Computational methods for micromechanics characterisation: Application to and experimental investigation of metal additive manufacturing and heat treatment processes
Tu, Yuhui
Tu, Yuhui
Loading...
Identifiers
http://hdl.handle.net/10379/17526
https://doi.org/10.13025/16680
https://doi.org/10.13025/16680
Repository DOI
Publication Date
2022-11-23
Type
Thesis
Downloads
Citation
Abstract
Powder bed fusion (PBF) is an additive manufacturing process in which finely focused thermal energy is used to selectively melt target regions within thin layers of metal powders. Due to the complex thermal conditions associated with the rapidly moving micro-melt pool and the layer-by-layer deposition of material, the resulting printed product in, for example, CoCr, Ti-6Al-4V, and 17-4PH stainless steel have unique and anisotropic microstructure, compared to traditional manufacturing processes. Another significant process within industrial PBF manufacturing is post-built heat treatment (HT). This step modifies the microstructure (as well as relieves residual stress), adding complexity to the processing parameters design. It is thus necessary to develop a further understanding of the process-structure (HT) and structure-property (PBF) relationships in PBF manufacturing. Multi-physics computational models are developed, based on experimental characterisation, to investigate the process-structure-property relationship in PBF. Particular attention is paid to model construction and maintaining a faithful representation of real PBF grains and sub-grain features. Direct microscopy microstructural characterisation is performed using electron backscatter diffraction of as-built and post-heat treatment specimens, where measures of texture and grain morphology are extracted and used to construct Voronoi tessellation-based or real image-based micromechanics models. Crystal plasticity finite element (CPFE) modelling is a micro-scale computational method to predict mechanical performance based on microstructure and crystallographic properties. CPFE model with physical dislocation mechanisms is employed to quantify the effect of the PBF microstructure variation (grain size, phase, morphology, and crystallographic orientation) on mechanical properties (tensile and fatigue). Phase-field method (PFM) is implemented to investigate the effect of heat treatment on grain growth and is then integrated with CPFE to offer an efficient method for an in-depth understanding of the relationship between thermo-processing, microstructural evolution and mechanical properties. This thesis also utilizes a large database of input-output samples from CPFE modelling to develop a trained deep neural network (DNN) model which instantly estimates the output (strength prediction) associated with a given input (microstructure) of multi-phase PBF stainless steels. The EBSD-based method for CPFE model generation is shown to give approximately 10% improved agreement for fatigue life prediction, compared with the more commonly-used Voronoi tessellation method. The effect of PBF inhomogeneity and post-built HT on the structure-property relationship was investigated using a dual-phase strain gradient crystal plasticity model based on physical dislocation mechanisms, based on the as-built and HT microstructure of Ti-6Al-4V, with a notable increase in lath width post HT. The effect of as-built lath width gradient in a single component on bulk stress-strain relationship exists but is minimal (1% variation in yield strength), whereas a greater effect (9% reduction in yield strength) is found in the post-heat treatment specimen. The PFM model is utilized for analysing the grain growth behaviour during the post-built heat treatment process through the simulation of grain boundary migration. It predicts five times grain growth in lath area from 0.59 to 3.0 𝑢𝑚2, after 100 minutes of annealing at 1127 K. The evolved lath prediction reaches close agreement compared to the EBSD measurement. Finally, the CPFE-based deep learning (DL) model exhibits high accuracy for the structure-property relationship as a surrogate predicting tool compared to CPFE while significantly reducing the computational cost to a few seconds.
Funder
Publisher
NUI Galway