Novel Sensitivity Study for Biomass Directional Devolatilization by Random Forest Models

Abstract

Devolatilization is always the primary process in biomass thermal conversion, and directional devolatilization has caught considerable attention in recent decades for producing certain fuels and raw chemical materials. In the present study, we report a novel sensitivity study for biomass directional devolatilization using random forest models, which shows obvious advantages in the parameter range, analysis time, and cost compared with the experimental approach. First, a biomass devolatilization product database is constructed with a detailed mechanism for various biomass types under different operation conditions. Then random forest models are developed from the constructed database to accelerate the Sobol sensitivity analysis for obtaining the fullparameter-effect phase diagram. The phase diagram shows that the cellulose fraction holds the maximum influence for the CH 4 , C 2 H 4 , CO, and tar yields, while it has has limited effects on the H 2 O, CO 2 , and solid residue (SR) yields. The final temperature has the maximum effect on the H 2 yield, and the LIG-C fraction shows the dominating effect on the SR yield. The final temperature and the LIG-C fraction have comparable and considerable effects on the H 2 O yield. This full-parameter phase diagram provides an efficient way to directionally choose the biomass types and alter the operation conditions to produce certain devolatilization products.

Nondestructive Detection for Egg Freshness Based on Hyperspectral Scattering Image Combined with Ensemble Learning

“…Then, 71.43% of the samples were randomly selected as the training set, and the remaining 28.57% 0% as the test set. We compared the prediction of egg freshness using the following six models: support vector machine (SVM) [ 36 ], k-nearest neighbor (KNN) [ 37 ], random forest (RF) [ 38 ], Naive Bayes (NB) [ 39 ], discriminant analysis classifier (DAC) [ 40 ], and latent Dirichlet allocation (LDA) [ 41 ]. In order to further improve the accuracy and the generalization ability of the egg freshness classification model, multiple weak classifiers were merged into a strong classifier by stacking ensemble learning [ 42 ].…”

Combustion of Biomass in Fluidized Beds: A Review of Key Phenomena and Future Perspectives

“…In an ideal scenario, a set of kinetics, independent of the type of biomass, can be obtained, similarly as it was done by Smith for coal chars . Since the variety of biomass is vast, the perspective research can turn into new data analysis techniques, for example, employing machine learning, which has been demonstrated useful in other areas of combustion research, such as predicting heating values of biomass, or the product yields from devolatilization …”

“…172 Since the variety of biomass is vast, the perspective research can turn into new data analysis techniques, for example, employing machine learning, which has been demonstrated useful in other areas of combustion research, such as predicting heating values of biomass, 337 or the product yields from devolatilization. 338 Information on the primary CO/CO 2 ratios from biomass combustion is scarce; hence, most researchers use correlations obtained from coal chars. This is clearly a significant research gap, since biomass and coals differ significantly in the ash content and metals in ash are known catalysts of combustion, it would be expected that the CO/CO 2 ratio also differs.…”

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

“…Hence, there is an urgent need to find other possible uses for tobacco instead of producing cigarettes. Tobacco could be utilized by thermochemical and biochemical routes, such as gasification, combustion, and carbonation, in which pyrolysis is always the primary step. − Therefore, it is necessary to study the pyrolysis behavior of tobacco and then establish accurate pyrolysis models, which is significant for developing healthy utilization technologies of tobacco.…”