Varying degrees Of incomplete filter regeneration lead to a progressive shortening of filtration cycles, which was both measured and modeled as a function of the time interval t(c) between cleaning pulses (our definition of "cake residence time"). Of particular interest was the effect of temperature. Experiments were performed with pressure-pulse cleaned ceramic filter elements typically used in high-temperature gas filtration, for up to 200 cycles, at two temperature levels (200 and 300 degrees C), at two filtration velocities (3 and 5 cm/s), for several values of cake thickness, and with two different particle materials. The cake residence time was varied by adjusting the initial cycle length between 6 and 87 min without changing the cake thickness. A 2-stage model is proposed to describe the observed patterns of decrease in cycle duration as a function of cycle length t(c). For cycles exceeding a certain critical length t(c)*, the rate of decrease is an exponential function of (t(c)-t(c)*), where t(c)* and the other fit parameters was found to depend mainly on particle material as well as temperature. Below t(c)* the rate of decrease becomes constant and independent of the actual cycle length. The constant was found to be a weak function of flow velocity and cake thickness as long as the cake was not too thin; a temperature dependence was not discernible for Phase 2. The results support the uniform-residual-layer hypothesis as a mechanistic explanation of the filtration behavior during both phases: a thin incremental layer of cake is left behind after each regeneration pulse, which drives the loss of cycle length. During Phase I the incremental layer is a temperature dependent function of cake residence time (in excess of t(c)*); during Phase 2 it is not. This suggests that the thickness of this residual cake layer is related to the solidification rate in the cake. For cake residence times below a critical value, the degree of solidification is too low to affect the stability. The regeneration efficiency is then controlled by other factors such as filtration velocity and regeneration intensity (cake thickness), as in any typical bag house filter. (C) 2009 Elsevier Ltd. All rights reserved.