Kinetic and thermodynamic modelling of thermal decomposition of bitumen under high pressure enhanced with simulated annealing and artificial intelligence

Abstract

Thermogravimetric analysis (TGA) of energy resources including bitumen provides kinetic parameters, which can be applied in design and simulation of processes for thermal recovery and conversion. In this investigation, pressurized non‐isothermal decomposition of bitumen under inert environment has been studied. The kinetic parameters were calculated at different pressures (0.1, 0.5, and 1 MPa) and heating rates (10, 20, and 30°C/min) using the differential method. The parameters were later optimized using simulated annealing (SA) optimization algorithm. The results were subsequently validated by comparing the predicted conversion (α) with those of artificial neural network (ANN). Thermodynamic parameters (enthalpy, entropy, and Gibb's free energy) were also calculated using the optimized kinetic parameters. The TGA results show that weight loss and thermal conversion decreased as the total pressure increased from 0.1–1 MPa, at all heating rates. Conversely, the thermal conversion rate (dα/dT) was observed to first decrease with increasing pressure (0.1–1 MPa) within the low temperature oxidation (LTO) and fuel deposition (FD) regions. In contrast, it increased with increasing total pressure within the high temperature oxidation (HTO) region. Furthermore, it was observed that the activation energy (Ea) increased with increasing pressures at all heating rates, while the frequency factor (A) was independent of the pressure or the heating rates. In addition, the thermodynamic parameters tend to increase with increasing pressure. Ultimately, the results established that the SA algorithm could be used to enhance the performance of the differential modelling method in calculating kinetic parameters and predicting the conversion.

Research Advances in Machine Learning Techniques in Gas Hydrate Applications

“…This is a feed-forward artificial neural network model. In the feed-forward neural networks, the movement is only possible in the forward direction . A typical MLP consists of many layers of nodes in a directed graph, with each layer connected to the next one, which can be represented mathematically as: y i = w 0 + w 1 w 1 + w 2 w 2 + ... + w n x n The equation is the transfer function in a neural network.…”

“…In the feed-forward neural networks, the movement is only possible in the forward direction. 37 A typical MLP consists of many layers of nodes in a directed graph, with each layer connected to the next one, which can be represented mathematically as: The equation is the transfer function in a neural network. This linear weight sum would be a threshold at some value, so that the output of the neuron would be either 1 or 0.…”

Prediction of Thermogravimetric Data for Asphaltenes Extracted from Deasphalted Oil Using Machine Learning Techniques

“…They successfully implemented the bi-Gauss method, which is an asymmetric multipeak fitting method based on which the kinetic mechanism and activation energy were obtained through an integral isoconversional nonlinear (NL-INT) method. In a recent interesting work, Alade et al 20 conducted a two-stage analysis for the thermal conversion of bitumen: (i) they optimized kinetic and thermodynamic parameters at different pressures and heating rates using a simulated annealing (SA) algorithm and (ii) an artificial neural network (ANN) was used to predict thermal conversion and compared against the conversion obtained from the optimal kinetic parameters in the first step. As a result, they found that thermal conversion decreased with increasing operating pressure of inert gases for the low-temperature oxidation region but vice versa for the high-temperature oxidation region.…”

Preface for the Special Issue Honouring Professor Hisham Nasr‐El‐Din

“…The Canadian Journal of Chemical Engineering is acknowledged for dedicating this special issue to the living memory of Dr. Nasr‐El‐Din. This acknowledgment is also extended to the authors who contributed to this special issue [ 1–28 ] as well as the guest editors:…”