Correlating Asphaltene Dimerization with Its Molecular Structure by Potential of Mean Force Calculation and Data Mining

Abstract

Asphaltene aggregation affects the entire production chain of the petrochemical industry, which also poses environmental challenges for oil pollution remediation. The aggregation process has been investigated for decades, but it remains unclear how the free energy of asphaltene association in solvents is correlated to its molecular structure. In this study, dimerization energies of 28 types of asphaltenes in water and toluene were calculated using the umbrella sampling method. Structural parameters related to the atom types and functional groups were screened to identify the factors most influencing the dimerization energy using multiple linear regression, multilayer perceptron, and support vector regression. Results demonstrated that the influence of molecular structure on asphaltene association in water was nonlinear, while attempts to capture the relationship using linear regression had larger error. The linkage per aromatic ring, number of aromatic carbons, and number of aliphatic chains were the top three factors accounting for 52% of the dimerization energy variation in water. Asphaltene dimerization in toluene was dominated by the content of sulfur in aromatic rings and the number of aromatic carbons which contributed to 55% of the energy variation. To the best of our knowledge, this was the first study successfully predicting asphaltene dimerization using molecular structure (R > 0.9) and quantifying simultaneously the relative importance of each structural parameter. The proposed modeling approach supported the decision making on the number of structural parameters to investigate for predicting asphaltene aggregation.

Aggregation Behavior of Model Asphaltenes Revealed from Large-Scale Coarse-Grained Molecular Simulations

“…Further literature and discussion on asphaltene structure elucidation, aggregation modeling, and adsorption are presented in the reviews of References 2,7,8,15,19,20,[25][26][27] Regarding the use of molecular simulations to study asphaltenic systems, all-atom and united-atom (grouping each carbon with the corresponding bonded hydrogens) descriptions have been implemented to study asphaltene systems of different complexity to calculate dimerization energies, research the correlation between functionality and aggregation behavior, study adsorption or interfacial behavior, as well as to obtain a variety of properties, like viscosity, structure factor, nanoaggregate properties, etc. [28][29][30][31][32][33][34][35][36][37][38][39][40][41][42] . As a consequence of the level of detail of the atomistic representation and the corresponding large computational requirements, the simulated systems usually span only a few nanometers in size and explore events that occur in the order of tens of nanoseconds 28,31,37 , which limits the study of the aggregation phenomenon.…”

Application of Life Cycle Assessment and Machine Learning for High-Throughput Screening of Green Chemical Substitutes

“…The collected data was randomly divided into three parts, including training group (70%), validation group (20%), and test group (10%). The data in the training group was used to build models with ANN, in which ReLU or Sigmoid activation functions were applied to train the models through adjusting the weights of connections between neurons in different layers with back-propagation algorithm. , The ANN model architecture was composed by an input layer, multiple hidden layers, and an output layer . The descriptors extracted by PCA were the variables in the input layer, while the corresponding LCIA value was the information in the output layer.…”

“…The data in training group was used to build models with ReLU activation functions and train the models through adjusting the weights of connections between neurons in different layers with back-propagation algorithm. 33,34 Thereafter, the validation dataset was introduced into the fitted model to perform an unbiased evaluation and meanwhile adjust the accuracy of models by tuning the hyper-parameters such as the number of hidden layers and the number of neurons in each hidden layer. 32 The data in test group was used to evaluate the final model as external validation.…”

Cross-Dimension Mining Model of Public Opinion Data in Online Education Based on Fuzzy Association Rules

“…This paper mainly uses the Pearson correlation coefficient (PCC) to judge the correlation. Pearson correlation coefficient is often used to weigh the linear correlation between two random variables [19], and the correlation between the training data z s i and test data z t j can be obtained, which is calculated as follows:…”