This work investigates how the temperature dependence of soybean oil viscosity varies with frying conditions and chemical composition, to develop a model that predicts oil degradation directly from process parameters. The temperature dependence of frying oil viscosity is described using Lioumbas et al. (2012) non-linear model, which is confirmed with satisfactory accuracy (r(2) > 0.9998). Two models, predicting viscosity separately from frying conditions and chemical composition, are established, both with an overall accuracy of greater than 95.0%. Using the viscosity parameter as a medium, multiple quasi-linear regression analysis was applied to derive an expression that predicts fatty acid composition and total polar materials from the average oil temperature and the frying time (r(2) = 0.930). This final expression quantifies the relationship between the frying conditions and the chemical composition of the oil, which is convenient for on-site quality control. (C) 2015 Elsevier Ltd. All rights reserved.