78
ALRAQABA . ISSUE 15
Manufacturing refers to the process of making
products from raw materials using manual
labour or machinery and typically performed
systematically with a division of labour. The
objective of the thesis is to help manufacturing
industries identify potential defects in product
quality as early as possible (even from the first
batch) without additional costs to manufacturing.
Python software was used to create a code (using
an artificial intelligence method called Random
Forest) to help audit the manufacturing process
from a small dataset. From a sample, the code
can predict the quality and performance of the
output with an accuracy of %95–92 (where the
previous accuracy was around %87–82).
The whole idea of the thesis is simple. Consider
a company that manufactures 1,000 pens, where
the defective pens account for about %7–5 of the
whole batch when using current machine learning
methods with different constraints, such as the
data being no less than 10,000 points, which
needs domain experts with a background in data
manipulation and requiring much time to predict,
sometimes more than a year.
Therefore, the author created a new method
called «MRF» to overcome those constraints
with fewer defects of only around %4–2, which
provides a better solution for future manufacturing
by modifying the output quality .
Furthermore, much material is involved in
manufacturing pens (rubber, ink, plastic, etc.).
With the MRF method, the materials can be
modified to increase the pens’ quality with
fewer defects.
Thesis Summary
Manufacturing
Process Causal
Knowledge
Discovery using
a Modified
Random Forest-
based Predictive
Model
Mishari Abdul-majeed Ibrahim
Associate Engineer - Mechanical Engi-
neering Technical Support Division
Theses
