چكيده به لاتين
To remove unsaturated aliphatic compounds from aromatic feedstocks, a catalytic method utilizing acid-activated clay as a catalyst is commonly employed. In this process, raw bentonite clay is typically activated using dilute acid under specific conditions of temperature and time. One of the key determinants of catalyst quality, is its specific surface area, is influenced by various factors, including the characteristics of the initial clay and multiple process variables. in this study, a machine learning approach was adopted due to its relatively high accuracy, generalizability, and efficiency in terms of time and cost savings. The machine learning model was initially trained using 463 samples collected from diverse sources, encompassing 19 variables comprising initial clay properties, process factors, and the final specific surface area. The most effective factors and their optimal ranges were determined. Subsequently, based on these optimal values, a Taguchi experimental design was conducted with five factors at three levels, resulting in a total of 27 runs. The machine learning model was further trained using new data. Among the linear, tree-based, and ensemble machine learning models, the "Gradient Boosting" model exhibited the best performance, achieving a final prediction accuracy of 74%. Model analysis revealed that the "acid concentration," "time," "temperature," and "weight ratio of clay to acid volume" were the most important process factors, while the properties of "octahedral metal content sum", "interlayer distance" , "silica-to-alumina ratio", and "initial surface area" were the most influential characteristics of the initial bentonite clay in determining the specific surface area of the activated clay. An activated clay sample was produced based on the optimal factor values obtained from the Gradient Boosting model, and its specific surface area was measured as 204 m²/g, which is within the predicted range of the machine learning model (205 ± 19). The total pore volume of the optimal sample was measured as 0.28 cm³/g. The total content of octahedral sheet metal oxides was found to be 13.4 % .wt in atomic absorption spectroscopy analysis, and the total acidity of the optimal sample in NH3-TPD analysis was determined as 9.1 mmol NH3/g. Finally, results from FTIR, XRD, and FESEM analyses indicated structural alteration in the clay during the activation process.
