CETvol87


 
 

 

                                                                    DOI: 10.3303/CET2187035 
 

 
 

 
 
 
 

 
 
 
 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Paper Received: 25 August 2020; Revised: 17 February 2021; Accepted: 12 April 2021 
Please cite this article as: Fanari F., Naue I.F., Desogus F., Grosso M., Wilhelm M., 2021, Durum Wheat Dough Torque Measurements: 
Characterization and Study of the Mixing Process Parameters as a Function of Water and Salt Amounts, Chemical Engineering Transactions, 
87, 205-210  DOI:10.3303/CET2187035 

 CHEMICAL ENGINEERING TRANSACTIONS 
VOL. 87, 2021 

A publication of 

The Italian Association 
of Chemical Engineering 
Online at www.cetjournal.it 

Guest Editors: Laura Piazza, Mauro Moresi, Francesco Donsì
Copyright © 2021, AIDIC Servizi S.r.l. 
ISBN 978-88-95608-85-3; ISSN 2283-9216

Durum Wheat Dough Torque Measurements: 
Characterization and Study of the Mixing Process Parameters 

as a Function of Water and Salt Amounts 

Fabio Fanaria, Ingo F. C. Naueb, Francesco Desogusa, Massimiliano Grossoa,*, 
Manfred Wilhelmb 
a
Department of Mechanical, Chemical and Materials Engineering, University of Cagliari, 09123 Cagliari, Italy 

b
Institute for Chemical Technology and Polymer Chemistry, Karlsruhe Institute of Technology (KIT), 76131 Karlsruhe, 

Germany 
massimiliano.grosso@dimcm.unica.it 

Different physicochemical phenomena are involved in breadmaking, which produce perceptible changes in the 
rheological properties of dough. In this regard, the kneading phase is crucial for the assessment of the quality 
of the final product. Indeed, the determination of the optimal time in the mixing process allows to optimize the 
rheological properties of the dough and the energy consumption. In this work, torque measurements were 
carried out on durum wheat commercial semolina using a Brabender Mixer 350 (Brabender® GmbH, 
Duisburg, D) where semolina was mixed with water in different amounts (from 40% to 60% of semolina 
weight). Each sample was mixed for 20 minutes. The salt effect was instead investigated by adding a variable 
amount of salt (from 0% to 2.5% of semolina weight). The optimal mixing time value, which occurs at the 
maximum torque value, was determined and studied with respect to salt and water quantity. Additionally, 
spectrograms of the torque time series data were computed in order to provide the magnitude (and frequency) 
of the torque oscillations with time. The results showed a significant influence of water amount on the torque 
maximum value and on the mixing time, with the latter that linearly depends on water amount. Also, the 
amplitude oscillations resulted to be related to the water amount, as their amplitude decreases when the water 
amount increases. On the other hand, the salt effect is less evident but an influence on optimal mixing time is 
however detected. 

1. Introduction

The kneading process is crucial for the establishment of the rheological properties of dough since it promotes 
the gluten network development in addition to allowing the mixing of the ingredients and the formation of a 
homogeneous and coherent mass (Fanari et al., 2020a). In mixing and kneading, the dough is stretched 
through the application of mechanical energy (Zheng et al., 2000). Such energy allows different 
conformational arrangements of the key biopolymers (particularly gluten proteins) which are present in the 
system and promotes interactions among the constituents (proteins, starch, and water) of the dough (Fanari et 
al., 2020b), leading to a three-dimensional gluten network structure, in which the starch globules are 
embedded (Park et al., 2009). This network is responsible for the viscoelastic properties of the dough and 
determines its machining properties and the final quality of the product (Contamine et al., 1995). Mixing time 
significantly affects the structures of the gluten networks in wheat doughs, which in turn can influence the 
dough baking quality. Cappelli et al. (Cappelli et al., 2020) recently wrote a review about the kneading process 
in breadmaking. In this work, they state that the identification, study, and comprehension of the key process 
parameters are the first steps to achieve in order to improve the quality of the final product. 
To study the kneading process and related changes in dough rheological properties, torque measurements 
can be used to quantify the strength opposed by the dough as a function of time. These measurements are 
empirical and follow the same basic principle, that is, the resistance opposed by the dough at the kneader 
shaft is transferred to a dynamometer and recorded. Torque curves provide information about the rheological 

205



characteristics of dough on which the quality of the flour is usually evaluated.  Additionally, from this curve, it is 
possible to find other interesting parameters, depending on the device used for the measurement: time of 
formation or development of the dough, time of stability, maximum degree of softening of the dough, dough 
elasticity, hydration capacity, and flour strength (Miś et al., 2012). These parameters are useful to optimize 
formulations for baking quality especially in terms of determining the optimum water amount for a specific kind 
of flour, optimizing the time of mixing to obtain a well-developed dough with optimal rheology and gas holding 
capacity (Yazar et al., 2016). Optimal mixing time is the time required for the optimal development of the 
dough, which occurs when the gluten fibrils spread into a 3D network encompassing the whole dough mass. 
This value is determined as the maximum of the torque curve, which also corresponds to the maximum dough 
strength (Watanabe et al., 1992). It is commonly assumed that the mechanical work input (i.e. mixing energy) 
is the most significant factor affecting the optimal dough development, together with the mixing time length 
(Baudouin et al., 2020). It was observed that an increase in the protein amount leads, in general, to a 
decrease in the optimal mixing time (Munteanu et al., 2016). Depending on the strength of the flour, the curve 
can present a maximum defined as a peak or as a plateau. For times longer than the optimal mixing time, a 
decrease in the mixing curve is usually observed. This feature is due to the breakdown of the gluten protein 
network. The network is held together by secondary bonding interactions which are susceptible to breakdown 
when subjected to high strains (Amemiya & Menjivar, 1992) and, at the same time, the starch granules appear 
to get more space to move in the dough matrix (Li et al., 2020). In industrial bakery plants, a common practice 
is to adjust the mixing time according to the mixing speed in order to deliver the specified amount of 
mechanical energy (Baudouin et al., 2020). If the aim is to improve dough rheology and resulting bread 
characteristics, the most important requirements are to select the best kneading machine, to manage the 
addition of water (quantity and temperature), and to select the most suitable improvers (organic acid, NaCl 
substitute, enzymes, hydrocolloids, and/or emulsifiers) (Cappelli et al., 2020). This should be evaluated on a 
case-by-case basis, according to the kind of product at issue. 
To summarize, it is possible to state that the quality of bread and bakery products dough strongly depends on 
the mixing conditions (mixer type, rotation speed, and mixing time), on the and water content, on the 
characteristics of the used flour (Ortolan & Steel, 2017) and, to a lesser extent, on the other ingredients and 
additives mixed in the dough (Fanari et al., 2019a; 2019b). The aim of the present study is to characterize the 
dough during mixing by torque measurements and, subsequently, to address empirical modelling of the 
mechanical properties varying during the process. The study is focused on durum wheat commercial semolina 
used in the production of a typical Sardinian bread named “Pani Carasau”. 

2. Materials and methods

Each dough sample was prepared by mixing 300 g of commercial semolina with distilled water and salt (when 
added) in a Brabender Measuring Mixer 350 (Brabender® GmbH, Duisburg, D) for 20 min, using a rotational 
mixing speed of 20 rpm. The value of 20 rpm was chosen to avoid important structure breaks in the dough 
structure. In Table 1 the properties of the semolina used in the work are reported. To study the water effect on 
the dough, semolina was mixed with water in different amounts (40%, 45%, 50%, 55%, and 60% of semolina 
weight).  

Table 1: Properties of semolina used in this study 

Proteins (%) Carbohydrates (%) Fats (%) 
Commercial Semolina 12.0 69.0 (0.8 are sugars) 0.8 

To study the salt effect, on the other side, the dough was prepared using an amount of water equal to 50% of 
the semolina weight and a variable amount of salt (0.5%, 1.0%, 1.5%, 2.0%, and 2.5% of the semolina 
weight). These samples were compared with W50 sample taken as reference sample (50% water and 0% 
salt). Table 2 reports the mass percentage of ingredients with respect to the semolina weight of the samples 
and their identification acronym. For each sample, three replicate measurements were performed, and the 
mean value was considered for the study. 
The torque curves obtained by the measurements were filtered with a wavelet decomposition using order 5 
Daubechies wavelets (Daubechies, 1992). The optimal mixing time value was evaluated as the time 
corresponding to the maximum torque value, finding the maximum of the smoothed curve for each sample. 
These values were analysed as a function of water and salt content to quantify their impact.  
Additionally, spectrograms were carried out on the residuals of the time series data obtained as the 
differences between original raw data and the filtered ones. The spectrogram is a visual representation of the 
time evolution of the signal power spectrum. A Fourier Transform is applied on finite windows of the signal, 

206



deriving a time-localized estimation of amplitude and frequency. For details on the method see, for example, 
the work by Flandrin (1989). With this procedure, it is then possible to evaluate the amplitude of the 
oscillations and the related frequencies as a function of time.  

Table 2: Samples percentage composition referred to semolina weight 

Water (%) Salt (%) 
W40 40.0 0.0 
W45 45.0 0.0 
W50 50.0 0.0 
W55 55.0 0.0 
W60 60.0 0.0 
S0.5 50.0 0.5 
S1.0 50.0 1.0 
S1.5 50.0 1.5 
S2.0 50.0 2.0 
S2.5 50.0 2.5 

3. Results

Figure 1 shows the comparison between torque curves and smoothed curves for samples with different water 
amount (Figure 1a) and samples with the same water amount, but different amount of salt (Figure 1b). 

Figure 1: Torque curves comparison among samples with different water (smoothed curves are represented 
as black lines) (a) and salt amounts (smoothed curves are represented as reported in the legend) (b) 

Data smoothing is necessary to address an average trend of the torque, removing the oscillations that 
obstacle the identification of the maximum torque points. Results are shown in Table 3, where the peak time 
and the corresponding values of maximum torque, calculated for the smoothed curves, are reported in 
columns 3 and 2, respectively. Figure 2 shows the parameters peak time and torque as a function of water 
content (Figures 2a and 2b) and the peak time as a function of salt amount (Figure 2c). The maximum 
strength (corresponding to the maximum torque value) decreases with the water amount, whereas the peak 
time increases. The peak time can be modelled with a linear dependence on the water content. Results of the 
linear regression are reported with the red line in Figure 2a (R

2
adj = 0.956, p-value = 3∙10

-3). Maximum torque 
also shows a linear dependence on the water content (R2adj = 0.973, p-value = 2∙10

-3) but with a negative 
slope (red line on Figure 2b). On the other hand, salt addition results in a linear increase of the peak time 
(R2adj = 0.987, p-value = 4∙10

-5) when added up to 2.5% (Figure 2c). Regarding torque values, the addition of 
salt produces a variable decrease, according to the added quantity and, in general, the magnitude of these 
effects is lower compared to water induced effects. 

207



Figure 2: Peak time (a) and maximum torque (b), calculated from the smoothed torque curves, as a function of 
water amount in the dough (percentages based on semolina weight); peak time (c) as a function of salt 
content; the red line is the linear regression model; in the tables above, the statistical parameters of the 
regression Standard error (Std Error), 95% Lower Confidence Limit (LCL) and Upper confidence Limit (UCL), 
and Adjusted R2 (Adj R.s.), are reported  

Table 3: Maximum torque, peak times, oscillation amplitude and their relative peak times 

Sample 
Max torque
(Nm) 

Corresponding 
peak time 
(min) 

Max oscillation  
amplitude peak 
(Nm) 

Corresponding 
peak time 
(min) 

W40 0.872 4.9 0.573 3.2 
W45 0.620 5.7 0.485 4.2 
W50 0.392 6.2 0.284 6.5 
W55 0.150 7.4 0.114 7.2 
W60 0.059 7.8 0.036 8.8 
S0.5 0.272 7.5 0.240 7.2 
S1.0 0.251 8.2 0.223 7.2 
S1.5 0.275 9.5 0.331 6.8 
S2.0 0.346 10.6 0.265 10.8 
S2.5 0.268 8.6 0.255 11.2 

Concerning the analysis of the oscillations, Figure 3a reports the residuals computed for the reference sample 
(W50). Such residuals are the differences between the experimental data and their projections onto the 
smoothed curve, thus giving an estimation of the actual oscillation that the torque is experiencing around its 
mobile average value. For the case shown here, one can see that oscillations are more pronounced at t ~ 7 
min (near the optimal mixing time) and then they slightly decrease with time. This behavior is more apparent in 
Figure 3b, where the maximum oscillation amplitudes, evaluated with the spectrogram at the maximum 
frequency, are reported as a function of time. Therefore, values of maximum oscillation amplitude and the 
corresponding peak time for all the samples are reported in Table 3 (columns 4-5). As it is possible to observe 
from data in Table 3, for some samples the maximum oscillation amplitude and maximum torque occur 
approximately at the same values reported in Figure 2. The time at which the maximum oscillations occur 
increases as water amount increases (p-value equal to 6∙10-4, R2adj = 0.982), as shown in Figure 4a. 
Oscillation amplitudes, on the contrary, decrease with the water amount in the dough, following a linear trend 
(p-value equal to 2∙10-3, R2adj = 0.964), as shown in Figure 4b. These results are analogue to those reported 
above for torque peaks and relative times. Regarding salt, on the other hand, there is no structure in the data, 
probably due to its less significant impact on the mixing parameters. Despite this, the peak time at which the 
maximum oscillation amplitudes occur still increases as a function of the salt amount in the dough, as shown 
in Figure 4c. A linear model reasonably describes the dependence (R2adj = 0.83, p-value = 7∙10

-3) thus 
confirming that maximum oscillation amplitude significantly increases with the salt amount. A visual inspection 
of the spectrograms seems to suggest that other peaks in maximum oscillation amplitude may occur at a time 
higher than the optimal mixing time (around 10-15 min depending on the sample). However, their exact 
quantification is not reliable, at least for the experimental data here considered. This second peak is more 
evident for samples with high amounts of water or salt which have a broader peak in the torque curve. 

208



Figure 3: Statistical residuals (a), maximum oscillation amplitudes (b), and oscillation spectrogram (c) as a 
function of time, for W50 sample 

Figure 4: Peak time (a) and relative maximum oscillations amplitude (b), as a function of water amount in the 
dough (percentages based on semolina weight); maximum oscillation corresponding time (c) as a function of 
salt content; the red line is the linear regression model; in the tables above, the statistical parameters of the 
regression Standard error (Std Error), 95% Lower Confidence Limit (LCL) and Upper confidence Limit (UCL), 
and Adjusted R2 (Adj R.s.), are reported  

4. Discussion and Conclusions

This works aimed to quantify the impact of salt and water addition on the mixing properties of dough, with 
particular regard to optimal mixing time and maximum dough strength. Water amount in the dough very 
significantly affects the optimal mixing time (it linearly increases as a function of water amount) and the 
maximum strength of dough (that linearly decreases with water amount). Increasing water amount in the 
dough increases the percentage of free water on it. This is likely responsible for its viscosity decrease and for 
the increase in mixing time due to the extension of the hydration process. The linear relationship which was 
found can be used to predict the dough mixing parameters starting with the water amount added to it. 
Moreover, looking at the oscillation amplitudes, it is possible to state that they decrease as water amount 
increases, due to the reduction of the dough strength, while the time at which they are higher increases, in 
analogy with torque peak time. Additionally, it was observed also another peak in amplitude oscillations at 
times higher than optimal mixing time. This probably can be explained as a consequence of structure break 

209



and rebuild following the dough overmixing. Concerning the salt effect, it is possible to state that its influence 
on peak time is significant while the influence on dough strength and on torque oscillations does not follow a 
specific trend and it is not very significant in most cases. When salt is added up to 2%, the peak time 
increases almost linearly. This is due to the ability of salt to slow the protein hydration process, as reported by 
McCann and Day (2013). This phenomenon is observable also from the homogenization of the amplitude 
oscillations that, for these samples, appear more concentrated around the optimal mixing time, so the dough 
seems to suffer the overmixing effects to a lesser extent. As a future work development, it would be interesting 
to test the same procedure with different kinds of flour and see if the modelling and the estimation of the 
parameters are applicable also in other cases. Moreover, the methodology could be applied to an industrial 
kneader to verify its potentiality in the process optimization, verifying if, and how, the laboratory procedure is 
applicable in the food industry. 

Acknowledgments 

This research was partly funded by Italian Government (Ministero dello Sviluppo Economico), Fondo per la 
Crescita Sostenibile – Sportello “Agrifood” PON I&C 2014-2020, Prog. n. F/200133/01-03/X45. 

References 

Amemiya J. I., Menjivar J. A., 1992, Comparison of small and large deformation measurements to 
characterize the rheology of wheat flour doughs, Journal of Food Engineering, 16(1–2), 91-108. 

Baudouin F., Nogueira T. L., van der Mijnsbrugge A., Frederix S., Redl A., Morel M.-H., 2020, 
Mechanochemical activation of gluten network development during dough mixing, Journal of Food 
Engineering, 283, 110035. 

Cappelli A., Bettaccini L., Cini E., 2020, The kneading process: a systematic review of the effects on dough 
rheology and resulting bread characteristics, including improvement strategies, Trends in Food Science 
and Technology, 104, 91-101. 

Contamine A. S., Abecassis J., Morel M.-H., Vergnes B., Verel A., 1995, Effect of mixing conditions on the 
quality of dough and biscuits, Cereal Chemistry, 72, 516-522. 

Daubechies I., 1992, Ten Lectures on Wavelets, Society for Industrial and Applied Mathematics, Philadelphia, 
USA. 

Fanari F., Desogus F., Scano E. A., Carboni G., Grosso M., 2020a, The effect of the relative amount of 
Ingredients on the rheological properties of semolina doughs, Sustainability, 12(7), 2705. 

Fanari, F., Carboni, G., Grosso, M., Desogus, F., 2020b, Thermal properties of semolina doughs with different 
relative amount of ingredients. Sustainability, 12(6), 2235. 

Fanari F., Desogus F., Scano E. A., Carboni G., Grosso M., 2019a, The rheological properties of semolina 
doughs: Influence of the relative amount of ingredients, Chemical Engineering Transactions, 76, 703-708. 

Fanari F., Carboni G., Grosso M., Desogus F., 2019b, Effect of the Relative Amount of Ingredients on the 
Thermal Properties of Semolina Doughs, Chemical Engineering Transactions, 76, 1207-1212. 

Flandrin P., 1999, Time-frequency/time scale analysis, Academic Press, San Diego, USA. 
Li F., Zhang Y., Guo B., Wei Y., Don C., Pan W., Zhang B., 2020, The mesoscopic structure in wheat flour 

dough development, Journal of Cereal Science, 95, 103087. 
McCann T. H., Day L., 2013, Effect of sodium chloride on gluten network formation, dough microstructure and 

rheology in relation to breadmaking, Journal of Cereal Science, 57(3), 444-452. 
Miś A., Grundas S., Dziki D., Laskowski J., 2012, Use of farinograph measurements for predicting 

extensograph traits of bread dough enriched with carob fibre and oat wholemeal, Journal of Food 
Engineering, 108(1), 1-12. 

Munteanu M.-G., Voicu G., Ungureanu N., Zăbavă B.-Ș., Ionescu M., Constantin G., Istrate I., 2016, Methods 
for determining the characteristics of flour and dough, 5th International Conference on Thermal Equipment, 
Renewable Energy and Rural Development, At Golden Sands, Bulgaria, 349-355. 

Ortolan F., Steel C. J., 2017, Protein characteristics that affect the quality of vital wheat gluten to be used in 
baking: a review, Comprehensive Reviews in Food Science and Food Safety, 16(3), 369-381.  

Park S. H., Wilson J. D., Seabourn B. W., 2009, Starch granule size distribution of hard red winter and hard 
red spring wheat: Its effects on mixing and breadmaking quality, Journal of Cereal Science, 49(1), 98-105. 

Watanabe E., Bell A. E., Brockway B. E., 1992, Rheological studies on wheat flour dough. II. Relationships 
between frequency, storage modulus and torque amplitude, Food Control, 3(2), 102-104. 

Yazar G., Duvarci O., Tavman S., Kokini J. L., 2016, Non-linear rheological properties of soft wheat flour 
dough at different stages of farinograph mixing, Applied Rheology, 26(5), 1-11. 

Zheng H., Morgenstern M. P., Campanella O. H., Larsen N. G., 2000, Rheological properties of dough during 
mechanical dough development, Journal of Cereal Science, 32(3), 293-306. 

210