Developing gasification process of polyethylene waste by utilization of response surface methodology as a machine learning technique and multi-objective optimizer approach

This study set out to evaluate the performance of response surface methodology as a machine learning technique on gasification process of polyethylene waste. Different models were developed for predicting gas yield, cold gas efficiency, carbon dioxide emission and lower heating value of syngas in gasification of polyethylene waste using response surface methodology. The accuracy and validity of these models were checked in comparison with the results obtained from the validated model. Most studies in the field of response surface methodology have only focused on its application for multi-objective optimization and largely have ignored its utilization as a machine learning technique. Central composite design was utilized to develop a model between the variables and the responses. Pressure and temperature of the gasifier, moisture content of polyethylene and equivalence ratio were the variables and the responses were gas yield, cold gas efficiency, carbon dioxide emission and lower heating value of syngas. The findings revealed that root mean square errors of the models developed by response surface methodology were 0.235, 0.438, 0.294 and 1.999 indicating their high validity. Finally, multi-objective optimization of polyethylene waste gasification was carried out using response surface methodology resulting in gas yield of 96.29 g/mol, cold gas efficiency of 76.22%, carbon dioxide emission of 4.66 g/mol and lower heating value of 493.44 kJ/mol. The optimum responses were predicted by response surface methodology with errors smaller than 5%.