COMPARISON OF DIFFERENT MODELLING TECHNIQUES FOR PREDICTING WELD STRENGTH IN LASER TRANSMISSION WELDS OF OAK WOOD POWDER-REINFORCED POLYPROPYLENE PARTS

This paper investigates the feasibility of laser transmission welding to join a one hundred percent homopolypropylene transparent part with a fifteen percent by-weight white oak wood fibre-reinforced homopolypropylene absorbent part in a lap-joint configuration. The effects of laser power, welding speed, stand-off distance and clamp pressure on the weld strength were examined, alongside the use of response surface methodology (RSM), artificial neural networks (ANN), and adaptive neuro-fuzzy inference system (ANFIS) to predict weld strength. Root mean square error (RMSE), mean absolute error (MAE) and coefficient of determination (R²) were used to evaluate the models, considering the impact of fibre orientation, moisture content and heat conductivity. Results showed that stand-off distance was the most significant parameter which affects weld strength, followed by welding speed, while laser power and clamp pressure had minimal effects. R² values were 0.90, 0.93, and 0.99 for RSM, ANN and ANFIS, respectively, with RMSE values of 0.61, 0.48 and 0.29, and MAE percentages of 8.20, 6.10 and 3.90. These results suggest that all models effectively predicted weld strength, with ANFIS providing the highest accuracy, followed by ANN and RSM.