مقاله


کد مقاله : 13980228179482

عنوان مقاله : Hybrid Neural network-GA modeling for separation of linear and branched paraffins by adsorption process for gasoline octane number improvement

نشریه شماره : 12 Autumn-Winter 2021

مشاهده شده : 132

فایل های مقاله :


نویسندگان

  نام و نام خانوادگی پست الکترونیک مرتبه علمی مدرک تحصیلی مسئول
1 Niloufar Fatourehchi n.fatourehchi@gmail.com Associate Professor PhD
2 Zahra Mashayekhi mashayekhiz@ripi.ir Associate Professor PhD
3 Saeed Sadeghpour Galooyak sadeghpours@ripi.ir Associate Professor PhD
4 Majid Masoumi masoumim@ripi.ir Faculty Member M.Sc

چکیده مقاله

After recognizing the toxic and carcinogenic effects of Lead organic compounds, production of compounds such as Methyl tertiary butyl ether as an additive to ordinary hydro carbonate gases was proposed. As a result, development of a new process for producing gas with high octane from complex compounds of light petroleum distillates was initiated. This method is based on separating C5-C8 linear and branched alkanes according to their absorption properties, chain length and the number of branches. In this study, the hybrid neural network model based on experimental data in the database has been used as an alternative model for predicting the separation rate of linear and branched paraffin through absorption process. Absorption temperature, absorption time, hydrocarbons' octane number, and hydrocarbon density are considered as four input parameters, and the ratio of linear paraffin concentration to total as the output parameter of neural network. The neural network model was successfully generalized by experimental database and then was investigated with the help of test data. The results of modeling for the test data indicated the success of neural network model in predicting the rate of linear paraffin separation from non-linear ones. Therefore, the developed neural network model can be used for determining the C/C0 with confidence in absorption process. According the obtained results for test data, the minimum mean squared error is 0/0518, which is a satisfactory measure. The model and experimental data were compared and regression coefficient 0.990 shows good matching between modeling results and experimental results.