Prediction of Sound Speed of Ionic Liquids Based on Multiple Linear Regression and Back-Propagation Artificial Neural Networks

Sound speed of ionic liquids (ILs) can be obtained using various methods, including experimental measurements, semi-empirical models, and theoretical research. Among these methods, the quantitative structure-property relationship (QSPR) model has gained wide attention. However, constructing an effective QSPR model depends on the selection of appropriate molecular descriptors. In this study, there are two descriptors for model construction, namely, the charge density distribution area of ILs at a specific interval obtained using the conductor-like screening model for the segment activity coefficient (COSMO-SAC) and the cavity volume of ILs. Using the multiple linear regression (MLR) approach and the back-propagation artificial neural network (BP-ANN) method, two quantitative structure-property relationship (QSPR) models, namely u-MLR and u-ANN, were proposed respectively to describe the sound speed (u) of ILs. These models incorporated temperature and molecular weight of ILs_. The evaluated data set comprised 171 ILs with 5114 data points. For the entire data set, u-MLR and u-ANN exhibited determination coefficients (R²) of 0.9706 and 0.9995, respectively. Furthermore, the calculated average absolute relative deviations (AARD) were 1.59% and 0.10% with root mean square errors (RMSE) of 30.68 m/s and 4.12 m/s, respectively. Therefore, the prediction performance of the u-ANN model based on back-propagation artificial neural networks is significantly better than that of the u-MLR based on multiple linear regression.