Toba-CPD: An Extended Chemical Percolation Devolatilization Model for Tobacco Pyrolysis.

Tobacco features chemical compositions different from that of raw lignocellulosic biomass. Currently, the performance of network models, like Bio-Chemical Percolation Devolatilization (Bio-CPD), on tobacco pyrolysis is unclear, and only global kinetics have been proposed for tobacco devolatilization, which does not have the versatility for a wide range of heating conditions and tobacco types. To address this issue, the present work first assessed the performance of the Bio-CPD model on tobacco pyrolysis through an a priori study, which showed large deviations. Afterward, an extended Chemical Percolation Devolatilization model for tobacco pyrolysis (Toba-CPD) was developed by modifying the kinetic parameters using a grid-search optimization strategy. The process of grid-search optimization strategy is based on the kinetic parameters of the Bio-CPD model and modified with experimental results of 11 tobacco types under a wide range of heating rates. Finally, the performance of Toba-CPD was measured with experimental results which were not used during parameters optimization. Results demonstrated that the Toba-CPD models could well reproduce the pyrolysis of various tobacco types under a wide range of heating rates (R 2 > 0.957).

Chemistry-Informed Neural Networks modelling of lignocellulosic biomass pyrolysis.

Biomass pyrolysis is a complicated reaction process that involves complex components and reaction pathways. Due to measurement limitations, the intermediate components are difficult to be detected, therefore their detailed kinetics are still not well established. To address this issue, novel Chemistry-Informed Neural Networks (CINNs) were developed to derive the lignocellulosic biomass pyrolysis kinetics from the thermogravimetric analysis (TGA) measurements in published literature. The derived pyrolysis kinetics, involving eight species and eleven reactions, could accurately reproduce the pyrolysis process for both the seen and unseen samples with R2>0.95. The comparisons with the CRECK multi-step and Bio-CPD models also demonstrated the advantages of the derived kinetics in predicting both the final volatiles yield and the pyrolysis process for various biomass types. This study explored a new tool for establishing solid fuel conversion kinetics from TGA measurements using chemistry-informed machine learning approaches.

Experimental and Kinetic Studies on Tobacco Pyrolysis under a Wide Range of Heating Rates.

In the present work, experimental and kinetic studies are conducted to explore and model tobacco pyrolysis characteristics under a wide range of heating conditions. First, thermal decomposition processes of a tobacco sample were investigated using thermogravimetric analysis/difference thermogravimetry (TGA/DTG) experiments under a wide range of heating rates (10-500 K/min), and the TGA/DTG profiles were compared to highlight the effect of heating rate on the pyrolysis characteristics. The results showed that the tobacco sample was sufficiently devolatilized at 1173.15 K (900 °C) and the final volatiles yields were not sensitive to the heating rate. Moreover, it was illustrated that the DTG curve presents a polymerization trend with the increase in heating rate. Then, kinetic parameters, including total component mass fraction, preexponential factor, and activation energy, were derived by deconvolution from TG/DTG profiles for each component with a one-step kinetic framework, and the correlations between kinetic parameters and heating rates were further explored and modeled. The results illustrated that four subpeaks can be found in the deconvolution, indicating the four components (volatile components, hemicellulose, cellulose, and lignin). In addition, the activation energy of each component was found to be insensitive with heating rate (with standard deviation less than 20%). Therefore, an average activation energy was used for each component to avoid the compensation effect and a power correlation between the heating rate and the preexponential factor could be found. A posteriori analysis also confirmed the validity of this correlation.

Estimating biomass major chemical constituents from ultimate analysis using a random forest model.

Chemical constituents are important properties for utilization of biomass, and experimental approaches are always expensive and time-consuming to determinate those properties. Here, a novel random forest (RF) model is developed for accurately predicting biomass major chemical constituents from the much-easier available ultimate analysis, and compared with the previous correlation as well as the experimental data. Two databases are constructed for training and application of the RF model from available literature. The training results show that the determination coefficients (R2) of the RF model predictions are 0.954, 0.933 and 0.968 for cellulose, hemicellulose and lignin, respectively. The application results show that the present RF model can give accurate predictions on chemical constituents for various biomasses with MAPE<20%, and R2 are 0.862, 0.904 and 0.962 for predictions of cellulose, hemicellulose and lignin, respectively. While the previous correlation only works for a narrow range used to develop the correlation, and gives unrealistic negative predictions with MAPE>500% for outside samples.