Modeling the density of acid gases at extensive ranges of pressure and temperature conditions

Abstract Carbon dioxide remains in the atmosphere for long time and has harmful effect on environment. Knowledge of the thermophysical properties of pure and impure carbon dioxide in this process can be very useful for monitoring this kind of processes. In this study, some intelligent models were developed to precisely predict the density of acid gases (i.e., carbon dioxide, hydrogen sulfide, and sulfur dioxide) under wide ranges of pressure (0.03–125.86 MPa), temperature, (220.204–450.052 K), mole fraction of carbon dioxide (50–100%), mole fraction of hydrogen sulfide (0–49%), and mole fraction of sulfur dioxide (0–5.2%). Two least square vector machine models optimized with harris-hawks optimization and coupled simulated annealing, two radial basis function neural network models optimized with harris-hawks optimization and cuckoo search algorithm, and a multilayer perceptron optimized with Levenberg-Marquardet algorithm were developed for the estimation of the density of acid gases based on 1911 experimental data points. Besides, by using the best models, a committee machine intelligent system was extended. Furthermore, seven equations of state were used to predict the density of the systems. The results revealed that the developed models present reliable prediction abilities. In addition, it was found that the committee machine intelligent system and multilayer perceptron are the fittest paradigms with coefficient of determination values of 0.985 for both paradigms. Lastly, the performance of the best-developed model was compared with equations of state. Although a good match between equations of state's predictions and experimental data was observed at low pressure, as pressure increased, the deviation between equations of state's prediction and experimental data increased. However, at all ranges of pressure, our best-developed model has good consistency with the experimental data. Trend analysis showed that our best model estimates the trend of variation of density as a function of pressure perfectly.