1. Introduction
Virgin olive oil (VOO) is obtained exclusively from the fruit of the olive tree (Olea europaea L.) using mechanical and physical processes, such as washing, decantation, centrifugation and filtration. The latter does not lead to alterations in the oil, exclusion of oils obtained using solvents or re-esterification processes and of any mixture with other oils (IOOC, 2006).
Currently, the so-called two phase continuous extraction system is the most widespread used in the VOO extraction process (Jiménez et al., 1995 and Piacqadio et al., 1998). In this system, the horizontal screw solid bowl, so-called ‘decanter’, is used for a primary separation of the oil fraction from the water and solid pomace (Altieri, 2010 and Altieri et al., 2013). However, the oil resulting from the ‘decanter’ still contains moisture and suspended solid particles, which both have to be removed (Uceda et al., 2006 and Uceda et al., 2008). If not, this can lead to the alteration of VOO since it can induce anaerobic fermentation, hydrolysis and oxidation reactions during storage, which reduces the oil quality and sensory characteristics causing the emergence of undesired off-flavors such as fusty or muddy sediment (Ranalli, 1989, Ambrosone et al., 2002, Tsimidou et al., 2005, Gómez-Alonso et al., 2007, Jiménez and Carpio, 2008, Di Giovacchino et al., 2002, Morelló et al., 2004 and Baiano et al., 2014).
Traditionally, after liquids are sieved, the separation of liquid phases with different density is performed by vertical centrifuges. These centrifuges are characterized by a high water consumption, high wastewater production and a considerable energy consumption (Masella et al., 2009 and Masella et al., 2012). For these reasons, a growing trend appeared in the olive oil industry to substitute the centrifuges by conical bottom settling tanks in the clarification step. These tanks have a cone angle between 45° and 60° and a capacity between 400 and 10,000 L. They are equipped with a purge system (manual or automatic) and can be used both for batch and continuous operation (Humanes and Humanes, 2011 and Altieri et al., 2014). Despite the fact that settling tanks are being implemented, little knowledge on the settling process of VOO from HSSB is available. The main factors affecting this separation processes are the density difference between liquid and solid partióles, the partióle size, the liquid viscosity, among others (Davis, 2010). Some of these parameters, such as oil density and viscosity, have already been studied for several vegetable oils, including olive oil (Esteban et al., 2012, Bonnet et al., 2011 and Fasina et al., 2006). These physical properties depend on the VOO fatty acid composition and are strongly affected by temperature (Gila et al., 2015). However, for other essential parameters, such as solids density and solid particle size distribution, no data was found in the literature.
Direct measurement of these separation processes are the best way to understand the behavior of liquid and solid particles flows, however they cannot be ascertained until the tank is available. Computational fluid dynamics (CFD) offers an alternative way to predict the behavior of this separation process. This tool permits to save costs and time, because it is less expensive than experiments, physical modifications are not necessary, and it can also predict which design changes are most crucial to improve performance. CFD techniques have already been applied in other separation processes (Romaní and Nirschl, 2013, Vieira and Barrozo, 2014 and Narasimha et al., 2005) including wastewater treatment that uses settling tanks to separate suspended solid sludge particles from the purified water (Stamou, 2008, Goula et al., 2008a, Goula et al., 2008b and Patziger et al., 2012). For these reasons, introduction of CFD technology in the study of oil clarification step can be very useful to obtain extra information and to improve some aspects of the extraction process in olive oil mill.
The aim of this work was the analysis of the clarification process of VOO from HSSB in settling columns at three different temperatures (15, 20 and 30 °C). In addition, a CFD model was developed for the settling column that represents and simulates this separation step.
2. Material and methods
2.1 Origin of the oil
For this work, around 9 L of VOO were taken directly from the HSSB after the vibratory sieve (with 1 mm of diameter). The oil was obtained from the ‘Picual’ variety olive fruits in the experimental oil mill of the research center IFAPA ‘Venta del Llano’ in Mengíbar, Jaén (Spain). The experimental plant pilot was equipped with a metallic hammer mill with a 6 mm sieve, a thermobeater formed by three containers (Pieralisi, Spain) of 600 kg each, a two-three phases horizontal centrifuge Pieralisi SC-90 (working at two phases way) with a theoretical processing capacity of 45,000 kg/day. The olive paste was loaded at 750 kg/h in the horizontal centrifuge, being controlled by the automation system. The operation work conditions were: crusher velocity of 2250 rpm; kneading temperature of 28 ± 2 °C and a kneading time of 45 min.
2.2 Experimental assays
Laboratory graduated cylinders (with a volume capacity of 1000 mL) were used as settling column, with an internal diameter and height of 6 cm and 35.37 cm respectively. An aliquot of 1000 mL of VOO sample was poured into of each settling column. Prior to starting the assay, the column was shaken, in order to obtain a homogeneous suspension. The assays were carried out simultaneously at three different room temperatures (15, 20 and 30 °C) in different temperature controlled chambers. Assays were performed in triplicate. Oil samples were taken from each column with a pipette (5 mL), at 10 cm of depth from the oil upper level (Fig. 1) at different times: 0, 5, 10, 20, 40, 80, 160, 320, 640 and 1280 min.
Fig. 1. Experimental device. Laboratory test tube with pipette coupled
2.3 Experimental analyses
2.3.1 Moisture and solid particle content
To determine the moisture content, approximately 5 mL of olive oil was weighed in a ceramic capsule with filter paper. Subsequently, the oil sample was dried in an oven at 105 °C until weight stabilization. The loss of weight yields humidity (%) and volatile matter (UNE 55-020-73, 1973). After drying, impurities were retained by the paper filter and the oil was extracted with petroleum ether in a Soxhlet system. Finally, the filter paper was dried and weighed to determine impurities amount (%) (UNE 55002:1962, 1962).
2.3.2 Particle-size distribution
Particle size distribution analysis was performed using a gravitational method adopted from the ISO 13317-2, 2001 and Farmer and Beckman, 1984). The latter uses a device consisting of a laboratory cylinder with pipette coupled, as described in Section 2.2 (Fig. 1). Similar to the experimental assays, an aliquot of 1000 mL of VOO sample was poured into each of the settling columns and shaken to obtain a homogeneous suspension. Subsequently, during solid particle settling at 20 °C, a volume was pipetted at a fixed depth (10 cm) and at different settling times. Next, the solid particle content of the sample was determined (Section 2.3.3), each time measuring the quantity of the still suspended particles in the suspension. Using these obtained results, combined with the known values of VOO density and viscosity according to Gila et al. (2015), the solid particle density of 1025 kg/m3 (Alba, 2008), the sampling time and depth, the gravity value (9.81 m/s2) and applying the equation based on the Stokes law’s (Eq. (1)), the distribution curve (Fig. 2) can be calculated:
where dp is the particle diameter (m), h is the depth of sampling point (m), j is the olive oil viscosity (Pa s) at 20 °C, t is the sampling time (s) after starting the assay, g is the gravity (m/s2), ps is the solid particle density (kg/m3) and p¡ is the liquid density (kg/m3) at 20°C.
Fig. 2. Particle size distribution of a VOO sample from HSSB.
2.3.3 Settling efficiency
The settling efficiency at 80 min was estimated using following relationships (Eq. (2)):
where n80 is the settling efficiency at 80 min expressed in %, c¡ is the initial moisture and solid particle content (MSPC) expressed in % and c80 is the MSPC at 80 min expressed in %.
2.4. CFD simulation
2.4.1. Problem description
The primary phase was the olive oil and the secondary phase was the solid particles (Table 1). The range of the suspended solids was divided into 7 distinct classes (Table 2) of particles based on the discretization of the measured size distribution obtained previously (Fig. 2). The settling column contained a uniform solids volume fraction of 0.36% at the beginning of the simulation (uniform agitation of the column). Then the particles started settling due to gravity and difference between densities.
Table 1. Physical properties of virgin olive oil and solid partióles used in the simulations.
2.4.2 Geometry, meshing and boundary cells
The design software GAMBIT 2.2.30 was used to build and to mesh the geometry of the model. The 3D settling column was 0.3537 m in height and 0.06 m in diameter, this is not a closed container, so the upper surface of the tube was modeled as ‘free wall’ whereas all others zones were modeled as ‘wall zone’. The selected grid was comprised of 20,492 hexagonal elements. Two other grids (one finer with 11,475 elements and one coarser with 39,889 elements) were also used to determine the effect of the overall grid resolution on predictions. While the predictions obtained using the coarse grid were found to be different from those resulting from the selected one, the difference between the predictions made by the selected and fine grids were insignificant. As a result, the solutions from the grid of 20,942 hexagonal elements were considered to be grid independent.
Table 2. Classes of solid particles used to account for the total suspended solids in the settling column
2.4.3 Physics of the model
The computational fluid dynamics code FLUENT 6.3.21 was used to carry out the simulations (ANSYS-Fluent, 2006). The flow was modeled as laminar, (Re < 2100). This case presents a liquid—solid interaction (olive oil and solid particles), so the Eulerian granular multiphase model was activated. The granular viscosity model chosen for this case was Syamlal-O’Brien (Syamlal and O’Brien, 1988), since this model was more appropriate for low volume fractions (less than 5% at the beginning of the simulation). No special boundary conditions have been defined for the walls.
2.5 Statistical analysis
For parameters b and m in Table 3, means and standard deviations were calculated using the statistical package Statistix, Version 9.1 (USA). The Statistix software was used to perform an analysis of variance (ANOVA) and Tukey’s honest significant difference test at a 95% confidence level (p < 0.05) in order to identify differences among groups.