Study of Inclusion Re-Entrainment in a Filter Bed D. KOCAEFE, A. MURRAY-CHIASSON, Y. KOCAEFE, and P. WAITE Filtration is widely used in various processes such as water treatment and aluminum casting for the removal of unwanted impurities called inclusions. The deposited inclusions can re-enter the flow as a result of the unfavorable hydrodynamic conditions within the system or flow instabilities, such as flow stop/start periods during casting cycles in the aluminum industry. In this project, the re-entrainment of inclusions was studied as a function of filter-bed length, particle size, inlet inclusion concentration, and inlet velocity. A physical model using water containing PVC particles as inclusions was built. Experi- ments were carried out under continuous-flow as well as interrupted-flow conditions in this pilot-scale filter. It was found that the smaller bed particles and longer bed length enhance the deposition and reduce the re-entrainment of inclusions. Increasing the inlet velocity has a negative effect on the deposition and increases the re-entrainment. A one-dimensional mathematical filtration model has also been developed, and its predictions were compared with the experimental data from the pilot filter and the plant. This article presents the experimental study, its results, and the comparison between model pre- dictions and experimental data. Applicability of the model to aluminum filtration is also illustrated. I. INTRODUCTION FILTRATION is used in many industrial processes to eliminate the undesired impurities called inclusions. In the aluminum industry, for example, the quality of the product depends on the cleanliness of the metal; the presence of inclusions can result in costly defects such as small pinholes in beverage cans or tears in aluminum foil. The filtration process has a complex mechanism influenced by hydrody- namic factors such as fluid flow, turbulence, surface and body forces, as well as chemical and metallurgical interac- tions between the inclusions, the bed media, and the liquid metal. These factors play a role not only in the deposition of inclusions in the bed media, but also in the re-entrainment of the inclusions back to the main flow. Inclusion re-entrainment affects the final quality of the product significantly; however, its mechanism is not well understood. The filter media can not retain all the deposited inclusions. There is re-entrainment of the inclusions due to the hydrodynamic effects during the continuous operation of the filter. In addition, the aluminum filtration process is semi- continuous. There is a period of continuous filtration until the casting is completed for a given number of ingots. Then, the process is stopped, and the ingots are removed. When the sub- sequent batch is ready, the filtration and the casting processes are restarted. During the cast stop/start periods, a significant quantity of the deposited inclusions is released back to the metal due to the flow instabilities. Therefore, the deposition rate is decreased and, consequently, filter efficiency is reduced. The experimental research reported in the literature on the re-entrainment in a continuous mode of filtration, to our knowledge, is very rare. Studies on the re-entrainment mech- anism during flow disturbances are almost inexistent. Inclu- sion re-entrainment has been represented with empirical equa- tions in mathematical models. These formulations are mostly valid only for a given specific experimental system for which the empirical constants are determined. Such a formulation was first suggested by Mints. [1] Subsequently, Eckert et al., [2] Adin and Rajapopalan, [3] and Kocaefe et al. [4] all suggested an empirical representation of re-entrainment. Tien [5] gives a good review of the literature on granular filtration, possible mecha- nisms of deposition, and empirical formulations used by dif- ferent researchers. Destephen and Choi [6] suggested a limiting capture-velocity concept to incorporate the re-entrainment in their statistical mathematical model. They stated that the mechanism of the re-entrainment should be studied before it can be represented well in their model. One of the rare experimental studies on the re-entrainment of inclusions was carried out by Moran et al. [7,8] for the filtra- tion of effluent in a drinking-water treatment plant. The impurities were inorganic solids, organic matter, and micro- organisms. They reported two distinct stages of filtration. One stage is called ripening, during which the effluent quality improves, usually during the initial period of filtration. The second stage is the breakthrough period where the effluent quality decreases with time, and this takes place at longer filtration times. The re-entrainment of inclusions, or detach- ment, as it is called by the authors, is usually the major fac- tor for the observation of the breakthrough period. The effects of media size and filtration velocity were considered. The effect of impurity detachment (re-entrainment) during the break- through period was studied experimentally for a size range corresponding to a microorganism called Cryptosporidium oocysts. The study offers no theoretical analysis of the re- entrainment phenomenon. Mahmood [9] carried out a study on the adhesion and the re-entrainment of kaolite particles on a bed of glass balls. They studied mainly the effect of carrier-liquid composition changes on the attraction and repulsion forces between the bed and kaolite particles. The study on the re-entrainment was limited to the backwash period only, and they observed that the quantity of particles released increased with increas- ing liquid velocity, as expected. METALLURGICAL AND MATERIALS TRANSACTIONS B VOLUME 35B, OCTOBER 2004—999 D. KOCAEFE, Professor, is with the Department of Applied Sciences, University of Quebec at Chicoutimi, Chicoutimi, PQ, Canada G7H 2B1. Contact e-mail: dkocaefe@uqac.ca A. MURRAY-CHIASSON, former Master Student, and Y. KOCAEFE, former Research Scientist, formerly with the Department of Applied Sciences, University of Quebec at Chicoutimi, are R&D Engineer-in-Training and Senior Scientist, respectively, Alcan Inter- national Ltd., Jonquière, PQ, Canada G7S 4K8. P. WAITE, Principal Scientist, is with Alcan International Ltd. Manuscript submitted October 29, 2003.