects: Engineering::Materials::Mechanical strength of materials
• Online Access: https://hdl.handle.net/10356/139817
• Institution: Nanyang Technological University

Laser powder bed fusion (L-PBF) is an additive manufacturing (AM) process that uses high power lasers to selectively melt metal powders in a layerwise manner. It introduces unique process-structure-property relationships, understanding of which is essential for producing structural parts with consistent acceptable properties. In view of the ongoing efforts towards the development of standardised approaches for the determination of suitable processing conditions and part quality for L-PBF-processed parts, this PhD research aims to clarify links of the process-structure-property triad and develop methods for material property assessment. Particular attention was paid to the high cycle fatigue properties because of the strong influence of manufacturing defects on fatigue crack initiation and the high cycle fatigue resistance. Using stainless steel 316L as the model material, the critical crack-initiating defects under controlled changes in processing parameters were examined. The influence of processing on defect formation and the mechanisms of fracture were elucidated. Several guidelines for assessing the criticality of defects and the fatigue properties were highlighted. The effects of fatigue loading condition were considered for analysing the process-structure-property relationships. In the presence of tensile mean stress, variations in the plastic deformation properties, as demonstrated by the effects of build orientation and post-processing treatments (annealing or hot isostatic pressing), were found to induce cyclic plasticity, which led to strong fatigue-ratcheting interactions in the high cycle fatigue regime. The simultaneous actions of ratcheting and fatigue generate complex nonlinear relationships between the stress amplitude and mean stress parameters, such that the fatigue properties could not be effectively predicted using traditional stress-based models. A modification to the Goodman relation was proposed to account for the added effects of cyclic plasticity. Understanding of the fatigue and fracture behaviours served as the physical basis for the development and validation of predictive models via the ‘process-property’ and ‘structure-property’ routes. Statistical models, based on the response surface method, were developed for approximating the process-property relationship, while the structure-property relationship was modelled by considering the fatigue failure modes. The latter approach allowed the specification of a reduction factor on fatigue life to account for the intrinsic effects of microstructure inhomogeneity, as well as the construction of a modified Basquin equation that incorporates the defect fraction parameter for predicting the effects of lack of fusion defects on the S-N properties. Lastly, considering the numerous influencing factors arising from the process and the associated failure behaviours, a neuro-fuzzy-based machine learning method was applied to provide an effective unifying approach for high cycle fatigue life prediction. A dataset, consisting of most of the experimental fatigue data obtained in this study, was constructed for simulating a complex nonlinear input-output environment. Two models, by employing the processing/post-processing parameters and the static tensile properties respectively as the inputs, were developed from the training data. Despite the diverse fatigue and fracture properties, the models produced good prediction accuracy when checked against the test data. Moreover, the method demonstrated good generalisation capability on applying to literature data. The strong predictive power exemplifies the use of data science as a convenient and effective alternative to the traditional fatigue modelling approaches.