Journal of Food Engineering 54 (2002) 63–73 www.elsevier.com/locate/jfoodeng Filled snack production by co-extrusion-cooking: 2. Effect of processing on cereal mixtures Bruno de Cindio a,*, Domenico Gabriele a, Claudio Maria Pollini b, Donatella Peressini c, Alessandro Sensidoni c a Department of Chemical Engineering and Materials, Laboratory of Rheology, University of Calabria, I-87030 Arcavacata di Rende (CS), Italy b Pavan S.p.A., via Monte Grappa 8, I-35015 Galliera Veneta (PD), Italy c Department of Food Science, University of Udine, via Marangoni 97, I-33100 Udine, Italy Received 17 November 2000; accepted 19 September 2001 Abstract Filled snack production has been considered from several point of views. In this work, the effect of the operational variables on the cereal mixture constituting the external part of this product has been studied. A typical flour blend was characterised and used in a previous developed model to simulate the extrusion conditions and to show process damage. Chemical and rheological methods are proposed as experimental techniques capable of measuring the damage induced by the process. The link between chemical contents and mechanical damage was proposed on the basis of the thermal history and of the mechanical power input. The simulation results obtained by a new proposed extrusion model applied to a real twin co-extruder, made it possible to predict damage such as the lysine decrease, and to show the effect of different screw speeds. It appeared that the computed results are in an acceptable agreement with industrial evidence. Ó 2002 Elsevier Science Ltd. All rights reserved. Keywords: Extrusion modelling; Rheology; Snack characterisation; Cooking–extrusion process 1. Introduction In the first work of this series (de Cindio, Gabriele, Pollini, Peressini, & Sensidoni, 2001), filled snack production has been considered from a theoretical process design point of view. A model based on the rheological properties of both dough and filler was developed to design an extruder capable of making a filled product with desired characteristics. It is worth noting that extrusion–cooking technology is widely used for manufacturing cereal foods such as pastas, snacks, breakfast cereals, baby foods and pet foods. Despite the increasing use of extrusion–cooking technology, there are still few published papers on the physical and chemical reactions taking place during the process and the rheological and nutritional implications associated with these changes. Molecular rearrangements and heat-induced changes * Corresponding author. Tel.: +39-098-4492080/2035; fax: +39-0984492058. E-mail addresses: rheo.lab@unical.it (B. de Cindio), pollini.c@ pavan.com (C. Maria Pollini), donatella.peressini@dsa.uniud.it (D. Peressini). among the different components of the food system make cooking–extrusion a complex technology. Consequently, snack production suffers from a certain degree of empiricism due to a limited understanding of the phenomena occurring. In the above cited paper, attention was focussed on the effect of the relatively complex rheological behaviour of the two components on the extruder fluid-dynamics and on its rheological modelling. In this second paper the simulation model was used to predict the damage that may be induced in the final product as a consequence of the imposed process conditions. In fact, the chemical and physical characteristics of the products strongly depend upon process variables such as extrusion temperature, screw speed and moisture content (Falcone & Phillips, 1988; Mercier & Feillet, 1975; Mercier, Charbonniere, Grebaut, & de la Gueriviere, 1980). At high temperature, starch granules undergo gelatinisation and melting causing an increase in dough viscosity. Mechanical degradation of starch, which enhanced the susceptibility to amylase hydrolysis, was also reported (Davidson, Paton, Diosady, & Larocque, 1984; Mercier, 1977). Heat and shear induce denaturation of 0260-8774/02/$ - see front matter Ó 2002 Elsevier Science Ltd. All rights reserved. PII: S 0 2 6 0 - 8 7 7 4 ( 0 1 ) 0 0 1 8 6 - 8 64 B. de Cindio et al. / Journal of Food Engineering 54 (2002) 63–73 Nomenclature aw c c0 E G G0 G00 H H0 k k100 L m n R water activity concentration of available lysine after the process (mg/g) concentration of available lysine before processing (mg/g) activation energy (kcal/mol) coded level of glucose storage modulus (Pa) loss modulus (Pa) extruder channel depth co-ordinate (m) channel height (m) frequency factor (min1 ) reference reaction rate (min1 ) extruded samples length (mm) consistency index (Pa sn ) flow index gas constant (kcal/mol °K) proteins, which unravel and are subjected to crosslinking reactions. The result is the formation of a new molecular aggregate structure. Chemical changes may also be derived from the hydrolysis of starch and sucrose increasing the reducing sugar content (Bj€ orck & Asp, 1983; Noguchi, Mosso, Aymard, Jeunink, & Cheftel, 1982). During heat processing, such a reducing sugar may react with free amino groups following the so-called ‘‘Maillard reaction’’ (MR). This is a complex reaction that is often desired for the development of golden brown colour and caramel aroma in bakery products and cooked foods, but it is also generally responsible for a reduction of the nutritional value (Mauron, 1981). The main purpose of this work was to investigate the changes and damages eventually induced into the product as a consequence of the process conditions. In fact, processing implies a change in the physical and chemical properties of the two components. Because the cereal dough, that constitutes the external part of the new co-extruded filled snack, seems to be more sensitive to processing than the filler, in the following only the process effects on the dough has been considered. However, the consequences of using different fillers will be been considered in another paper of this series (Peressini et al., 2001). The considered effects have been divided into nutritional damages and thermo-mechanical property changes. In order to fulfil the objective of this work, the first question to be answered was how to assess suitable experimental methods capable of making a quantitative measurement of the considered changes. Concerning the nutritional damage, some chemical components were considered which are present already in the dough, i.e. T t vb vx vz Greeks x c_ / U d g gs absolute temperature (°K) time (s) barrel velocity (m/s) component of the fluid velocity along x axis (m/s) component of the fluid velocity along z axis (m/s) frequency (Hz) shear rate (s1 ) extruded samples diameter (mm) rheometer plate diameter (mm) phase angle complex viscosity ðPa sÞ shear viscosity ðPa sÞ lysine, or are obtained by a reaction, i.e. furosine. Thus some typical chemical methods are discussed and adapted to this specific case. To quantify the thermomechanical changes of a cereal mixture, it is possible to use a rheological test. Therefore a time cure, i.e. a temperature scanning of the dynamic moduli measured during a small amplitude oscillation test at a fixed frequency, was proposed because it seemed to be very helpful for this purpose. Therefore in the following, these tests are discussed and the results obtained by means of the proposed techniques are presented for a typical snack. The second question to be answered was how to predict and control the process. The rheological characterisation was the basis of the simulation model developed elsewhere, but some modifications were necessary to obtain predictions of the snack properties from a typical industrial twin-screw extruder. 2. Materials and methods 2.1. Formulation and sample preparation Two flour blends, kindly supplied by Pavan (Galliera Veneta, Italy) and thereafter indicated as blend 1 and blend 2, were tested; their composition is reported in Table 1. Dry blends were prepared by means of a traditional flour ribbon mixer, and the minor components were previously added and mixed with a portion of floury raw materials. The mixing time was about 15 min, then the material was fed to the industrial extruder that is described below. The moisture content of the blends was adjusted to 20.5% by injecting the water supplied 65 B. de Cindio et al. / Journal of Food Engineering 54 (2002) 63–73 Table 1 Flour composition (% dry basis) Ingredients Blend 1 Blend 2 Oat flour Whole wheat flour Maize starch Bran Salt Sugar Dry maltose 52.0 26.0 15.0 3.4 1.7 1.6 0.3 – 56.4 32.5 7.4 3.7 – – directly by a micrometric pump into the first zone of the extruder (see Fig. 1). An intermeshing co-rotating twin-screw extruder (Model TT58W, Pavan, Italy) with double flight was used as the extrusion-cooker. This equipment was built as a modular extruder of five zones obtained by assembling together 25 different components. The choice of the characteristics and the number of components of each zone is essentially based on industrial experience, however, the main geometrical characteristics are reported in Table 2. Screw speed could be varied between 0 and 500 rpm by directly operating on a c.c. motor. Barrel temperature was controlled in the five zones by means of electrical heating resistors and a water cooling circuit. The 25 components of the screw had different pitches and lengths, and as a consequence they had different helix angles. In addition, four of them were dynamic mixers capable of kneading the mixture. To increase pressure, some other elements were mounted with a reverse angle to obtain a flight acting as a barrier Table 2 Geometrical characteristics of the twin-screw extrusion cooker TT58W (Pavan, Galliera Veneta, Italy) Screw root diameters (m) Barrel diameters (m) Screw inter-axis distance (m) Channel height (m) Number of flights Number of sections Number of elements 0.0370 0.0576 0.0480 0.0103 2 5 25 to the flow. At the end of the extruder there is a head with a die capable of making products of the desired shape. A pressure of 100 atm and a temperature of 155 °C were measured during production at the head inlet. The product coming out from the extruder die was cylindrically shaped and was cut into short pieces, that were cooled at room temperature (21 °C) on a belt conveyor with the aid of a forced ventilation filtered-air tunnel and then roasted. In such a way, small cylindrical samples (/ ¼ 13 mm and L ¼ 20 mm) were obtained and tested. 2.2. Chemical characterisation Blend 1 samples were subjected to a complete chemical characterisation with the following methods and techniques. Moisture, protein and ash were determined according to official methods (AOAC, 1980), while pH was evaluated by AACC (1995) Method 02-52. Fig. 1. Sketch of the Extruder–Cooking equipment, with the indicated temperature set points. 66 B. de Cindio et al. / Journal of Food Engineering 54 (2002) 63–73 Total reducing sugars (g glucose/100 g d.b.) of blend 1 were evaluated as previously described (Sensidoni, Peressini, Pollini, & Munari, 1996). Water activity (aw ) of blend 1 dough with a total moisture content of 30% (w.b.) was measured with an AquaLab CX-2 hygrometer (Decagon Devices, Pullman, Washington, USA) at 25 °C. Fat content was measured using a Soxhlet extraction, after an acid hydrolysis of the sample. A 5 g sample, previously ground by means of a laboratory mortar, was added to 50 ml HCl (16% w/v), and the obtained solution was maintained for 40 min at 70 °C in a thermostatic bath. Then the hydrolysed matter was filtered by means of previously wetted paper filter. After drying the filter in an oven at 105 °C for 30 min, fat was extracted (6 h) from the filter using the Soxhlet method, and thus determined. The degree of gelatinisation was determined according to Chiang and Johnson (1977), the amino acid composition was determined by gas-chromatographic analysis, and the furosine content (mg/100 g of protein) was determined according to Sensidoni et al. (1996). When foods contain reducing sugars and free e-amino groups of lysine, the early stage of MR leads to the formation of stable Amadori compounds, which may be partially converted into furosine (e-N-2-furoylmethyl-L -lysine) by subjecting them to an acid hydrolysis. It was possible to use the acidic hydrolysed matter to determine the amino acid profile. The hydrolysed solution (1 ml) and 90 ml of anhydrous ethyl alcohol were mixed in a rotavapor flask and dried at 40 °C. Distilled water (4 ml) and 50 ll of D L -2-aminobutyric acid solution (1.18 mg/ml, Riedel-de Ha€en, Germany) were added to the dry matter as internal standard. Then the sample was cleaned up and amino acid determination was made as suggested by Pirini, Conte, Francioso, and Lercker (1992). Capillary gas chromatography was performed by using a Varian 3300 chromatograph equipped with a FID detector, an integrator (Carlo Erba Mega Series, Italy) and a 25 m  0:32 mm i.d. fused silica capillary column, coated with 0:2 lm thick OV 1701 film (Chrompack, Holland). Helium was used as carrier gas at a flow rate of 1.6 ml/min; 1 ll of sample was injected with a splitting system 1:80. Injector and detector temperatures were set at 280 °C. The column was heated from 80 to 280 °C, at a 8 °C/ min rate. Pure amino acids were used as standards for the quantitative determination (Amino acid kit LAA-21, Sigma, USA). The coefficient of variation for the internal standard was calculated to be 12%. All the experimental determinations were at least triplicated. Thus the quantity of the amino acid (mg/gN) was found. For lysine content, because the sample preparation implies the formation of a certain amount of lysine from the Amadori compound hydrolisis, it was necessary to correct the data by considering the amount of blocked lysine (Mauron, 1981). 2.3. Rheological measurements The rheological measurements were performed by means of an RFS-Rheometrics Fluid Analyser (Rheometrics, USA) using the oscillatory mode. Dough samples at 30% of moisture content were prepared and tested using the ‘‘serrated plates’’ ðU ¼ 2 cmÞ geometry, this was necessary to avoid slippage at the test cell solid surfaces. The apparatus was equipped with an auto calibrating system to adjust the gap to the desired set value of 1.5 mm. In such a way it was possible to compensate for the errors due to temperature changes. To avoid moisture loss during the test, the edge of the sample was covered with a thin layer of silicon oil, and the test cell was enclosed inside a home-made saturatedair chamber. The time cure test was performed at a 1 °C/min heating rate in the temperature range between 25 and 140 °C, at a fixed frequency with an amplitude small enough to ensure linear viscoelasticity. Before starting the experiment, the sample was conditioned in order to have the same temperature of the test cell. The storage and loss moduli (G0 and G00 ) and the loss tangent ðtan d ¼ G00 = G0 Þ were measured during the oscillatory experiments. 3. Results and discussion 3.1. Chemical properties The chemical and physical properties of the snacks obtained from blend 1, measured according to the previously described procedure, are shown in Table 3. It appears that the final product contains a very small amount of water and that the degree of gelatinisation is very high. As said earlier, one of the main phenomenon induced by heating cereal dough is the MR, that produces the desirable cooked flavour and coloured components in this kind of food (Bredie, Mottram, & Guy, 1998). Some neocompounds of MR have been used as molecular markers for evaluating raw materials, processing and storage conditions (Resmini & Pellegrino, 1994; Resmini, Pellegrino, Pagani, & Noni, 1993; Sensidoni & Table 3 Properties of blend 1 snacks Moisture (g/100 g) Ash (g/100 g d.b.) Protein (g/100 g d.b.) Fat (g/100 g d.b.) Degree of gelatinisation (%) 3.36 2.22 8.68 5.80 98.2 B. de Cindio et al. / Journal of Food Engineering 54 (2002) 63–73 Table 4 Furosine content (mg/100 g protein) of various cereal products Food product Furosine (mg/100 g protein) Reference Blend 1 snack Snacks 116 375–504 Baby cereals 293–3180 Dried pasta 61–582 Baked products 227–667 This work Resmini, Pellegrino, and Battelli (1990b) Guerra-Hernandez and Corzo (1996) Resmini, Pagani, and Pellegrino (1990a), Sensidoni, Peressini, and Pollini (1994) and Sensidoni et al. (1996) Resmini et al. (1990a) Peressini, 1998; Sensidoni et al., 1996). In this work, furosine was chosen as the molecular marker because its content in different processed cereal foods appears to be a good index of the process conditions. As shown in Table 4, a wide range of furosine values was found in both dried pasta and baby cereals due to different processing conditions and reducing sugar availability (Guerra-Hernandez & Corzo, 1996; Sensidoni et al., 1996). It appears that more the process is severe more the furosine content is greater. The low furosine content of 116 mg/100 g protein (standard deviation 2) measured for blend 1 snacks, indicates that in this case a rather mild operating condition has been realised. In Fig. 2, the influence of the process on the amino acid content is shown. With the exception of lysine, extrusion–cooking and roasting did not change significantly the amino acid quantity of the ingredients. The loss of available lysine was 31% due to both mechanical and heat damage, therefore the lysine loss can be suggested as an index of the technological damage made to the raw material during processing. In the final product, the total lysine included 7% of unavailable lysine, which was blocked as an Amadori compound in the MR, therefore the lysine content de- Fig. 2. Comparison between cereal blend 1 and corresponding snack amino acid composition. 67 crease is linked to the process and is influenced by many factors such as temperature, residence time, screw speed and moisture content. It is well accepted that high temperature strongly favours lysine loss (Noguchi et al., 1982). Concerning the effect of increasing screw speed, a combined effect was noticed: high shear rates induced hydrolysis of saccarose (Noguchi et al., 1982) and starch (Bj€ orck & Asp, 1983), giving reducing sugars that can participate to the MR that consumes lysine. On the other hand, some other authors (Bj€ orck, Asp, & Dahlqvist, 1984) observed that lysine retention increased when increasing screw speed, as a consequence of the reduction of the residence time in the extruder. For water content, Noguchi et al. (1982) suggested that water shows a protective effect because it is a product of MR, therefore an increase in water content from 13% to 18% implies a positive effect on the lysine retention (Noguchi et al., 1982; Bj€ orck et al., 1985). With the imposed process conditions, Noguchi et al. (1982) reported losses of available lysine of up to 40% in an extruded cereal product. Thus, it is reasonable to assume that an available lysine reduction of only 31% as measured in our case, represents a good result for a product obtained by extrusion–cooking and roasting. 3.2. Rheological properties The chemical characterisation discussed above is capable of quantifying more or less roughly the process severity, but nothing is said about changes to macrostructure that is crucial in determining snack texture. To perform this, it was suggested that the development of G0 and G00 be followed during a time cure test. This technique can substitute for classical DSC calorimetric technique, with the advantage of giving directly the value of the mechanical properties. The dynamic viscoelastic parameters (G0 , G00 and d) measured during heating blend 1 dough at 1 Hz, are reported in Fig. 3. The storage and loss moduli increased to a maximum as Fig. 3. Storage ( ), loss ( ) moduli and phase degree () changes during a time cure test for the cereal blend 1 at 30% moisture content (1 °C/min; 1 Hz). 68 B. de Cindio et al. / Journal of Food Engineering 54 (2002) 63–73 Fig. 4. Storage ( ), loss ( ) moduli and phase degree () changes during a time cure test for the cereal blend 1 at 30% moisture content (10 °C/min; 10 Hz). the temperature was raised from 25 to 75 °C, then a continuous decrease was observed till a minimum value at 110 °C was attained. By further increasing temperature, G0 and G00 again increased. When looking at the phase angle, it appeared that the sample becomes more solid-like by increasing temperature up to about 90 °C. By further increasing the temperature a sharp increase in d was observed as denoted by the more liquid like behaviour, and a peak was attained at 110 °C, thereafter d decreases becoming lower and lower. A test done at 10 °C/min and 10 Hz up to 100 °C did not show large differences compared to the previous data (Fig. 4). The interpretation of the rheological results requires a previous structural description of the system under investigation. The cereal dispersion with 30% moisture content can be sketched as a dispersed phase of starch granules entrapped within a protein network and surrounded by the aqueous phase (see Fig. 5). Generally speaking heating induces starch gelatinisation, which is a hydrothermal phenomenon that implies both swelling and melting of starch granules and consequently a loss of the ordered structure (Roos, 1995). The presence of a three-dimensional protein network entrapping starch granules inside the dough, both decreases swelling because it hinders water diffusion and provides a mechanical resistance, but also it has the effect of reducing granule–granule interactions (Champenois, Rao, & Walker, 1998; Derby, Miller, Miller, & Trimbo, 1975). Therefore starch gelatinisation should be less favoured, and a slower reaction velocity and higher transition temperature are then expected. It is worth noting also that sucrose and sodium chloride content tend to increase the gelatinisation temperature (Ghiasi, Hoseney, & Varriano-Marston, 1983; Peressini, Sensidoni, Pollini, & de Cindio, 1999). During heating, when considering the viscoelastic properties of a concentrated starch dispersion, it is well accepted that they depend on the volume occupied by the granules and by their deformability (Rolee & Le Meste, 1997, 1999). It was found that higher values of starch volume fraction are responsible for an increase in the mechanical properties during gelatinisation (Rolee & Le Meste, 1997). Therefore the increase in the storage modulus observed up to 75 °C (Fig. 3) might be analogously attributed to the swelling of starch granules, Fig. 5. Sketch of starch granules embedded in a protein network. B. de Cindio et al. / Journal of Food Engineering 54 (2002) 63–73 which implies an increase in granule–granule interactions. Doublier, Paton, and Llamas (1987) showed that wheat, maize and oat starches show an increasing order of swelling power. Dreese, Faubion, and Hoseney (1988) and Kokini, Cocero, and Madeka (1995) observed that proteins such as glutenin and gliadin gave slight changes in the viscoelastic properties up to 90 °C and therefore starch gelatinisation played the main role in determining the thermo-mechanical behaviour of the sample before the second transition. Heating above 75 °C induced a large decrease in G0 and G00 (Fig. 3), which should be due to starch deformability increase (Rolee & Le Meste, 1997, 1999), that was explained by the dissociation of ordered areas of starch (Tester & Morrison, 1990) and by the melting of the more stable amylopectin crystals (Evans & Haisman, 1982). The trend of G0 and G00 to increase above 110 °C could have different explanations. It was reported that gliadin, glutenin and zein in this range of temperature are responsible for the onset of a network structure due to cross linking/aggregation reactions that increased the elastic component (Kokini et al., 1995). In addition, lipid-amylose complexes start to melt only near 110 °C (Doublier et al., 1987). Some investigations showed that the presence of lipids in starch granules caused a delay or a suppression of swelling up to 85 °C (Tester & Morrison, 1990; Wang & White, 1994). It is well known that oat starch has a larger amyloselipid complex and a greater swelling power than that of other cereals (Doublier et al., 1987; Sowa & White, 1992). It is possible that melting of this complex favours swelling of the residual starch granules and promotes the increase in G0 , G00 and therefore the solid-like characteristics. Therefore, the second transition should be mainly dependent on the presence of oat flour. To confirm this, another cereal blend (blend 2) without this component was tested. The composition of blend 2 is reported in Table 1 and the rheological behaviour of the two doughs prepared with the same procedure is compared in terms of storage modulus G0 in Fig. 6. As expected it appeared that the second transition is much smaller, thus confirming the suggested explanations. In the following, only blend 1 was used throughout the simulations, being a typical industrial product used in snack preparation. Blend 2 was used for comparison just to have a better understanding of the temperature behaviour of snack dough. 69 Fig. 6. Comparison between the storage moduli obtained by blend 1 (j) and blend 2 () at 30% moisture content: time cure test 1 °C/min and 1 Hz. the first paper of this series (de Cindio et al., 2001), it was previously necessary to obtain a flow curve and a damage kinetic equation. The first one was computed by means of the rheological characterisation presented above, that was used also to produce an activation temperature dependent factor, capable of shifting the flow curve with temperature. This was necessary to obtain the temperature dependent rheological properties that were inserted into the extruder model. Shear viscosity gs was approximated by the complex viscosity g defined as: pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi G02 þ G002 : ð1Þ g ¼ x By assuming x ¼ c_ it was possible to shift the power law constitutive equation to any temperature according to the time cure test data. Because in the previous experiments the frequency was chosen 1 Hz, this means that the shift factor was evaluated at c_ ¼ 1 s1 that in the proposed power law equation (de Cindio et al., 2001) represents just the value of the m-constant:  n1  n1     gs ¼ mc_  ¼ g ð1 Hz; T Þc_  : ð2Þ The values of m were computed from the rheological data and are reported in Table 5; the exponent n was assumed to be constant with respect to both c_ and T, and a value of 0.1 was used throughout. According to Wolf, Thompson, and Reineccius (1978) a simple first-order kinetic equation was used for the lysine loss reaction: 3.3. Simulation results dc ¼ kc: dt The results obtained by the determinations done on the industrial equipment, may be compared to the data predicted by means of a simulative model. To perform process simulations according to the model developed in The frequency factor was assumed to follow an Arrhenius law:    E 1 1 k ¼ k100 exp   : ð4Þ R T 373:15 ð3Þ 70 B. de Cindio et al. / Journal of Food Engineering 54 (2002) 63–73 Table 5 Geometrical and operational characteristics of the extruded components Extruder component Section length (mm) 6 7 8 9 10a 11 12 13 14 15a 16 17 18 19 20a 21 22 23 24 25 25 74.7 49.9 49.75 74.6 25 37.3 50 50 74.87 25 50 50 25 50 50 50 50 50 99.93 a Helix angle (°) Pitch (mm) Tin (°C) Pin (atm) m (kPa s) 18.6 26.8 18.6 26.8 50 75 50 50 )18.6 26.8 18.6 18.6 50 75 50 50 )18.6 18.6 18.6 18.6 50 50 50 50 18.6 18.6 18.6 18.6 18.6 50 50 50 50 50 60 64.4 77.5 86.3 95 135 135 135 135 135 140 140 140 140 140 140 140 140 140 150 20 18.1 12.5 8.7 5 25 65 55.5 42.7 30 60 100 96 92 90 115 112.5 110 107.5 105 7.030 9.497 50.55 30.71 24.61 202.7 202.7 202.7 202.7 202.7 208.0 208.0 208.0 208.0 208.0 208.0 208.0 208.0 208.0 202.1 The indicated element is a kneader. Both k100 ðmin1 Þ and the activation energy E (kcal/ mol) depend on the glucose content, pH and water activity. The following fitting equations were used and are valid for a food model system according to the above quoted reference: E ¼ 32:9 þ 1:1G3  1:77aw G  1:32G2 þ 1:5pHG; ð5Þ k100 ¼ 0:0122 þ 0:00175G3 þ 0:00104Gaw ; ð6Þ where G and pH stands for the coded levels of glucose and pH, respectively, and aw is the water activity. The simulations were performed following a very simple analysis of an extruder by using the Tadmor model and assuming the power law constitutive equation (Eq. 2) (Gabriele, Curcio, & de Cindio, 2001). In addition, steady flow, no slip at wall, body and negligible inertia forces were assumed. Since in this work the industrial results were obtained from a double flight twin-screw extruder, this was approximated to a single screw extruder with the same length but a mass flow rate one fourth of the total one. In this way each flight of the real apparatus was considered as a single channel with one fourth of the total mass flow rate. As with the single equipment components, the first five, corresponding to the complete first zone and a part of the second one, were not considered because they are almost half empty and show a solid bed transport. It is worth noting that for the considered case, during simulations it is important to predict the lysine loss, that is mainly temperature dependent. Therefore it is reasonable to neglect units where no high temperature is found. As a consequence only the completely filled channels were considered starting from the sixth unit. The geometrical characteristics and the operational variables were fixed for every component and are reported in Table 5. Temperature and pressure data were taken according to industrial experience. With this approach the geometry was simplified by assuming the screw channel as a rectangular duct (flat plate model), with an almost infinite width (width to deep ratio higher than 10) and constant height. The screw was considered stationary and the barrel rotates with a velocity vb . The balance equations were solved step by step for the considered components, by means of a numerical method and it was possible to evaluate the velocity profile inside the channel as a function of the height, both along the channel direction ðvz Þ and across the channel direction ðvx Þ (Gabriele et al., 2001). The computations were done at three different values of the screw speeds for the blend 1 dough. The temperature was assumed constant inside the component and no radial gradient was considered. In Table 5 the inlet and outlet temperatures and pressure of every component are also reported. The cross-channel velocity computed profile of component number 11 is shown in Fig. 7: velocity values are negative in the lower part of the channel, while they become positive in the upper part; therefore as expected no net flow in the x direction is present. The down channel velocity profile of the same component is reported in Fig. 8. It showed well the effect of the pressure flow that is opposite to the drag flow causing a reduction of the net output. It is worth noting that at high temperature values drag flow may become lower than pressure flow, owing to the reduction in the 71 B. de Cindio et al. / Journal of Food Engineering 54 (2002) 63–73 Fig. 7. Cross channel distribution computed for extruder component number 11 (reverse helix angle): H0 ¼ 0:0103 m. Fig. 9. Specific mechanical power computed along the extruder (j 150 rpm, 300 rpm, 500 rpm). Fig. 8. Down channel distribution computed for extruder component number 11 (reverse helix angle): H0 ¼ 0:0103 m. Fig. 10. Fraction of the available lysine computed along the extruder (c0 initial lysine concentration, j 150 rpm, 300 rpm, 500 rpm). viscosity, and because the material tends to back circulate inside the extruder channel. The model was run with three different speeds and in Fig. 9 the total specific mechanical power, that is four times the computed one to account that industrial apparatus is a double flight twin-screw extruder, is reported. It is quite obvious that the greater is the screw speed, the greater is the required power supply. But it also appeared that it is almost constant along the extruder regardless the considered screw speed. By integrating over the extruder length, the total mechanical power was found at any considered screw speed, and therefore the corresponding mass flow rates were computed (see Table 6). The simulative model developed so far was capable also to give the lysine loss according to Eqs. (3)–(6), and the operational conditions of Table 5. All the properties have been computed at the indicated temperatures, the value of pH was fixed at 5.9, the water activity of blend 1 dough with a total moisture content of 30% was 0.967 and the total reducing sugar content, expressed as equivalent glucose, was assumed to be 0.5% (w/w). The results reported in Fig. 10 show that by increasing residence time (slow speeds) a rather large increase of the lysine loss is predicted. This is in agreement with the fact that lysine content was considered essentially temperature dependent, and therefore the longer the time inside the extruder, the greater is the loss. It may be seen that the loss increases more than linearly when increasing speed. In Table 6, both the remaining lysine and the residence times are reported for the investigated velocities. Table 6 Summary of the process simulation results Screw speed (rpm) Power (kW) Mass flow rate (kg/h) Lysine fraction (%) Residence time (s) 150 300 500 18.2 39.1 68.6 117 229 384 50.0 69.3 79.7 52.6 28.1 18.0 72 B. de Cindio et al. / Journal of Food Engineering 54 (2002) 63–73 These values are considered in a good agreement with the industrial results. 4. Conclusions It was shown above how it is possible to treat a rather difficult problem using a new approach based essentially on rheological modelling. It was found that two bioindicators were very helpful in defining process damage in terms of dough components, but it was also seen that the rheological measurements are capable of accounting for structural changes in a useful way that can be inserted into constitutive equations. In fact it was possible to follow these changes even if not directly, along the whole process. Concerning nutritional decay, it should be remarked that these products are more considered by consumers for their sensorial aspects (flavour, colour and texture) than for their nutritional properties. Since snack foods are not considered a major source of dietary protein in Europe, a decrease in available lysine in such foods might not be of nutritional importance and should not be considered as an index of a poor quality. Nevertheless, lysine sensitivity to the thermal treatment was useful in following the process. The changes induced directly by the mechanical power should be considered, because in the current model this latter influences only the flow rate and, as a consequence, the residence time, and therefore the heat treatment. On the other hand, it is well known that mechanical power tends to decrease the molecular weight of the proteins, implying that rheological properties becomes poorer. Finally the results obtained so far represent a necessary preliminary step in developing a deeper understanding of the entire phenomenon of extrusion–cooking. The proposed model and the verified analytical methods represent a significant step in approaching in a less empirical way both the designing and the conduction of snack production processes. Acknowledgements This work was supported by an MURST – IMI grant for the applied research. References AACC-American Association of Cereal Chemists. (1995). Approved Methods of the AACC (9th ed.). St. Paul, MN: The Association. AOAC. (1980). Official Methods of Analysis (13th ed.). Washington, DC: Association of Official Analytical Chemists. Bredie, W. L. P., Mottram, D. S., & Guy, R. C. E. (1998). Aroma volatiles generated during extrusion cooking of maize flour. Journal of Agricultural and Food Chemistry, 46(4), 1479–1487. Bj€ orck, I., & Asp, N. G. (1983). The effects of extrusion cooking on nutritional value. A literature review. Journal of Food Engineering, 2(4), 281–308. Bj€ orck, I., Asp, N. G., & Dahlqvist, A. (1984). Protein nutritional value of extrusion-cooked wheat flours. Food Chemistry, 15(3), 203–214. Bj€ orck, I., Matoba, T., & Nair, B. M. (1985). In vitro enzymatic determination of protein nutritional value and the amount of available lysine in extruded cereal-based products. Agricultural and Biological Chemistry, 49(4), 945–951. Champenois, Y., Rao, M. A., & Walker, L. P. (1998). Influence of a-amylase on the viscoelastic properties of starch-gluten pastes and gels. Journal of the Science of Food and Agricultural, 78(1), 127–133. Chiang, B. Y., & Johnson, J. A. (1977). Measurement of total and gelatinized starch by glucoamylase and o-toluidine reagent. Cereal Chemistry, 54(3), 429–435. Davidson, V. J., Paton, D., Diosady, L. L., & Larocque, G. (1984). Degradation of wheat starch in a single screw extruder: characteristics of extruded starch polymers. Journal of Food Science, 49(2), 453–458. de Cindio, B., Gabriele, D., Pollini, C. M., Peressini, D., & Sensidoni, A. (2001). Filled snack production by cooking-extrusion: 1. Rheological modelling of the process. Journal of Food Engineering (in press). Derby, R. I., Miller, B. S., Miller, B. F., & Trimbo, H. B. (1975). Visual observation of wheat-starch gelatinization in limited water systems. Cereal Chemistry, 52(5), 702–713. Doublier, J. L., Paton, D., & Llamas, G. (1987). A rheological investigation of oat starch pastes. Cereal Chemistry, 64(1), 21–26. Dreese, P. C., Faubion, J. M., & Hoseney, R. C. (1988). Dynamic rheological properties of flour, gluten, and gluten-starch doughs. I. Temperature-dependent changes during heating. Cereal Chemistry, 65(4), 348–353. Evans, I. D., & Haisman, D. R. (1982). The effect of solutes on the gelatinization temperature range of potato starch. Starch/Staerke, 34(7), 224–231. Falcone, R. G., & Phillips, R. D. (1988). Effects of feed composition, feed moisture and barrel temperature on the physical and rheological properties of snack-like products prepared from cowpea and sorghum flours by extrusion. Journal of Food Science, 53(5), 1464–1469. Gabriele, D., Curcio, S., & de Cindio, B. (2001). Optimal design of single screw extruder for liquorice candy production: A rheology based approach. Journal of Food Engineering, 48(1), 33–44. Ghiasi, K., Hoseney, R. C., & Varriano-Marston, E. (1983). Effects of flour components and dough ingredients on starch gelatinization. Cereal Chemistry, 60(1), 58–61. Guerra-Hernandez, E., & Corzo, N. (1996). Furosine determination in baby cereals by ion-pair reversed-phase liquid chromatography. Cereal Chemistry, 73(6), 729–731. Sowa, S. M., & White, P. J. (1992). Hartunian Characterization of starch isolated from oat groats with different amounts of lipid. Cereal Chemistry, 69(5), 521–527. Kokini, J. L., Cocero, A. M., & Madeka, H. (1995). State diagrams help predict rheology of cereal proteins. Food Technology, 49(10), 74–82. Mauron, J. (1981). The Maillard reaction in food: a critical review from the nutritional standpoint. In C. E. Eriksson (Ed.), Progress in Food and Nutrition Science (Vol. 5). Maillard reaction in food. Chemical, physiological and technological aspects (pp. 5–35). Oxford: Pergamon Press. Mercier, C. (1977). Effect of extrusion cooking on potato starch using a twin-screw French extruder. Staerke, 29(2), 48–52. Mercier, C., & Feillet, P. (1975). Modification of carbohydrate components by extrusion-cooking of cereal products. Cereal Chemistry, 52(3), 283–297. B. de Cindio et al. / Journal of Food Engineering 54 (2002) 63–73 Mercier, C., Charbonniere, R., Grebaut, J., & de la Gueriviere, J. F. (1980). Formation of amylose-lipid complexes by twin screw extrusion cooking of manioc starch. Cereal Chemistry, 57(1), 4–9. Noguchi, A., Mosso, K., Aymard, C., Jeunink, J., & Cheftel, J. C. (1982). Maillard reactions during extrusion–cooking of proteinenriched biscuits. Lebensmittel-Wissenschaft und-Technologie, 15(2), 105–110. Peressini, D., Sensidoni, A., Pollini, C. M., & de Cindio, B. (1999). Improvement of wheat fresh pasta-making quality: influence of sodium chloride on dough rheological properties. Italian Food and Beverage Technology (17-September), 20–23. Peressini, D., Sensidoni, A., Pollini, C. M., Gabriele, D., Migliori, M., & de Cindio, B. (2001). Filled-snacks production by coextrusion– cooking: 3. A rheological based method to compare filler processing properties. Journal of Food Engineering (in press). Pirini, A., Conte, L. S., Francioso, O., & Lercker, G. (1992). Capillary gas chromatographic determination of free amino acids in honey as a means of discrimination between different botanical sources. Journal of High Resolution Chromatography, 15(3), 165–170. Resmini, P., Pagani, M. A., & Pellegrino, L. (1990a). Valutazione del danno termico nella pasta alimentare mediante determinazione per HPLC della e-furoilmetil-lisina (furosina). Tecnica Molitoria, 41(10), 821–826. Resmini, P., & Pellegrino, L. (1994). Occurrence of protein-bound lysylpyrrolaldehyde in dried pasta. Cereal Chemistry, 71(3), 254– 262. Resmini, P., Pellegrino, L., & Battelli, G. (1990b). Accurate quantification of furosine in milk and dairy products by a direct HPLC method. Italian Journal of Food Science, 2(3), 173–183. Resmini, P., Pellegrino, L., Pagani, M. A., & Noni, I. (1993). De Formation of 2-acetyl-3D-glucopyranosylfuran (glucosylisomaltol) 73 from nonenzymatic browning in pasta drying. Italian Journal of Food Science, 5(4), 341–353. Rolee, A., & Le Meste, M. (1997). Thermomechanical behavior of concentrated starch-water preparations. Cereal Chemistry, 74(5), 581–588. Rolee, A., & Le Meste, M. (1999). Effect of moisture content on thermomechanical behavior of concentrated wheat starch-water preparations. Cereal Chemistry, 76(3), 452–458. Roos, Y. H. (1995). Phase transitions in foods. New York: Academic Press. Sensidoni, A., & Peressini, D. (1998). Riduzione della reazione di Maillard nella produzione di paste essiccate Studio di trattamenti preventivi sulla materia prima. Tecnica Molitoria, 49(3), 233– 240. Sensidoni, A., Peressini, D., & Pollini, C. M. (1994). Analisi della furosina in un processo di pastificazione che comprende un’essiccazione VHT. Tecnica Molitoria, 45(12), 1297–1302. Sensidoni, A., Peressini, D., Pollini, C. M., & Munari, M. (1996). Effect of mechanical stress and reducing sugar content of semolina on pasta damage during drying. Italian Food and Beverage Technology, 7(May), 9–14. Tester, R. F., & Morrison, W. R. (1990). Swelling and gelatinization of cereal starches. I. Effects of amylopectin amylose and lipids. Cereal Chemistry, 67(6), 551–557. Wang, L. Z., & White, P. J. (1994). Structure and physicochemical properties of starches from oats with different lipid contents. Cereal Chemistry, 71(5), 443–450. Wolf, J. C., Thompson, D. R., & Reineccius, G. A. (1978). Comparisons between model predictions and measured values for available lysine losses in a model food system. Journal of Food Science, 43(5), 1486–1490.