Digitization of farinogram plots and estimation of mixing stability

Abstract

Twenty-four farinograms and accompanying flour characteristics obtained from a bakery were used to get additional information for baking characteristics of flours. Farinograms were digitized and four novel parameters were included for comparison: a and b were extracted from an equation of the form y = ae^−bt; the height of the upper curve and the width of the farinograph curve at a time value equal to the dough development time. Stepwise multiple regressions were carried out relating bread volume to novel and existing parameters (water absorption, development time, arrival time, departure time, stability and degree of softening). Results indicated that four farinogram parameters, resistance, water absorption, a and b were related to bake height with an overall value of 61%. A relatively weak correlation (R = 0.44, P < 0.05) was detected between specific loaf volume and bake height.

Introduction

The predictive relationship between physical dough test results made in the laboratory using the farinograph and extensograph and commercial bakery experience is poor (Oliver and Allen, 1992). A number of authors have attempted to extract additional information from the traditional rheological equipment used for flours. Most of the models use empirical and regression methods to predict essential bread making properties and they are useful for the bakers as a general assessment of baking. It has been believed for a long time that more information is available or can be obtained from the mixing curve than is visibly apparent, i.e., we can measure more than mixing time and tolerance, water absorption, and general dough strength. Attempts have been made to use both farinograph and mixograph to predict baking properties. Cuevas and Puche (1986) used farinograph data to obtain an indication of the rheological behaviour of corn dough. They obtained the indices of apparent viscosity by expressing the geometry of the mixing system and the fluid dynamics in dimensionless form. Bettge et al. (1989) produced a multivariable model with high correlation coefficient using combinations of protein, hardness and alveograph values to predict loaf volume. A new parameter based on the minimum point on the first derivate of the alveograph curve defined as DM was developed by Addo et al. (1990) and related to loaf volume. Their model with DM, a parameter P from the alveogram and water absorption data was produced using stepwise regression analysis, and it explained over 90% of the variability in loaf volume potential. Weegels and Hamer (1989) developed a model using stepwise multilinear regression analysis to predict the baking quality of gluten. The equation predicted loaf volume as a function of extensograph resistance and protein content. Wikstrom (1997) extracted 12 parameters from each mixogram and correlated them to the loaf volume with multivariate partial least squares (PLS) analysis. Five mixogram parameters were found optimal for the final evaluation, explaining 90% of the variance in bread volume. Chung et al. (2001) reported that computer-analyzed mixograph parameters could be affiliated to develop prediction models to be used for flour quality evaluation in wheat breeding programs. They obtained additional information from the mixograph and reported that the prediction models from the mixograph for bake mixing time, bake water absorption and bread loaf volume performed well enough to use in quality evaluation and industrial use. The models were developed using multivariate statistical method, continuum regression, a new multivariate calibration technique that encompasses multiple linear regression, principal component regression and partial least square analysis using mixograph parameters to predict experimental baking parameters. Dobraszczyk and Schofield (2002) derived a number of parameters from the mixograph (peak area, the time in minutes at midline peak height, peak bandwidth, the height between the upper and lower envelopes at the peak) using the PLS multiple regression method. Prediction of loaf volume was improved when data of protein content were included in mixograph parameters. Later, Dobraszczyk et al. (2005) found that using mixograph parameters was much better in prediction of baking performance than using protein content or quality scores based on high molecular weight glutenin subunit composition.

The aim of the current study was to digitize farinograms and extract a number of additional parameters from the plots and to correlate these parameters with baking performance using data from a nine month period of commercial flour production.