Prediction of mechanical and thermal properties in bronze-filled polyamide 66 composites using artificial neural network

Microcomposites based on polyamide 66 (PA66) reinforced with bronze powder in low contents (3, 5 and, 7 wt%) were prepared in a co-rotating twin-screw extruder. Mechanical performance, including tensile characteristics, impact resistance, and Shore D hardness, was evaluated. The results indicated that the elongation at break and impact strength decreased with the increase in bronze loading, while the hardness reached a maximum (15% enhancement) when using 7 wt% of bronze powder. Scanning electron microscopy (SEM) was utilized to analyze the fracture surface and study the toughening mechanisms. The thermal expansion coefficient, as a good indicator of dimensional stability, was measured by applying thermomechanical analysis (TMA). The experimentally measured mechanical and thermal properties were modeled by an artificial neural network (ANN) method. The network was trained by Levenberg–Marquardt back-propagation (LMBP) in a single hidden layer which is consist of five neurons. Based on the excellent consistency between the ANN predictions and empirical results, ANN models can be considered as a reliable tool to estimate and evaluate material properties before synthesis and manufacturing.