Prediction of speed of sound of deep eutectic solvents using artificial neural network coupled with group contribution approach

Predicting the physiochemical properties of deep eutectic solvents (DESs) is crucial for designing new solvents. Heat capacity and speed of sound are important thermodynamic properties in chemical processes. However, experimental data on the speed of sound in DESs is limited. Consequently, a thermodynamic model is needed to estimate the speed of sound in DESs over a wide range of pressures and temperatures. A key challenge in these models is accurately estimating the ideal gas heat capacity. Since the ideal gas heat capacity of DESs is often unavailable, a machine learning (ML) approach, using artificial neural networks (ANNs) coupled with a Group Contribution (GC) method, is a promising technique. The GC approach will be used to estimate critical temperature, volume, and acentric factor of DESs, which can then be input into the ANN model to predict the speed of sound. The results show that using a combination of a GC method and ANNs or CatBoost ML provides a highly accurate prediction of the speed of sound in DESs. Input parameters to the ANN + GC include temperature, acentric factor, molecular weight, and critical volume. The absolute relative deviation (ARD%) and R² values of correlated speed of sound for the ANN + GC model have been obtained 0.032% and 0.998, respectively. The ARD% for both the ANN + GC and ML + GC approaches was substantially lower than that of the correlation-based models. Furthermore, cumulative frequency diagrams and the leverage approach were implemented to validate the quality and reliability of the proposed model. The leverage analysis confirmed the accuracy of the data used and the high reliability of the ANN + GC model for estimating the speed of sound in DESs. This analysis indicates that the ANN + GC and ML + GC methods can effectively estimate the speed of sound in DESs based on molecular structure. Therefore, these approaches offer a promising tool for predicting the speed of sound of newly designed DESs when experimental data is unavailable.