COTABARREN Ivana Maria
congresos y reuniones científicas
Modeling of an Industrial Vibrating Double-Deck Screen of a Urea Granulation Circuit
COTABARREN, IVANA; ROSSIT, JOSÉ; PIÑA, JULIANA; BUCALÁ, VERÓNICA
Congreso; 2008 AIChE Annual Meeting; 2008
American Institute of Chemical Engineers
Granulation is a key particle size enlargement process, widely used in the pharmaceutical, food, mining and fertilizer industries (Adetayo et al., 1995). Approximately 60% of the products in the chemical industry are produced in granular form (Balliu, 2005). Granulation converts fine particles and/or atomizable liquids (suspensions, solutions or melts) into granular material with more desirable properties than the original feed. The granulation process is considered as one of the most significant advances in the fertilizers industry, providing products with higher resistance and lower tendency to caking and lump formation. About 40% of the worlds population depend indirectly on fertilizers for their daily bread (Balliu, 2005). On a worldwide basis urea is the most popular solid nitrogen fertilizer and its use grows much more rapidly than that of other products (Fertilizer Manual, 1998). Urea granulation is a complex operation that cannot be carried out in a single unit; it is rather achieved by a combination of process units with specific functions constituting the granulation circuit (Figure 1). The main unit is the granulator, where small urea particles known as seeds (generally product out of specification) are continuously introduced and sprayed with a concentrated solution of the fertilizer. The seeds grow through deposition of the fertilizer solution droplets on the solids surface followed by water evaporation (Bertin et al., 2007). There are different types of granulators such as fluidized beds and fluidized drums. The granules that leave the enlargement size unit are classified in double-deck screens into product, oversize and undersize streams. The product is transported to storage facilities, while the oversize fraction is fed to a double-roll crusher for size reduction. The crushed oversize particles are then combined with the undersize granules and returned to the granulator as seeds (Cotabarren et al., 2008). Generally, in fertilizer granulation plants only a small fraction of the material leaving the granulator is in the specified product size range; therefore high recycle ratios are common. The characteristics of the recycle, which are the consequence of what has happened previously in the granulator, influence what will happen later on in that unit. Thus, cycling surging and drifting of particles make take place. In extreme cases, these periodical oscillations coupled with the large dead time, can result in plant shut down or permanent variations in the plant capacity as well as product quality (Adetayo et al., 1995). To minimize these problems it is necessary to have a fundamental understanding of the effects of the recycling of material, which is nearly always indispensable in the granulation process, on the behavior of the circuit. Many authors found that the operation of the screening and crushing section has a decisive influence on the recycle stream and hence on the circuit stability (among others, Zhang et al., 2000; Heinrich et al., 2003; Drechsler et al., 2005; Radichkov et al., 2006). In view of this and the recognized important role of the plant simulation to predict and optimize the granulation circuit operation (Adetayo et al., 1995; Wildeboer, 1998; Drechsler et al., 2005), reliable models for all the process units should be available. Recently, a validated mathematical model for the industrial double-roll crusher of a urea granulation circuit has been reported (Cotabarren et al., 2008). As for the crusher, the screen modeling requires the knowledge of some parameters that have to be determined from experimental data. Two different approaches, the kinetic and probabilistic, have been used for modeling screening operations. Karra (1979) developed a mathematical model for describing the performance of a vibrating screen, based in terms of the oversize partition curve. This is a curve that represents the percentage of material recovered in the oversize stream, normalized respect to the fifty percent passing size (d50). Hence, the d50 or cut size, is defined as the particle size for which the oversize partition curve has a value of 0.5. The cut size is a parameter affected by a combination of operating variables and material conditions such us the screen feed rate, location of the deck, throughfall aperture (h) percent of oversize (%>h), half-size under (%<0.5h) and near-mesh (0.75h<%<1.25h) material in the feed among others. Determination of the d50 is required in order to calculate the partition curve for different operating conditions. It is worth to mentioned that the Karra model (Karra, 1979) has been implemented in commercial solid simulators such as MODSIMTM (www.mineraltech.com/MODSIM) and has special application in describing the performance of industrial vibrating screens (Standish et al. 1986). In the present work, and with the overall goal of developing a reliable simulator of industrial urea granulation plants, the Karra model (Karra, 1979) has been adapted to represent the sieving process in large-scale double-deck screens. The screen classification parameters were fitted using industrial data from a plant of high capacity (1 million tons of granulated urea/year). The experimental data were collected in two large-scale double-deck screens of identical characteristics belonging to one of the two parallel circuits of the industrial plant. Samples, by duplicate, of the feed (F), oversize (O), product (P) and undersize (U) streams were collected and granulometrically analyzed every 4 and 12 hours for two independent experiments that lasted 36 and 78 hours, respectively. The industrial particle size distributions were represented on thirteen size intervals defined by the experimental sieve set. The undersize mass flow is frequently measured. However, as usual happens in many industrial plants, gross errors were detected in this measurement. The only reliable and available experimental mass flow was that corresponding to the product stream; value indirectly calculated from the mass flow and urea concentration of the urea solution fed into the granulator. Therefore, to fit the double-deck screen model, the calculation of the feed, oversize and undersize streams was necessary. Obviously, knowing the particle size distributions of all the screen streams and one mass flow, just three equations are needed to compute the flowrates of the remaining streams. This can be easily done, however the use of new techniques such as data reconciliation allows determining the unmeasured variables and satisfying the material balances for each size interval, simultaneously. For this reason and also to prevent the influence of the unavoidable errors of the measured values on the model parameter adjustment, the measured data were first reconciled. The data reconciliation was formulated as a constrain optimization problem and solved by means of the Athena Visual Studio Software (www.AthenaVisual.com). The objective of this optimization problem was to minimize the total difference between measured and estimated mass flows for each size interval, weighted by the variance of the measurements using the least square method (Reimers et al., 2007). In a second step, the parameters of the classification model were fitted by using all the reconciled data (i.e., the mass flows and particle size distributions of all the streams that enter or leave the screen). The cut size of the upper deck was found to be very close to the screen aperture, indicating that the oversize separation was highly efficient. The lower deck had the hard task of separating near-sized particles, for this reason the corresponding cut size was found to be significantly lower than the screen aperture. In fact, on average, d50 was 23% smaller than the mesh aperture. Correlations of d50 as a function of the feed mass flowrate, screen aperture, location of the deck, throughfall aperture (h), percent of oversize (%>h), half-size under (%<0.5h) and near-mesh (0.75h<%<1.25h) material of the feed, as well as material bulk density, were performed to obtain a fully predictive screen model. In conclusion, a reliable double-deck screen model for an industrial plant is provided in this work. The calculated data indicate that this mathematical model, which only requires the feed mass flow, the particle size distribution, and screen physical characteristics as input data, accurately represents the experimental information collected from a large-scale urea granulation plant.