4. Conclusions
In this paper, three groups of norm descriptors were established by the structure of ILs, anion and anion. Based on 9020 ρ data points, 7324 η data points and 608 λ data points, three QSPR models were established to calculate the properties of ILs (density, viscosity and thermal conductivity) at variable temperatures and pressures. The performance of the f -T -P QSPR models was improved by introducing norm descriptors into temperature and pressure terms. The statistical results indicate that these models have good prediction accuracy and credibility with the high R2 of 0.9966, 0.9676 and 0.9847, the low AARD of 0.48 %, 3.91 % and 1.02 % for predicting the density, viscosity and thermal conductivity, respectively. Validation analysis shows that these QSPR models have good prediction ability and robustness. Moreover, parameters (α , β , γ and χ ) of thef-T-P model for 16329 ILs are predicted by QSPR models and most of the predictions of density and viscosity values of ILs at variable temperature and pressure are within a reasonable range, except for [bmim][FS], [Hprthiur][I], [mC10im][Pf2N], [N113(eOH)][FAP] and [odbu][C8SO3]. At variable temperature and atmospheric pressure, the predicted thermal conductivities of most ILs are in a reasonable range. At variable temperature and pressure, the predicted thermal conductivity of these ILs corresponding to seven anions are within the reasonable range, and the predictions of these ILs corresponding to four cations and six anions are not in the reasonable range. The prediction results show that the model is stable and widely used.