Prediction carbonization yields and the sensitivity analyses using deep learning neural networks and support vector machines

The goal of this study is to use the deep artificial neural network (DLNN) and support vector machines (SVM) methods to predict the biochar, bio-oil, and synthesis gas yields of the Atriplex nitens S. plant. Furthermore, the study's sensitivity analysis predicted the importance levels of factors influencing carbonization yields. Three different carbonization temperatures (400 degrees C, 500 degrees C, and 600 degrees C), two different holding times (30 and 60 min), and two different gas flow rate rates (0.2 and 0.5 L min(-1)) were used as input parameters in the DLNN and SVM models for this purpose. The DLNN method employed two hidden layers. The learning function was Levenberg-Marquardt. As activation functions, the hyperbolic tangent sigmoid transfer function and the linear transfer function were used. The network was tested with various numbers of neurons, and 21 different architectures were created using the number of neurons that produced the lowest MAE value. In the SVM model, the L1QB solver was used. According to the study's findings, the DLNN method made more accurate predictions than the SVM. The DLNN11 (R-2 = 0.987), DLNN19 (R-2 = 0.983), and DLNN7 (R-2 = 0.960) network architectures produced the best results in this method for biochar, bio-oil, and synthesis gas yield, respectively. As a result of sensitivity analysis, the effect of holding time on biochar and bio-oil yields was found to be greater than other parameters. The synthesis gas yield was most affected by the carbonization temperature.