Page 22 - eBook_Proceedings of the International Conference on Digital Manufacturing V2

Proceedings of the International Conference on Digital Manufacturing –
Volume 2

modelling the dynamic response of non-linear structure system
with varying mechanical properties. Hashemi, Jang & Beheshti
(2023) developed surrogate FEA-ML based model to predict time-
varying displacement response of a two-dimensional 31 bar planar
truss structures. Four ML algorithms have been used to train ML
model; however, Extreme Gradient Boosting regression
outperforms other ML algorithms during model performance
evaluation. Badarinath, Chierichetti & Kakhki (2021) also applied
regression machine learning techniques as surrogate models for
finite element analysis to estimate real-time stress distribution in
one-dimensional beam structures.

In conclusion, multiple studies have been conducted to
optimise the design of PET bottles through finite element method
and experimental works; however, limited studies are available
regarding the structural and weight optimisation of HDPE bottles.
Moreover, the potential of Artificial Intelligence within structural
design is still under the phase of exploration. Therefore, this study
primarily focuses on establishing the standard operating
procedure (SOP) to optimise the overall weight of HDPE lubricant
bottles under stacking condition, which can be applied to all kinds
of bottles available in the local market. To outline the novelty of
this study, we have first conducted the non-linear static hyper-
elastic analysis of the HDPE lubricant bottle via two approaches:
optimising the can’s structure and varying bottle thickness using
parison controller programming by sectioning the bottle profile
into 100 points. FEA simulation is performed to generate the
dataset by varying the point thickness and then, the technique of
supervised machine learning is integrated by feeding that data to
predict the maximum deformation and equivalent stress.

METHODOLOGY

The flow of the present study is to establish the SOP for structural
optimisation of lubricants comprising of two key methodologies:
Finite Element Analysis modelling and Machine Learning. These
two sections act as a bridge for setting an innovative approach
towards structural and design optimisation of lubricant cans. As

6