Page 52 - FDMAsia NovDec 2024

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

Andrii Shyp geometric constraints to the model in order to control how
The software allows the application of several types of

the material is removed.
All of them were tested in the early models, but, since

some of them did not show a significant optimisation or were
not pertinent to the tested model, only three were chosen for
the final analysis: “Planar symmetry”, “Rotational symmetry”,
“Point symmetry”. The “Planar symmetry” constraint forces
the optimised model to be symmetric about a specified plane,
the “Rota-tional symmetry” constraint forces the optimised
model to be symmetric about a specified axis and the “Point
symmetry” constraint forces the optimised model to be
symmetric about a specified point.

160 mm × 40 mm. The number of wooden samples used Based on the laboratory tests carried out within the
is 23. The discretisation of the model can be analysed in current project, a Monte Carlo analysis was carried out using
“discretisation of the model”. a proprietary script in which the material parameters were
For the optimisation, the topological optimisation method varied using the mean and the standard deviation resulting
was chosen, blocking the areas of application of the conditions from the laboratory tests.
on the contour and the area of application of the uniform For each model to which different geometric constraints
pressure. were applied, the script was applied to run 30 models with

In the optimisation process, the finite element density different mechanical parameters. The results were collected
was allowed to vary between 1 and 0.001 with a maximum in a database compared and contrasted.
variation per analysis cycle of 0.25. Three de-sign response In order to incorporate as close as possible the geometric
functions have been defined, namely “Energy Stiffness Measure”, results obtained by Monte-Carlo simulation, it was initially
“Volume” and “Signed Von Mises Stress”. proposed to use a generative machine learning model
The objective of the optimisation was to minimise the design (diffuser) which, starting from the synthetic data of each
response values for “Volume”. The optimisation task initially beam (opening, spans, loads), can generate topologically
applied to the model was to limit the volume variation below optimised glued laminated timber elements in elevation. The
45 percent. Unfortunately, the software applied the 45 percent first attempts resulted in a lack of alignment of the machine

material reduction directly to the model and then reshaped learning model.
it in order to withstand the increased interior stresses, so it Noting the lack of alignment of the generative model,
was not a correct approach. another machine learning method took the place of the current
Due to this shortcoming, in the final models, the optimisation one, namely an artificial neural network (ANN) that would
task was to have displacements under the allowable limits embed the results of topological optimisations and be able
specified in the current norms (beam span/200). This instructed to indicate, based on the data of the problem, a measure of
the software to gradually remove material until the allowable the stiffness of a topologically optimised beam in order to

displacement is reached, and then remodel it to create a be able to identify whether or not such optimisation would
more uniform distribution of stresses. produce satisfactory results from the design point of view.