Page 37 - IJAMD-1-1

International Journal of AI
for Material and Design ML for quality improvement in L-PBF

optimize hyperparameters for four ML models separately.
This approach proved effective in enhancing the models’
predictive capabilities for determining the minimum creep
rate of alloy 718 samples in L-PBF. 18
2.2.3. Genetic algorithm
A genetic algorithm is an evolutionary search algorithm
that mimics the process of natural selection, wherein the
algorithm’s data structure resembles a chromosome. The
process of the genetic algorithm begins with a randomly
generated population of chromosomes, followed by
population evaluation. If an optimal solution is identified,
the process concludes; otherwise, the algorithm undergoes
selection, crossover, and mutation operations to obtain a
new generation. 23,26 A brief flowchart outlining the genetic
algorithm is presented in Figure 8. In a study by Rong-Ji
et al., a genetic algorithm was employed to optimize
hyperparameters and apply an artificial NN (ANN) for
simulating the shrinkage ratios of L-PBF components under
different process parameters. Another study by Zouhri
27
et al. used a genetic algorithm to fine-tune hyperparameters
for SVMs, achieving a remarkable 93% prediction accuracy
for the density in L-PBF. In addition, they tested 8,190
different hyperparameter combinations using grid search
to identify an optimal multiple linear process ANN. This
model demonstrated superior performance and accuracy
over SVM in predicting L-PBF density. 28

3. Quality improvement
Quality improvement in the L-PBF refers to the
implementation of a series of measures and methods aimed
at enhancing the quality of printed products. The objective
of these improvement initiatives is to reduce defects and
inconsistencies in the manufacturing process, ultimately
yielding high-quality and reliable products. Such quality
enhancements are crucial for compliance with industry
standards, fulfilling customer requirements, and ensuring Figure 8. The flowchart of the genetic algorithm.
the reliability of the manufactured components. 29
mechanical strength requirements. Finally, it is imperative
The efforts for quality improvement encompass a to ensure that the internal structure of the product is dense,
range of strategies, including process optimization, quality reducing porosity and increasing relative density, all while
inspections and testing, real-time process monitoring maintaining residual stresses within appropriate limits. 30
and feedback, post-processing, and surface treatment.
Numerous criteria are employed to evaluate the quality In the actual fabrication process, the detection and
of printed components, and an excellently crafted printed adjustment of process parameters rely on the operator’s
part should adhere to the following standards. First, it must experience and do not provide real-time process
ensure dimensional accuracy, aligning precisely with the monitoring and feedback to assess product quality. In
design specifications. Second, there is a need to enhance addition, L-PBF involves a multitude of parameters, and
the surface quality of the printed parts by minimizing traditional methods struggle to manage this complexity.
surface roughness and preventing the occurrence of surface Figure 9 summarizes the parameters related to each quality
defects or indentations. Third, it is essential to improve indicator. Therefore, to effectively improve product quality,
the mechanical properties of the printed parts, such as ML has become an increasingly mainstream approach. ML
tensile strength, hardness, and fatigue life, to meet specific is better equipped to handle and analyze large-scale data

Volume 1 Issue 1 (2024) 31 https://doi.org/10.36922/ijamd.2301