Iranian Journal of Chemical Engineering (IJChE)

Abstract Time-consuming and costly experiments to measure the cetane number (CN) of biodiesel make computations even more valuable. In the current study, two artificial intelligence (AI) models have been used to predict the biodiesel CN by using comprehensive datasets (440 datasets). They were the gradient-based artificial neural network (GB-ANN) and the multi-layer-perceptron ANN optimized by the genetic algorithm (GA-ANN) for the first time. The three input variablesof the model for predicting the target variable of the biodiesel CN are the average number of carbon atoms, average number of double bonds, and average molecular weight of the fatty acid methyl esters. The learning function, transfer function, number of hidden layers, and number of neurons in the hidden layers are some of the optimized parameters in the current AI-models. The developed models were compared using statistical criteria such as the coefficient of determination (R²), mean square error (MSE), average absolute relative deviation (AARD), standard deviation (STD) and mean absolute percentage error (MAPE). The resulting outcomes revealed that the highest R² and the lowest MSE were related to the GB-ANN model with two hidden layers, trainbfg learning method and logsig-tansig-purelin transfer function. The R² and MSE for the optimized model are equal to 0.9296 and 0.0005 respectively. Although the GA-ANN achieved acceptable outcomes, its statistical analyses produced weaker outcomes than the AI-model based on GB-ANN.

Abstract The artificial neural network (ANN) approach was applied to develop simple correlations for predicting the thermal conductivity of nitrogen-methane and carbon dioxide-methane mixtures. The genetic algorithm method was used to obtain global optimum parameters (weights and biases) of the ANNs. The methane mole fraction, temperature, pressure, and density as effective parameters on thermal conductivity were network input variables. 171 and 180 data points related to the nitrogen-methane and carbon dioxide-methane gas mixtures, respectively, divided to test and train datasets. Simple correlations were obtained due to the small number of optimal neurons in the ANN structures. The mean relative errors of 0.206% and 0.199% for the testing dataset indicate the high accuracy and validation of the correlations. The work indicates that artificial intelligence approaches are very useful for thermal conductivity modeling in natural gases. A sensitivity analysis was performed on all input variables that indicates that the gas mixture density has the greatest impact on the thermal conductivity.

Abstract Two main objectives have been considered in this paper: providing a good model to predict the critical temperature and pressure of binary hydrocarbon mixtures, and comparing the efficiency of the artificial neural network algorithms and the support vector regression as two commonly used soft computing methods. In order to have a fair comparison and to achieve the highest efficiency, a comprehensive search method is used in neural network modeling, and a particle swram optimization algorithm for SVM modeling. To compare the accuracy of the models, various criteria such as ARD, MAE, MSE, RAE and R2 are used. The simulation results show that the ARD for the prediction of the true critical temperature and pressure of the binary hydrocarbon mixtures for the final optimized ANN-based model is equal to 0.0161 and 0.0387, respectively. The corressponding ARD value for the SVM-based model is equal to 0.0086 and 0.0091 for critical temperature and pressure, respectively. Simulation results show that although both models have a very high predictive accuracy, the SVM has higher learning speed and accuracy than ANN.

Abstract Solubility data of solid in aqueous and different organic solvents are very important physicochemical properties considered in the design of the industrial processes and the theoretical studies. In this study, experimental solubility data of 666 pharmaceutical compounds in water and 712 pharmaceutical compounds in organic solvents were collected from different sources. Three different artificial neural networks including multilayer perceptron, radial basis function and support vector machine were constructed to predict the solubility of these different pharmaceutical compounds in water and different solvents. Molecular weight, melting point, temperature and the number of each functional group in the pharmaceutical compound and organic solvents were selected as the input variables of these three different neural network models. The neural network predictions were compared with the experimental data and the SVR-PSO model with the Average Absolute Relative Deviation equal to 0.0166 for the solubility in water and 0.0707 for solubility in organic compounds was selected as the most accurate model.

Abstract Due to wide application of styrene for production of different materials, it is considered as an important product in industry. Therefore, optimizing styrene production conditions is of great importance in petrochemical industry. In this paper, styrene production reactors of Tabriz Petrochemical Complex are modeled using Artificial Neural Network (ANN) model and Adaptive Neuro Fuzzy Inference System (ANFIS). Comparison of two models revealed that the neural networks are more reliable. The process of design and evaluation of models are carried out using industrial data which show credibility of designed models. The neural networks are designed to predict the styrene output from reactors as a function of effective input parameters on the styrene production. Predictions of designed neural networks were used to study the effect of each variable, such as oxygen flow rate and steam oil ratio, on the amount of styrene produced. Also, the optimal values of effective variables for maximum production of styrene were obtained. Furthermore, in order to obtain accurate results, catalyst deactivation of styrene reactors has been modeled using Fuzzy Inference System. As a result, catalyst activity as a function of time is obtained.

Abstract Stuck pipe is one of the most serious drilling problems, estimated to cost the petroleum industry hundreds of millions of dollars annually. One way to avoid stuck pipe risks is to predict the stuck pipe with the available drilling parameters which can be employed to modify drilling variables. In this work, Artificial Neural Network (ANN) was used for stuck pipe prediction according to the fact that this method is applicable when relationships of parameters are too complicated. Based on the drilling fluid condition from one of the Iranian oil fields, stuck pipe instances were divided into static and dynamic types. The results of this study show more than 90% accuracy for stuck pipe prediction in the investigated oilfield. The methodology presented in this paper enables the Iranian drilling industry to estimate the risk of stuck pipe occurrenc during the well planning procedure.