Novel molten powder interface layer (MPIL) method for multiphase simulation of directed energy deposition
Directed Energy Deposition (DED), is an additive manufacturing process, that is rapidly gaining traction in the industry for creating intricate geometries. However, due to complex physical phenomena, including melting, solidification, heat, and mass transfer, predicting mechanical properties in DED-...
Saved in:
Main Author: | |
---|---|
Other Authors: | |
Format: | Thesis-Doctor of Philosophy |
Language: | English |
Published: |
Nanyang Technological University
2024
|
Subjects: | |
Online Access: | https://hdl.handle.net/10356/173382 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Nanyang Technological University |
Language: | English |
Summary: | Directed Energy Deposition (DED), is an additive manufacturing process, that is rapidly gaining traction in the industry for creating intricate geometries. However, due to complex physical phenomena, including melting, solidification, heat, and mass transfer, predicting mechanical properties in DED-produced parts remains challenging. One of the critical phenomena that is not well known is the influence of molten powder distributive properties on melt pool characteristics. This is primarily due to the lack of technique and difficulties in incorporating molten powder information into powder-based DED modeling as detailed in section 2.6. Therefore, this research mainly focuses on developing a novel method that can incorporate molten powder distributive properties into the melt-pool simulation.
A novel method called Molten Powder Interface Layer (MPIL) was developed using Computational Fluid Dynamics (CFD) as detailed in section 3.3. The MPIL strategy enables the study of molten powder distributive properties such as mass, velocity, and temperature on melt pool characteristics of the DED process. The MPIL strategy demonstrated reasonable accuracy and trend between the simulation and experimental outcomes, indicating the validity of the proposed model as detailed in section 3.4. Using MPIL, the influence of laser power, powder mass flowrate, and scan speed on track geometry was established with trends consistent with the literature as detailed in section 3.5.
Furthermore, molten powder distributive properties are also affected by the nozzle geometry, hence, an integrated framework using MPIL that takes into consideration nozzle-influenced molten powder distributive properties was developed as detailed in section 4.3. The main contribution of this study consists of the integration of nozzle-influenced molten powder distributive properties. Information from validated powder flow simulation and experimental results from the literature was incorporated into the MPIL method. Using the integrated framework, the effect of powder mass flowrate, axial gas, and shielding gas on the melt pool geometry was established, with results consistent with the literature as detailed in section 4.4. A phenomenon was observed on the nozzle design impact on melt pool characteristics where a nozzle that produces a narrower bell-shaped distribution of molten powder mass is preferred as detailed in section 4.5.
In addition, physics-based simulations focusing on in-process information are integrated with experiments focusing on pre- and post-process information. Training a machine learning model with only experimental data that captures pre- and post-process information can reveal the general relationship of the DED process. However, the trained model cannot capture critical information derived from physics-based simulation that captures in-process information. Hence, numerical simulation was used to generate a high-quality physics-based dataset (in-process information) that cannot be measured experimentally. The in-process information was integrated with pre-process and post-process information as detailed in section 5.4.2, where the trained deep learning model exhibited enhanced prediction capability through the integration. |
---|