Optimization of wear characteristic of epoxy composites by response surface and artificial neural network methodologies

Epoxy is widely used in many industries including automotive and construction. In thesis work, ferrochrome carbide and glass reinforcements have been used to enhance wear resistance of epoxy material. Total number of 54 specimen is produced on 18 different experimental points determined by employing Central Composite Design (CCD) approach. For each specimen, abrasive wear test is performed in laboratory environment. Findings are interpreted and certain conclusions have been drawn from the results via response surface and neural network approach. Glass reinforcement is found to have a negative correlation with wear rate of epoxy which means that the more glass is added to epoxy, the more wear resistance epoxy gets. Minimum wear rate is obtained when glass reinforcement is at its maximum level (17,07%). Ferrochromium reinforcement increases wear rate when small sized particles (0-50 microns) is concerned whereas it decreases wear rate for large sized particles (50-75 microns). Compare to large sized particles, small sized particles have a significant effect on wear resistance in a positive way. Hardness and production cost of samples are included to the work as additional response variables. Optimization with multiple response is conducted to find out optimum design points for different scenarios predetermined. Many artificial neural network alternatives are trained and tested. Among all the alternatives, the best predictive network is selected and compared with response surface model (RSM) prediction in terms of prediction accuracy. According to mean absolute percentage error (MAPE), neural network outperforms RSM model. When it comes to R2 and mean square error (MSE) performance indicators, RSM model achieves better results.