Download as pdf or txt
Download as pdf or txt
You are on page 1of 15

Food and Bioproducts Processing 135 (2022) 190–204

Available online at www.sciencedirect.com

Food and Bioproducts Processing

journal homepage: www.elsevier.com/locate/fbp

Extrusion of pea snack foods and control of


biopolymer changes aided by rheology and
]]
]]]]]]
]]

simulation

I. Jebalia, G. Della Valle, M. Kristiawan
INRAE, UR 1268 Biopolymers Interactions and Assemblies (BIA), 44316 Nantes, France

a r t i c l e i n f o a b s t r a c t

Article history: Pea flour and a blend of pea starch and protein isolate with a starch-to-protein ratio close
Received 11 May 2022 to 2/1 were processed using co-rotating twin-screw extruders. Different extruder scales
Received in revised form 2 August operated at a moisture content of 18–35% and a screw speed of 120–700 rpm resulted in
2022 large intervals of melt temperature T (95–165 °C) and specific mechanical energy SME
Accepted 4 August 2022 (150–2000 kJ/kg). With increasing T and SME, starch solubility in water increased due to
Available online 8 August 2022 starch melting and depolymerisation, and protein solubility in SDS decreased due to
formation of protein aggregates linked by disulphide bonds. Extruded foods’ morphology
Keywords: was studied by CLSM. Protein cross-linking increased the median size of protein ag­
Morphology gregates; starch destructuration increased the area of the starch phase to the detriment of
Protein aggregates the protein one. Melt shear viscosity was determined using a pre-shearing capillary rhe­
Starch ometer. The melts exhibited shear-thinning behaviour according to a power-law model,
Solid foams with values close to those of model parameters observed for pea flour and SP 2/1 blend.
1D twin-screw extrusion model The rheological model was implemented in a 1D global extrusion model to simulate pea
Viscosity ingredient extrusion. Satisfactory correlations between predicted extrusion variables (T,
SME) and biopolymer transformation suggested that products with target structure can be
obtained using simulation to tune extrusion conditions.
© 2022 Institution of Chemical Engineers. Published by Elsevier Ltd. All rights reserved.

Extrusion is an efficient, versatile, simple and safe process


1. Introduction
that is widely used to develop various products (Ek and
Ganjyal, 2020). It can be applied to incorporate protein-rich
Pulse crops (pea, lentil, bean, etc.) are an excellent source of
plant ingredients into foods. The extrusion process consists
protein (20–30% dry basis), dietary fibre (10–30%), starch
of a succession of two phenomena: (1) transition from a di­
(40–55%), vitamins and minerals, and they are low in sugar,
vided solid medium (powder) to a deformable continuous
sodium and fat (Boye et al., 2010). Contrary to cereals, pulse
one (melt) due to thermo-mechanical treatment and che­
proteins are high in lysine and low in sulphur-containing
mical action; (2) expansion of continuous medium to a solid
amino acids: cysteine, methionine, and tryptophan. There­
cellular matrix due to a sudden pressure drop at the die exit.
fore, they are good candidates to complement cereals in ex­
The density of products depends on extrusion conditions,
truded foods. The formulation of expanded snacks made
e.g., high feed moisture content and low melt temperature
entirely from pulses is an interesting way to promote the
inside the die results in dense materials (Kristiawan and
consumption of healthy legume-based foods in the young
Della Valle, 2020). The use of physical blowing agents, such
population.
as N2 gas, can increase food expansion accompanied by
small and uniform cells (Luo et al., 2020).
Elevated extrusion temperatures and shear forces in­

Corresponding author. crease biopolymer transformation such as starch melting
E-mail address: and depolymerisation, protein denaturation and aggregation,
magdalena.kristiawan@inrae.fr (M. Kristiawan).
https://doi.org/10.1016/j.fbp.2022.08.001
0960-3085/© 2022 Institution of Chemical Engineers. Published by Elsevier Ltd. All rights reserved.
Food and Bioproducts Processing 135 (2022) 190–204 191

Nomenclature flour or %), and gain in WSIsolids of extruded


product (%).
α water plasticisation coefficient (di­ WSIstarch, ΔWSIstarch WSI of starch (in kg/kg dry starch
mensionless). in flour or %), and gain in WSIstarch of extruded
C, D, n0, ai parameters in fitting equations. product (%).
ANOVA analysis of variance.
aMC moisture content-based shift factor (di­
mensionless).
aT temperature-based shift factor (di­ and product browning (Ek et al., 2020; Kristiawan and Della
mensionless). Valle, 2020). The melting dataset is widely available for
CSLM confocal laser scanning microscopy. starches but there is no equivalent entity for proteins. The
DTE dithioerythritol. knowledge of starch melting temperature at various
DSC differential scanning calorimetry. moisture contents is primordial for setting extrusion condi­
D50 median protein aggregate size (μm). tions to obtain fully melting matrices adapted for texturing.
E activation energy (J/mol). Increasing dissipated heat promotes molecular changes in
η viscosity (Pa.s). proteins that take place in several steps: association, dis­
ηG superposed viscosity after time-temperature sociation and aggregation of subunits by non-covalent (hy­
translation and time-moisture content drophobic, ionic, H) and covalent (disulphide, isopeptide)
translation (Pa.s). bonds (Camire, 1991; Day and Swanson, 2013; Stanley, 1989).
, app true and apparent shear rates (1/s). Moreover, increasing shear energy promotes protein un­
Ii starch-protein interface index (di­ folding, resulting in protein re-association, which means that
mensionless). protein unfolding and cross-linking occur simultaneously
K, K0 flow consistency index (Pa.sn) and its value during extrusion (Della Valle et al., 1994; Fang et al., 2013).
at reference temperature T0 and reference These structural changes of biopolymers affect melt
moisture content MC0. rheology, which in turn influences the process itself and melt
L/D length-to-diameter ratio. expansion and, thus, the final structure of the food (Beck
LSD leas significant different test. et al., 2017; Kristiawan et al., 2018, 2020; Kristiawan and Della
MC, MC0 moisture content and its reference value (% Valle, 2020; Xie et al., 2012). In addition, an intense protein
in wet basis). cross-linking can lead to extruded products with the ap­
N screw speed (rpm). pearance of surface irregularities and breakup (Pommet
Nu Nusselt number. et al., 2003). The structure of expanded solid foods can be
n flow behaviour index (dimensionless). described at various structural levels by density, cellular
P pressure (Pa), measured at die entrance of structure and morphology, or the starch-protein organisation
extruder and Rheoplast®. in the cell walls of these foods (Guessasma et al., 2011). The
PAR protein area ratio. cell wall material can be considered as a composite of starch
PF pea flour. and protein whose morphology depends on composition,
PPI pea protein isolate. extrusion conditions and biopolymer transformation
P_DTE, ΔP_DTE protein solubility in DTE (%) and gain in (Habeych et al., 2008; Jebalia et al., 2019; Kristiawan et al.,
protein solubility in DTE (%). 2018). In turn, the structure of extruded foods, expanded or
P_SDS, ΔP_SDS protein solubility in SDS (%) and loss of not, controls their texture and end-use properties, including
protein solubility in SDS (%). nutritional ones (Brennan et al., 2013; Carbonaro et al., 2012;
QF mass flow rate of powdery raw material and Laleg et al., 2016; Philipp et al., 2017, 2018).
water (kg/h). Most of the research on the extrusion of pulse legumes
QV volumetric flow rate of the material (mm3/s), has focused on experimental studies that link operating
R gas constant (J/mol.K). conditions and recipes to extrusion variables, food structure
Ri radius of the capillary (mm). and functional properties using statistical modelling (Beck
S cross-section of the central piston (mm2). et al., 2018; Berrios et al., 2008; Della Valle et al., 1994; Félix-
SDS sodium dodecyl sulphate. Medina et al., 2020; Rangira et al., 2020; Sumargo et al., 2016).
SE-HPLC size-exclusion high-performance liquid Since extrusion trials are laborious and time-consuming, the
chromatography. challenges in extrusion optimisation, scale-up and trouble­
SP 2/1 blend of pea starch and pea protein isolate shooting can be tackled using extrusion modelling. A 1D
with starch-to-protein ratio = 2 /1. global model that calculates the changes in the main ther­
SME, SME_mes, SME_com specific mechanical energy, momechanical parameters such as dissipated energy, tem­
retrieved experimentally and predicted (kJ/kg). perature, viscosity and pressure has been proposed by Della
τw wall shear stress (Pa). Valle et al. (1993) and Vergnes et al. (1998). Knowledge of the
T, T0 melt temperature and its reference viscous behaviour of molten starchy products obtained in
value (°C). extrusion conditions is pivotal for modelling the material
T_mes, T_com measured and predicted melt tem­ flow in an extruder. However, this data is currently lacking in
perature at die entrance (°C). the industrial field, mainly because classical rheometers are
Td, Tb imposed die and last barrel temperature (°C). not relevant and measurements are difficult. Using an off-
Tm melting temperature of starch (°C). line pre-shearing capillary rheometer (Rheoplast®) and spe­
w.b. wet basis. cific in-line slit die rheometers, not widely available, makes it
WSIsolids, ΔWSIsolids WSI of total solids (in kg/kg dry possible to produce structural changes in starch and proteins
192 Food and Bioproducts Processing 135 (2022) 190–204

Table 1 – Chemical composition of raw material (% dry basis).


Raw material Proteins Starch Ash Lipid Sugar Others*
c c b
Pea flour (PF) 24 ± 0.2 46 ± 0.3 2 ± 0.04 2 ± 0.02 ng 26 ± 0.3a
Pea starch 0.5 ± 0.04d 98 ± 2a 0.1 ± 0.01c – ng 1.4 ± 0.05d
Pea protein isolates (PPI) 88 ± 0.5a 0.4 ± 0.01d 4 ± 0.09a – ng 7.6 ± 0.04b
SP 2/1 32 ± 0.4b 63 ± 0.4b 2 ± 0.04b – ng 3 ± 0.05c

*Fibre and other components, determined by difference method. ng: not determined. Means with the different superscripts within a column
are significantly different (significance level of 0.05).

before rheological measurement (Horvat et al., 2013; Vergnes defined as the maximum temperature of the melting en­
and Villemaire, 1987; Xie et al., 2012). dothermic peak. Some results are reported in Table 2.
There is still a lack of knowledge and data, particularly in
relation to melt rheology, necessary to foresee the structural 2.2. Extrusion experiments
changes of the biopolymeric components that govern snack
features. Consequently, to date, the design of pulse based- The extrusion trials for manufacturing dense composites
extruded products at the industrial level is still based on a from pea flour and SP 2/1 were conducted using a laboratory-
trial-and-error approach. Given this context, this work firstly scale co-rotating twin-screw extruder (Thermo Scientific™
aims to ascertain the effect of extrusion variables such as Process 11, Germany), as described in detail by Jebalia et al.
temperature and specific mechanical energy on starch and (2019). The screw diameter was 11 mm, and the length-to-
protein structural changes of snacks and composite mate­ diameter (L/D) ratio was 24.5. The screw profile was com­
rials, represented as snack cell walls, and the impact of posed of right-handed screw (conveying) elements with a
biopolymer transformation on composite morphology. In pitch of 11 mm, followed by a block of four kneading discs
contrast with “pea snacks” (Kristiawan et al. 2018), the da­ with a staggering angle of 90 ° and a total length of 11 mm,
taset of the impact of measured extrusion variables on starch and one left-handed screw element of 5.5 mm (supplemen­
and protein structural change of “pea composites” and, in tary material, Fig. S.1a). The design of the experiments was
turn, composites morphology has not been published before. described as follows (Table 2). For all materials, the flow rate
Then, by determining the viscosity of molten pea ingredients of powder (QF) was kept constant at 0.24 kg/h, and the die
under extrusion conditions, our second goal is to predict temperature was regulated at 95 °C to avoid expansion and to
structural changes using extrusion modelling. For this pur­ be thus able to process dense composites. The moisture
pose, we will use extrusion data available from the literature content (MC) of feed was varied at three levels: 25%, 30%, and
and the present work that focuses on two raw materials: pea 35%. For each value of MC, the rotation speed (N) was
flour and a blend of pea starch and pea protein isolate (SP 2/1) changed at three levels: 120, 300, and 500 rpm. For each N,
with a starch-to-protein ratio of 2/1 (dry basis). This work is the last barrel temperature (Tb) was set at Tm–20, Tm and
original because the determination of viscous behaviour of Tm+20 °C, where Tm is the melting temperature of the feed.
pulses-based ingredients and the design of pulse-based ex­ The setting of operating conditions (MC, N, Tb) led to a wide
truded foods and composites using rheology-based process interval of measured extrusion variables, which can generate
simulation never be accomplished anywhere. the composites with a wide range of structural changes. For
further analysis, three representative composites of pea flour
2. Materials and methods and SP 2/1, extruded at similar three SME levels (≈ 200, 1000,
2000 kJ/g), were selected. Only extrusion trials that generated
2.1. Raw materials the chosen samples were performed in triplicate. For the
selected trials, the reported values of measured SME
Pea starch (median particle diameter: D50 = 27 µm) and pea (SME_mes), melt temperature (T_mes) and die entrance
protein isolates (PPI; D50 = 72 µm) were supplied by Roquette pressure (P) were the average values with a standard devia­
(France). Pea flour (D50 = 480 µm) was purchased from tion. These trials were significantly different in terms of ex­
Sotexpro (France). A blend with a starch/protein ratio of 2/1 trusion variables at the significance level of 0.05 (Table 2).
(weight/weight, dry basis), called SP 2/1, was obtained by The name of selected samples was given in Table 2 and
mixing pea starch and PPI using a Kenwood mixer. The pea supplementary material (Table S1). The sample names in­
flour had a starch-to-protein ratio similar to that of SP 2/1 dicate the SME level (L, M, H) and the formulation (PF: pea
(1.92/1). The chemical composition of the raw materials was flour, SP 2/1: starch-protein blend).
determined by standard enzymatic and Kjeldahl methods, Expanded snacks (“solid foams”) from pea flour were
described in detail elsewhere (Kristiawan et al., 2018). All processed using a pre-pilot co-rotating twin-screw extruder
measurements were conducted in triplicate. The composi­ (Coperion Werner & Pfleiderer ZSK 26Mc) with a screw dia­
tions are presented in Table 1. The reported values are the meter of 25.5 mm and a L/D of 29, as formerly indicated by
average values with a standard deviation. All raw materials Kristiawan et al. (2018). All trials were performed using the
were significantly different at the significance level of 0.05 same last barrel temperature: Tb = 90 °C. Several die geome­
regarding the content of starch, protein and other con­ tries were used: (1) a circular die with a diameter of 3 mm and
stituents such as fibre. a length of 15 mm was directly attached to the extruder die
The melting temperature (Tm) of hydrated pea in­ head; and (2) a multi-step slit die rheometer combined with a
gredients was retrieved from previous works (Jebalia et al., circular die (D = 3 mm, with three length options L: 5, 10,
2019; Kristiawan et al., 2018). It was measured using a DSC 15 mm) mounted in the extruder die head. The geometry of
apparatus following the method of Logié et al. (2017) and was the slit die was described elsewhere (Horvat et al., 2013). The
Food and Bioproducts Processing 135 (2022) 190–204 193

Table 2 – Design of experiments for manufacturing of SP 2/1-based composites.


Trial Operating conditions Extrusion variables Sample

MC (%) N (rpm) Tb Tm (°C) T_mes (°C) SME_mes (kJ/kg) P (MPa) code*

1 35% 500 Tm–20 112 97 978 0.9


2 Tm 99 913 0.6
3 Tm ± 20 104 963 0.6
4 300 Tm–20 100 470 0.3
5 Tm 96 520 0.4
6 Tm ± 20 101 590 0.6
7 120 Tm–20 93 ± 2c 192 ± 13c 1.8 ± 0.3c SP 2/1_L
8 Tm 97 210 2.4
9 Tm ± 20 100 285 1.2
10 30% 500 Tm–20 122 101 1295 1
11 Tm 103 1051 0.5
12 Tm ± 20 111 1198 0.8
13 300 Tm–20 105 628 1.2
14 Tm 105 736 1.4
15 Tm ± 20 108 595 0.6
16 120 Tm–20 103 305 1.8
17 Tm 100 287 3.3
18 Tm ± 20 104 288 2.3
19 25% 500 Tm–20 134 122 ± 5b 1914 ± 284a 2.6 ± 0.2b SP 2/1_H
20 Tm 118 1866 1.1
21 Tm ± 20 124 1561 1.3
22 300 Tm–20 116 953 1
23 Tm 113 933 1.8
24 Tm ± 20 135 ± 5a 990 ± 59b 3.7 ± 0.4a SP 2/1_M
25 120 Tm–20 110 361 1.4
26 Tm 106 790 3
27 Tm ± 20 113 427 3.3

*Selected samples for further analysis. Means with the different superscripts within a column, for the selected samples, are significantly
different (significance level of 0.05).

temperature of the slit die was set at 130 °C and 150 °C, while 2.3.2. WSI of total solids (WSIsolids)
the circular one was not regulated. The screw profile was The extruded products were ground (< 250 µm) and dispersed
different from that used for dense composite extrusion: in deionized water (30 °C, 30 min), and then centrifuged. The
conveying elements with decreasing pitch followed by three supernatant was dried overnight in a Chopin oven at 105 °C.
kneading discs staggered at 45 °, three left-handed screw WSIsolids was determined as the percentage of the mass of
elements followed by the last conveying elements (supple­ supernatant dry solid in the original dry sample.
mentary material, Fig. S.1b).
For both extrusion systems, the raw material was fed by a 2.3.3. WSI of starch (WSIstarch)
gravimetrically controlled feeder and water was added to the The ground samples first underwent the same extraction and
second barrel by a volumetric pump to obtain feed with the centrifugation procedure as described above. Thereafter, the
desired moisture contents. Both extruders operated at sev­ amount of soluble starch in the supernatant was determined
eral levels of moisture content MC (18–35% w.b.) and screw by titration of the supernatant with orcinol in an automatic
speed (120–700 rpm). spectrocolorimeter.

2.3.4. Protein solubility


2.3. Extruded product analysis
Protein solubility was determined by triplicate measure­
ments using the method of Kristiawan et al. (2018). The
The experimental database of pea snack features was re­
ground extruded products (< 250 µm) underwent two suc­
trieved from the literature (Kristiawan et al., 2018). Con­
cessive protein extractions. The first step consisted in ex­
cerning pea composite characteristics, the structural
tracting the proteins held by weak interactions (non-covalent
changes of starch and protein were determined in this study,
bonds) in a buffer containing sodium dodecyl sulphate (SDS).
whereas their morphological features were analysed in a
In the second step, the proteins insoluble in SDS were ex­
previous study (Jebalia et al., 2019). The methods used for the
tracted at 60 °C under sonication in a buffer containing an
physical-chemical characterisation were described else­
agent that reduces disulphide (S-S) bonds such as dithioer­
where, so this section recalls them briefly. The data of
ythritol (DTE). Both SDS and DTE extracts were analysed by a
composite features is summarized in Table 3. The reported
high-performance size exclusion chromatography (SE-HPLC)
values were the average values with a standard deviation.
apparatus to determine the molar mass distribution of so­
luble protein aggregates. Protein solubility either in SDS
2.3.1. Water solubility index (WSI) (P_SDS) or DTE (P_DTE) was then expressed as a percentage
The measurements of WSI were carried out in triplicate fol­ of the integrated chromatographic area relative to all ex­
lowing the method of Kristiawan et al. (2018). tractable proteins.
194 Food and Bioproducts Processing 135 (2022) 190–204

Table 3 – Solubility of biopolymers and morphological features of composites samples.


Samples WSIsolids WSIstarch Protein solubility (%) Morphological features*

(%) (%) in SDS in DTE D50 of protein Protein area

aggregates ratio

(µm)
e e a f
Native flour 15.0 ± 0.5 7.50 ± 0.6 92.0 ± 2.7 8.00 ± 0.5
Native PPI 17.5 ± 0.6d 6.00 ± 0.7ef 76.0 ± 1.5c 24.0 ± 0.5c
Native starch 2.00 ± 0.5 g 2.50 ± 0.7 g
PF_L 17.1 ± 0.6d 23.7 ± 1.0b 87.0 ± 1.8b 9.00 ± 0.9f 11.0 ± 1.0c 0.31 ± 0.035a
PF_M 21.0 ± 1.0c 17.3 ± 0.9c 78.0 ± 1.1c 16.4 ± 1.0e 16.5 ± 3.0bc 0.24 ± 0.020ab
PF_H 38.6 ± 1.0a 30.7 ± 1.1a 71.0 ± 2.1d 21.0 ± 0.8d 15.0 ± 2.0bc 0.19 ± 0.020b
SP2/1_L 6.56 ± 0.5 f 4.11 ± 0.6fg 69.0 ± 0.9d 29.0 ± 0.9b 35.0 ± 10a 0.31 ± 0.065a
SP2/1_M 15.4 ± 0.6de 13.8 ± 0.7d 56.0 ± 0.9e 39.0 ± 0.9a 40.0 ± 4.0a 0.24 ± 0.035ab
SP2/1_H 25.8 ± 1.0b 17.9 ± 1.4c 58.0 ± 0.8e 29.0 ± 0.8b 27.0 ± 4.0ab 0.17 ± 0.035b

*Data was retrieved from a previous study (Jebalia et al., 2019). Means with the different superscripts within a column are significantly
different (significance level of 0.05).

The final results of WSIsolids, WSIstarch and protein solu­ conditions is moisture content (MC) between 20% and 35%
bility were presented either as their original value or as the wet basis, temperature (T) between 85 and 175 °C, and spe­
gain or loss in solubility: ΔWSIsolids, ΔWSIstarch, ΔP_SDS, cific mechanical energy (SME) up to 250 kJ/kg, modulated by
ΔP_DTE, computed as the difference between the solubility of the rotational speed of the central piston (50 or 200 rpm).
extruded products and that of native material. The untreated The raw materials were mixed with water using a
flour and PPI were taken as reference for extruded flour and Kenwood mixer to obtain selected moisture contents. The
SP 2/1, respectively. As shown in Table 3, the properties of mixture was then maintained at 4 °C for 24 h for moisture
extruded composites were significantly different from those equilibration, before measuring the final moisture content.
of native materials (significance level of 0.05). For each run, about 15 g of raw material were processed in
the Rheoplast®. After the material was pre-sheared in the
Couette cell by piston rotation for 20 s at a controlled tem­
2.3.5. Starch-protein morphology
perature (T), the piston was moved downward to push the
The microstructure of pea composites was observed using
molten material through the capillary at decreasing velocity,
confocal laser scanning microscopy (CLSM) and quantified by
so that the apparent shear rate in the capillary varied from
image analysis following the method established by Jebalia
256 to 0.5 s–1. The sample temperature (T) was measured
et al. (2019).
before the capillary die. The pressure upstream of the ca­
Specimen slices with a thickness of 20 µm were obtained
pillary was recorded during measurements for each shear
by cutting the extruded sample perpendicularly to the ex­
rate. Capillaries of the same diameter (2 mm) but different
trusion flow using a cryotome at − 20 °C. For protein labelling,
lengths (0.1, 8, 16 and 32 mm) were used. Each pressure
one part (by mass) of 0.01% (weight volume/volume) fuchsine
profile was then measured for four L/D ratios (0.05, 4, 8, 16),
acid in 1% (volume/volume) acetic acid was mixed with one
with three repetitions leading to a measurement error of up
part of Kaiser’s glycerol/gelatine solution at 40 °C.
to 15%.
The slices were investigated using a CLSM apparatus
(Nikon A1, Germany) in the epifluorescence mode excited by
a green laser beam at 561 nm. A long-pass filter selected the
2.4.2. Data treatment
emitted light at 570–620 nm. Thereafter, image analysis was
Data analysis was carried out according to the basic princi­
done using Matlab software for at least three plane projec­
ples of capillary rheometry, including the Bagley correction.
tion CLSM images of three spots in the composites. The CLSM
The apparent shear rate ( app ) was computed from the
images were digitized to obtain binary images. Each finite
displacement speed of the central piston (injection speed) as
particle in the image was labelled and the following features
follows:
were determined: the median protein aggregate size (D50)
and the protein area fraction. The variation of the digitiza­ 4Q v 4VS
tion threshold ( ± 10%) induced uncertainty in the particle app = = (1)
Ri3 Ri3
area values of about 15%.
where Q v is the volumetric flow rate of the material (mm3/s),
2.4. Determination of viscous behaviour Ri is the radius of the capillary (mm), V is the downward
speed of the piston (mm/s), and S is the cross-section of the
2.4.1. Measurements central piston (mm2).
Rheological measurements were performed using a pre- The pressure drop variation against the L/D capillary ratio
shearing rheometer (Rheoplast®), as described elsewhere was then plotted for each apparent shear rate and fitted by
(Vergnes & Villemaire, 1987). It combines the features of a linear functions (R2 > 0.95) to obtain Bagley plots (supple­
Couette rotational system and a capillary rheometer. Starch mentary material, Fig. S.2a).
and protein transformation took place in the Couette cell due The variation of wall shear stress ( w) against true shear
to shearing, imposed by the rotation of a central piston and rate was then plotted (Fig. S.2b). On the basis of knowledge
the heating by heating collars. The domain of measurement about Bagley plots, w can be computed as follows:
Food and Bioproducts Processing 135 (2022) 190–204 195

P exponential functions. C and D are model parameters, and


w = (2)
4(L / D) the products of these parameters can result in parameter K0
In a more approximate approach, disregarding the en­ of Eq. 7a.
trance effect, w was computed from the slope of a straight
line passing through the pressure at L / D = 16 and the co­ 2.5. Extrusion simulation
ordinate origin (0, 0).
The flow behaviour index (dimensionless), n, is then de­ The twin-screw extrusion of pea ingredients was simulated
fined by: using Ludovic® software. This simulating software was based
d (log w ) on a 1D global model developed by Della Valle et al. (1993)
n= (3)
d (log app ) and Vergnes et al. (1998). This model uses an iterative
method to compute the energy balance and the flow rate/
The true wall shear rate ( ), for each measurement con­ pressure relationship. Both drag flow and pressure-driven
dition (MC, T) was determined using the Rabinowitch cor­ flow are considered, involving a non-Newtonian viscosity
rection: model of melt. Computation of various parameters is carried
3n + 1 out separately for each type of screw element and for the die
= app (4)
4n components. The elementary models are linked together to
The shear viscosity ( ) is then equal to: obtain a global description of the flow field, such as melt
temperature, SME, viscosity, pressure and filling ratio along
w
= (5) the extruder. Melting is assumed to be instantaneous and
takes place before the first restrictive screw element, such as
Finally, flow curves expressing shear viscosity=f (true wall the left-handed screw element and kneading blocks, in­
shear rate) were built (Fig. S.2c) and fitted with a power dicated by T > Tm. Afterwards, it is assumed that the ma­
function (R2 > 0.95), indicating that the viscous behaviour can terial is fully molten and that it fills the screw channel,
be represented by a power law model. according to local geometry and flow conditions. Since the
final product temperature is unknown and the screws are
2.4.3. Development of viscosity model starve-fed, the filling ratio of the system is not known. Thus,
It was checked that the Ostwald-de Waele power-law model the computation starts from the die and proceeds backwards
could be used to determine starchy melt viscous behaviour in towards the hopper and follows an iterative procedure.
shear flow: In addition to a viscosity model, the input data include the
=K n 1 thermal and physical properties of feeds (Choi and Okos,
(6)
1986; Della Valle and Vergnes, 1994; Vergnes et al., 1998), the
where K is the consistency index and n is the flow behaviour heat transfer coefficient between melt and equipment de­
index. fined using the Nusselt number (Nu), the extruder config­
According to Xie et al. (2012), K and n depend on the uration, screw profile and extruder operating conditions
temperature (T ) and moisture content (MC): (screw speed, flow rate, barrel and die temperature profile).
E 1 1 For all simulations, Nu was set to 20. Other extrusion para­
K = K 0exp (MC MC0 ) (7a)
R T T0 meters were set according to an experimental extrusion da­
tabase of composites and snacks (Section 2.2).
n = n0 + a1 T + a2 MC + a3 T MC (7b) The workflow of predicting food features using extrusion
simulation is as follows. First, the calculated extrusion vari­
where T is the melt temperature in K (Eq. 7a) and °C (Eq. 7b); E
ables such as melt temperature (T_com) and specific me­
is the activation energy (J.mol–1); R is the gas constant
chanical energy (SME_com) are compared to the measured
(J.mol–1. K–1); K0 is the consistency (Pa.sn) at the reference
ones to confirm the validity of extrusion simulation. Then, in
temperature (T0 =80 °C = 353.15 K) and the reference moisture
a second step, the measured structural changes of starch and
content (MC0 = 0.1); α is the water plasticization coefficient;
proteins are linked to the calculated extrusion variables. On
and n0 and ai are the flow index parameters.
the basis of these relationships, sets of extrusion variables
The parameters of Eq. 6, K and n can be derived from the
(T_com, SME_com) and extrusion parameters such as screw
viscosity master curve using time-temperature and time-
speed, flow rate and barrel temperature profile that lead to
plasticizer superposition approaches (Krauklis et al., 2019).
targeted product structure are defined.
The master curve, which includes both effects of tempera­
ture and moisture content, can be calculated as follows:
= f (aT aMC ) 2.6. Statistical analysis
G (8)

E1 The reported values were the average values with a standard


aT = C exp (9)
RT deviation. Some data, such as chemical composition of raw
materials, composites feature, and flow curves of some
aMC = D exp ( MC) (10) melts, were analyzed using one-way analysis of variance
(ANOVA) followed by the least significant difference test
n = n0 = constant (11) (Fisher LSD) at the significance level of 0.05. One-way
where G is the reduced viscosity, and aT and aMC are ANOVA, Fisher LSD, and correlation tests (Pearson and
temperature and moisture content shift factors, respectively. Spearman) between variables (processing and structural
The data of aT =f(1/T ) and aMC =f(MC) can be fitted by changes) were performed by XLSTAT 2022 (Addinsoft,
196 Food and Bioproducts Processing 135 (2022) 190–204

Fig. 1 – Relationship between extrusion variables and operating parameters.


Extrusion experiments to manufacture pea composites were conducted at different moisture contents: 25% (white), 30%
(grey) and 35% (black), using pea flour (square) and SP 2/1 (circle) as raw material. The pea snack data (crosses) were retrieved
from Kristiawan et al. (2018).

(a) Variation of specific mechanical energy (SME_mes) with filling ratio (Q/N);
(b) Variation of melt temperature (T_mes) with SME_mes.
The solid lines represent data fitting:

(a) Composites: SME_mes= 0.69 × (Q/N)–0.97; R2 = 0.80;


Snacks: SME_mes= 133.9 × (Q/N)–0.51; R2 = 0.87;
(b) Composites (pea flour): T_mes= 0.019 ×SME_mes+ 95.25; R2 = 0.64;
Composites (SP2/1): T_mes= 0.011 ×SME_mes+ 98.72; R2 = 0.36;.
Snacks: T_mes= 27.24 ×SME_mes0.26; R2 = 0.76.

France). Pearson stands for linear correlation while friction and melt viscous dissipation rather than conduction
Spearman for monotonic correlation. from the barrel, and that they are enhanced when screw
configuration is more restrictive.
3. Results and discussion
3.1.2. Impact of extrusion on starch structural change
3.1. Extrusion experimental results WSIsolids are positively and linearly correlated with WSIstarch,
regardless of raw materials, extrusion systems and condi­
3.1.1. Variations of extrusion variables with operating tions (supplementary material, Fig. S.3, R2 = 0.81). The values
parameters of WSIsolids are generally higher than those of WSIstarch be­
Combining two extrusion datasets (composites and ex­ cause, in addition to starch, WSIsolids considers the solubility
panded snacks) led to a wide range of extrusion variables: of other components such as fibres that may be solubilised
measured SME (SME_mes = 150–2000 kJ/kg) and melt tem­ during extrusion. Concerning protein aggregates resulting
perature at die entrance (T_mes = 95–165 °C) (Fig. 1a,b), which from the cross-linking of proteins, they are not soluble in
leads to a large panel of product transformation and struc­ water (Stanley, 1989). Consequently, the greater the WSIsolids
ture changes. is, the greater is its gain (ΔWSIsolids) compared to the solu­
As indicated in Fig. 1a, good correlations were found be­ bility of untreated ingredients. Therefore, the larger the
tween SME_mes and the ratio of total flow rate to screw starch macromolecules depolymerised by extrusion will be.
speed (filling ratio: Q/N), meaning that SME can be easily An increase in specific mechanical energy (SME_mes) in­
controlled by tuning the filling ratio of the screw. The cor­ creased starch solubility in water (Fig. 2a), with a distinct
relations were distinct for composites and snacks (R2 = correlation observed for each extrusion system (snacks and
0.8–0.87) due to the differences in extruder screw profiles. composites, R2 = 0.86–0.88). Pearson correlation analysis re­
The use of a more restrictive screw profile resulted in a sult supported that SME_mes is very strongly and positively
higher melt temperature (maximum T_mes: 165 °C vs. 135 °C, correlated with ΔWSIsolids, and thus with starch solubility
snacks vs. composites, Fig. 1b). Acceptable positive correla­ (r = 0.94, significance level = 0.05, supplementary material,
tions between T_mes and SME_mes were found (R2 = Table S2). The difference between the trend observed for the
0.64–0.76) for each product, except for SP 2/1 composites, two materials is probably due to the difference in screw
regardless of the operating conditions and moisture content. profile, less restrictive for pea composites and more re­
Pearson correlation analysis result supported that T_mes is strictive for pea snacks. The gain in starch solubility of pea
strongly and positively correlated with SME_mes (r = 0.78, snack systems extruded at higher temperatures (120–165 °C)
significance level = 0.05, supplementary material, Table S2). abruptly increased from 6.5% to 25% in a narrow SME domain
Indeed, an increase in moisture content had a negative im­ (450–550 kJ/kg) before it levelled at 30%. Conversely, pea
pact on T_mes and SME_mes. These results indicated that composite systems extruded at lower temperatures
heating phenomena in extrusion were dominated by solid (95–135 °C) required a much larger SME (up to 2000 kJ/kg) to
Food and Bioproducts Processing 135 (2022) 190–204 197

Fig. 2 – Impact of measured melt temperature or SME on biopolymer solubilities: gain in WSIsolids (a), loss of protein
solubility in SDS (b), and gain in protein solubility in DTE (c). The curves correspond to the data fitting: (a) Snacks: ΔWSIsolids
= 27/(1 + exp(−0.0237 ×(SME_mes − 483))); R2 = 0.86; (b) Composites: ΔWSIsolids = 0.011 ×SME_mes + 1.58; R2 = 0.88; (c)
ΔP_SDS = 278008/(1 + exp(−0.041 ×(T_mes − 373))); R2 = 0.72; (d) ΔP_DTE = 92965/(1 + exp(−0.0358 ×(T_mes − 387))); R2 = 0.76.

achieve the same state of final starch destructuration increasing melt temperature (T_mes) from 95° to 165°C in­
(ΔWSIsolid ≈ 25%). Both the heat, generated by viscous dis­ creased ΔP_SDS from 5% to 60% (R2 = 0.72), whereas ΔP_DTE
sipation, and shear (mechanical energy), contributed to increased from 1.5% to 40% (R2 = 0.76) (Fig. 2b,c). Spearman
starch destructuring mechanisms at low moisture (up to correlation analysis result supported that T_mes is strongly
35%), a common trend for extruded starchy products. The and positively correlated with ΔP_SDS and ΔP_DTE (r = 0.77
SME itself took thermal and mechanical energy into account. and 0.66, respectively, significance level = 0.05, supplemen­
The heat induced crystallite melting and internal dis­ tary material, Table S3). The loss of protein solubility in SDS
organisation of the granular structure, whereas mechanical started at 137 °C, while the onset of formation of DTE-soluble
energy induced granule fragmentation (Barron et al., 2002; Li proteins occurred at 130 °C, regardless of raw material and
et al., 2014; Xie et al., 2012). Combination of thermal and extrusion systems. As specific mechanical energy (SME)
mechanical energy increased the breakdown and solubilisa­ correlated positively with melt temperature (T) (Fig. 1b), the
tion of starch granules and macromolecules to a greater ex­ same trend was found as in the case of T: SME impacted
tent, including insoluble fibres, and subsequently increased ΔP_SDS and ΔP_DTE positively (as also shown by Spearman
the amount of water-soluble components. correlation matrix, supplementary material, Table S3).
However, the temperature was better correlated to protein
3.1.3. Impact of extrusion on protein structural change solubility changes than SME: R2 0.76 vs. 0.23 for ΔP_DTE of
After extrusion, an increase in loss of protein solubility in composites, and R2 0.76 vs. 0.62 for ΔP_DTE of snacks (results
SDS (ΔP_SDS), indicating the degree of non-covalent protein- not shown). This result indicates that heat is a dominant
protein interactions, occurs concomitantly with the increase factor in creating S-S bonds during extrusion.
in gain in protein solubility in DTE (ΔP_DTE), regardless of The mechanisms of protein insolubilisation during ex­
raw material, extrusion conditions and systems trusion can be drawn from our results and the literature.
(Supplementary material, Fig. S.4). The very strong and po­ Increasing heat and shear during extrusion led to the protein
sitive correlation between ΔP_SDS and ΔP_DTE was also denaturation and creation of SDS-insoluble protein ag­
shown by Pearson correlation test (r = 0.82, significance level gregates via protein cross-linking by covalent bonds. This
= 0.05, supplementary material, Table S2). The ΔP_DTE re­ trend has been reported for the extrusion of soya, legumes
flects the quantity of protein cross-linked by S-S bonds. This and glutens (Alonso et al., 2000; Camire, 1991; Laleg et al.,
result therefore confirmed that the protein insolubilisation is 2016; Pommet et al., 2003; Stanley, 1989). However, heat
due to protein cross-linking by covalent bonds. Furthermore, seems to play a crucial role in forming S-S bonds. On the
198 Food and Bioproducts Processing 135 (2022) 190–204

Fig. 3 – Variations of composite morphological features with starch and protein solubilities. (a) Change of protein area ratio
with the gain in WSIsolids. The curve stands for data fitting: PAR= –0.01 × ΔWSIsolids + 0.34; R2 = 0.95; (b) Variation of median
size of protein aggregates with protein solubility in DTE. The trend is as follows: D50 = 1.04 × P_DTE – 0.57; R2 = 0.88.

other hand, shear can enhance the unfolding of protein ag­ agreement with preceding results showing that SME nega­
gregates (Della Valle et al., 1994; Fang et al., 2013), after which tively impacted the protein area ratio (R2 = 0.86) (supple­
the re-association and re-cross-linking of the protein sub- mentary material, Fig. S.5) (Jebalia et al., 2019). The median
units occur under heat. size protein aggregate (D50) is positively correlated with
The combination of solubility data set of pea composites protein solubility in DTE (R2 = 0.88, Fig. 3b). The protein cross-
with the one of “pea snacks”, published in Kristiawan et al. linking by covalent bonds (S-S) can lead to the formation of
(2018), extends the interval of experimental domain for ex­ large protein aggregates. Pearson correlation analysis result
trusion variable and product characteristics. The overlapping (supplementary material, Table S2) supported that PAR is
of several data points of starch and protein solubility for very strongly and negatively correlated with ΔWSIsolids
snacks and composites (‘snack’s cell wall’) means that the (r = −0.88); in contrast, D50 is positively correlated with P_DTE
selected composites can be a representative of the cell wall. (r = 0.94) at the significance level of 0.05.
This work emphasizes that, for the first time, the mor­
phology of composites (‘snack’s cell wall’) can be modulated
3.1.4. Impact of biopolymer structural changes on composite by protein structural changes tuned by the variation of pro­
morphology cessing variables. The previous work (Jebalia et al., 2019) fo­
The pea composites had the morphology of protein ag­ cussed on the impact of morphological features on
gregates dispersed in a continuous and amorphous starch composite mechanical properties but did not really deal with
matrix, regardless of raw material and moisture content the relations between extrusion processing and morpholo­
(supplementary material, Fig. S.5, snapshots of CLSM image), gical features.
as described by Jebalia et al. (2019). Depending on SME, the
protein aggregates varied greatly in size (D50 = 10–50 µm) and
protein area ratio (0.15–0.30) and exhibited various shapes 3.2. Viscous behaviour
(spherical, ellipsoidal, fibrillar) (see Table 3). At the sig­
nificance level of 0.05, protein aggregates of pea flour-based Viscosity of pea flour and SP 2/1 decreased with increasing
composites are significantly smaller than SP 2/1-based shear rate, thus manifesting shear-thinning behaviour. The
composites (D50 11–16.5 vs 27–40 µm). This difference can be flow curves can be fitted by the power-law model (Eq. 6),
likely explained by different morphologies of native mate­ regardless of T, MC and SME conditions (Fig. 4 and supple­
rials. Indeed, the native SP blends presented assembled mentary material, Fig. S.6). The SME had little impact on the
proteins due to the wet process extraction of PPI (Jebalia shear viscosity, even under the lowest temperature and
et al., 2019), whereas native PF proteins are distributed moisture content (see supplementary material, Fig. S.6).
around starch granules according to smaller domains. In Statistical analysis, conducted at the significance level of
contrast to D50, a significant difference in protein area ratio 0.05, proved that at the same shear rate, there is no sig­
(PAR) was observed regarding the SME level of composites nificant difference between the viscosity of SP 2/1 melt is­
manufacturing. At the same SME level, regardless of raw sued of different SME (supplementary material, Table S4a,b).
material, PAR is similar. This result may appear as surprising regarding the me­
These morphological features were correlated with bio­ chanical energy sensitivity on molten starch viscosity (Della
polymer structural changes, as shown by their correlation Valle et al., 1996). It can be attributed to the protective effect
with ΔWSIsolids and protein solubility in DTE (Fig. 3a,b). The exerted by other constituents (fibres, protein aggregates) on
protein area ratio is negatively and linearly correlated with starch. Therefore, the rheological model was built without
the gain in WSIsolids (Fig. 3a, R2 =0.95) and, thus, with starch taking SME into account.
destructuration. By favouring the swelling of amorphous As shown by the flow curves of pea flour processed with a
starch, the increase of thermo-mechanical energy (T, SME) moisture content of 35%, a noticeable decrease in viscosity
could enlarge the volume (surface area on microscopy was observed by increasing the temperature from 105° to
images) of the continuous starch phase to the detriment of 145°C, i.e., Tm, Tm+20, Tm+40 °C (Fig. 4a). Viscosity varied
that of dispersed protein aggregates. This result is in little when the product was heated from 85° to 105°C (Tm–20,
Food and Bioproducts Processing 135 (2022) 190–204 199

Fig. 4 – Flow curves of pea flour. (a) Flow curves at different temperatures at constant MC (35%); (b) Master curves for different
MC at T0 = 105 °C; (c) Comparison of pea flour and SP2/1 master curves at MC0 = 30% and T0 = 105 °C; (d) Comparison with pea
starch flow curves (MC = 30%, T = 115 °C).

The curves stand for data fitting using the power function.
Table 4 – Parameters of viscosity model (Eq. 7a, 11).
The sensitivity of the MC effect (21%, 25%, 30%, and 35%)
Coefficients Pea ingredients
was therefore studied at the same reference temperature
Pea flour SP 2/1 (105 °C) for pea flour (Fig. 4b). The moisture content had a
negative effect on the viscosity, as expected, and suggested
K0 (Pa.sn) 8.05E+05 1.39E+06
that water acts as a plasticizer. By adapting the time-tem­
E/R (K) 4 210 4 237
α 11.4 14.6 perature superposition principle to the plasticizer (water)
n0 0.288 0.292 content, the master curves obtained at four MC values were
shifted to build a final master curve at a reference MC of 30%
(Fig. S7b). The fitting yielded the values of the parameters of
Tm). This result was also observed for other moisture con­ the rheological model: K0 and n. The variation of shift factor
tents. It can be explained by the competition, in the interval aMC with MC followed exponential functions (R2 = 0.77 and
[Tm–20, Tm °C], between the decrease of the volume fraction 0.86 for SP 2/1 and pea flour, respectively (Fig. S.7b), allowing
of solid particles (fibre, protein aggregates, residual starch us to determine the water plasticizing coefficient α. The va­
granules) and the increase of molten starch macromolecules lues of the viscosity model parameters (K0, n, α and E/R) for
in the continuous phase. pea products are reported in Table 4. Pea flour and SP 2/1 in
For each moisture content (21 < MC < 35%, on a wet basis), the molten state had similar values of thermal sensitivity E/R
a master curve was obtained from flow curves measured at (4210 K vs. 4240 K), water sensitivity α (11.4 vs. 14.6), and flow
different temperatures. Here, the viscosities (η) were ex­ behaviour index n (0.29), respectively, and their final master
pressed as a function of the reduced shear rate (aT× ) using curves were very similar to each other (Fig. 4c). Statistical
the time-temperature superposition principle (Krauklis et al., analysis, conducted at the significance level of 0.05, proved
2019) in order to more accurately determine the parameters that at the same shear rate, there is no significant difference
of the viscosity model. This is illustrated for pea flour by the between the viscosity of pea flour and SP 2/1 melts. (sup­
master curves built at reference temperature (T0 = 105 °C) plementary material, Table S5). This result suggests that the
(supplementary material, Fig. S.7a). The changes in shift viscous behaviour of these materials depends little on the
factor aT with 1/T can be fitted by an exponential function (R2 protein and fibre content. To obtain more insight into the
≈ 0.75, for pea flour and SP 2/1 (Fig. S.7b), leading to a para­ effect of ingredients on the viscous behaviour of pulses, the
meter called thermal sensitivity E/R. flow curves of SP 2/1 and pea flour were compared to those of
200 Food and Bioproducts Processing 135 (2022) 190–204

Fig. 5 – Comparison of extrusion simulation results with measured ones for SME (a) and melt temperature at die entrance (b).
For composites, the melt temperature was studied at three levels of last barrel temperature (Tb): Tm–20 °C (□,○), Tm °C ( , )
and Tm+ 20 °C (■, •), where rectangular and circle symbols stand for pea flour and SP 2/1, respectively. The grey dashed lines
indicate the equity between measured and predicted variables, while the solid lines stand for the linear correlation between
them: (a) SME_com= 0.657 ×SME_mes; R2 = 0.87; (b) Snacks: T_com= 1.031 ×T_mes; R2 = 0.90; Composites (pea flour):
Tb= Tm–20; T_com= 0.57 ×T_mes+ 49.3; R2 = 0.41; TbTm; T_com= 0.79 ×T_mes+ 31.8; ]R2 = 0.92; TbTm+ 20;
T_com= 0.69 ×T_mes+ 50; ]R2 = 0.96. Composites (SP 2/1): Tb= Tm–20; T_com= 1.46 ×T_mes–39.9; R2 = 0.80; TbTm;
T_com= 1.58 ×T_mes–43.4; ]R2 = 0.92; TbTm+ 20; T_com= 1.27 ×T_mes–7.23; ]R2 = 0.99.

pea starch found in the literature (Barron et al., 2002; Logié, For most simulations of composite extrusion, the predicted
2017). Pea starch viscosity, measured using the same rhe­ temperature at die entrance correlates well with the mea­
ometer (Rheoplast®) at similar conditions (MC = 30%, sured temperature, regardless of moisture content for each
T = 115 °C, SME < 250 kJ/kg), is only slightly higher than that of material and each barrel temperature (Tb) (R2 > 0.80, except
pea flour and SP 2/1 (Fig. 4d). Statistical analysis, conducted for pea flour at Tb = Tm–20 °C) (Fig. 5b). For snack extrusion
at the significance level of 0.05, proved that there is no sig­ simulation, a good agreement between the computed and
nificant difference between the viscosity of pea starch, pea measured melt temperature was found regardless of extru­
flour, and SP 2/1 melts at a shear rate higher than 95 1/s sion conditions and die geometry (R2 = 0.90, Fig. 5b). Contrary
(supplementary material, Table S5). These results are in to composites, the snack extrusion simulations revealed a
agreement with the finding of Chanvrier et al. (2015), unique correlation between T_com and T_mes because all
showing that molten maize flour, starch/zein blend (85/15%) trials were performed using the same barrel temperature
and maize starch exhibited shear viscosities very similar to (Tb). The predicted temperature, in most cases, is slightly
each other. However, the pea starch processed at higher SME overestimated, which may be partly due to the inaccuracy of
(1000 kJ/kg) exhibited lower viscosity than pea ingredients the measurements. Indeed, using a temperature sensor
processed at lower SME (< 250 kJ/kg) (Fig. 4d). For the whole flush-mounted at the wall of the die entrance may result in a
shear rate, the viscosity of this ultra-processed pea starch is temperature that is different from the actual temperature in
significantly different from other materials at the sig­ the product core. The results for the legume case confirm the
nificance level of 0.05 (supplementary material, Table S5). proven capacity of a 1D global extrusion model (Ludovic®) to
Higher starch depolymerisation can lower system viscosity simulate extrusion processing of starchy products (Berzin
(Della Valle et al., 1996). et al., 2010; Della Valle et al., 1993; Robin et al., 2010).
In conclusion, regardless of the minor fractions (lipids, The variations in structural changes of biopolymers were
proteins, fibre), the shear viscosity of pulses and their bio­ then correlated with the melt temperature and SME. They
polymer blends in the molten state is governed by the be­ were calculated using the extrusion model (Fig. 6), in agree­
haviour of the starch phase and its transformation. ment with experimental results (Fig. 2). Regarding starch
transformation, an acceptable correlation was obtained be­
3.3. Extrusion simulation tween the gain in the water solubility index of solids
(ΔWSIsolids) and the predicted SME: R2 = 0.77 and 0.82, for
First, the predicted extrusion variables were compared with composites and snacks, respectively (Fig. 6a). Similar to ex­
values measured during experiments. A good and unique perimental results, ΔWSIsolids was better correlated to pre­
correlation between predicted and measured SME was ob­ dicted SME than predicted temperature (results not shown).
tained (R2 = 0.87), regardless of extrusion system, formula­ It should be kept in mind that the intensity of starch de­
tion and operating conditions, although the computed SME structuration, assessed by WSIstarch can be deduced from
was slightly underestimated (Fig. 5a). This slight difference is WSIsolids through the strong correlation between them (sup­
probably because the Ludovic® flow model only considers the plementary material, Fig. S.3).
melting and viscous dissipation energies, and neglects fric­ Concerning protein transformation, the gain in DTE-so­
tion (inter-particle, particle-metal, metal-metal). luble proteins (ΔP_DTE) was positively correlated with the
Food and Bioproducts Processing 135 (2022) 190–204 201

Fig. 6 – Variations of structural characteristics of starch: ΔWSIsolids (a) and proteins: ΔP_DTE (b), with computed temperature
and SME. The curves correspond to the data fitting: (a) Pea snacks: ΔWSIsolids = 27/(1 + exp(−0.0359 ×(SME_com − 294))); R2
= 0.82; Pea composites: ΔWSIsolids = 0.015 ×SME_com+ 2.53; R2 = 0.77. (b) ΔP_DTE = 101/(1 + exp(−0.0373 × (T_com − 194)));
R2 = 0.81.

Fig. 7 – 2D contour plots of extruder operating chart. Study case: Manufacturing of pea snacks using a Coperion Werner &
Pfleiderer ZSK 26Mc extruder equipped with a circular die. Pea flour was extruded at MC = 21% and Q = 20 kg/h. The hatched
area indicates the range of extrusion conditions (N, Tb) leading to predicted melt temperature (155 < T_com < 162.5 °C) and
specific mechanical energy (425 < SME_com < 625 kJ/kg) in order to achieve the goals: ΔP_DTE= 20 ± 5% and ΔWSIsolids
= 27.5 ± 2.5%.

computed temperature (R2 = 0.81, Fig. 6b), in accordance with predicted by temperature than SME (result not shown).
the experimental results. Indeed, ΔP_DTE, indicating the Hence, extrusion can be simulated for predicting extruded
amount of protein cross-linked by S-S bonds, was better product structure from the knowledge of extrusion variables.
202 Food and Bioproducts Processing 135 (2022) 190–204

In the last step, a sensitivity study of the effects of ex­ as a tool in designing extruded foods with targeted struc­
trusion conditions (screw speed, flow rate, barrel tempera­ tures, and opens the way for the prediction of the structural
ture) on predicted processing variables was conducted features of extruded pulse crops. In the future, a mechanistic
through extrusion simulation. As an example of a study case, model for predicting the variation of melt microstructure
pea flour was processed using a pre-pilot extruder equipped along the extruder screw and die will be established through
with a circular die at MC = 21% and Q = 20 kg/h. It resulted in an in-depth study of the local interfacial stress and local
2D contour plots, expressing computed extrusion variables, structural changes of starch and proteins along the screws
i.e., melt temperature at die entrance T_com and specific and die. Complete characterisations of extruded products
mechanical energy SME_com, as a function of two extrusion issued of dead-stop operation will be performed. Further
operating parameters (i.e., screw speed N and last barrel improvements involve determining the cellular structure of
temperature Tb). The product temperature at die entrance expanded snacks from the knowledge of morphology, rheo­
(T_com) increased from 130° to 178°C and SME_com from 448 logical behaviour and extrusion variables to predict their
to 624 kJ/kg when barrel temperature and screw speed in­ texture.
creased from 90° to 145°C and 200–700 rpm, respectively. The
higher value of T_com compared to Tb was indicative of the Declaration of Competing Interest
importance of viscous dissipation (Fig. 7). As an example of
these plot applications, let us consider the manufacturing of The authors declare that they have no known competing fi­
expanded pea snacks with the following characteristics: nancial interests or personal relationships that could have
maximum ΔWSIsolids (27.5 ± 2.5%) and an average gain in appeared to influence the work reported in this paper.
protein solubility in DTE = 20 ± 5%. Note that, in turn, this
gain in protein solubility in DTE can lead to a composite (cell
Acknowledgements
wall) with a morphology characterised by an average of D50
of protein aggregates of ≈ 20 µm (Fig. 3b). On the basis of the
We are thankful to Joëlle Bonicel and Valérie Micard (INRAE-
correlations between extrusion variables (T_com, SME_com)
IATE) for protein SE-HPLC analysis and protein solubility
and transformation (Fig. 6), it follows that this product can be
determination, and to Mathieu Bretesche (INAE-BIA) for
obtained by processing pea flour at T_com = 155–162.5 °C and
rheological measurements. We are grateful to Azad Emin
SME_com = 425–625 kJ/kg. The hatched area in Fig. 7 gives the
(KIT) for pea snack extrusion, and to the Pays de La Loire
particular combinations of N and Tb leading to desired melt
Region (France) for financially supporting Imen Jebalia’s PhD
temperature and SME. The same rationale can be applied to
thesis.
other extrusion parameters (MC, Q) to deliver a complete set
of operating conditions.
Using the present extrusion database with an appropriate
Appendix A. Supporting information
viscosity model of the feeds, the extrusion simulation can
Supplementary data associated with this article can be found
help us to predict the transformation of biopolymers and the
in the online version at doi:10.1016/j.fbp.2022.08.001.
structure of extruded foods.

4. Conclusions References

This study has shown that rheology and process simulation Alonso, R., Orue, E., Zabalza, M.J., Grant, G., Marzo, F., 2000. Effect
of extrusion cooking on structure and functional properties of
could help to design pulse-based extruded foods. The struc­
pea and kidney bean proteins. J. Sci. Food Agricult. 80 (3),
tural changes in pea biopolymers, e.g., starch depolymer­
397–403.
isation and protein aggregation that occur during the Barron, C., Della Valle, G., Colonna, P., Vergnes, B., 2002. Energy
extrusion process, were assessed by solubility indices. These balance of low hydrated starches transition under shear.
changes govern the morphological features of pea-based J. Food Sci. 67 (4), 1426–1437.
starch-protein composites, such as protein area ratio and Beck, S.M., Knoerzer, K., Foerster, M., Mayo, S., Philipp, C., Arcot,
median width of protein aggregates. These results suggest J., 2018. Low moisture extrusion of pea protein and pea fibre
fortified rice starch blends. J. Food Eng. 231, 61–71. https://doi.
that thermal and shear energies work synergistically to
org/10.1016/j.jfoodeng.2018.03.004
perform starch transformation, whereas thermal energy Beck, S.M., Knoerzer, K., Sellahewa, J., Emin, M.A., Arcot, J., 2017.
probably controls protein cross-linking by forming S-S bonds. Effect of different heat-treatment times and applied shear on
These structural changes can therefore be tuned by the secondary structure, molecular weight distribution, solubility
proper setting of extrusion variables such as melt tempera­ and rheological properties of pea protein isolate as in­
ture (T) and SME. These extrusion variables can be predicted vestigated by capillary rheometry. J. Food Eng. 208, 66–76.
by a 1D global model of co-rotating twin-screw extrusion Berrios, J.J., Tang, J., & Swanson, B.G. (2008). Extruded legumes
(Patent No. US 2008/0145483 A1). The United States of
implemented in simulation software (Ludovic®). One of the
America.
model inputs, the viscous behaviour of pea flour and starch/ Berzin, F., Tara, A., Tighzert, L., Vergnes, B., 2010. Importance of
protein blend, was determined using a capillary pre-shearing coupling between specific energy and viscosity in the mod­
rheometer under conditions similar to extrusion. The visc­ eling of twin-screw extrusion of starchy products.
osity of the molten hydrated protein-starch blend and pea Polym. Eng. Sci. 50 (9), 1758–1766. https://doi.org/10.1002/pen.
flour was governed by the starch viscosity, regardless of the 21702
Boye, J., Zare, F., Pletch, A., 2010. Pulse proteins: processing,
minority fraction (proteins, fibres) present. The extrusion
characterization, functional properties and applications in
simulations allowed a fair prediction of extrusion variables (T
food and feed. Food Res. Int. 43 (2), 414–431.
and SME) from extrusion parameters (feed rate, screw speed, Brennan, M.A., Derbyshire, E., Tiwari, B.K., Brennan, C.S., 2013.
barrel temperature). This study reveals how extrusion si­ Ready-to-eat snack products: the role of extrusion technology
mulation empowered with a rheological model can be used in developing consumer acceptable and nutritious snacks. Int.
Food and Bioproducts Processing 135 (2022) 190–204 203

J. Food Sci. Technol. 48 (5), 893–902. https://doi.org/10.1111/ Horvat, M., Emin, M.A., Hochstein, B., Willenbacher, N.,
ijfs.12055 Schuchmann, H.P., 2013. A multiple-step slit die rheometer for
Camire, M.E., 1991. Protein functionality modification by extru­ rheological characterization of extruded starch melts. J. Food
sion cooking. J. Am. Oil Chem. Soc. 68 (3), 200–205. https://doi. Eng. 116 (2), 398–403.
org/10.1007/BF02657770 Jebalia, I., Maigret, J.-E., Réguerre, A.-L., Novales, B., Guessasma,
Carbonaro, M., Maselli, P., Nucara, A., 2012. Relationship between S., Lourdin, D., Della Valle, G., Kristiawan, M., 2019.
digestibility and secondary structure of raw and thermally Morphology and mechanical behaviour of pea-based starch-
treated legume proteins: a Fourier transform infrared (FT-IR) protein composites obtained by extrusion. Carbohydr. Polym.
spectroscopic study. Amino Acids 43 (2), 911–921. https://doi. 223, 115086.
org/10.1007/s00726-011-1151-4 Krauklis, A.E., Akulichev, A.G., Gagani, A.I., Echtermeyer, A.T.,
Chanvrier, H., Chaunier, L., Della Valle, G., Lourdin, D., 2015. Flow 2019. Time-temperature-plasticization superposition prin­
and foam properties of extruded maize flour and its biopo­ ciple: Predicting creep of a plasticized epoxy. Polymers
lymer blends expanded by microwave. Food Res. Int. 76, ((Basel)) 11 (11), 1848.
567–575.〈http://www.sciencedirect.com/science/article/pii/ Kristiawan, M., Della Valle, G., 2020. Chapter 6 - Transport phe­
S0963996915301083〉. nomena and material changes during extrusion. In: Ganjyal,
Choi, Y., Okos, M.R., 1986. Effects of temperature and composi­ G.M. (Ed.), Extrusion Cooking: Cereal Grains Processing.
tion on the thermal properties of foods. In: Maguer, M.L., Jelen, Woodhead Publishing, pp. 179–204. https://doi.org/10.1016/
P. (Eds.), Food engineering and process applications, Vol. 1. B978-0-12-815360-4.00006-7
Transport phenomena. Elsevier Applied Science Publishers, Kristiawan, M., Guessasma, S., Della Valle, G., 2020. Chapter 10 -
pp. 93–101. Extrusion cooking modeling, control, and optimization. In:
Day, L., Swanson, B.G., 2013. Functionality of protein-fortified Ganjyal, G.M. (Ed.), Extrusion Cooking: Cereal Grains
extrudates. Comp. Rev. Food Sci. Food Saf. 12 (5), 546–564. Processing. Woodhead Publishing, pp. 295–330. https://doi.
https://doi.org/10.1111/1541-4337.12023 org/10.1016/B978-0-12-815360-4.00010-9
Della Valle, G., Barrès, C., Plewa, J., Tayeb, J., Vergnes, B., 1993. Kristiawan, M., Micard, V., Maladira, P., Alchamieh, C., Maigret, J.-
Computer simulation of starchy products’ transformation by E., Réguerre, A.-L., Emin, M.A., Della Valle, G., 2018. Multi-scale
twin-screw extrusion. J. Food Eng. 19 (1), 1–31. https://doi.org/ structural changes of starch and proteins during pea flour
10.1016/0260-8774(93)90059-S extrusion. Food Res. Int. 108, 203–215. https://doi.org/10.1016/
Della Valle, G., Colonna, P., Patria, A., Vergnes, B., 1996. Influence j.foodres.2018.03.027
of amylose content on the viscous behavior of low hydrated Laleg, K., Cassan, D., Barron, C., Prabhasankar, P., Micard, V.,
molten starches. J. Rheol. 40 (3), 347–362. https://doi.org/10. 2016. Structural, culinary, nutritional and anti-nutritional
1122/1.550747 properties of high protein, gluten Free, 100% legume Pasta.
Della Valle, G., Quillien, L., Gueguen, J., 1994. Relationships be­ Plos One 11 (9), 1–19.
tween processing conditions and starch and protein mod­ Li, M., Hasjim, J., Xie, F., Halley, P.J., Gilbert, R.G., 2014. Shear
ifications during extrusion-cooking of pea flour. degradation of molecular, crystalline, and granular structures
J. Sci. Food Agricult. 64 (4), 509–517. https://doi.org/10.1002/ of starch during extrusion. Starch - Stärke 66 (7–8), 595–605.
jsfa.2740640418 https://doi.org/10.1002/star.201300201
Della Valle, G., Vergnes, B., 1994. Propriétés thermophysiques et Logié, N. (2017). Modifications structurales et comportement
rhéologiques des substrats utilisés en cuisson-extrusion. In: rhéologique d′amidons faiblement hydratés sous traitement
Colonna, P., Della Valle, G. (Eds.), La Cuisson-Extrusion. thermomécanique. In Thèse de doctorat en Molécules et
Lavoisier Tec & Doc, pp. 440–467. Matière Condensée. PhD Thesis, Université de Lille 1, France.
Ek, P., Ganjyal, G.M., 2020. Chapter 1 - Basics of extrusion pro­ Luo, S., Chan, E., Masatcioglu, M.T., Erkinbaev, C., Paliwal, J.,
cessing. In: Ganjyal, G.M. (Ed.), Extrusion Cooking: Cereal Koksel, F., 2020. Effects of extrusion conditions and nitrogen
Grains Processing. Woodhead Publishing, pp. 1–28. https://doi. injection on physical, mechanical, and microstructural prop­
org/10.1016/B978-0-12-815360-4.00001-8 erties of red lentil puffed snacks. Food Bioprod. Process. 121,
Ek, P., Kowalski, R.J., Ganjyal, G.M., 2020. Chapter 4 - Raw material 143–153. https://doi.org/10.1016/j.fbp.2020.02.002
behaviors in extrusion processing I (Carbohydrates). In: Philipp, C., Emin, M.A., Buckow, R., Silcock, P., Oey, I., 2018. Pea
Ganjyal, G.M. (Ed.), Extrusion Cooking: Cereal Grains protein-fortified extruded snacks: Linking melt viscosity and
Processing. Woodhead Publishing, pp. 119–152. https://doi. glass transition temperature with expansion behaviour.
org/10.1016/B978-0-12-815360-4.00004-3 J. Food Eng. 217, 93–100. https://doi.org/10.1016/j.jfoodeng.
Fang, Y., Zhang, B., Wei, Y., Li, S., 2013. Effects of specific me­ 2017.08.022
chanical energy on soy protein aggregation during extrusion Philipp, C., Oey, I., Silcock, P., Beck, S.M., Buckow, R., 2017. Impact
process studied by size exclusion chromatography coupled of protein content on physical and microstructural properties
with multi-angle laser light scattering. J. Food Eng. 115 (2), of extruded rice starch-pea protein snacks. J. Food Eng. 212,
220–225.〈http://www.sciencedirect.com/science/article/pii/ 165–173. https://doi.org/10.1016/j.jfoodeng.2017.05.024
S0260877412004736〉. Pommet, M., Redl, A., Morel, M.-H., Domenek, S., Guilbert, S.,
Félix-Medina, J.V., Montes-Ávila, J., Reyes-Moreno, C., Perales- 2003. Thermoplastic processing of protein-based bioplastics:
Sánchez, J.X.K., Gómez-Favela, M.A., Aguilar-Palazuelos, E., chemical engineering aspects of mixing, extrusion and hot
Gutiérrez-Dorado, R., 2020. Second-generation snacks with molding. Macromol. Symposia 197 (1), 207–218. https://doi.org/
high nutritional and antioxidant value produced by an opti­ 10.1002/masy.200350719
mized extrusion process from corn/common bean flours Rangira, I., Gu, B.-J., Ek, P., Ganjyal, G.M., 2020. Pea starch exhibits
mixtures. LWT 124, 109172. https://doi.org/10.1016/j.lwt.2020. good expansion characteristics under relatively lower tem­
109172 peratures during extrusion cooking. J. Food Sci. 85 (10),
Guessasma, S., Chaunier, L., Della Valle, G., Lourdin, D., 2011. 3333–3344. https://doi.org/10.1111/1750-3841.15450
Mechanical modelling of cereal solid foods. Trends Food Robin, F., Engmann, J., Pineau, N., Chanvrier, H., Bovet, N., Valle,
Sci. Technol. 22 (4), 142–153. https://doi.org/10.1016/j.tifs.2011. G.Della, 2010. Extrusion, structure and mechanical properties
01.005 of complex starchy foams. J. Food Eng. 98 (1), 19–27. https://
Habeych, E., Dekkers, B., van der Goot, A.J., Boom, R., 2008. doi.org/10.1016/j.jfoodeng.2009.11.016
Starch–zein blends formed by shear flow. Chem. Eng. Sci. 63 Stanley, W.D., 1989. Protein reactions during extrusion proces­
(21), 5229–5238. https://doi.org/10.1016/j.ces.2008.07.008 sing. In: Mercier, C., Linko, P., Harper, J.M. (Eds.), Extrusion
204 Food and Bioproducts Processing 135 (2022) 190–204

cooking. American Association of Cereal Chemists, Inc., pp. Polym. Eng. Sci. 38 (11), 1781–1792. https://doi.org/10.1002/pen.
321–341. 10348
Sumargo, F., Gulati, P., Weier, S.A., Clarke, J., Rose, D.J., 2016. Vergnes, B., Villemaire, J.P., 1987. Rheological behaviour of low
Effects of processing moisture on the physical properties and moisture molten maize starch. Rheologica Acta 26 (6),
in vitro digestibility of starch and protein in extruded brown 570–576. https://doi.org/10.1007/BF01333742
rice and pinto bean composite flours. Food Chem. 211, Xie, F., Halley, P.J., Avérous, L., 2012. Rheology to understand and
726–733. https://doi.org/10.1016/j.foodchem.2016.05.097 optimize processibility, structures and properties of starch
Vergnes, B., Della Valle, G., Delamare, L., 1998. A global computer polymeric materials. Prog. Polym. Sci. 37 (4), 595–623. https://
software for polymer flows in corotating twin screw extruders. doi.org/10.1016/j.progpolymsci.2011.07.002

You might also like