- Email: [email protected]
water levels and differential speed during olive oil extraction
Journal of Food Engineering 119 (2013) 561–572
Contents lists available at SciVerse ScienceDirect
Journal of Food Engineering journal homepage: www.elsevier.com/locate/jfoodeng
Horizontal centrifuge with screw conveyor (decanter): Optimization of oil/water levels and differential speed during olive oil extraction Giuseppe Altieri ⇑, Giovanni Carlo Di Renzo, Francesco Genovese SAFE – Scuola di Scienze Agrarie, Forestali, Alimentari e Ambientali, Università degli Studi della Basilicata, Viale dell’Ateneo Lucano, 10, 85100 Potenza, Italy
a r t i c l e
i n f o
Article history: Received 27 February 2013 Received in revised form 29 May 2013 Accepted 16 June 2013 Available online 2 July 2013 Keywords: Olive oil extraction Quality preservation Decanter centrifuge Feedback control
a b s t r a c t Olive oil extraction using a horizontal centrifuge with a screw conveyor is an essential operation to reduce production costs. However, proper control of the plant is required to maintain a high level of extraction yield and olive oil quality. Rheological characteristics of the olive paste, which change in relation to water content, fruit variety, maturity level and seasonal temperature variations, greatly affect the efficiency of centrifugal extraction. If olive paste is fed to a decanter without automatic control, then nonoptimal extraction is performed. After successfully testing a suitable flow mass sensor in order to set up an automatic system to control the olive paste mass flow rate fed to a decanter centrifuge during olive oil extraction, a feedback control system was tested in the laboratory and built in-line in an industrial processing plant. This allowed trials to be carried out at constant mass flow rate for both the paste and the added water, on a decanter centrifuge with variable differential speed between bowl and screw conveyor (i.e. variable DN) and with regulation capability of oil–water ring levels. Constant quality olives (cv. Coratina) were used for the trials; the correlation was evaluated between oil/water ring levels and DN with respect to the extraction yield, husk fat content and vegetable water fat content. The ‘‘oil recovery efficiency’’ (go) and ‘‘separation coefficient’’ S peaked with maximum residence time, using DN = 13.3 and RING = 284.8, while go minimum was achieved at higher DN. More specifically, for DN values higher than about 18 or lesser than 12, go becomes independent of RING. Furthermore, go and S values are strictly related to the geometry of the decanter centrifuge, whereas the paste dilution ratio determines the preservation quality of the oil extracted, and the maximum oil recovery efficiency represents the remaining objective related to the quantity of extracted oil. From experimental data some interesting relations were found linking decanter centrifuge parameters; their relationship depends on a complete quadratic interaction model constituted by the input variables that leads to highly correlated polynomials between DN vs. Uopdb (‘‘mass fraction of oil in the fed olive paste’’ (dry basis)) and go optimal vs. Uopdb. These relations allow optimization of DN and RING values such that maximum oil recovery efficiency is achieved. Indeed, knowledge of Uopdb by an on-line method allows the decanter DN and RING operating parameters, to be calculated and modified at once, obtaining optimal maximum oil recovery efficiency. Ó 2013 Elsevier Ltd. All rights reserved.
1. Introduction Olive oil is the most widely used fat in the diet of Mediterranean countries, thanks especially to its healthy properties, unique aroma and long shelf life, and due to its natural antioxidant content, which make it different from other vegetable oils (Salvador et al., 2003). Evolution of the oil extraction process has led to the replacement of traditional discontinuous lines, using the pressure system
⇑ Corresponding author. Tel.: +39 0971 205468; fax: +39 0971 205429. E-mail address: [email protected] (G. Altieri). 0260-8774/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.jfoodeng.2013.06.033
extraction, with continuous ones, using centrifugal extraction. In particular, the horizontal centrifuge with a screw conveyor (i.e. decanter) is widely used in olive oil extraction, especially if large amounts of olives have to be processed in a short time (Ranalli et al., 1997; Piacqadio et al., 1998). However, olive oil extraction by centrifuge is dramatically affected by changes in the rheological characteristics of the olive paste in relation to water content, fruit variety, maturity level and seasonal temperature variations. Therefore, correct control of the process is essential in order to achieve high levels of both extraction yield and olive oil quality (such as content in natural antioxidants and aromatic compounds) (Amirante et al., 1995; Di Renzo and Colelli, 1997; Piacqadio et al., 1998; Ranalli et al., 2001; Altieri, 2010).
562
G. Altieri et al. / Journal of Food Engineering 119 (2013) 561–572
Nomenclature R% DN RING SS K SR n
‘‘percent recovery parameter’’ (%/s) differential speed between bowl and screw conveyor of decanter centrifuge in rpm. water/oil ring level shear stress (Pa) consistency coefficient (Pa sn) shear rate (s1) flow behaviour index
Uwvw Uovw Usvw Ql Uwl
Decanter inlet Qp mass flow rate of the fed olive paste (kg/h) Qw process water mass flow rate for olive paste dilution (kg/h) a ‘‘dilution coefficient’’ as the ratio between the process water mass flow rate and the olive paste mass flow rate fed to the decanter (Qw/Qp) Uwp mass fraction of water in the fed olive paste Uop mass fraction of oil in the fed olive paste (wet basis) Uopdb mass fraction of oil in the fed olive paste (dry basis) (Uopdb = Uop/Usp) Usp mass fraction of solids in the fed olive paste Decanter Qh Uwh Uoh Ush Fo Fvw
outlet discharged husk mass flow rate (kg/h) mass fraction of water in the discharged husk mass fraction of oil in the discharged husk (wet basis) (1 Uwh Uoh) mass fraction of solids in the discharged husk ‘‘oil out flow’’ as the mass flow rate of the extracted oil (kg/h) ‘‘vegetable water’’ mass out flow rate (kg/h)
In the continuous extraction process by decanter centrifuge, adding warm water to the olive paste, during malaxation or extraction, improves the separation of olive oil from water and husk (Amirante et al., 1995). Other studies underline the need for technologies to be gentle to preserve both the sensory and nutritional value of olive oil (Ranalli et al., 2001; Monteleone et al., 1998; Caponio et al., 2003), as opposed to enhancing extraction yield, reducing processing costs and environmental impact (Roig et al., 2006). Therefore, centrifugal extraction plant manufacturers concentrate their attention on limiting the addition of processing water, which represents the main cause of oxidative damage to olive paste and polyphenols reduction. With a view to optimizing separation efficiency and extraction performance using a decanter centrifuge, theoretical and empirical studies have been carried out to evaluate the incidence of several process parameters on decanter operation. Such studies have underlined the importance of paste properties and continuous control of parameters in olive oil processing (Amirante and Catalano, 1993, 2000; Amirante et al., 1995; Di Renzo and Colelli, 1997). Even more research effort has been spent gaining insights into the interaction between the decanter and processing parameters namely: incidence of oil/water ring level thickness, paste mass flow rate, dilution water mass flow rate, screw conveyor torque and conveyor/bowl differential speed DN with respect to extraction yield and fat content in discharged husks (Amirante et al., 1993; Amirante and Catalano, 1993, 2000; Leung, 1998; CornerWalker, 2000; Corner-Walker and Records, 2000; Catalano et al.,
Uol Usl
mass fraction of water in the discharged vegetable water (Fvw) mass fraction of oil in the discharged vegetable water (Fvw) (wet basis) (1 Uwvw Upvw) mass fraction of solids in the discharged vegetable water (Fvw) ‘‘liquid out flow’’ as the mass flow rate (Fo + Fvw) at the decanter exit mass fraction of water in the discharged overall liquid (Ql) mass fraction of oil in the discharged overall liquid (Ql) (wet basis) (1 Uwl Uol) mass fraction of solids in the discharged overall liquid (Ql)
Ratio parameters go ‘‘oil recovery efficiency’’ as the ratio between extracted oil and the overall oil of the olive paste D ‘‘distribution coefficient’’ as the ratio between the overall liquid mass flow rate (water + oil) in the husk and overall liquid mass flow rate (water + oil) at the decanter entry S ‘‘separation coefficient’’ as (1 D) e the ratio between discharged husk mass flow rate and mass flow rate of the fed olive paste (Qh/Qp) c the ratio between ‘‘vegetable water’’ mass out flow rate and mass flow rate of the fed olive paste (Fvw/Qp) k the ratio between ‘‘liquid out flow’’ and mass flow rate of the fed olive paste (Ql/Qp = c + x) x the ratio between ‘‘oil out flow’’ and mass flow rate of the fed olive paste (Fo/Qp)
2003; Anlauf, 2007; Daou et al., 2007; Boncinelli et al., 2009; Altieri, 2010). Therefore, in order to evaluate the relationship involving oil/ water ring levels and conveyor/bowl differential speed DN vs. centrifuge extraction yield, in this study an automatic feedback control system was designed and set up to control the paste mass flow rate fed to the decanter centrifuge. Both laboratory and industrial tests were performed. Subsequently the feedback control system was used in olive oil mill operations to regulate at constant value the olive paste flow rate fed to the decanter centrifuge, also the added water mass flow rate was held at a constant value. A decanter centrifuge was employed with variable differential speed between bowl and screw conveyor (i.e. variable DN) and with regulation capability of oil– water ring levels. Samples of husks and vegetable water, collected using different processing conditions, were used to evaluate residual fat and water content in order to measure extraction efficiency against DN and oil–water ring levels.
2. Materials and methods The first step of our research entailed defining the adequate mass flow rate sensor and then designing and building the feedback control system in order to control the mass flow rate of olive paste fed to the decanter centrifuge. The Corimass G300+ sensor (by Khrone) proved the most suitable as it measures the olive paste mass flow rate by determining
G. Altieri et al. / Journal of Food Engineering 119 (2013) 561–572
the Coriolis force along a vibrating titanium pipe. It can also measure the volumetric flow rate, density and temperature of the flowing material. In order to verify its accuracy and reliability, the G300+ sensor was tested in a laboratory by means of 16 trials carried out by manually weighing the collected amount of olive paste flowing in 180 s through the sensor, the olive paste temperature varying in the range from 12 °C to 30 °C. The pump electric motor (1.5 kW or 2 HP) was connected directly on line and its speed was 1410 rpm, the manual variable gearbox that connected the three-phase motor to the pump was regulated in order to have a measured paste mass flow rate of 1100 kg/h with the paste temperature at 26 °C. The control system was based on a software PID standard feedback control system (see Fig. 1) consisting of: a notebook computer equipped with data acquisition board DAQCard AI-16X-E50 (16 ADC channels with 16-bit resolution, by National Instruments); in-house management software built using LabView 6.0.2 (by National Instruments) constituting the feedback control software and data acquisition system; Corimass G300+ mass flow rate sensor (by Khrone); the connecting interface was through an industrial two-wire RS485 slave serial interface (Modbus protocol); olive paste pump coupled with a three-phase motor (1.5 kW of mechanical power at the lab and 3.0 kW at the industrial plant); variable-frequency drive (VFD) (Siemens – Master Vector, 3 kW); the connecting interface was through an industrial two-wire RS485 slave serial interface (proprietary USS protocol by Siemens). Three constant mass flow rates (1500, 2000 and 2500 kg/h) were tested when the feedback control system was controlling the pump three-phase motor by VFD. The feedback control system allows the regulation of the olive paste mass flow rate fed to decanter centrifuge by varying the rotational speed of the olive paste pump three-phase motor using the VFD. The entire system was first tested in the laboratory and then in the olive oil mill.
2.1. Laboratory tests In order to test the feedback control system in the laboratory, a tank with reversed pyramidal bottom was equipped with an eccen-
tric screw pump (Moineau pump) coupled with a variable gearbox speed reducer (maximum mass flow rate of 5000 kg/h at 200 rpm); the tank was subsequently filled with olive paste recirculated by the pump. The mass flow rate sensor was installed on the pump outlet; the data were collected by a notebook computer. The response time of the feedback system depended on the VFD delay in controlling the three-phase motor, which amounted to about 5 s. Several trials were carried out to assess the robustness and responsiveness of the control system. 2.2. Industrial tests The experiments on the feedback control system in the industrial plant were carried out using an extraction plant equipped with one hammer crusher, two stacked malaxer units, one rotary lobe pump for olive paste feeding to the decanter and a Barracane – Megala 450 decanter having 1700 mm bowl length, 450 mm bowl inner diameter, bowl cylinder/cone height ratio equal to 1.0, 3500 rpm bowl rotational speed, 130 mm screw conveyor helical step. This decanter allows the regulation of the dam plates, which regulate the oil/water levels, even when the decanter is running. The parameters monitored during the tests were: olive paste temperature; mass flow rate of olive paste; mechanical power delivered by the three-phase motor to the pump as a percent of maximum motor nominal mechanical power (%). After the first test in the laboratory, the feedback control system was installed permanently in the industrial plant. Homogeneous batches of olive drupes (Olea europaea L., cv. Coratina), harvested in olive groves in Puglia (southern Italy), were selected and used to perform an experimental design of three trials (at different RING regulation) using approximately 12,500 kg each day. The trials were divided into five tests (at different DN) of approximately 2500 kg each, to ensure an average working time of about 85 min, in order to guarantee the following test protocol: 15 min to stabilize the functioning of the decanter centrifuge, 70 min available for sampling. For each test a sample was taken of olive paste and four samples (one every 15 min) of husks and vegetable water, in order to evaluate both moisture and fat content. During the trials the differential speed between the screw conveyor and bowl (DN) was varied at 12, 15, 16, 17 and 19 rpm. For each DN the oil–water ring level was varied at 283.6, 285.2 and 286.8 mm corresponding to levels A, B, C, respectively. The paste produced by the olives had 44.9% moisture content (std. dev. of 2.8%) and 18.9% fat content (wet basis) (std. dev. of 1.0%). Apparent viscosity of the olive paste was measured by a rotational viscosity meter (HB-DV II+ – Brookfield), using Brookfield disk sensors (spindles) typed 3, 4, 5 and 6. Mitschka’s method (Mitschka, 1982) was used to characterize the olive paste as reported in Steffe (1996) and in Briggs and Steffe (1997). The power law model described by Eq. (1) applies to olive paste:
SS ¼ K SRn
Fig. 1. Feedback control system used to control the olive paste mass flow 1 rate supplied to the decanter centrifuge.
563
ð1Þ
Olive paste had the consistency coefficient K = 1.277 (std. dev. of 0.0966) and the flow behaviour index n = 0.131 (std. dev. of 0.0073) measured at 25 °C. Paste was processed according with the following operation flow sheet: malaxing for 30 min at 28 °C; centrifugal oil extraction
564
G. Altieri et al. / Journal of Food Engineering 119 (2013) 561–572
of different treatments, of about 2500 kg of olive paste, using a triple-phase decanter centrifuge (vegetable water, oil and husks). The paste flow rate was fixed at a constant value (1800 kg/h) and water flow rate at 900 kg/h obtaining a dilution factor (a) of 0.50. Data were analysed using a non-parametric analysis of variance by Friedman’s test with respect to DN and ring levels, followed up by a multiple comparison test (MCT) under Tukey’s ‘‘honestly significant difference’’ (HSD) criterion. MCT on overall paired samples were analysed by the Mann–Whitney U-test and the familywise error rate (FWER) was controlled by Hommel’s method, adjusting the p-value of each comparison (Hommel, 1988, 1989). In each case, FWER was set to a significance level of 0.10, given the great variability of vegetable products albeit produced in the same orchard. Since a high level of variance could be expected, use of the 0.10 FWER level is felt justified. Moisture content of olive paste, husk and vegetable water was based on weight difference after sample drying, according to the official method. Fat content was assessed using a solvent extractor based on the Randall extraction technique (Ser 148/3 – VELP SCIENTIFICA). All analyses were carried out in triplicate. The parameters used for the mass balance of the decanter centrifuge are depicted in Fig. 2. The known parameters are: a, Qp, Uop, Uwp, Uoh, Uwh, Uovw, and Uwvw. Given the mass balance of solids, the following equation holds:
S, depending on machine capacity to separate solids from liquids; husk water content (Uwh), which depends on machine capacity to discriminate between the two liquid phases and particularly to separate efficiently the oil phase from water discharged with husk. These parameters result from Eqs. (8) and (9) below, where the dependent parameters are all known:
Qp Usp ¼ Qh Ush þ Fvw Usvw
go ¼ 1 e
ð2Þ
Moreover, from the definition of Fo and Fvw, Eqs. (3) and (4) also hold:
Fo ¼ Qp Uop Qh Uoh Fvw Uovw
ð3Þ
Fvw Uwvw ¼ a Qp þ Qp Uwp Qh Uwh
ð4Þ
Eqs. (2)–(4) can be solved as functions of known and measured parameters. Hence the solution is:
Qh ¼
Qp½Usvwða þ UwpÞ UspUwvw ðUsvwUwh UshUwvwÞ
Qp½Ushða þ UwpÞ UspUwh Fvw ¼ ðUsvwUwh UshUwvwÞ Fo ¼ Qp Uop
ð5Þ
ð6Þ
Qp½Usvwða þ UwpÞ UspUwvw Uoh ðUsvwUwh UshUwvwÞ
Qp½Ushða þ UwpÞ UspUwh Uovw ðUsvwUwh Ush UwvwÞ
ð7Þ
Furthermore, the following decanter functioning parameters were calculated (Altieri, 2010): oil recovery efficiency go, that is an index of overall extraction process; the separation coefficient
Fig. 2. Diagram of decanter centrifuge with the main mass flow rate parameters used in mass balance equations.
go ¼
Fo Qp Uop
S¼1
QhðUwh þ UohÞ Qpða þ Uwp þ UopÞ
ð8Þ
ð9Þ
Starting from their definition the following parameters are found: e from Eq. (5), c from Eq. (6), x from Eq. (7) and k as c + x. Therefore, the equations can be written in terms of ratio parameters as follows:
e¼
Usvwða þ UwpÞ UspUwvw UsvwUwh Ush Uwvw
ð10Þ
c¼
Ushða þ UwpÞ þ UspUwh Usvw Uwh Ush Uwvw
ð11Þ
x ¼ Uop e Uoh c Uovw
S¼1e
Uoh Uovw x c ¼ Uop Uop Uop Uwh þ Uoh
a þ Uwp þ Uop
k¼cþx
ð12Þ ð13Þ
ð14Þ ð15Þ
3. Results and discussion 3.1. Laboratory tests of equipment Fig. 3 shows the results of laboratory tests comparing the effective mass flow rate of the pump and the measure performed with the G300+ sensor against temperature variation, the data measured by Corimass G300+ sensor were used for polynomial fitting. The motor pump was connected directly on-line (DOL) and its gearbox adjusted in order to have an operating point of roughly 1100 kg/h at 26 °C of paste temperature. The tank was then filled with olive paste at approximately 12 °C. The results (see Fig. 3) show a variation in paste mass flow rate with paste temperature even if the pump motor is connected DOL and the pump-motor gearbox ratio was fixed. Indeed, when the temperature rises from 12 °C to 30 °C, then the mass flow rate increases by 17% (see Fig. 3). This means an increase related to the decrease in olive paste viscosity. This confirmed that the mass flow rate is strictly dependent on paste temperature (and viscosity) and not only on pump properties. Use of the G300+ sensor permits exact control of the mass flow rate, which makes the sensor suitable and very effective for measuring the olive paste mass flow rate, given that the mass flow rate varies according to paste temperature, paste viscosity and pump properties. Measurement of the paste mass flow rate was affected by less than 1% relative error when using the G300+ sensor. Fig. 4 shows the feedback control system start-up at fixed setpoints of 1500, 2000 and 2500 kg/h of paste mass flow rate. The mass flow rate varied continuously during pumping due to changes in paste viscosity. The measured time delay of 4 s is due to the delayed control of VFD over the three-phase motor. The measured
G. Altieri et al. / Journal of Food Engineering 119 (2013) 561–572
565
Fig. 3. Laboratory test of the G300+ sensor carried out by weighing a collected amount of olive paste flowing through the sensor in 180 s, varying the olive paste temperature from 12 °C to 30 °C. The motor pump is connected directly on-line at a fixed gearbox ratio in order to have an operating point of approximately 1100 kg/h at 26 °C of paste temperature. The Corimass G300+ sensor measured data were used for polynomial fitting.
Fig. 4. Feedback control system static performance when controlling the motor pump by VFD at three set point values of paste mass flow rate (paste temperature was 29 °C).
Fig. 5. Feedback control system dynamic performance when controlling the motor pump by VFD at a set point of 2500 kg/h of paste mass flow rate with sudden changes in pump gearbox transmission ratio (paste temperature was 29 °C).
566
G. Altieri et al. / Journal of Food Engineering 119 (2013) 561–572
rise time (at 90% of set-point) (15 s at 1500 kg/h, 23 s at 2000 kg/h and 30 s at 2500 kg/h) is a measure of the feedback control system static performance, which means a rise in average mass flow rate speed of 90.1 kg/h per second. Fig. 5 shows the feedback control system reaction caused by sudden variation in mass flow rate when changing the pump gearbox transmission ratio. After an increase in flow rate (point 1 in Fig. 5) a reduction occurs in mechanical power supplied (point 2 in Fig. 5) to the pump (which means a reduction in the number of revolutions), and after 50 s the original flow rate (point 3 in Fig. 5) was re-established. On analysing the response time, a parameter typed R% (%/s) and called the ‘‘percent recovery parameter’’ was introduced to measure the dynamic performance of the feedback control system. This parameter is defined as the ratio between the relative change of the mass flow rate (as percent of set point value) and the time required by the control system to restore the set point, after a mass flow rate sudden change. Therefore, the parameter measures the mass flow rate variation, as a percent of the fixed set point, that the feedback system is able to control in a time period of 1 s, so it is related to the responsiveness of the feedback system. The average R% was found equal to ±0.24%. Fig. 6 compares the ‘‘regulated’’ feedback control system with the ‘‘unregulated’’ system when operating on highly viscous paste. Although the average mass flow rate is almost the same (1543.08 kg/h ‘‘unregulated’’ vs. 1520.23 kg/h ‘‘regulated’’), the difference is conspicuous when comparing data variance: 415476.7 ‘‘unregulated’’ vs. 31946.7 ‘‘regulated’’. The ratio between variance of ‘‘unregulated’’ data and that of ‘‘regulated’’ data (RV) is approximately 13.01. This demonstrates the effectiveness of the feedback system in controlling the paste mass flow rate. The parameter that shows the dynamic regulating performance of the feedback system is the average absolute relative error percent (AAREP) related to the imposed set-point that shows the precision of the feedback system. The AAREP is 9.5% when regulated and 37.8% when unregulated, the ratio being 3.96. This represents the improvement in precision while the feedback control system is running. Fig. 7 shows the feedback control system when operating on paste held at 18 °C and 26 °C. The dynamic performance indices of the feedback system are also tabulated in Fig. 7. With paste at higher temperatures the regulating precision of the feedback system is enhanced due to the decreased viscosity
of the treated paste; the AAREP is 9.5% at 18 °C and 3.6% at 26 °C. Hypothesizing a linear decrease in AAREP with temperature, then the measured rate of the decrease is 0.74% per degree centigrade. 3.2. Equipment tested in the industrial plant The feedback control system was preliminarily operated in an industrial plant working at a constant flow rate of about 1000 kg/ h, in order to assess and characterize its dynamic behaviour on long running trials. Fig. 8 shows that sharp changes in olive paste mass flow rate are very frequent when operating in an olive oil mill. Furthermore, there are periods (e.g. from 670 to 1362 s, equivalent to 11.5 min of work) when the average mass flow rate (867 kg/h) is less than the set-point value (1000 kg/h). In this case, the olive paste fed to the decanter could prove more diluted as dilution water remains constant. Hence this causes a variation in decanter centrifuge regulation that affects separated oil quality and quantity. Mass flow rate variation delivered by the paste pump is due to hydraulic pressure variation in the malaxer unit during olive paste discharge. The level of olive paste in the unit affects hydraulic pressure over the olive paste fed into the pump. This leads to a change in pump hydraulic efficiency and represents, together with variations in olive paste temperature and viscosity, additional interference during the extraction process using the decanter centrifuge. Therefore, use of the feedback control system in order to control the mass flow rate of olive paste fed to the decanter centrifuge is mandatory in order to operate at optimal conditions without significant sharp variations in paste mass flow rates. Moreover, Fig. 8 shows that sudden changes are less frequent when feedback control is operating. Indeed, from 8.6% without regulation the AAREP decreases to 2.4% when operating with regulation. 3.3. Industrial plant trials Figs. 9–11 show iso-curves of go, Uwh and S respectively, obtained from experimental data (black dots). The iso-curves were obtained using the Matlab function ‘‘interp2’’ using a surface cubic splines interpolation. In order to show the experimental points that are significantly different, were defined the ‘‘not significantly different neighbouring values’’ points, in statistical sense, that are those experimental
Fig. 6. Feedback control system dynamic performance when controlling the motor pump by VFD at 1500 kg/h of paste mass flow rate was compared with the pump without VFD control (paste temperature was 18 °C).
G. Altieri et al. / Journal of Food Engineering 119 (2013) 561–572
567
Fig. 7. Feedback control system dynamic performance when controlling the motor pump by VFD at 1500 kg/h of paste mass flow rate was compared when paste temperature was set at 18 °C and 26 °C.
Fig. 8. Feedback control system dynamic performance when controlling the motor pump by VFD at 1000 kg/h of paste mass flow rate was compared with the pump without VFD control (paste temperature was 24 °C).
points (black dots) in Figs. 9–11 that are ‘‘not significantly different’’ and ‘‘adjacent’’ to each experimental point, these points are connected with a greyed area. More specifically this happens only in Fig. 10 because the two points at DN = 12 and DN = 15 (both with RING = 283.5) show a value that is not significantly different at the stated significance level of 0.10. Fig. 9 shows iso-curves obtained from experimental data and Eq. (8) for go; go absolute maximum (0.972) is achieved with DN = 13.3 and RING = 284.8. As expected go maximum is achieved at lower DN which guarantees a higher residence time for blending inside the decanter. In this case the maximum is slightly dependent on ring level. Further, go minimum is achieved at the highest DN where residence time is too short to allow efficient oil extraction. Moreover, for DN values higher than about 18 (see Fig. 9) the go value becomes independent of RING, because the iso-curves are roughly parallel to RING axis, but depends uniquely on DN. In addition, at these high values of DN, the go decrease rate becomes very high. Therefore, possible use of the go secondary maximum as a working point, located in the inner part of the iso-curve equal to 0.954, could be really difficult due to the high go decreasing rate that makes extraction optimization difficult.
The same condition occurs when considering the working point at go absolute maximum, but this is less problematic because, as DN decreases, the reduction in go is lower than in the previous case. Once again, for DN values lower than 12 (see Fig. 9), go becomes independent of RING because iso-curves are roughly parallel to the RING axis, but it depends uniquely on DN for this specific decanter. Fig. 10 shows iso-curves obtained from Uwh experimental data; Uwh maximum (0.596) is achieved with DN = 13.4 and RING = 285.6 whereas Uwh minimum (0.437) is achieved with DN = 13.2 and RING = 283.5. The minimum is not well statistically identified as it can span over DN values from 12 to 15. Uwh is correlated with the machine’s capability to separate two liquid phases and especially with efficient separation of oil from discharged water into husk: higher Uwh values allow the machine to accomplish a higher separation grade between oil and vegetable water. Fig. 11 shows iso-curves obtained from the experimental data and Eq. (9) for S; S maximum (0.801) is achieved with DN = 13.2 and RING = 284.4; as expected this is achieved at the lowest DN that guarantees a higher residence time for the blending inside the decanter. In this case the maximum is heavily dependent on RING.
568
G. Altieri et al. / Journal of Food Engineering 119 (2013) 561–572
Fig. 9. Iso-curves obtained from experimental data and Eq. (8) for go by surface cubic spline interpolation; greyed area represents not significantly different neighbouring values; black dots are the experimental points.
Fig. 10. Iso-curves obtained from experimental data of Uwh by surface cubic spline interpolation; the greyed area represents not significantly different neighbouring values; black dots are the experimental points.
Further, S minimum is achieved at the highest DN where the residence time is too short to allow efficient liquid–solid separa-
tion. Moreover, for DN values higher than about 18 (see Fig. 11), S becomes only slightly dependent on RING, because the iso-curves
569
G. Altieri et al. / Journal of Food Engineering 119 (2013) 561–572
Fig. 11. Iso-curves obtained from experimental data and Eq. (9) for S by surface cubic spline interpolation; the greyed area represents not significantly different neighbouring values; black dots are the experimental points.
are almost parallel to the RING axis, but it depends uniquely on DN. Further; at such high DN values, the S decrease rate becomes very high. Separation coefficient S is correlated to the ability of the centrifuge to separate the solid phase from liquids. A decanter centrifuge that separates the solid phase from liquids as far as possible is preferable. A higher separation coefficient guarantees a higher oil recovery efficiency even at lower paste dilution: this is preferable to maintain extracted oil quality. Furthermore, the similarity between Fig. 9 and Fig. 11 shows that go and S are strictly related to decanter centrifuge geometric characteristics because maximum and minimum values almost coincide. It has been pointed out a positive correlation (i.e. R2 = 0.74) between these two parameters. Nevertheless, the relation between these parameters is more complex than the simply linear one. Moreover, once the decanter working point has been selected with reference to go maximum, then Uwh is an outcome. This means that Uwh cannot be selected whereas the S value is very close to its maximum. Hence, optimization of go allows at the same time optimization of S which, as previously outlined, guarantees extracted oil quality if operating at low paste dilution. Therefore, the main objective is to operate at go maximum as quality of extracted oil is achieved by the paste dilution ratio set at 0.5. 3.4. Ratio parameters of the decanter centrifuge (a, go, S, e, c, k, x) and their relations with DN and RING On the basis of our experimental results there were shown to be relationships between ratio parameters of the decanter centrifuge. They depend uniquely on decanter geometry and input variables Uopdb, DN and RING, through a complete quadratic interaction model. The model is based on a ‘‘black box’’ approach with
independent variables combined in a multilinear model of full quadratic surface response analysis. The ‘‘x2fx’’ Matlab function has been used to perform the variables handling. More specifically the model consists of a linear function of 21 terms formed by: a constant term; linear terms Uopdb, DN, DN0.5, DN1, RING; interaction terms constituted by all pairwise products of linear terms and squared terms Uopdb2, DN2, DN1, DN2, RING2. Each model variable is a column vector and when the complete quadratic interaction model is assembled, then a matrix M is built with the model variables as columns. The solution is found by multiple linear regression linking each ratio parameter to matrix M through a vector of constant terms found as system least squares solution. This constant vector represents the signature of the ratio parameter in question against input variables: Uopdb, DN and RING. Results of this processing method are reported in Table 1 which shows that from experimental data, given the high correlation level, the solution found can be used to draw go iso-curves as a function of DN and RING parametrically against Uopdb. Consequently, if the Uopdb value of olive paste is known from chemical or spectrophotometric analytical methods, this
Table 1 Evaluation of multiple linear regression of ratio parameters go, S, e, c, x and k as functions of input variables Uopdb, DN and RING. Parameter
R2
Average relative error%
Minimum relative error%
Maximum relative error%
go
0.9964 0.9918 0.9898 0.9896 0.9997 0.9898
0.09 0.81 1.22 1.39 0.09 1.09
0.03 0.25 0.45 0.40 0.03 0.33
0.19 1.62 2.99 2.81 0.19 2.24
S
e c x k
570
G. Altieri et al. / Journal of Food Engineering 119 (2013) 561–572
dependence can be used to fine tune decanter DN and RING, such that maximum oil recovery efficiency is achieved. In order to find the relations that allow identification of DN and RING optimal values which guarantee go maximum, Uopdb value was stepped from 22% to 22% with respect to the Uopdb average
value equal to 0.525, and then DN and RING optimal values were taken at go maximum. Samples of these steps are shown in Fig. 12. Fig. 12 shows iso-curves of go as a function of DN and RING when Uopdb is fixed at 22% (Fig. 12a) and +22% (Fig. 12b) of the Uopdb average value. The go absolute maximum and secondary
Fig. 12. Iso-curves obtained from multiple linear regression over experimental data for go vs. DN and RING. Uopdb is fixed at 22% (case a) and +22% (case b) of average Uopdb value equal to 0.525.
G. Altieri et al. / Journal of Food Engineering 119 (2013) 561–572
maximum move, increasing their values, from low DN (11.9) to high DN (13.3) as Uopdb increases and olive paste moisture content decreases. Fig. 12a shows that go absolute maximum is dependent on both RING and DN when considering pastes with high moisture content and therefore with a low Uopdb. Dependence on RING decreases as the iso-curves become almost parallel to the RING axis, more specifically at DN from 12.5 to 14 and at DN above 18. Fig. 12b shows that go absolute maximum is dependent on DN but very slightly on RING when considering olive pastes with low moisture content and therefore having a high Uopdb. Dependence on RING diminishes as the iso-curves become roughly parallel to the RING axis, specifically at DN from 12 to 15.5. Fig. 13 shows the curves that allow identification of DN and RING optimal values which guarantee go maximum. Highly correlated polynomials are found that link the DN optimal value (thirdorder) and RING optimal value (second-order) to Uopdb values.
571
Therefore, if the Uopdb of olive paste is known by an on-line method, then the polynomials found allow to calculate and modify, in real time, decanter DN and RING optimal values in order to achieve maximum oil recovery efficiency. Fig. 14 shows the relation between go absolute maximum vs. Uopdb. The curve was obtained from multiple linear regression of go with regard to experimental data. The Uopdb value was stepped from 22% to 22% with respect to the Uopdb average. The go absolute maximum was then taken. Once again, a highly correlated fourth-order polynomial is found that links go optimal value to Uopdb values. The go optimal value increases when Uopdb increases up to the go maximum value, after which go decreases. Therefore, this demonstrates that the go optimal value is determined by the olive paste moisture content, and, further, that it is a maximum value of Uopdb (0.61) which guarantees the maximum optimal go (0.982).
Fig. 13. Curves obtained from multiple linear regression of go over experimental data. The fixed Uopdb value was stepped from 22% to 22% with respect to the Uopdb average value equal to 0.525. DN and RING optimal values were then taken at maximum go. Highly correlated polynomials exist that link optimal value of DN (third-order) and RING (second-order) to Uopdb values.
Fig. 14. Curve obtained from multiple linear regression of go over experimental data. The fixed Uopdb value was stepped from 22% to 22% with respect to the Uopdb average value equal to 0.525. The maximum go was then determined. There is a highly correlated fourth-order polynomial that links the optimal value of go to Uopdb values. Moreover, the optimal go value increases when Uopdb increases up to a maximum of 0.9820 at a Uopdb of 0.61, after which go decreases.
572
G. Altieri et al. / Journal of Food Engineering 119 (2013) 561–572
4. Conclusions Olive oil centrifugal extraction strictly depends on both olive paste temperature and water content. Therefore, if olive paste is fed to a decanter without an automatic control, then non-optimal extraction is performed. In order to set up an automatic system for the control of the decanter during olive oil extraction, a feedback control system was tested in both laboratory and industrial processing plant after successfully testing a suitable flow mass sensor. A set of trials were carried out, using constant quality olives (cv. Coratina), and an on-line system was built in order to control the olive paste mass flow rate fed to a decanter centrifuge. This control system allowed assessment of the incidence of several parameters on extraction yield, husk fat content and vegetable water fat content. Our findings confirmed that olive paste mass flow rate was dependent on pump characteristics, olive paste temperature and viscosity. The feedback control system was able to guarantee a constant feed to the decanter centrifuge even in the presence of olive paste physical changes. Both ‘‘oil recovery efficiency’’ (go) and the ‘‘separation coefficient’’ S peaked with the highest residence time, using DN = 13.3 and RING = 284.8, while go minimum was achieved at higher DN. Moreover, for DN values higher than about 18 or less than 12, the go value becomes independent of RING. The go and S values are strictly related to decanter centrifuge geometry, such that operating at go maximum is the main objective because paste dilution ratio guarantees extracted oil quality. Interesting relations were found between decanter centrifuge ratio parameters depending on a complete quadratic interaction model from the input variables leading to highly correlated polynomial relations between DN vs. Uopdb and go optimal vs. Uopdb. These relations could be used to optimize DN and RING in order to achieve maximum oil recovery efficiency. Indeed, if the olive paste Uopdb value is known by an on-line method, then the polynomials found enable decanter DN and RING optimal values to be calculated and modified, in real time, so as to achieve maximum oil recovery efficiency. Acknowledgements The authors thank both the ing. Lorusso and dott.ssa Barracane of the firm Barracane s.r.l. for their valuable help and to have made possible the carrying out of these trials. References Altieri, G., 2010. Comparative trials and an empirical model to assess throughput indices in olive oil extraction by decanter centrifuge. Journal of Food Engineering 97, 46–56.
Amirante, P., Catalano, P., 1993. Analisi teorica e sperimentale dell’estrazione dell’olio d’oliva per centrifugazione [theoretical and experimental analysis of olive oil extraction by centrifugation]. La Rivista Italiana Delle Sostanze Grasse LXX, 329–335 (in Italian). Amirante, P., Di Renzo, G.C., Colelli, G., 1995. Estrazione centrifuga con controllo continuo dei parametri operativi [extraction by centrifuge with continuous control of the operating parameters]. Rivista di Ingegneria Agraria, Quaderno n 17, 344–351 (in Italian). Amirante, R., Baccioni, L., Catalano, P., Montel, G.L., 1993. Nuove tecnologie per l’estrazione dell’olio di oliva: decanter con cono a pressione dinamica variabile e controllo della velocità differenziale tamburo-coclea [new technologies for the extraction of olive oil: decanters with a dynamic variable pressure cone and control of the differential speed conveyor-bowl]. La Rivista Italiana Delle Sostanze Grasse LXXVI, 129–140 (in Italian). Amirante, R., Catalano, P., 2000. Fluid dynamic analysis of the solid–liquid separation process by centrifugation. Journal of Agricultural Engineering Research 77 (2), 193–201. Anlauf, H., 2007. Recent developments in centrifuge technology. Separation and Purification Technology 58 (2), 242–246. Boncinelli, P., Daou, M., Cini, E., Catalano, P., 2009. A simplified model for designing and regulating centrifugal decanters for olive oil production. Transactions of the ASABE, American Society of Agricultural and Biological Engineers 52 (6), 1961– 1968. Briggs, J.L., Steffe, J.F., 1997. Using brookfield data and the mitschka method to evaluate power law foods. Journal of Texture Studies 28 (5), 517–522. Caponio, F., Gomes, T., Summo, C., Pasqualone, A., 2003. Influence of the type of olive-crusher used on the quality of extra virgin olive oils. European Journal of Lipid Science Technology 105, 201–206. Catalano, P., Pipitone, F., Calafatello, A., Leone, A., 2003. Productive efficiency of decanters with short and variable dynamic pressure cones. Biosystems Engineering 86 (4), 459–464. Corner-Walker, N., 2000. The dry solids decanter centrifuge: capacity scaling. Filtration and Separation 37/4 (May), 28–32. Corner-Walker, N., Records, F.A., 2000. The dry solids decanter centrifuge: conveyor torque and differential. Filtration and Separation 18 (October), 23. Daou, M., Furferi, R., Recchia, L., Cini, E., 2007. A modelling approach to extra virgin olive oil extraction. Journal of Agricultural Engineering 38 (4), 11–20. http:// dx.doi.org/10.4081/jae.2007.4.11. Di Renzo, G.C., Colelli, G., 1997. Flow behavior of olive paste. Applied Engineering in Agriculture 13 (6), 751–755. Hommel, G., 1988. A stagewise rejective multiple test procedure on a modified Bonferroni test. Biometrika 75, 383–386. Hommel, G., 1989. A comparison of two modified Bonferoni procedures. Biometrika 76, 624–625. Leung, W.W.-F., 1998. Torque requirement for high-solids centrifugal sludge dewatering. Filtration and Separation 883 (November), 887. Mitschka, P., 1982. Simple conversion of Brookfield R.V.T. readings into viscosity functions. Rheologica Acta 12, 207–209. Monteleone, E., Caporale, G., Carlucci, A., Paglierini, E., 1998. Optimisation of extra virgin olive oil quality. Journal of the Science of Food and Agriculture 77, 31–37. Piacqadio, P., De Stefano, G., Sciancalepore, V., 1998. Quality of virgin olive oil extracted with new centrifugation system using a two phase decanter. Fett/ Lipid 10, 472–474. Ranalli, A., Cabras, P., Iannucci, E., Contento, S., 2001. Lipochromes, vitamins, aromas and other components of virgin olive oil are affected by processing technology. Food Chemistry 73, 445–451. Ranalli, A., De Mattia, G., Ferrante, M.L., 1997. Comparative evaluation of the olive oil given by a new processing system. International Journal of Food Science and Technology 32, 289–297. Roig, A., Cayuela, M.L., Sanchez-Monedero, M.A., 2006. An overview on olive mill wastes and their valorisation methods. Waste Management 26, 960–969. Salvador, M.D., Aranda, F., Gomez-Alonso, S., Fregapane, G., 2003. Influence of extraction system, production year and area on Cornicabra virgin olive oil: a study of five crop season. Food Chemistry 80, 359–366. Steffe, J.F., 1996. Rheological Methods in Food Process Engineering, Second ed. Freeman Press, East Lansing, MI.
Contents lists available at SciVerse ScienceDirect
Journal of Food Engineering journal homepage: www.elsevier.com/locate/jfoodeng
Horizontal centrifuge with screw conveyor (decanter): Optimization of oil/water levels and differential speed during olive oil extraction Giuseppe Altieri ⇑, Giovanni Carlo Di Renzo, Francesco Genovese SAFE – Scuola di Scienze Agrarie, Forestali, Alimentari e Ambientali, Università degli Studi della Basilicata, Viale dell’Ateneo Lucano, 10, 85100 Potenza, Italy
a r t i c l e
i n f o
Article history: Received 27 February 2013 Received in revised form 29 May 2013 Accepted 16 June 2013 Available online 2 July 2013 Keywords: Olive oil extraction Quality preservation Decanter centrifuge Feedback control
a b s t r a c t Olive oil extraction using a horizontal centrifuge with a screw conveyor is an essential operation to reduce production costs. However, proper control of the plant is required to maintain a high level of extraction yield and olive oil quality. Rheological characteristics of the olive paste, which change in relation to water content, fruit variety, maturity level and seasonal temperature variations, greatly affect the efficiency of centrifugal extraction. If olive paste is fed to a decanter without automatic control, then nonoptimal extraction is performed. After successfully testing a suitable flow mass sensor in order to set up an automatic system to control the olive paste mass flow rate fed to a decanter centrifuge during olive oil extraction, a feedback control system was tested in the laboratory and built in-line in an industrial processing plant. This allowed trials to be carried out at constant mass flow rate for both the paste and the added water, on a decanter centrifuge with variable differential speed between bowl and screw conveyor (i.e. variable DN) and with regulation capability of oil–water ring levels. Constant quality olives (cv. Coratina) were used for the trials; the correlation was evaluated between oil/water ring levels and DN with respect to the extraction yield, husk fat content and vegetable water fat content. The ‘‘oil recovery efficiency’’ (go) and ‘‘separation coefficient’’ S peaked with maximum residence time, using DN = 13.3 and RING = 284.8, while go minimum was achieved at higher DN. More specifically, for DN values higher than about 18 or lesser than 12, go becomes independent of RING. Furthermore, go and S values are strictly related to the geometry of the decanter centrifuge, whereas the paste dilution ratio determines the preservation quality of the oil extracted, and the maximum oil recovery efficiency represents the remaining objective related to the quantity of extracted oil. From experimental data some interesting relations were found linking decanter centrifuge parameters; their relationship depends on a complete quadratic interaction model constituted by the input variables that leads to highly correlated polynomials between DN vs. Uopdb (‘‘mass fraction of oil in the fed olive paste’’ (dry basis)) and go optimal vs. Uopdb. These relations allow optimization of DN and RING values such that maximum oil recovery efficiency is achieved. Indeed, knowledge of Uopdb by an on-line method allows the decanter DN and RING operating parameters, to be calculated and modified at once, obtaining optimal maximum oil recovery efficiency. Ó 2013 Elsevier Ltd. All rights reserved.
1. Introduction Olive oil is the most widely used fat in the diet of Mediterranean countries, thanks especially to its healthy properties, unique aroma and long shelf life, and due to its natural antioxidant content, which make it different from other vegetable oils (Salvador et al., 2003). Evolution of the oil extraction process has led to the replacement of traditional discontinuous lines, using the pressure system
⇑ Corresponding author. Tel.: +39 0971 205468; fax: +39 0971 205429. E-mail address: [email protected] (G. Altieri). 0260-8774/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.jfoodeng.2013.06.033
extraction, with continuous ones, using centrifugal extraction. In particular, the horizontal centrifuge with a screw conveyor (i.e. decanter) is widely used in olive oil extraction, especially if large amounts of olives have to be processed in a short time (Ranalli et al., 1997; Piacqadio et al., 1998). However, olive oil extraction by centrifuge is dramatically affected by changes in the rheological characteristics of the olive paste in relation to water content, fruit variety, maturity level and seasonal temperature variations. Therefore, correct control of the process is essential in order to achieve high levels of both extraction yield and olive oil quality (such as content in natural antioxidants and aromatic compounds) (Amirante et al., 1995; Di Renzo and Colelli, 1997; Piacqadio et al., 1998; Ranalli et al., 2001; Altieri, 2010).
562
G. Altieri et al. / Journal of Food Engineering 119 (2013) 561–572
Nomenclature R% DN RING SS K SR n
‘‘percent recovery parameter’’ (%/s) differential speed between bowl and screw conveyor of decanter centrifuge in rpm. water/oil ring level shear stress (Pa) consistency coefficient (Pa sn) shear rate (s1) flow behaviour index
Uwvw Uovw Usvw Ql Uwl
Decanter inlet Qp mass flow rate of the fed olive paste (kg/h) Qw process water mass flow rate for olive paste dilution (kg/h) a ‘‘dilution coefficient’’ as the ratio between the process water mass flow rate and the olive paste mass flow rate fed to the decanter (Qw/Qp) Uwp mass fraction of water in the fed olive paste Uop mass fraction of oil in the fed olive paste (wet basis) Uopdb mass fraction of oil in the fed olive paste (dry basis) (Uopdb = Uop/Usp) Usp mass fraction of solids in the fed olive paste Decanter Qh Uwh Uoh Ush Fo Fvw
outlet discharged husk mass flow rate (kg/h) mass fraction of water in the discharged husk mass fraction of oil in the discharged husk (wet basis) (1 Uwh Uoh) mass fraction of solids in the discharged husk ‘‘oil out flow’’ as the mass flow rate of the extracted oil (kg/h) ‘‘vegetable water’’ mass out flow rate (kg/h)
In the continuous extraction process by decanter centrifuge, adding warm water to the olive paste, during malaxation or extraction, improves the separation of olive oil from water and husk (Amirante et al., 1995). Other studies underline the need for technologies to be gentle to preserve both the sensory and nutritional value of olive oil (Ranalli et al., 2001; Monteleone et al., 1998; Caponio et al., 2003), as opposed to enhancing extraction yield, reducing processing costs and environmental impact (Roig et al., 2006). Therefore, centrifugal extraction plant manufacturers concentrate their attention on limiting the addition of processing water, which represents the main cause of oxidative damage to olive paste and polyphenols reduction. With a view to optimizing separation efficiency and extraction performance using a decanter centrifuge, theoretical and empirical studies have been carried out to evaluate the incidence of several process parameters on decanter operation. Such studies have underlined the importance of paste properties and continuous control of parameters in olive oil processing (Amirante and Catalano, 1993, 2000; Amirante et al., 1995; Di Renzo and Colelli, 1997). Even more research effort has been spent gaining insights into the interaction between the decanter and processing parameters namely: incidence of oil/water ring level thickness, paste mass flow rate, dilution water mass flow rate, screw conveyor torque and conveyor/bowl differential speed DN with respect to extraction yield and fat content in discharged husks (Amirante et al., 1993; Amirante and Catalano, 1993, 2000; Leung, 1998; CornerWalker, 2000; Corner-Walker and Records, 2000; Catalano et al.,
Uol Usl
mass fraction of water in the discharged vegetable water (Fvw) mass fraction of oil in the discharged vegetable water (Fvw) (wet basis) (1 Uwvw Upvw) mass fraction of solids in the discharged vegetable water (Fvw) ‘‘liquid out flow’’ as the mass flow rate (Fo + Fvw) at the decanter exit mass fraction of water in the discharged overall liquid (Ql) mass fraction of oil in the discharged overall liquid (Ql) (wet basis) (1 Uwl Uol) mass fraction of solids in the discharged overall liquid (Ql)
Ratio parameters go ‘‘oil recovery efficiency’’ as the ratio between extracted oil and the overall oil of the olive paste D ‘‘distribution coefficient’’ as the ratio between the overall liquid mass flow rate (water + oil) in the husk and overall liquid mass flow rate (water + oil) at the decanter entry S ‘‘separation coefficient’’ as (1 D) e the ratio between discharged husk mass flow rate and mass flow rate of the fed olive paste (Qh/Qp) c the ratio between ‘‘vegetable water’’ mass out flow rate and mass flow rate of the fed olive paste (Fvw/Qp) k the ratio between ‘‘liquid out flow’’ and mass flow rate of the fed olive paste (Ql/Qp = c + x) x the ratio between ‘‘oil out flow’’ and mass flow rate of the fed olive paste (Fo/Qp)
2003; Anlauf, 2007; Daou et al., 2007; Boncinelli et al., 2009; Altieri, 2010). Therefore, in order to evaluate the relationship involving oil/ water ring levels and conveyor/bowl differential speed DN vs. centrifuge extraction yield, in this study an automatic feedback control system was designed and set up to control the paste mass flow rate fed to the decanter centrifuge. Both laboratory and industrial tests were performed. Subsequently the feedback control system was used in olive oil mill operations to regulate at constant value the olive paste flow rate fed to the decanter centrifuge, also the added water mass flow rate was held at a constant value. A decanter centrifuge was employed with variable differential speed between bowl and screw conveyor (i.e. variable DN) and with regulation capability of oil– water ring levels. Samples of husks and vegetable water, collected using different processing conditions, were used to evaluate residual fat and water content in order to measure extraction efficiency against DN and oil–water ring levels.
2. Materials and methods The first step of our research entailed defining the adequate mass flow rate sensor and then designing and building the feedback control system in order to control the mass flow rate of olive paste fed to the decanter centrifuge. The Corimass G300+ sensor (by Khrone) proved the most suitable as it measures the olive paste mass flow rate by determining
G. Altieri et al. / Journal of Food Engineering 119 (2013) 561–572
the Coriolis force along a vibrating titanium pipe. It can also measure the volumetric flow rate, density and temperature of the flowing material. In order to verify its accuracy and reliability, the G300+ sensor was tested in a laboratory by means of 16 trials carried out by manually weighing the collected amount of olive paste flowing in 180 s through the sensor, the olive paste temperature varying in the range from 12 °C to 30 °C. The pump electric motor (1.5 kW or 2 HP) was connected directly on line and its speed was 1410 rpm, the manual variable gearbox that connected the three-phase motor to the pump was regulated in order to have a measured paste mass flow rate of 1100 kg/h with the paste temperature at 26 °C. The control system was based on a software PID standard feedback control system (see Fig. 1) consisting of: a notebook computer equipped with data acquisition board DAQCard AI-16X-E50 (16 ADC channels with 16-bit resolution, by National Instruments); in-house management software built using LabView 6.0.2 (by National Instruments) constituting the feedback control software and data acquisition system; Corimass G300+ mass flow rate sensor (by Khrone); the connecting interface was through an industrial two-wire RS485 slave serial interface (Modbus protocol); olive paste pump coupled with a three-phase motor (1.5 kW of mechanical power at the lab and 3.0 kW at the industrial plant); variable-frequency drive (VFD) (Siemens – Master Vector, 3 kW); the connecting interface was through an industrial two-wire RS485 slave serial interface (proprietary USS protocol by Siemens). Three constant mass flow rates (1500, 2000 and 2500 kg/h) were tested when the feedback control system was controlling the pump three-phase motor by VFD. The feedback control system allows the regulation of the olive paste mass flow rate fed to decanter centrifuge by varying the rotational speed of the olive paste pump three-phase motor using the VFD. The entire system was first tested in the laboratory and then in the olive oil mill.
2.1. Laboratory tests In order to test the feedback control system in the laboratory, a tank with reversed pyramidal bottom was equipped with an eccen-
tric screw pump (Moineau pump) coupled with a variable gearbox speed reducer (maximum mass flow rate of 5000 kg/h at 200 rpm); the tank was subsequently filled with olive paste recirculated by the pump. The mass flow rate sensor was installed on the pump outlet; the data were collected by a notebook computer. The response time of the feedback system depended on the VFD delay in controlling the three-phase motor, which amounted to about 5 s. Several trials were carried out to assess the robustness and responsiveness of the control system. 2.2. Industrial tests The experiments on the feedback control system in the industrial plant were carried out using an extraction plant equipped with one hammer crusher, two stacked malaxer units, one rotary lobe pump for olive paste feeding to the decanter and a Barracane – Megala 450 decanter having 1700 mm bowl length, 450 mm bowl inner diameter, bowl cylinder/cone height ratio equal to 1.0, 3500 rpm bowl rotational speed, 130 mm screw conveyor helical step. This decanter allows the regulation of the dam plates, which regulate the oil/water levels, even when the decanter is running. The parameters monitored during the tests were: olive paste temperature; mass flow rate of olive paste; mechanical power delivered by the three-phase motor to the pump as a percent of maximum motor nominal mechanical power (%). After the first test in the laboratory, the feedback control system was installed permanently in the industrial plant. Homogeneous batches of olive drupes (Olea europaea L., cv. Coratina), harvested in olive groves in Puglia (southern Italy), were selected and used to perform an experimental design of three trials (at different RING regulation) using approximately 12,500 kg each day. The trials were divided into five tests (at different DN) of approximately 2500 kg each, to ensure an average working time of about 85 min, in order to guarantee the following test protocol: 15 min to stabilize the functioning of the decanter centrifuge, 70 min available for sampling. For each test a sample was taken of olive paste and four samples (one every 15 min) of husks and vegetable water, in order to evaluate both moisture and fat content. During the trials the differential speed between the screw conveyor and bowl (DN) was varied at 12, 15, 16, 17 and 19 rpm. For each DN the oil–water ring level was varied at 283.6, 285.2 and 286.8 mm corresponding to levels A, B, C, respectively. The paste produced by the olives had 44.9% moisture content (std. dev. of 2.8%) and 18.9% fat content (wet basis) (std. dev. of 1.0%). Apparent viscosity of the olive paste was measured by a rotational viscosity meter (HB-DV II+ – Brookfield), using Brookfield disk sensors (spindles) typed 3, 4, 5 and 6. Mitschka’s method (Mitschka, 1982) was used to characterize the olive paste as reported in Steffe (1996) and in Briggs and Steffe (1997). The power law model described by Eq. (1) applies to olive paste:
SS ¼ K SRn
Fig. 1. Feedback control system used to control the olive paste mass flow 1 rate supplied to the decanter centrifuge.
563
ð1Þ
Olive paste had the consistency coefficient K = 1.277 (std. dev. of 0.0966) and the flow behaviour index n = 0.131 (std. dev. of 0.0073) measured at 25 °C. Paste was processed according with the following operation flow sheet: malaxing for 30 min at 28 °C; centrifugal oil extraction
564
G. Altieri et al. / Journal of Food Engineering 119 (2013) 561–572
of different treatments, of about 2500 kg of olive paste, using a triple-phase decanter centrifuge (vegetable water, oil and husks). The paste flow rate was fixed at a constant value (1800 kg/h) and water flow rate at 900 kg/h obtaining a dilution factor (a) of 0.50. Data were analysed using a non-parametric analysis of variance by Friedman’s test with respect to DN and ring levels, followed up by a multiple comparison test (MCT) under Tukey’s ‘‘honestly significant difference’’ (HSD) criterion. MCT on overall paired samples were analysed by the Mann–Whitney U-test and the familywise error rate (FWER) was controlled by Hommel’s method, adjusting the p-value of each comparison (Hommel, 1988, 1989). In each case, FWER was set to a significance level of 0.10, given the great variability of vegetable products albeit produced in the same orchard. Since a high level of variance could be expected, use of the 0.10 FWER level is felt justified. Moisture content of olive paste, husk and vegetable water was based on weight difference after sample drying, according to the official method. Fat content was assessed using a solvent extractor based on the Randall extraction technique (Ser 148/3 – VELP SCIENTIFICA). All analyses were carried out in triplicate. The parameters used for the mass balance of the decanter centrifuge are depicted in Fig. 2. The known parameters are: a, Qp, Uop, Uwp, Uoh, Uwh, Uovw, and Uwvw. Given the mass balance of solids, the following equation holds:
S, depending on machine capacity to separate solids from liquids; husk water content (Uwh), which depends on machine capacity to discriminate between the two liquid phases and particularly to separate efficiently the oil phase from water discharged with husk. These parameters result from Eqs. (8) and (9) below, where the dependent parameters are all known:
Qp Usp ¼ Qh Ush þ Fvw Usvw
go ¼ 1 e
ð2Þ
Moreover, from the definition of Fo and Fvw, Eqs. (3) and (4) also hold:
Fo ¼ Qp Uop Qh Uoh Fvw Uovw
ð3Þ
Fvw Uwvw ¼ a Qp þ Qp Uwp Qh Uwh
ð4Þ
Eqs. (2)–(4) can be solved as functions of known and measured parameters. Hence the solution is:
Qh ¼
Qp½Usvwða þ UwpÞ UspUwvw ðUsvwUwh UshUwvwÞ
Qp½Ushða þ UwpÞ UspUwh Fvw ¼ ðUsvwUwh UshUwvwÞ Fo ¼ Qp Uop
ð5Þ
ð6Þ
Qp½Usvwða þ UwpÞ UspUwvw Uoh ðUsvwUwh UshUwvwÞ
Qp½Ushða þ UwpÞ UspUwh Uovw ðUsvwUwh Ush UwvwÞ
ð7Þ
Furthermore, the following decanter functioning parameters were calculated (Altieri, 2010): oil recovery efficiency go, that is an index of overall extraction process; the separation coefficient
Fig. 2. Diagram of decanter centrifuge with the main mass flow rate parameters used in mass balance equations.
go ¼
Fo Qp Uop
S¼1
QhðUwh þ UohÞ Qpða þ Uwp þ UopÞ
ð8Þ
ð9Þ
Starting from their definition the following parameters are found: e from Eq. (5), c from Eq. (6), x from Eq. (7) and k as c + x. Therefore, the equations can be written in terms of ratio parameters as follows:
e¼
Usvwða þ UwpÞ UspUwvw UsvwUwh Ush Uwvw
ð10Þ
c¼
Ushða þ UwpÞ þ UspUwh Usvw Uwh Ush Uwvw
ð11Þ
x ¼ Uop e Uoh c Uovw
S¼1e
Uoh Uovw x c ¼ Uop Uop Uop Uwh þ Uoh
a þ Uwp þ Uop
k¼cþx
ð12Þ ð13Þ
ð14Þ ð15Þ
3. Results and discussion 3.1. Laboratory tests of equipment Fig. 3 shows the results of laboratory tests comparing the effective mass flow rate of the pump and the measure performed with the G300+ sensor against temperature variation, the data measured by Corimass G300+ sensor were used for polynomial fitting. The motor pump was connected directly on-line (DOL) and its gearbox adjusted in order to have an operating point of roughly 1100 kg/h at 26 °C of paste temperature. The tank was then filled with olive paste at approximately 12 °C. The results (see Fig. 3) show a variation in paste mass flow rate with paste temperature even if the pump motor is connected DOL and the pump-motor gearbox ratio was fixed. Indeed, when the temperature rises from 12 °C to 30 °C, then the mass flow rate increases by 17% (see Fig. 3). This means an increase related to the decrease in olive paste viscosity. This confirmed that the mass flow rate is strictly dependent on paste temperature (and viscosity) and not only on pump properties. Use of the G300+ sensor permits exact control of the mass flow rate, which makes the sensor suitable and very effective for measuring the olive paste mass flow rate, given that the mass flow rate varies according to paste temperature, paste viscosity and pump properties. Measurement of the paste mass flow rate was affected by less than 1% relative error when using the G300+ sensor. Fig. 4 shows the feedback control system start-up at fixed setpoints of 1500, 2000 and 2500 kg/h of paste mass flow rate. The mass flow rate varied continuously during pumping due to changes in paste viscosity. The measured time delay of 4 s is due to the delayed control of VFD over the three-phase motor. The measured
G. Altieri et al. / Journal of Food Engineering 119 (2013) 561–572
565
Fig. 3. Laboratory test of the G300+ sensor carried out by weighing a collected amount of olive paste flowing through the sensor in 180 s, varying the olive paste temperature from 12 °C to 30 °C. The motor pump is connected directly on-line at a fixed gearbox ratio in order to have an operating point of approximately 1100 kg/h at 26 °C of paste temperature. The Corimass G300+ sensor measured data were used for polynomial fitting.
Fig. 4. Feedback control system static performance when controlling the motor pump by VFD at three set point values of paste mass flow rate (paste temperature was 29 °C).
Fig. 5. Feedback control system dynamic performance when controlling the motor pump by VFD at a set point of 2500 kg/h of paste mass flow rate with sudden changes in pump gearbox transmission ratio (paste temperature was 29 °C).
566
G. Altieri et al. / Journal of Food Engineering 119 (2013) 561–572
rise time (at 90% of set-point) (15 s at 1500 kg/h, 23 s at 2000 kg/h and 30 s at 2500 kg/h) is a measure of the feedback control system static performance, which means a rise in average mass flow rate speed of 90.1 kg/h per second. Fig. 5 shows the feedback control system reaction caused by sudden variation in mass flow rate when changing the pump gearbox transmission ratio. After an increase in flow rate (point 1 in Fig. 5) a reduction occurs in mechanical power supplied (point 2 in Fig. 5) to the pump (which means a reduction in the number of revolutions), and after 50 s the original flow rate (point 3 in Fig. 5) was re-established. On analysing the response time, a parameter typed R% (%/s) and called the ‘‘percent recovery parameter’’ was introduced to measure the dynamic performance of the feedback control system. This parameter is defined as the ratio between the relative change of the mass flow rate (as percent of set point value) and the time required by the control system to restore the set point, after a mass flow rate sudden change. Therefore, the parameter measures the mass flow rate variation, as a percent of the fixed set point, that the feedback system is able to control in a time period of 1 s, so it is related to the responsiveness of the feedback system. The average R% was found equal to ±0.24%. Fig. 6 compares the ‘‘regulated’’ feedback control system with the ‘‘unregulated’’ system when operating on highly viscous paste. Although the average mass flow rate is almost the same (1543.08 kg/h ‘‘unregulated’’ vs. 1520.23 kg/h ‘‘regulated’’), the difference is conspicuous when comparing data variance: 415476.7 ‘‘unregulated’’ vs. 31946.7 ‘‘regulated’’. The ratio between variance of ‘‘unregulated’’ data and that of ‘‘regulated’’ data (RV) is approximately 13.01. This demonstrates the effectiveness of the feedback system in controlling the paste mass flow rate. The parameter that shows the dynamic regulating performance of the feedback system is the average absolute relative error percent (AAREP) related to the imposed set-point that shows the precision of the feedback system. The AAREP is 9.5% when regulated and 37.8% when unregulated, the ratio being 3.96. This represents the improvement in precision while the feedback control system is running. Fig. 7 shows the feedback control system when operating on paste held at 18 °C and 26 °C. The dynamic performance indices of the feedback system are also tabulated in Fig. 7. With paste at higher temperatures the regulating precision of the feedback system is enhanced due to the decreased viscosity
of the treated paste; the AAREP is 9.5% at 18 °C and 3.6% at 26 °C. Hypothesizing a linear decrease in AAREP with temperature, then the measured rate of the decrease is 0.74% per degree centigrade. 3.2. Equipment tested in the industrial plant The feedback control system was preliminarily operated in an industrial plant working at a constant flow rate of about 1000 kg/ h, in order to assess and characterize its dynamic behaviour on long running trials. Fig. 8 shows that sharp changes in olive paste mass flow rate are very frequent when operating in an olive oil mill. Furthermore, there are periods (e.g. from 670 to 1362 s, equivalent to 11.5 min of work) when the average mass flow rate (867 kg/h) is less than the set-point value (1000 kg/h). In this case, the olive paste fed to the decanter could prove more diluted as dilution water remains constant. Hence this causes a variation in decanter centrifuge regulation that affects separated oil quality and quantity. Mass flow rate variation delivered by the paste pump is due to hydraulic pressure variation in the malaxer unit during olive paste discharge. The level of olive paste in the unit affects hydraulic pressure over the olive paste fed into the pump. This leads to a change in pump hydraulic efficiency and represents, together with variations in olive paste temperature and viscosity, additional interference during the extraction process using the decanter centrifuge. Therefore, use of the feedback control system in order to control the mass flow rate of olive paste fed to the decanter centrifuge is mandatory in order to operate at optimal conditions without significant sharp variations in paste mass flow rates. Moreover, Fig. 8 shows that sudden changes are less frequent when feedback control is operating. Indeed, from 8.6% without regulation the AAREP decreases to 2.4% when operating with regulation. 3.3. Industrial plant trials Figs. 9–11 show iso-curves of go, Uwh and S respectively, obtained from experimental data (black dots). The iso-curves were obtained using the Matlab function ‘‘interp2’’ using a surface cubic splines interpolation. In order to show the experimental points that are significantly different, were defined the ‘‘not significantly different neighbouring values’’ points, in statistical sense, that are those experimental
Fig. 6. Feedback control system dynamic performance when controlling the motor pump by VFD at 1500 kg/h of paste mass flow rate was compared with the pump without VFD control (paste temperature was 18 °C).
G. Altieri et al. / Journal of Food Engineering 119 (2013) 561–572
567
Fig. 7. Feedback control system dynamic performance when controlling the motor pump by VFD at 1500 kg/h of paste mass flow rate was compared when paste temperature was set at 18 °C and 26 °C.
Fig. 8. Feedback control system dynamic performance when controlling the motor pump by VFD at 1000 kg/h of paste mass flow rate was compared with the pump without VFD control (paste temperature was 24 °C).
points (black dots) in Figs. 9–11 that are ‘‘not significantly different’’ and ‘‘adjacent’’ to each experimental point, these points are connected with a greyed area. More specifically this happens only in Fig. 10 because the two points at DN = 12 and DN = 15 (both with RING = 283.5) show a value that is not significantly different at the stated significance level of 0.10. Fig. 9 shows iso-curves obtained from experimental data and Eq. (8) for go; go absolute maximum (0.972) is achieved with DN = 13.3 and RING = 284.8. As expected go maximum is achieved at lower DN which guarantees a higher residence time for blending inside the decanter. In this case the maximum is slightly dependent on ring level. Further, go minimum is achieved at the highest DN where residence time is too short to allow efficient oil extraction. Moreover, for DN values higher than about 18 (see Fig. 9) the go value becomes independent of RING, because the iso-curves are roughly parallel to RING axis, but depends uniquely on DN. In addition, at these high values of DN, the go decrease rate becomes very high. Therefore, possible use of the go secondary maximum as a working point, located in the inner part of the iso-curve equal to 0.954, could be really difficult due to the high go decreasing rate that makes extraction optimization difficult.
The same condition occurs when considering the working point at go absolute maximum, but this is less problematic because, as DN decreases, the reduction in go is lower than in the previous case. Once again, for DN values lower than 12 (see Fig. 9), go becomes independent of RING because iso-curves are roughly parallel to the RING axis, but it depends uniquely on DN for this specific decanter. Fig. 10 shows iso-curves obtained from Uwh experimental data; Uwh maximum (0.596) is achieved with DN = 13.4 and RING = 285.6 whereas Uwh minimum (0.437) is achieved with DN = 13.2 and RING = 283.5. The minimum is not well statistically identified as it can span over DN values from 12 to 15. Uwh is correlated with the machine’s capability to separate two liquid phases and especially with efficient separation of oil from discharged water into husk: higher Uwh values allow the machine to accomplish a higher separation grade between oil and vegetable water. Fig. 11 shows iso-curves obtained from the experimental data and Eq. (9) for S; S maximum (0.801) is achieved with DN = 13.2 and RING = 284.4; as expected this is achieved at the lowest DN that guarantees a higher residence time for the blending inside the decanter. In this case the maximum is heavily dependent on RING.
568
G. Altieri et al. / Journal of Food Engineering 119 (2013) 561–572
Fig. 9. Iso-curves obtained from experimental data and Eq. (8) for go by surface cubic spline interpolation; greyed area represents not significantly different neighbouring values; black dots are the experimental points.
Fig. 10. Iso-curves obtained from experimental data of Uwh by surface cubic spline interpolation; the greyed area represents not significantly different neighbouring values; black dots are the experimental points.
Further, S minimum is achieved at the highest DN where the residence time is too short to allow efficient liquid–solid separa-
tion. Moreover, for DN values higher than about 18 (see Fig. 11), S becomes only slightly dependent on RING, because the iso-curves
569
G. Altieri et al. / Journal of Food Engineering 119 (2013) 561–572
Fig. 11. Iso-curves obtained from experimental data and Eq. (9) for S by surface cubic spline interpolation; the greyed area represents not significantly different neighbouring values; black dots are the experimental points.
are almost parallel to the RING axis, but it depends uniquely on DN. Further; at such high DN values, the S decrease rate becomes very high. Separation coefficient S is correlated to the ability of the centrifuge to separate the solid phase from liquids. A decanter centrifuge that separates the solid phase from liquids as far as possible is preferable. A higher separation coefficient guarantees a higher oil recovery efficiency even at lower paste dilution: this is preferable to maintain extracted oil quality. Furthermore, the similarity between Fig. 9 and Fig. 11 shows that go and S are strictly related to decanter centrifuge geometric characteristics because maximum and minimum values almost coincide. It has been pointed out a positive correlation (i.e. R2 = 0.74) between these two parameters. Nevertheless, the relation between these parameters is more complex than the simply linear one. Moreover, once the decanter working point has been selected with reference to go maximum, then Uwh is an outcome. This means that Uwh cannot be selected whereas the S value is very close to its maximum. Hence, optimization of go allows at the same time optimization of S which, as previously outlined, guarantees extracted oil quality if operating at low paste dilution. Therefore, the main objective is to operate at go maximum as quality of extracted oil is achieved by the paste dilution ratio set at 0.5. 3.4. Ratio parameters of the decanter centrifuge (a, go, S, e, c, k, x) and their relations with DN and RING On the basis of our experimental results there were shown to be relationships between ratio parameters of the decanter centrifuge. They depend uniquely on decanter geometry and input variables Uopdb, DN and RING, through a complete quadratic interaction model. The model is based on a ‘‘black box’’ approach with
independent variables combined in a multilinear model of full quadratic surface response analysis. The ‘‘x2fx’’ Matlab function has been used to perform the variables handling. More specifically the model consists of a linear function of 21 terms formed by: a constant term; linear terms Uopdb, DN, DN0.5, DN1, RING; interaction terms constituted by all pairwise products of linear terms and squared terms Uopdb2, DN2, DN1, DN2, RING2. Each model variable is a column vector and when the complete quadratic interaction model is assembled, then a matrix M is built with the model variables as columns. The solution is found by multiple linear regression linking each ratio parameter to matrix M through a vector of constant terms found as system least squares solution. This constant vector represents the signature of the ratio parameter in question against input variables: Uopdb, DN and RING. Results of this processing method are reported in Table 1 which shows that from experimental data, given the high correlation level, the solution found can be used to draw go iso-curves as a function of DN and RING parametrically against Uopdb. Consequently, if the Uopdb value of olive paste is known from chemical or spectrophotometric analytical methods, this
Table 1 Evaluation of multiple linear regression of ratio parameters go, S, e, c, x and k as functions of input variables Uopdb, DN and RING. Parameter
R2
Average relative error%
Minimum relative error%
Maximum relative error%
go
0.9964 0.9918 0.9898 0.9896 0.9997 0.9898
0.09 0.81 1.22 1.39 0.09 1.09
0.03 0.25 0.45 0.40 0.03 0.33
0.19 1.62 2.99 2.81 0.19 2.24
S
e c x k
570
G. Altieri et al. / Journal of Food Engineering 119 (2013) 561–572
dependence can be used to fine tune decanter DN and RING, such that maximum oil recovery efficiency is achieved. In order to find the relations that allow identification of DN and RING optimal values which guarantee go maximum, Uopdb value was stepped from 22% to 22% with respect to the Uopdb average
value equal to 0.525, and then DN and RING optimal values were taken at go maximum. Samples of these steps are shown in Fig. 12. Fig. 12 shows iso-curves of go as a function of DN and RING when Uopdb is fixed at 22% (Fig. 12a) and +22% (Fig. 12b) of the Uopdb average value. The go absolute maximum and secondary
Fig. 12. Iso-curves obtained from multiple linear regression over experimental data for go vs. DN and RING. Uopdb is fixed at 22% (case a) and +22% (case b) of average Uopdb value equal to 0.525.
G. Altieri et al. / Journal of Food Engineering 119 (2013) 561–572
maximum move, increasing their values, from low DN (11.9) to high DN (13.3) as Uopdb increases and olive paste moisture content decreases. Fig. 12a shows that go absolute maximum is dependent on both RING and DN when considering pastes with high moisture content and therefore with a low Uopdb. Dependence on RING decreases as the iso-curves become almost parallel to the RING axis, more specifically at DN from 12.5 to 14 and at DN above 18. Fig. 12b shows that go absolute maximum is dependent on DN but very slightly on RING when considering olive pastes with low moisture content and therefore having a high Uopdb. Dependence on RING diminishes as the iso-curves become roughly parallel to the RING axis, specifically at DN from 12 to 15.5. Fig. 13 shows the curves that allow identification of DN and RING optimal values which guarantee go maximum. Highly correlated polynomials are found that link the DN optimal value (thirdorder) and RING optimal value (second-order) to Uopdb values.
571
Therefore, if the Uopdb of olive paste is known by an on-line method, then the polynomials found allow to calculate and modify, in real time, decanter DN and RING optimal values in order to achieve maximum oil recovery efficiency. Fig. 14 shows the relation between go absolute maximum vs. Uopdb. The curve was obtained from multiple linear regression of go with regard to experimental data. The Uopdb value was stepped from 22% to 22% with respect to the Uopdb average. The go absolute maximum was then taken. Once again, a highly correlated fourth-order polynomial is found that links go optimal value to Uopdb values. The go optimal value increases when Uopdb increases up to the go maximum value, after which go decreases. Therefore, this demonstrates that the go optimal value is determined by the olive paste moisture content, and, further, that it is a maximum value of Uopdb (0.61) which guarantees the maximum optimal go (0.982).
Fig. 13. Curves obtained from multiple linear regression of go over experimental data. The fixed Uopdb value was stepped from 22% to 22% with respect to the Uopdb average value equal to 0.525. DN and RING optimal values were then taken at maximum go. Highly correlated polynomials exist that link optimal value of DN (third-order) and RING (second-order) to Uopdb values.
Fig. 14. Curve obtained from multiple linear regression of go over experimental data. The fixed Uopdb value was stepped from 22% to 22% with respect to the Uopdb average value equal to 0.525. The maximum go was then determined. There is a highly correlated fourth-order polynomial that links the optimal value of go to Uopdb values. Moreover, the optimal go value increases when Uopdb increases up to a maximum of 0.9820 at a Uopdb of 0.61, after which go decreases.
572
G. Altieri et al. / Journal of Food Engineering 119 (2013) 561–572
4. Conclusions Olive oil centrifugal extraction strictly depends on both olive paste temperature and water content. Therefore, if olive paste is fed to a decanter without an automatic control, then non-optimal extraction is performed. In order to set up an automatic system for the control of the decanter during olive oil extraction, a feedback control system was tested in both laboratory and industrial processing plant after successfully testing a suitable flow mass sensor. A set of trials were carried out, using constant quality olives (cv. Coratina), and an on-line system was built in order to control the olive paste mass flow rate fed to a decanter centrifuge. This control system allowed assessment of the incidence of several parameters on extraction yield, husk fat content and vegetable water fat content. Our findings confirmed that olive paste mass flow rate was dependent on pump characteristics, olive paste temperature and viscosity. The feedback control system was able to guarantee a constant feed to the decanter centrifuge even in the presence of olive paste physical changes. Both ‘‘oil recovery efficiency’’ (go) and the ‘‘separation coefficient’’ S peaked with the highest residence time, using DN = 13.3 and RING = 284.8, while go minimum was achieved at higher DN. Moreover, for DN values higher than about 18 or less than 12, the go value becomes independent of RING. The go and S values are strictly related to decanter centrifuge geometry, such that operating at go maximum is the main objective because paste dilution ratio guarantees extracted oil quality. Interesting relations were found between decanter centrifuge ratio parameters depending on a complete quadratic interaction model from the input variables leading to highly correlated polynomial relations between DN vs. Uopdb and go optimal vs. Uopdb. These relations could be used to optimize DN and RING in order to achieve maximum oil recovery efficiency. Indeed, if the olive paste Uopdb value is known by an on-line method, then the polynomials found enable decanter DN and RING optimal values to be calculated and modified, in real time, so as to achieve maximum oil recovery efficiency. Acknowledgements The authors thank both the ing. Lorusso and dott.ssa Barracane of the firm Barracane s.r.l. for their valuable help and to have made possible the carrying out of these trials. References Altieri, G., 2010. Comparative trials and an empirical model to assess throughput indices in olive oil extraction by decanter centrifuge. Journal of Food Engineering 97, 46–56.
Amirante, P., Catalano, P., 1993. Analisi teorica e sperimentale dell’estrazione dell’olio d’oliva per centrifugazione [theoretical and experimental analysis of olive oil extraction by centrifugation]. La Rivista Italiana Delle Sostanze Grasse LXX, 329–335 (in Italian). Amirante, P., Di Renzo, G.C., Colelli, G., 1995. Estrazione centrifuga con controllo continuo dei parametri operativi [extraction by centrifuge with continuous control of the operating parameters]. Rivista di Ingegneria Agraria, Quaderno n 17, 344–351 (in Italian). Amirante, R., Baccioni, L., Catalano, P., Montel, G.L., 1993. Nuove tecnologie per l’estrazione dell’olio di oliva: decanter con cono a pressione dinamica variabile e controllo della velocità differenziale tamburo-coclea [new technologies for the extraction of olive oil: decanters with a dynamic variable pressure cone and control of the differential speed conveyor-bowl]. La Rivista Italiana Delle Sostanze Grasse LXXVI, 129–140 (in Italian). Amirante, R., Catalano, P., 2000. Fluid dynamic analysis of the solid–liquid separation process by centrifugation. Journal of Agricultural Engineering Research 77 (2), 193–201. Anlauf, H., 2007. Recent developments in centrifuge technology. Separation and Purification Technology 58 (2), 242–246. Boncinelli, P., Daou, M., Cini, E., Catalano, P., 2009. A simplified model for designing and regulating centrifugal decanters for olive oil production. Transactions of the ASABE, American Society of Agricultural and Biological Engineers 52 (6), 1961– 1968. Briggs, J.L., Steffe, J.F., 1997. Using brookfield data and the mitschka method to evaluate power law foods. Journal of Texture Studies 28 (5), 517–522. Caponio, F., Gomes, T., Summo, C., Pasqualone, A., 2003. Influence of the type of olive-crusher used on the quality of extra virgin olive oils. European Journal of Lipid Science Technology 105, 201–206. Catalano, P., Pipitone, F., Calafatello, A., Leone, A., 2003. Productive efficiency of decanters with short and variable dynamic pressure cones. Biosystems Engineering 86 (4), 459–464. Corner-Walker, N., 2000. The dry solids decanter centrifuge: capacity scaling. Filtration and Separation 37/4 (May), 28–32. Corner-Walker, N., Records, F.A., 2000. The dry solids decanter centrifuge: conveyor torque and differential. Filtration and Separation 18 (October), 23. Daou, M., Furferi, R., Recchia, L., Cini, E., 2007. A modelling approach to extra virgin olive oil extraction. Journal of Agricultural Engineering 38 (4), 11–20. http:// dx.doi.org/10.4081/jae.2007.4.11. Di Renzo, G.C., Colelli, G., 1997. Flow behavior of olive paste. Applied Engineering in Agriculture 13 (6), 751–755. Hommel, G., 1988. A stagewise rejective multiple test procedure on a modified Bonferroni test. Biometrika 75, 383–386. Hommel, G., 1989. A comparison of two modified Bonferoni procedures. Biometrika 76, 624–625. Leung, W.W.-F., 1998. Torque requirement for high-solids centrifugal sludge dewatering. Filtration and Separation 883 (November), 887. Mitschka, P., 1982. Simple conversion of Brookfield R.V.T. readings into viscosity functions. Rheologica Acta 12, 207–209. Monteleone, E., Caporale, G., Carlucci, A., Paglierini, E., 1998. Optimisation of extra virgin olive oil quality. Journal of the Science of Food and Agriculture 77, 31–37. Piacqadio, P., De Stefano, G., Sciancalepore, V., 1998. Quality of virgin olive oil extracted with new centrifugation system using a two phase decanter. Fett/ Lipid 10, 472–474. Ranalli, A., Cabras, P., Iannucci, E., Contento, S., 2001. Lipochromes, vitamins, aromas and other components of virgin olive oil are affected by processing technology. Food Chemistry 73, 445–451. Ranalli, A., De Mattia, G., Ferrante, M.L., 1997. Comparative evaluation of the olive oil given by a new processing system. International Journal of Food Science and Technology 32, 289–297. Roig, A., Cayuela, M.L., Sanchez-Monedero, M.A., 2006. An overview on olive mill wastes and their valorisation methods. Waste Management 26, 960–969. Salvador, M.D., Aranda, F., Gomez-Alonso, S., Fregapane, G., 2003. Influence of extraction system, production year and area on Cornicabra virgin olive oil: a study of five crop season. Food Chemistry 80, 359–366. Steffe, J.F., 1996. Rheological Methods in Food Process Engineering, Second ed. Freeman Press, East Lansing, MI.