Page 54 - FDMAsia NovDec 2024

52 MATERIALS NOV/DEC 2024 FDM ASIA | www.fdmasia.com

Freerange Stock chosen beforehand which means that further improvements
can be added for future models and experimental campaigns.
These parameters are related to the properties of the
wood material (type of wood, density, modulus of elasticity),

dimensions and then the discretisation of the model.
The results obtained herein indicate that in the case of
beam models, larger cross-section sizes and spans will benefit
even more from the topological optimisation.
Also, the topological optimisation task can be improved to
take into account the lamellar structure of glulam and optimise
the element such that the final geometry could rather be
obtained by placing the correctly cut lamella inside the element
Intelligence model is carried out based on numerical modelling without having to do a CNC machining step for the element.

using FEM analysis of a number of glued laminated wood Future developments will need to account for larger
samples and laboratory testing to confirm the correctness structural element sizes (cross-sections and span) and also
of the material model. take into account different loading conditions such as unevenly
The training of the AI model involves the use of bent distributed loads or point loads.
glulam beams characterised by a series of variable parameters, Also, it is clear that a great step in developing machine
obtaining a large number of topologically optimised models/ learning applications for this kind of task is to deploy
datasets (pairs of synthetic parameters—displacements). These a generative adversarial network architecture that can

models were used for ANN training and for the validation of eventually propose geometries that can then be optimised
the result. in fewer steps.
A deep neural network model is proposed herein with Besides, from this experiment we also found that
the goal of testing whether or not a useful alignment can be topological optimisation tasks usually require a large number
obtained from machine learning techniques for this kind of of iterations in order to reach the design goals. Therefore,
problem and this kind of optimisation task. embedding this information into machine learning models
The ANN network obtained can indicate based on the for structural elements belonging to the same topological
problem data a measure of the stiffness of a topologically groups will result in a faster design process since certain
optimised beam in order to be able to identify whether such aspects regarding structural behavior such as strength and

optimisation will produce satisfactory results from a design stiffness can be quickly estimated using Artificial Intelligence
point of view. techniques.
The resulting neural network can be used as it is for the Topologically optimised geometry propositions could be
task mentioned in the present work, being properly sized as obtained by employing generative machine learning model
to avoid overfitting issues. techniques which can propose geometries that are closer to
Further iteration on this neural network will be developed the topologically optimised results using FEM and as such
in order to improve the architecture in terms of size and present a starting point for the design analysis in a reduced

activation function for the task previously mentioned. amount of time. FDM
The models proposed for this paper follow certain parameters ENQUIRY NO. 6301