Article (Scientific journals)
Characterization, propagation, and sensitivity analysis of uncertainties in the directed energy deposition process using a deep learning-based surrogate model
Pham Quy Duc, Thinh; Truong Vinh Hoang; Xuan Van Tran et al.
2022In Probabilistic Engineering Mechanics, 69, p. 103297
Peer Reviewed verified by ORBi
 

Files


Full Text
Thinh Pham Probab Eng Mech J 2022.pdf
Publisher postprint (4.06 MB)
Download

All documents in ORBi are protected by a user license.

Send to



Details



Keywords :
Statistical and Nonlinear Physics; Directed Energy Deposition; Deep Learning; M4 Steel
Abstract :
[en] Uncertainties raised from process parameters, material properties, and environmental conditions significantly impact the quality of the printed parts in the directed energy deposition (DED) process. In this study, we perform the characterization, propagation, and sensitivity analysis of the uncertainties in the DED process using deep learning (DL)-based surrogate model to investigate the influence of the uncertain input parameters on the quality of the final printed product. A DL-based surrogate model is first constructed using the offline data obtained from a finite element (FE) model, which was validated against experiments. Sources of uncertainties are characterized using the probabilistic method and are propagated using the Monte-Carlo (MC) simulation. Moreover, we perform the sensitivity analysis (SA) to determine the most influential sources of uncertainty and two potential use cases based on uncertainty quantification results Owing to the fast execution time of the surrogate model, the MC simulation is significantly efficient in terms of computational resources as compared to the simulation that is solely based on the FE model. It is shown that the investigated sources of uncertainty contribute substantially to the inconsistency of the final product quality as they induce variations in temperature field, cooling rate at the liquidus point, and melting pool sizes. These quantities are also strongly dependent on the clad height. Based on the SA results, the laser power, the scanning speed, the heat convection, and the thermal conductivity induce the most uncertainties to the melting pool sizes. Two potential use cases are assessed to further demonstrate the usefulness of UQ results (i.e., uncertainty reduction). In general, these findings provide valuable insights for the process parameter optimization of the DED under uncertainty to improve the quality of the final printed parts.
Disciplines :
Materials science & engineering
Author, co-author :
Pham Quy Duc, Thinh ;  Université de Liège - ULiège > Aérospatiale et Mécanique (A&M)
Truong Vinh Hoang;  RWTH Aachen - École supérieure polytechnique de Rhénanie-Westphalie Aix-la-Chapelle [DE]
Xuan Van Tran;  Thu Dau Mot
Quoc Tuan Pham;  Ton Duc Thang University
Seif Fetni;  ULiège - University of Liège [BE]
Duchene, Laurent  ;  Université de Liège - ULiège > Département ArGEnCo > Analyse multi-échelles dans le domaine des matériaux et structures du génie civil
Tran Son Hoang;  ULiège - University of Liège [BE]
Habraken, Anne  ;  Université de Liège - ULiège > Département ArGEnCo > Département Argenco : Secteur MS2F
Language :
English
Title :
Characterization, propagation, and sensitivity analysis of uncertainties in the directed energy deposition process using a deep learning-based surrogate model
Alternative titles :
[fr] Caractérisation, propagation et analyse de sensibilité des incertitudes dans le processus de dépôt par énergie dirigée à l'aide d'un modèle de substitution basé sur l'apprentissage profond.
Original title :
[en] Characterization, propagation, and sensitivity analysis of uncertainties in the directed energy deposition process using a deep learning-based surrogate model
Publication date :
17 May 2022
Journal title :
Probabilistic Engineering Mechanics
ISSN :
0266-8920
eISSN :
1878-4275
Publisher :
Elsevier BV
Volume :
69
Pages :
103297
Peer reviewed :
Peer Reviewed verified by ORBi
Name of the research project :
EDPOMP
Funders :
Vingroup Innovation Foundation (VINIF)
Funding number :
project code VINIF.2020.DA15
Available on ORBi :
since 27 July 2022

Statistics


Number of views
106 (13 by ULiège)
Number of downloads
54 (4 by ULiège)

Scopus citations®
 
7
Scopus citations®
without self-citations
4
OpenCitations
 
6

Bibliography


Similar publications



Contact ORBi