Al-based optimization of surface roughness in wood band sawing using polynomial regression and differential evolution
Date Issued
2025-05
Author(s)
Selimi,Bujar
Jevtoska, Elena
Abstract
This study presents an artificial intelligence–assisted optimization framework
for minimizing surface roughness (Rₘₐₓ) in wood band sawing. A third-order
Polynomial Regression Model (PRM-3) was developed and trained on
experimentally collected data obtained under varied cutting conditions,
including angle, height, and feed rate. To ensure robust generalization,
configurations with a 60° cutting angle were deliberately excluded from training
and used solely for out-of-sample validation. PRM-3 was integrated with the
Differential Evolution (DE) algorithm to identify optimal process configurations.
For comparative purposes, a Gaussian Process Regression (GPR) model was also
implemented to evaluate the relative generalization capability. Results
confirmed the superior performance of PRM-3 in terms of accuracy, stability, and
generalization, compared to GPR, demonstrating high potential for deployment
in intelligent wood machining applications. The proposed framework represents
a valuable integration of interpretable modeling and automated optimization for
surface quality control in industrial settings
for minimizing surface roughness (Rₘₐₓ) in wood band sawing. A third-order
Polynomial Regression Model (PRM-3) was developed and trained on
experimentally collected data obtained under varied cutting conditions,
including angle, height, and feed rate. To ensure robust generalization,
configurations with a 60° cutting angle were deliberately excluded from training
and used solely for out-of-sample validation. PRM-3 was integrated with the
Differential Evolution (DE) algorithm to identify optimal process configurations.
For comparative purposes, a Gaussian Process Regression (GPR) model was also
implemented to evaluate the relative generalization capability. Results
confirmed the superior performance of PRM-3 in terms of accuracy, stability, and
generalization, compared to GPR, demonstrating high potential for deployment
in intelligent wood machining applications. The proposed framework represents
a valuable integration of interpretable modeling and automated optimization for
surface quality control in industrial settings
Subjects