Predictive modeling and multi-parametric optimization of catalytic pyrolysis of disposable face mask using a combinational approach of response surface methodology and machine learning: Insights into the influence of waste-derived catalyst

Abstract

The current study aims to model and optimize the catalytic pyrolysis of face masks incorporating a waste-derived catalyst to analyze the effect of process parameters (temperature, feed-to-catalyst ratio, and inert gas flow rate) on the oil yield of the process. An integrated approach of response surface methodology (R2-0.95) and machine learning (decision trees regression, R2-0.83) demonstrated a higher prediction accuracy and lower error margins. Explainable artificial intelligence tools spotlighted temperature to be the predominant parameter followed by feed-to-catalyst ratio. Experimental oil yield (13.5%) obtained at optimized parameters (516 °C temperature, 3:1 feed-to-catalyst ratio, and 163 mL/min inert gas flow rate) was compared with those predicted through response surface methodology (13.7%) and decision trees regression (13.12%), showcasing an absolute error range of 0.2–0.4 wt%. Gas chromatography-mass spectroscopy analysis of oil highlighted the presence of silica compounds that can be extracted as value-added chemicals. Further, the overall percentage of naphthene's, paraffins, and olefins in the oil were approximated to be around 40.5% based on the peak area. The presence of hydrocarbons in oil having carbon numbers predominantly in the range of gasoline and diesel as well as a high heating value of 36.56 MJ/kg demonstrated its potential to be used as a fuel. The present study demonstrates the feasibility of valorizing the face mask into energy-dense oil comparable to commercial range fuel using a spent adsorbent based catalyst.