Predicting melt pool behaviour in LPBF through high fidelity modelling

Reynolds, William J. (2022) Predicting melt pool behaviour in LPBF through high fidelity modelling. PhD thesis, University of Nottingham.

PDF (Submission after minor changes) (Thesis - as examined) - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader
Available under Licence Creative Commons Attribution.
Download (10MB) | Preview


Empirical process parameter optimisation for laser powder bed fusion (LPBF) is a long, arduous process that inefficiently locates process parameters that may or may not be optimal. It is often circumstantial; meaning it is dependent on the thermodynamic regime at a given location within a component. In order to obtain mastery of this process we must control the energy distribution delivered to the melt pool.

Reducing the porosity and cracking of components made through LPBF is still the largest barrier for entry in high integrity applications. Optimising laser power and speed are considered two of the most influential factors in LPBF, which drive track consolidation, reducing porosity. These parameters though, are difficult to link to desirable melt pool properties, without extensive experimental testing. Controlling the process via surface temperature has been shown to substantially improve processing, as well as provide a more direct link with the process. However, no literature examines which melt pool temperatures are optimal in producing highly dense components.

The primary novelty in this work is the derivation and validation of a novel optimisation method that monitors and modulates peak melt pool temperatures by varying the laser power administered to the melt pool through computational modelling. Understanding optimal melt pool peak temperatures is a valuable approach since increasingly, machine tools are able to monitor surface temperature and, in the future, attenuate energy density accordingly. This work provides a framework in which porosity and cracking defects can be eliminated through optimal process parameter calculations. This process parameter optimisation utilises high performance computing facilities to simulate LPBF at small time durations, with high resolution. This aspect has been shown to be critical in providing detailed information for the evaluation of cracking behaviours and the prediction of optimal processing conditions.

This work makes use of a ‘high fidelity model’ to accurately and precisely measure and evaluate conditions within LPBF, in addition to providing a platform for the process optimisation to run from. Model fidelity here describes the exactness of the degree to which something is copied or produced. For this model, this involves including fundamental physics that describe how photons in the laser beam are reflected and absorbed into the metal surface, rather than using approximations to generalise the heat input. For example, a low fidelity heat source would use a volumetric approach to heat the metal powder particles. This includes assumptions as to the absorptivity of the powder bed, how deep that particular laser would heat the powder bed, and therefore would have a larger error when comparing the temperature on the powder surface to experiments. A high fidelity heat source, that will be used in this work, replicates reality to a higher degree, so that temperatures can be modelled with greater accuracy. In this instance, ray tracing has been proven in literature to accurately reflect the behaviour of photons interacting with the metal surface, so that the absorptivity of the metal is not approximated, but calculated using the Fresnel equations [1]. A high fidelity model does not have to be non-linear, but often is due to temperature dependant properties. In this work, complex physical phenomena such as: melting and solidification, buoyancy effects, surface tension, temperature dependant surface tension (Marangoni flow), vaporisation phenomena including recoil pressure and evaporation flux, melt pool flow and variable absorptivity from the Fresnel equations, and a stochastic powder bed are used to replicate LPBF through numerical modelling. The model is designed to be flexible and user friendly, to allow the modification of every aspect of the laser, powder and other process conditions.

Three main research chapters are presented in this work, along with an extensive literature review covering the complex nature of LPBF, combined with a methodology to create a high fidelity model to simulate this process. Results have been validated with four different validation experimental to model comparisons. These validations include;

 Validation against a continuous laser system for 316L stainless steel in conduction mode

 Validation against a continuous laser system for 316L stainless steel in keyhole mode

 Validation against a ramp up laser profile for 316L stainless steel

 Validation against a pulsed laser system for AA204 for two different parameter sets.

These results include validation with a variable pulsed laser system for 316L stainless steel, to ensure that the model can be used with variable laser conditions, as well as a more typical continuous wave laser. In order to replicate LPBF single track deposition of high strength aluminium alloy AA2024, a revised refractive index was tailored to experimental conditions. Literature shows how literature values of absorptivity of aluminium do not represent the measured behaviour in LPBF. For the first time, this work combines a revised refractive index to simulate absorptivity of AA2024, whilst keeping a high fidelity ray tracing heat source. This allows for detailed analysis of the development of porosity, which is vital in determining track consolidation.

The main body of this work comprises of a predictive framework for LPBF parameter optimisation. This framework has been proven for two different material systems, 316L stainless steel and AA2024. Using an inverse solution, a link is created between laser power and melt pool temperature, allowing for stable melt pools to be established and maintained throughout a single track deposition. Links have been established with surface temperature and common LPBF defects, such as a lack of fusion defects and keyhole porosity. A new parameter, P-ratio, has been derived and shown to work as a single processing variable that LPBF can be optimised too. P-ratio has proven to be independent of laser speed, and provides a constant recoil pressure to the surface of a melt pool to ensure track consolidation. This work uses a case study on AA2024, to validate the inverse solution with a second material system. The inverse solution correctly identifies a laser power at a set point distance and exposure time that gives the highest density, compared to samples validated in experimental conditions.

Additionally, the model has successfully demonstrated its use in measuring and evaluating cracking for high strength aluminium alloy, AA2024. Data generated by the model clearly shows differences between cracked and crack free parameter sets. Large differences in cooling rates and temperature gradients show that the main drivers for cracking in LPBF can be evaluated with the high fidelity model. Results from this work will be used to predict and eliminate cracking in high strength aluminium alloys.

Item Type: Thesis (University of Nottingham only) (PhD)
Supervisors: Clare, Adam
Simonelli, Marco
Grant, David
Keywords: Additive manufacturing; Computer simulation; Stainless steel; Alloys; Aluminum; Melt pool
Subjects: T Technology > TS Manufactures
Faculties/Schools: UK Campuses > Faculty of Engineering
Item ID: 69201
Depositing User: Reynolds, William
Date Deposited: 31 Jul 2022 04:42
Last Modified: 31 Jul 2022 04:42

Actions (Archive Staff Only)

Edit View Edit View