Elsevier

Journal of Food Engineering

Volume 165, November 2015, Pages 166-171
Journal of Food Engineering

Optimal design of experiments and measurements of the water sorption process of wheat grains using a modified Peleg model

https://doi.org/10.1016/j.jfoodeng.2015.06.025Get rights and content

Highlights

  • A modified Peleg model was developed to account for two major components of the wheat grains.

  • Optimal design of experiments for the modified Peleg model has been carried out.

  • Optimal DoE has been performed based Cramer Rao lower bound method.

Abstract

Wheat grains consist of three major components, the bran layer, the endosperm and the germ, with very different water sorption kinetic. The original two parameter Peleg model cannot describe a water sorption process well for such heterogeneous compounds. Therefore, the model was modified to account for the two biggest fractions, the endosperm and the bran layer. This modified model has four parameters and can be used to accurately describe the hydration process of wheat grains.

Two experiments were carried out, an initial experiment to get rough parameter values and a second experiment, which was designed optimally by using the Cramer-Rao lower bond method. The percentage parameter errors for the four parameters of the modified model were reduced from 669%, 24%, 12%, and 2.4% to 38%, 5.4%, 4.5% and 1.9% respectively. The presented results demonstrate the advantage of optimal design of experiments.

Introduction

Hydration kinetics in food technology are key physicochemical properties. The knowledge of moisture content and moisturizing kinetics is vitally important to determine the chemical and physical properties of food products and their shelf life.

In this work, the water absorption kinetics of wheat grains were investigated and an optimal experimental design was applied. The actual water content of wheat grains due to storage is too low for optimal milling. During the milling process the bran and the germ have to be separated efficiently from the endosperm, because the economic values are related to their purity. The water addition exaggerates the difference of the different parts of the wheat kernel and simplifies the separation process. The hydration kinetics for the endosperm is rather slow compared to the germ and the bran (Delcour and Hoseney, 2010) and a hydrated bran is very flexible and will stay mostly intact during milling and can then be sieved out. Therefore, a hydration step, the tempering, is applied routinely before milling to produce optimal quality. A more detailed knowledge about the hydration kinetics of the components of the wheat grains is therefore required to prepare an optimal separation process.

There are fundamental approaches to water absorption kinetics. Cunningham et al. (2007) and Munson-McGee et al., 2011a, Munson-McGee et al., 2011b used models for water uptake based on Fick’s diffusion laws. However, models based on the diffusion law are usually very complex and not very convenient for computing in most situations. Therefore, Peleg (1988) suggested an empiric two parameter model to describe water absorption curves. For example Sopade et al. (1992) and Maskan (2001) used this model to accurately describe the water absorption of cereal grains and wheat respectively. It is still very common due to its simplicity and ideal for an optimal design of experiments due to its modest computational requirements.

Optimal experimental design is usually carried out to get the maximal amount of information from an experiment with the least amount of effort. This goal is achieved by variation of measurement times and measurement locations or other process variables such temperature or pH. One specific and also quite common goal of an optimal experimental design can be to minimize parameter estimation errors. Ataíde and Hitzmann (2009) used an optimal experimental design to analyze enzyme kinetics in a stirred tank reactor. They proved that fed-batch processes will always result in smaller parameter estimation errors than for normal batch processes. Franceschini and Macchietto (2008) presented a list of various recent model based optimal experimental design applications from chemical kinetics to biological modelling. Sánchez et al. (2012) presented a method for a pareto-optimal design. They showed that it is possible to get a non-optimal but still good experimental design regarding multiple criteria at the same time.

According to Dolan and Mishra (2013), the importance of accurate parameter error estimation is often underestimated in food science. In this work, the optimal design of experiments to improve the parameter estimation error of a modified Peleg model was investigated. To do so, an initial experiment for water sorption of wheat grains was carried out to roughly estimate the parameters. Afterwards these initial parameters were used to carry an optimal experimental designs using the Cramer-Rao lower bound method.

Section snippets

Performing optimal design of experiments

The goal of optimal design of experiments is to determine optimal measurement points or the evolution of process variables (such as batch or fed-batch mode) for an experiment, in a way that the variances of model parameters calculated from corresponding measurements are as small as possible. Therefore, one requirement is a theoretical model describing the process under consideration.

Experiments

Some constraints for the optimal design of experiments had to be specified. Here the maximum duration of the experiments was set to 48 h (2 days) and the number of measurements to be carried out specified to be 12.

Altogether two experiments were performed. First, the initial non-optimal experiment for water absorption whose measurement values are used for the rough estimation of the parameter values of the modified Peleg model. The 12 measurements (in triplicate) were taken at the following

Results & discussion

In Fig. 1 the results of the initial non-optimal water uptake measurements as well as the fitted original 2 parameter Peleg model and the modified 4 parameter Peleg model are shown. All water uptake values are relative to the initial wet weight of the grains. Also shown are the two individual water uptake progressions of the bran layer and the endosperm of the modified model, which correspond to the two major compounds of the grain. The root mean square error of the modified Peleg model is

Conclusion

For an overall process optimization, theoretical models as well as rough values for the model parameters must be known. In this investigation, the water absorption process of wheat grains was considered. At first an experiment with 12 moisture measurements in triplicate was performed and the measurement times were chosen based on experience. It could be shown that the modified Peleg model is very suitable to describe the measured moisturizing process of fresh wheat grains, where the normal

Cited by (19)

  • Effect of pearling on composition, microstructure, water migration and cooking quality of highland barley (Hordeum vulgare var. Coeleste Linnaeus)

    2022, Food Chemistry
    Citation Excerpt :

    In recent years, the Peleg model and Low-field nuclear magnetic resonance (LF-NMR) have become efficient tools to assess the water distribution and migration in grain kernels during soaking, but no one has applied these methods to HB. The Peleg model has been frequently used to represent water absorption in food products, which provides experimental data to predict the moisture balance in grain soaked for short times and is the most popular for simulating grain water absorption (Paquet-Durand, Zettel, Kohlus, & Hitzmann, 2015; Uriarte-Aceves & Sopade, 2021). LF-NMR spectroscopy is commonly applied in the food industry due to its rapid, nondestructive, and low-cost characteristics.

  • Modelling multiphasic starch digestograms with multiterm exponential and non-exponential equations

    2022, Carbohydrate Polymers
    Citation Excerpt :

    In association with the objective procedure of Sopade (2021) for multiphasic in vitro starch digestograms, henceforth referred to as the Sopade Objective Procedure, the present study examines multiterm exponential and non-exponential models for such digestograms and explores their practical boundary conditions or constraints. The theories behind the two-term exponential and the two-term non-exponential model in food research are classically discussed elsewhere (Bello-Perez et al., 2019; Li et al., 2019; Paquet-Durand et al., 2015). These theories can be extended to three-term exponential and non-exponential models, which the present study proposes.

  • Artificial neural network (ANNs) and mathematical modelling of hydration of green chickpea

    2021, Information Processing in Agriculture
    Citation Excerpt :

    Researchers reported the wide applicability of Page, Peleg and Weibul model for the prediction of hydration characteristics of food grains [2,16–18]. However other models such as First order, Ibarz et al. and Modified Peleg models were also used to define the hydration behavior of food grains [19–22]. Based on the hydration time, the fitted models describe the moisture content (MC) or moisture ratio (MR).

  • Modeling the effect of temperature on the hydration kinetic whole moong grain

    2018, Journal of the Saudi Society of Agricultural Sciences
View all citing articles on Scopus
View full text