Abstract: Air filter clogging causes a nonlinear relation between filtration pressure drop and particle mass loading. Mass loading and ambient dust conditions are difficult to measure directly. A layered modeling method was developed to address this problem. Initial permeability and the flow–resistance relation of the filter element were characterized through bench tests. A laboratory baseline pressure-drop model was then established. A discrete state-space model with mass loading and equivalent dust concentration was constructed. An unscented particle filter was applied for filter state estimation. Bench continuous clogging tests were performed for validation. The method was compared with particle filter, unscented Kalman filter, and extended Kalman filter. Under different flow-rate conditions, pressure-drop reconstruction achieved an average root mean square error (RMSE) of 20.46 Pa, a mean absolute error (MAE) of 13.75 Pa, and a mean absolute percentage error (MAPE) of 3.47%. Mass-loading estimation achieved an average RMSE of 11.36 g/m² and an MAE of 9.31 g/m². Validation with vehicle transient-condition data gave a pressure-drop reconstruction RMSE of 16.64 Pa and an MAE of 12.74 Pa. The results show that the proposed method enables stable and effective estimation of the filter clogging state under varying flow rates and fluctuating operating conditions.