IFAC PapersOnLine 52-13 (2019) 403–408
ScienceDirect ScienceDirect
Available online at www.sciencedirect.com
2405-8963 © 2019, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
Peer review under responsibility of International Federation of Automatic Control.
10.1016/j.ifacol.2019.11.083
© 2019, IFAC (International Federation of Automatic Control) Hosting by Elsevier Ltd. All rights reserved.
1. INTRODUCTION
Fused Deposition Modeling (FDM) is an additive
manufacturing technology that produces three dimensional
parts by vertically stacking layers of extruded semi-molten
materials. Since its inception in the 1980s, FDM has
stimulated great interest due to key drivers like product
customization, low cost and complex functional part
manufacture without tooling (Mohamed et al. (2015); Mishra
and Mahapatra (2018)). Despite these benefits, there are
challenges with FDM technology such as lack of high
accuracy and desired mechanical properties as well as its
heavy reliance on the adjustment of process parameters
(Mohamed et al. (2016)). In this regard, many research
studies have focused on understanding and optimizing FDM
process parameters for the enhancement of various
performance characteristics like part surface quality,
dimensional accuracy as well as mechanical properties in
order to extend the application areas of FDM technology
(Mohamed et al. (2016); Wu et al. (2018)).
In the study undertaken by Boschetto et al. (2013), a FDM
surface roughness prediction model was developed for all
ranges of the deposition angle. Multiple feed-forward
Artificial Neural Network (ANN) structures were generated
to fit experimental data by varying the number of neurons
and activation function type. Following this, an evaluation
function was utilized to select the best ANN model based on
performance criteria. Accordingly, validation of the proposed
surface roughness model was achieved using various
materials and different FDM machines, thus demonstrating
ANN suitability for surface quality prediction in FDM
fabrication.
Sood et al. (2009) performed experimental investigations on
the effect of FDM process parameters such as layer thickness,
part build orientation, raster angle, raster to raster air gap and
raster width on the dimensional accuracy of a standard
acrylonitrile styrene butadiene (ABS) test specimen.
Furthermore, the authors implemented grey Taguchi method
in order to generate a single response from the dimensional
performance characteristics: percentage change in length,
width and thickness. Finally, ANN model was developed for
prediction of the dimensional performance characteristics.
The proposed model proved its suitability for use due to a
small percentage of error between predicted and experimental
data.
Rodríguez-Panes et al. (2018) present a comparative study of
the tensile strength of test specimens produced by FDM
technology using two common polymer materials: ABS and
polyactic acid (PLA). The effect of process parameters such
as layer height, infill density and manufacturing orientation
on tensile yield stress, tensile strength, nominal strain at
break and the modulus of elasticity were investigated for each
material. It was observed that the adjustment of layer height
had little effect on mechanical strength for ABS whereas for
PLA, an increase in layer height resulted in a reduction of
tensile strength by 11%. Generally, infill density had the
greatest influence on the results, with a more apparent effect
on PLA. In essence, the test specimens manufactured with
PLA demonstrated more rigidity and higher tensile strength
than ABS.
A common trend from the existing literature, however,
reveals that studies mainly focused on the static mechanical
properties whereas practical applications of FDM parts in
machinery and transportation like gears, propellers and
Keywords: Fused Deposition Modeling, Artificial Neural Network, Machine Learning, Natural Frequency
__________________________________________________________________________________________________________
__________________________________________________________________________________________________________
Abstract: This research study demonstrates the use of machine learning tools for the prediction of
dynamic mechanical characteristics of parts produced by the Fused Deposition Modeling (FDM) process.
In this regard, I-optimal design of experiments was followed with raster angle, air gap, build orientation
and number of contours as independent variables together with natural frequency as the mechanical part
characteristic for investigation. Accordingly, a Artificial Neural Network (ANN) model was trained using
the Bayesian regularization function. Finally, the trained ANN model was validated by performing
multiple confirmation runs which provided predictions generally within 5% accuracy.
*Department of Mechanical and Manufacturing Engineering,
The University of the West Indies, St. Augustine, Trinidad and Tobago,
(e-mail:Fahraz.Ali@sta.uwi.edu, Boppana.Chowdary@sta.uwi.edu)
Fahraz Ali* Boppana V. Chowdary*
Natural Frequency prediction of FDM manufactured parts
using ANN approach