Damaged starch and dietary fibre content in Swedish wheat flour : PLS-modelling to predict baking volume

Abstract: SDmatic, SRC-CHOPIN 2 and Alveolab are used to evaluate flour, but not widely used in Sweden. This study aimed to evaluate the machines and see if they could be used to predict baking volume for bread baked on Swedish wheat flour. PLS-models were built with baking volume as the Y-variable. It was noticed that baking volume of breads made on winter wheats and spring wheats were explained by different parameters and as a result building separate PLS-models for these groups gave the best results. Damaged starch had negative impact on baking volume for spring wheats but not for bread baked on winter wheats. All PLS-models were optimised for Q2 by removal of X-variables. Variables from SDmatic, SRC-CHOPIN 2 or Alveolab were left in all PLS-models. The most promising model in this study was built on winter wheats and had a Root Mean Square Error of Prediction (RMSEP) at 75 ml, which can be compared to the average bread with a volume of 2032 ml. This model had only one parameter from these machines and it is thus unclear how useful these machines are when predicting baking volume of bread baked on Swedish wheat flour. Glucomannan was the most important parameter for this model based on Variable Importance in Projection (VIP)-scores and was positively correlated with baking volume. Baking volume was the only predicted quality parameter and future studies should analyse how these machines can predict other quality parameters, such as crumb structure, bread staling and consumer acceptability.