AyFait's Blog

Go to Medium

SimsLayer (Not Yet Functional)

FEA simulation of 3D-Printed Parts

This idea is birthed when I asked myself if there are efficient simulation tools specifically designed for 3D-Printed parts. Because there is no doubt of assumption that they have entirely different mechanical properties due to the manufacturing process involved.

Why 3D-Printed Parts needs specialized simulation tools:
While general Finite Element Analysis (FEA) software can be used, dedicated additive manufacturing (AM) software is far superior for this specific purpose because it understands:
= Layer-by-layer material addition rather than assuming the entire part exists at once.
= Directional material properties (anisotropy), especially for fiber-reinforced filaments.
= Thermal history of the part throughout the build.
= Structural integrity as it relates to directional loading due to layer-by-layer orientation of the part(s)

This is particularly vital for metal 3D printing, where material costs are high and failure is expensive!

Stages Involved:

> Geometry or model is prepared in a CAD software
> Geometry is either sliced using an external slicer or SimsLayer slicer
> Geometry is imported into SimsLayer
> Geometry is sliced if not already
> GNN optimizes layers
> Preprocessing is carried out
> Meshing
> GNN optimizes mesh
> FEA
> GNN suggests optimization for geometry

Read more...

This is the content for section 2.

This is the content for section 3.

One Click Data Cleaner Algorithm (DCA)

Wrote this code after getting frustrated over the repeated process of data cleaning while just starting my ML journey back in July 2024.

Domain knowledge of dataset is highly needed before using the algorithm.
Assumes that the .csv file has a single row of defined header/title (starting point, first row).

Scope:
> Takes raw dataset
> Goes through each cell, fixing using outlined instruction
> Gives back cleaned dataset

Structure:
> Fills empty cells using interpolation
> Detects categorical columns
> Maps non-numerical elements to numbers
> Detects alpha-numerical/object-numerical cells

Some columns get deleted or "dropped" from the dataset if it does not meet a threshold (which can be changed as deemed fit), like number of empty cells, number of full "char" cells. Resulting in smaller features in the output dataset.

Read more...