Numerical modelling of heat transfer and experimental validation in powder-bed fusion with the virtual domain approximation

Among metal additive manufacturing technologies, powder-bed fusion features very thin layers and rapid solidification rates, leading to long build jobs and a highly localized process. Many efforts are being devoted to accelerate simulation times for practical industrial applications. The new approac...

Full description

Saved in:
Bibliographic Details
Published in:Finite elements in analysis and design Vol. 168; p. 103343
Main Authors: Neiva, Eric, Chiumenti, Michele, Cervera, Miguel, Salsi, Emilio, Piscopo, Gabriele, Badia, Santiago, Martín, Alberto F., Chen, Zhuoer, Lee, Caroline, Davies, Christopher
Format: Journal Article Publication
Language:English
Published: Amsterdam Elsevier B.V 01.01.2020
Elsevier BV
Subjects:
Additive manufacturing (AM)
Anàlisi numèrica
Approximation
Computer simulation
Computing costs
Enginyeria dels materials
Fabricació
Finite element method
Finite elements (FE)
High performance computing (HPC)
Industrial applications
Manufacturing processes
Matemàtiques i estadística
Mathematical models
Metal·lúrgia
Models matemàtics
Mètodes en elements finits
Parameter estimation
Powder-bed fusion (PBF)
Rapid solidification
Selective laser melting (SLM)
Thermal Analysis
Thin films
Àrees temàtiques de la UPC
High performance computing (HPC)
Powder-bed fusion (PBF)
Finite elements (FE)
Additive manufacturing (AM)
Thermal analysis
Selective laser melting (SLM)
ISSN:0168-874X, 1872-6925
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:Among metal additive manufacturing technologies, powder-bed fusion features very thin layers and rapid solidification rates, leading to long build jobs and a highly localized process. Many efforts are being devoted to accelerate simulation times for practical industrial applications. The new approach suggested here, the virtual domain approximation, is a physics-based rationale for spatial reduction of the domain in the thermal finite-element analysis at the part scale. Computational experiments address, among others, validation against a large physical experiment of 17.5 [cm3] of deposited volume in 647 layers. For fast and automatic parameter estimation at such level of complexity, a high-performance computing framework is employed. It couples FEMPAR-AM, a specialized parallel finite-element software, with Dakota, for the parametric exploration. Compared to previous state-of-the-art, this formulation provides higher accuracy at the same computational cost. This sets the path to a fully virtualized model, considering an upwards-moving domain covering the last printed layers. •Novel physics-based rationale for domain virtualization of regions of low physical relevance (e.g. powder bed, build plate) in the part-scale thermal finite-element analysis of metal Additive Manufacturing (AM) processes.•New method offers better compromise between accuracy and efficiency (in predicting temperature values and average rates of change) than previous approaches for spatial reduction of the domain.•Application of a high-end HPC computational framework in the context of verification and validation in metal AM. The framework combines (1) a fully parallel FE model for the simulation of AM processes and (2) a suite of iterative mathematical and statistical methods for parametric exploration of computational models.•Application of the HPC framework enabled practical and automatic parameter estimation at the large scale of the physical experiment considered in the work: 17.5 [cm3] of deposited volume in 647 layers and 3.5 [h] of process time.•Potential to push ahead the proposed method for full model virtualization, where the domain of analysis is reduced to the last printed layers and it moves upwards in time to follow the growth of the geometry.
Bibliography:ObjectType-Article-1
SourceType-Scholarly Journals-1
ObjectType-Feature-2
content type line 14
ISSN:0168-874X
1872-6925
DOI:10.1016/j.finel.2019.103343