Modeling ultrasonic welding of polymers using an optimized artificial intelligence model using a gradient-based optimizer

In this study, a new hybrid artificial intelligence approach is proposed to model the ultrasonic welding of a polymeric material blend. The proposed approach is composed of an ensemble random vector functional link model (ERVFL) integrated with a gradient-based optimizer (GBO). First, welding experiments were conducted on acrylonitrile butadiene styrene (ABS) and polycarbonate (PC) blends produced by the injection molding method. The experiments were designed according to the L27 orthogonal array considering three process factors (applied pressure, welding time, and vibration amplitude) and two responses (average temperature and joint strength). Then, the obtained experimental data were used to train the developed model. To verify the accuracy of the model, it was compared with standalone ERVFL in addition to two fine-tuned ERVFL models (ERVFL-SCA and ERVFL-MRFO) in which ERVFL is incorporated with sine cosine algorithm (SCA) or Manta ray foraging optimization (MRFO). The four models were evaluated using five statistical tools. ERVFL-GBO has the highest coefficient of determination and the lowest root mean square error, mean relative error mean absolute error, and coefficient of variance compared with other models which indicate its high accuracy over other tested models.