Synergistic valorization of wheat husk-derived HZSM-5 catalyst in pyrolysis of polystyrene and polypropylene: sustainable waste-to-energy conversion enhanced by machine learning models

The current study aims to model and optimize the catalytic pyrolysis of plastics, incorporating an agricultural biomass waste-derived catalyst. Polystyrene (PSW) and polypropylene (PPW) are experimented with thermal and catalytic pyrolysis. Agricultural biomass waste (wheat husk) was selected and acid treated with sulfuric acid (HZSM-5_SA) and hydrochloric acid (HZSM-5_CA), and then used as catalyst. Thermal and catalytic pyrolysis were conducted in a semi batch reactor, with reaction temperature (500 ℃) and different ratios (10:1, 10:2 & 10:3). At a ratio 10:2, PSW with HZSM-5_SA produced 91.19 wt.% of oil yield and PPW with HZSM-5_SA produced 85.73 wt.% of oil yield. The catalyst HZSM-5_SA was effective in the reduction of reaction temperature and time, it decreased from 450 ℃ to 437 ℃ and 22 min to 14 min for PSW. Catalyst activity was also observed for PPW, the reaction temperature decreased from 471 ℃ to 456 ℃ and 34 min to 19 min. Oil properties were determined and it was found that the kinematic viscosity of oil obtained from PSW with HZSM-5_SA was 2.53 cSt, which coincide with the diesel Bharat Stage (BS VI 2020). Total conversion of pyrolysis products was predicted using six Machine Learning (ML) models such as Random Forest, Support Vector, K-Nearest Neighbor, Decision Tree, AdaBoost, and Gradient Boost. Among all the models, the Gradient Boost regressor model had a good evaluation metrics of R² value of 0.984 and RMSE of 0.019, respectively. This study illustrates the use of ML models to predict the total conversion and their correlation matrix with target and feature variables. This study also highlights that cost-effective catalyst can be prepared from biomass (wheat husk) and the use of ML models to train the datasets and evaluate the actual and predicted values.