A machine learning model to predict the pyrolytic kinetics of different types of feedstocks

An in-depth knowledge of pyrolytic kinetics is vital for understanding the thermal decomposition process. Numerous experimental studies have investigated the kinetic performance of the pyrolysis of different raw materials. An accurate prediction of pyrolysis kinetics could substantially reduce the efforts of researchers and decrease the cost of experiments. In this work, a model to predict the mean values of model-free activation energies of pyrolysis for five types of feedstocks was successfully constructed using the random forest machine learning method. The coefficient of determination of the fitting result reached a value as high as 0.9964, which indicates significant potential for making a quick initial pyrolytic kinetic estimation using machine learning methods. Specifically, from the results of a partial dependence analysis of the lignocellulose-type feedstock, the atomic ratios of H/C and O/C were found to have negative correlations with the pyrolytic activation energies. However, the effect of the ash content on the activation energy strongly depended on the organic component species present in the lignocellulose feedstocks. This work confirms the possibility of predicting model-free pyrolytic activation energies by utilizing machine learning methods, which can improve the efficiency and understanding of the kinetic analysis of pyrolysis for biomass and fossil investigations.

Automated summary

An in-depth understanding of pyrolytic kinetics is crucial for comprehending the process of thermal decomposition. This study successfully constructed a model to predict the mean values of model-free activation energies of pyrolysis for five different feedstocks using the random forest machine learning method. The results indicate the potential of machine learning methods for quick initial pyrolytic kinetic estimation. The study also revealed the correlations between atomic ratios and activation energies, as well as the influence of ash content on activation energy, depending on the organic component species present in the feedstocks.