Optimization of joint strength in friction stir welded wood plastic composites using ANFIS and Cheetah Optimizer

Abstract

This study focuses on modeling and optimizing the friction stir welding (FSW) process of wood-plastic composites (WPCs) made of low-density polyethylene reinforced with wood flour to improve joint performance. The input parameters considered are rotational speed (RS) and welding speed (WS), while the output properties are flexural strength and modulus, measured after tool entry and before tool exit. Three machine learning algorithms—multilayer perceptron (MLP), decision tree (DT), and adaptive neuro-fuzzy inference system (ANFIS)—were used to model the relationships between the input parameters and output responses. The ANFIS model showed the best predictive performance, with R2 values above 0.98 and minimal errors, indicating its reliability in modeling FSW properties. Optimization using the ANFIS model and the Cheetah Optimizer determined the optimal parameters of 1116 RPM for RS and 0.20 mm/s for WS, which resulted in a maximum flexural strength of 13.96 MPa after tool entry and 13.50 MPa before tool exit, along with consistent flexural modulus values. The study emphasizes the importance of uniform heat distribution to prevent polymer degradation and improve weld quality. Verification experiments showed the optimization model's effectiveness, with relative errors of 13.42% for entry strength and 5.50% for exit strength. This research demonstrates that combining machine learning and metaheuristic optimization can significantly enhance FSW joint performance in WPCs, offering insights for improving strength, consistency, and thermal stability in WPC welding technology.

Keywords

• Friction stir welding
• Wood-plastic composite
• Joint properties
• Machine learning
• Cheetah optimizer