Unlocking the synergistic effects in biomass-plastic Co-pyrolysis using a hybrid machine learning approach with blended rule descriptors

Co-pyrolysis presents a promising strategy for recycling organic solid waste from biomass and plastic, with thermogravimetric (TG) profiles provide valuable insights into thermo-kinetics and synergistic effects in the co-pyrolysis process. However, obtaining TG profiles through experimental methods is time-consuming and costly. This study systematically analyzed co-pyrolysis experiments, thermo-kinetics, and TG profiles using cotton straw (CS) and mulch film (MF). Then, a hybrid machine learning (ML) model was proposed that integrated a blended rule descriptor with a Gradient Boosting Decision Tree (GBDT) model to predict the TG profile of co-pyrolysis. Results revealed that co-pyrolysis not only enhanced biochar yield but also reduced the reaction activation energy compared to MF alone. The hybrid ML model provided excellent predictive accuracy for the co-pyrolysis process compared to the GBDT model, achieving R 2 = 0.998, RMSE = 1.732, and MAPE = 7.076 in the external validation dataset. In conclusion, this work systematically studied the co-pyrolysis process and proposed a hybrid ML model to guide the utilization of organic solid waste. To describe the TG profile of complex biomass-plastic co-pyrolysis, the blended rule is combined with a ML model to create a hybrid ML model, which exhibited promising performance across the training, testing, and external validation datasets. • A hybrid model is built using the blended rule as descriptor to predict TG profiles. • The product distribution in straw-plastic co-pyrolysis is experimentally studied. • A thermo-kinetic analysis reveals the reaction mechanisms of co-pyrolysis.