Experimental investigation and development of a support vector machine model for Fischer-Tropsch synthesis in the presence of Co (III)/Al2O3 catalyst

In present paper experimental data obtained from a batch autoclave reactor carrying out a Fischer–Tropsch synthesis in the presence of Co (III)/Al2O3 catalyst was used to develop an artificial intelligence-based kinetics models. Such kinetics models are best suited in cases for which conventional reaction kinetics models cannot be developed/fitted. The model is based on Support Vector Machine (SVM). Model input variables are temperature, pressure and duration of reaction and model output variables are synthesized gas composition that are for CO2, CO, CH4, and H2. Model development that consists of training, optimization and test was performed using randomly selected 80%, 10%, and 10% of the data respectively. Model predictions are in good agreement with experimental data and are compared with those obtained form an ANN-based model that has been developed using the same data set used for development of SVM-based model. Calculated squared correlation coefficients for the molar percentage of CH4, CO2 and CO show that SVM-based model can improve accuracy of predictions by about 26%. Based on the results of this case study SVM proved to be a reliable accurate estimation method that require more attention in the field of chemical engineering.

Keywords:

Catalyst , Fischer-Tropsch synthesis , Support vector machine