Comparison of Old and New Dough Mixing Methods and their Utility in Predicting Bread Quality

J.M.C. Dang* (1) and M.L. Bason (2)

1 Perten Instruments of Australia Pty. Ltd., Macquarie Park NSW 2113, Australia
2 Perten Instruments AB, Hägersten, SE-12609, Sweden
*Corresponding author:

Summary of the paper presented at AGSA 2015, 65th Australian Cereal Chemistry Conference, 16-18 September 2015, Crowne Plaza Coogee Beach Hotel, Coogee NSW.


The need for relevant and timely flour quality information in modern flour mills and bakeries is an ongoing challenge for instrument producers and users. High-energy commercial dough mixers, now in common use in bread bakeries around the world, provide better process relevance for rapid bread making systems. Traditional techniques to measure dough rheology have many drawbacks - slow tests, poor process relevance, limited sample types, and results that are usually limited to either the mixing or end-use quality of bread flour. Recent methods from Perten Instruments (doughLAB [dL] and micro-doughLAB [mdL]) and Chopin (Mixolab) have addressed some of these concerns. This study compared the recently developed mixing methods and their capabilities in predicting bread quality.

Materials and Methods

Samples and treatments
61 wheat flours with diverse properties were obtained from various Australian flour mills (Allied Mills, Ben Furney Flour Mills, Laucke Flour Mills, Manildra Group, Millers Foods, Tasmanian Flour Mills and Weston Milling). Samples were stored in air tight containers in a cool (18°C) dry environment until analysis.

Sample analyses
The moisture content of each flour was determined by AACCI Method 44-15.02 (AACC International, 2014). Mixing properties were determined on the Mixolab (Chopin Technologies, Villeneuve la Garenne, France), dL and mdL (Perten Instruments Australia, Macquarie Park, Australia). For Mixolab tests, samples (50.00 ± 0.01 g, 14% moisture basis, adjusted for moisture content and to a constant dough mass of 75.00 g) were mixed in duplicate to optimum consistency according to the manufacturer's "Chopin+" protocol (Table 1, AACCI Method 54-60.01). Samples for the dL (300.0 ± 0.01 g) and mdL (4.00 ± 0.01 g) were mixed in duplicate to optimum consistency with the manufacturer's protocols at slow (63 rpm, AACCI Method 54-21.01) and fast (120 rpm, AACCI Method 54-70.01) speeds (Table 1). In addition, samples were tested in duplicate on the mdL using the manufacturer's cooking protocol (Table 1) using optimum water absorption achieved in the mixing tests. Extensograph testing was performed according to AACCI Method 54-10.01, in duplicate, on doughs prepared from the dL. Duplicate volume measurements were determined on duplicate pup loaves (250 g, rapid dough protocol), according to AACCI Method 10-05.01.

Statistical analyses
Mixing properties were determined from the Mixolab, dL and mdL mixing curves as defined in Figure 1, and compared to Extensograph and baking data by univariate (UVA) and multivariate (MVA) analyses. Univariate comparisons of mixing properties between methods were performed by regression analyses (Minitab ver. 13) on all samples (n = 61). Method precision was evaluated from one-way analysis of variance of data as the root mean square of the error term (RMS) and the coefficient of variation (CV). Torque/time spectra at one-second intervals from the dL and mdL were exported from the dL for Windows (DLW) software (v. 1.4, Perten Instruments, Macquarie Park, Australia) in ASCII format into the Unscrambler software (v. 10.3, CAMO ASA, Oslo, Norway) for MVA against Mixolab, Extensograph and baking reference data. All samples (61) were used to generate partial least squares (PLS) regression calibration models between the reference values and dL/mdL analysis parameters for data regions between 20-600 s. Statistics calculated for each data region included the root mean square error of cross-validation (RMSECV) and the coefficient of determination (R2).

Table 1:
Dough mixing configurations1

1Chopin+, slow mix and fast mix configurations according to AACCI methods 54-60.01, 54-21.01 and 54-70.01 respectively. Sample weights were corrected to 14% moisture basis. End = end of test.

Figure 1: Commonly measured parameters from the Mixolab (A), dL/mdL standard mixing (B), and mdL cooking curves (C).

In Figure 1, C, T and T°C are torque, time and temperature values, respectively, at the five stages of the Mixolab curve: C1-dough development, C2-protein weakening, C3-starch gelatinization, C4-gel stability and amylase activity, and C5-starch retrogradation. dL/mdL mixing parameters include peak torque, dough development time (DDT), stability, softening at 5 min (120 rpm) or 12 min (63 rpm) after peak and accumulated energy at peak torque (Wh/kg). Water absorption (WA) is derived from the peak torque and amount of water added during the test. mdL cooking parameters (similar to those of the Mixolab) include Peak1, Hold1, Peak2, Hold2, Final torques and associated times at the five stages of the mdL cooking curve.

Results and Discussion

Mixing Methods
The faster dL and mdL methods (120 rpm) had better precision, with lower coefficients of variation (CV), than the Mixolab and slower (63 rpm) dL methods (Table 2). The high CV for C1 time (dough development time) in the Mixolab method was probably due to the inconsistency of the software in differentiating hydration (generally first) and true mixing (generally second) peaks in duplicate analyses, in cases where two peaks were observed.

The results show good univariate correlations between all mixing methods for water absorption (WA), with the dL 120 rpm method giving the closest correlation to the 63 rpm method (Table 3). Univariate correlations for dough development time (DDT) and stability were reasonable between dL and mdL, but poorer between these instruments and Mixolab (results not shown). Correlations to 63 rpm mixing parameters improved with MVA of data, with lower RMSECVs and higher R2s. Perten Instruments software (Prediction Pack) has been developed to allow these calibrations to be applied in the dL and mdL.

Table 2: Summary of mixing properties of wheat flour doughs tested on the Mixolab, dL and mdL using manufacturer's standard protocols, compared to the traditional mixing method (63 rpm).1

1RMS = root mean square of the error term. CV (%) = coefficient of variation (relative repeatability standard deviation).

Table 3: Univariate (linear) and multivariate (partial least squares) regressions of Mixolab, doughLAB and micro-doughLAB mixing results with those from the 63 rpm method.

aR2 = coefficient of determination, is the estimate of the variability accounted for by the regression; RMS = root mean square of residuals of the linear fit; n = 61;
bRMSECV = root mean standard error of cross-validation between analysis results from the various mixing methods and the 63 rpm method; PLS = partial least squares regression; n = 61.

Cooking Methods

The mdL cooking method had better precision - with generally lower CVs - than the Mixolab method (Table 4).

There was a good univariate correlation (R2 = 83.0%) between mdL and Mixolab cooking methods for WA. Weaker correlations were observed between mdL DDT and Mixolab C1 time (R2 = 50.7%), and mdL Hold1 and Mixolab C2 (R2 = 53.9%) (results not shown). For approximately 10% of the samples, the Mixolab curves showed cavitation of torque following the hot peak that were absent from mdL curves for the corresponding samples (Figure 2). In both methods, the cooked dough became rubbery and exhibited a ‘shark skin’ effect, this is indicative of flow fracture, where the dough loses the ability to comply to large deformations (Bason, et al., 2007). Torque cavitation in the Mixolab curve was mainly attributed to the higher maximum temperature (90°C) of the Mixolab method compared to that used in the mdL cooking method (80°C). When the maximum temperature in the Mixolab was reduced to 80°C, a smooth torque curve was observed with no cavitation (results not shown).

Figure 2: Mixing curve of a weak flour tested using the Mixolab method with 90°C maximum temperature (A) and micro-doughLAB cooking method with 80°C maximum temperature (B). The cavitation in torque (circled) in the Mixolab curve following the hot peak was absent in the micro-doughLAB curve where a lower maximum temperature was employed.

Table 4: Summary of cooking properties of wheat flour doughs tested on the Mixolab and mdL using manufacturer's protocols.1

1RMS = root mean square of the error term. CV (%) = coefficient of variation (relative repeatability standard deviation).

Correlation to Extensograph and Bake Volume

Mixing parameters DDT and energy at peak torque from the dL and mdL fast methods, and C1 time from the Mixolab method gave weak univariate correlations to Extensograph parameters (extensibility and maximum resistance at 45 and 135 minutes) and bake volume (R2 < 57.4%, results not shown). There was also no correlation (R2 < 21.9%) between Extensograph results and bake volume. The results suggest that mixing parameters of flour-water doughs are poor indicators of end-use quality. The correlations may improve with full-formulation doughs (Oliver & Allen, 1992).


This study used univariate and multivariate analyses of data from recently developed mixing methods to compare them with the traditional mixing method and to assess their capabilities in predicting end-use quality.

The doughLAB rapid (120 rpm) method had better precision among the methods, and produced results that best correlated to the traditional 63 rpm method. Correlations generally improved with multivariate analysis of data, regardless of the mixing method used. However, all methods and mixing parameters showed weak to poor correlation to Extensograph and bake volume. The results suggest that mixing parameters of flour-water doughs are poor indicators of bread making quality.

Further studies are being conducted using a stress-decay test on the micro-doughLAB to study the elasticity of dough, and using this information to predict bake volume. The mdL stress-decay test showed potential to provide data on mixing characteristics and viscoelastic properties of a dough (Dang & Bason, 2013). When combined with the predictive capability of the Prediction Pack software, the stress-decay test on the mdL can potentially be a useful one-stop instrument for bakers, breeders and researchers.


The authors would like to thank Sam Bason of Perten Instruments of Australia for assistance with sample testing.


AACC International. (2014), Approved Methods of Analysis, 11th Ed., Approved Methods 10-05.01, 44-15.02, 54-10.01, 54-21.01, 54-60.01, 54-70.01.AACC International, St. Paul, MN.

Bason, M. L., Charrié, C., Dang, J. M. C., & Booth, R. I. (2007). Can you mix hot dough? Proceedings of the C&E Spring Meeting, 2-4 May 2007, Montpellier, AACC International.

Dang, J. M. C., & Bason, M. L. (2013). Determining elasticity of dough in the micro-doughLAB. Perten Science World, 7:(2-3).

Oliver, J. R., & Allen, H. M. (1992). The prediction of bread baking performance using the farinograph and extensograph. Journal of Cereal Science, 15:(79-89).