Final Report Summary - FLOURPLUS (Intelligent and easy tool to categorise and characterise flour quality for consumer-driven wheat baked goods in European SME-bakery and cereal sector)
Executive Summary:
The project FLOURplus had the objective to provide a solution for bakery companies that helps reacting to quality variations of wheat flour.
The solution developed, the so-called "FLOURplus system", is an online IT system that gives recommendations on how to adapt the recipe and the bakery process, based on the analytical data of flour.
The FLOURplus system takes into account the variety of situations encountered in Europe's bakery companies: different degrees of automation, different recipes and processes, different analytical data available, and different skills of the production employees.
There are two levels of services provided by the FLOURplus system: level 1 is easy to use, allows a quick start and provides an interpretation of the analytical certificates delivered by the mills, based on extensive experimental data generated in the project. Level 2 but requires more effort from the user. Hence, the calculations made in level 2 are based on the user's own data. The user can define his own flour, recipes, processes and final products in the system. In daily production, data can be entered into the system, from which correlation models are calculated. When a new batch of a particular product is to be produced, the FLOURplus calculates the expectable properties of the final baked good and proposes recipe and process adjustment as required.
The concept of the FLOURplus system was developed in close cooperation with bakery companies to match their processes and wishes in the best possible way. It was implemented as an easy-to-handle prototype that offers an intuitive interface in 4 languages, a database able to host data both from the project and the users, powerful modelling algorithms and clear recommendations. In level 2, as the system learns from the user's data, its accuracy increases with each single production batch. It is possible to import data from past productions, in order not to start with an empty system.
The experimental work conducted in the project also allowed the formulation of recommendations on optimal flour analytical setups. This will help bakers and millers to challenge and improve their current flour analytical practice.
Next to the FLOURplus system, an inline kneading monitoring instrument, the "FLOURplus device" was developed. This system allows monitoring dough development in real time at stop kneading at the optimal time. The FLOURplus device can be used in combination with the FLOURplus system or independently.
In addition, the project resulted in recommendations on the use of alternative strains of yeasts and lactic acid bacteria in baking, including process specifications, which enables bakery companies producing breads and rolls with a wider diversity or sensory properties.
Numerous companies from various countries, belonging to the core target group of the project team (bakery companies and flour millers) but also to associated businesses (manufacturer of bakery ingredients) from France, Germany, Belgium, Spain and the Netherlands indicated to be highly interested in using the project's results. The expected benefits for them are an increase in product quality, an increase in productivity and a reduction of food waste.
Project Context and Objectives:
The project context includes several aspects. One is digitalization, which offers new perspectives in many facets of daily life but also in industrial production. The vision is that in a connected production, information available in real time allows to make faster and better decisions, ultimately in according to automated decision algorithms. This approach can be adapted in the food industry to the decision-making related to variations of raw material quality.
Hence, the food industry transforms agricultural materials from natural origins, which are subject to quality variations. This is especially true in the case of wheat flour, as many factors have an influence on the flour quality and its functionality: the wheat varieties, the climatic conditions, the culture parameters (including fertilization), the harvesting process, the storage, and the milling process.
To compensate for these variations, flour millers have the possibility to select grain batches and to blend them in an appropriate way. However, this allows only reducing variation, not to fully standardize flour quality. As a result, the quality delivered to bakeries vary from batch to batch. The highest variations are observed when switching from the grain of one season to the next one.
In the mills, analyses are made on each batch of flour and an analytical certificate is delivered to the baker. However, most bakery companies have no method for interpreting the analytical values. In addition, the analytical parameters measured vary from mill to mill and even more from country to country.
It was clearly shown in the first part of the project that almost all bakeries have one recipe and one process setting for each bakery product and that they always use those standard conditions when starting a production. However, due to the variability of the quality of the raw materials and especially of the flour, some problems may appear on the production line (e.g. sticky dough) of with the final product which may be out of spec (wrong size, volume, colour, structure of the crumb, sensory properties. When these problems occur, the production employees have to adapt the recipe or the process. Up to now, this is performed in a non-systematic way, based on the experience of the employees, and with a high risk to fail. The consequence is down-time, food waste and a suboptimal quality.
The project FLOURplus was based on the assumption that the relationship between flour quality as described by the analytical data, process parameters and baked good properties can be modelled. Based on those models, it should be possible to calculate the expectable baked good properties as a function of the analytical data and the process parameters. The pre-requisite is a database that contains enough production data, and appropriate modelling algorithms. In addition, it is necessary to make sure that the most relevant analytical properties and process parameters are measured and taken into consideration.
It is striking that in the food industry and especially in bakery companies, lots of raw material quality data, process data and final product data are monitored, but that these data are in most cases only compared with reference data and not analysed in relation to each other in a modelling approach.
Starting from this situation, it was the ambition of the project FLOURplus to deliver a system which helps bakers making decisions on recipe and process adaptions in a proactive way, based on the flour analytical data, along with recommendations on which analytical data are the most helpful. A further objective was to deliver an inline kneading monitoring device which allows process adjustments based on data generated in real-time from the semi-finished product. Finally, a third objective was to get knowledge on the potential of alternative strains of yeast and lactic acid bacteria to obtain a wider range of sensory properties of breads and rolls.
Reaching these objectives involved:
- Reviewing the wheat flour analytical practice and the bakery process, especially the management of quality and process data across Europe,
- Developing a datable appropriate to host flour quality data, process data, and baked good data provided by the experimental work performed in the project and by future users of the FLOURplus system.
- Developing modelling algorithms capable to take into consideration the fact that each user is willing to use its own combination of parameters for modelling. These algorithms need to be able to make predictions of the baked good quality and appropriate recommendations on recipe and process adjustments, depending on the flour analytical values and based on data from past productions.
- Developing an easy-to-use user interface based on typical processes of bakery companies, allowing a fast decision-making in production.
- Making flour analysis and baking trials with a high number of flour batches and process settings, and characterising the final baked products both by instrumental measurements and sensory analysis, in order to deliver data that feeds the database.
- Developing a dough kneading monitoring sensor.
- Making extensive process trials with selected alternative strains of yeast and lactic acid bacteria and characterising the final baked products both by instrumental measurements and sensory analysis, in order to identify the optimal process settings for each of them and to describe the obtainable sensory properties.
Project Results:
The main S&T results of the project FLOURplus are:
1/ The FLOURplus system for calculation of recipe and process adaptions based on flour analytical data
2/ The FLOURplus device for inline monitoring of dough development in the kneading process
3/ Recommendations on optimised flour analytical setups
4/ Recommendations on the use of alternative strains of yeasts and lactic acid bacteria in baking
This section gives publishable further details on these four main results.
1/ The FLOURplus system
The FLOURplus system was designed to be a practical help in bakery production in order to allow bakers to make fast and reliable decisions before starting processing a particular batch of flour. The following constrains were taken into consideration:
a/ Bakery companies have diverse IT infrastructures. The system should be able to run on most current IT systems without high investment costs.
b/ Analyses performed on flour are very diverse across Europe. It is not realistic to impose a particular analytical setup as a pre-requisite of the use of the FLOURplus system. The system should be customizable to take into account the analytical data available in the companies. Nevertheless, based on the recommendations formulated in the project (see point 3 below), bakers and millers have the possibility to optimize their analytical setup if they wish to do so, which will contribute to improve the prediction accuracy of the FLOURplus system.
c/ Although the main steps of bread and roll production are universal (kneading, cutting of dough pieces, proofing, baking), and the main ingredients are flour, water, yeast and salt, there is a wide diversity of recipes and processes across Europe.
d/ Some bakeries are more artisan-oriented and do not gather extensive quality and production data. These bakeries need a prediction tool that does not require the input of complex data, even at the price of prediction accuracy.
e/ Other bakeries already monitor numerous quality and production data, which are currently stored in diverse IT systems. These bakeries may wish to make use of the full potential of these data to develop prediction models based on their own data. This allows the highest prediction accuracy.
f/ The system needs to operate in the local language of the employees to be accepted.
The following features of the FLOURplus system address the above-mentioned constrains:
a/ The FLOURplus system was developed as an online system which can be operated using just a web browser on a desktop or a mobile device with internet access. The interface developed supports Firefox, Chrome, Edge and Safari. All data required by the system, both from the project and from the users, are stored on a central server. State-of-the art data encryption system may be use in the future commercial use of the system, to make sure that data is securely transferred and stored. All calculations performed by the FLOURplus system are run on the server, which means that no particular computing capacity is necessary from the user's side.
b/ The FLOURplus system currently supports 90 single analytical parameters, which were chosen to reflect the diversity of analytical methods used across Europe. When defining flour specifications in the system, the user can select freely which parameters he desires to use out of them. Only the selected parameters will appear and be taken into consideration in the following steps. The 90 values are based on the following methods and instruments:
Moisture (ICC 110/1), Ash content ICC 104/1, Protein content (ICC 105/2), Sedimentation value (ICC 116/1), Brabender Glutopeak (setting: 25°C, 3000 rpm), Perten Glutomatic (ICC 155), pH and acidity (ICC 145), Falling number (ICC 107/1), Rapid Visco Analyser, Brabender Micro Visco Amylograph, Chopin SDmatic (ICC 172), SRC-CHOPIN (AACC 56-11.2) Mütek Particle Charge Analyser, Brabender Farinograph (ICC 115/1), Chopin Mixolab (ICC 173), Brabender Extensograph (ICC 114/1), Chopin AlveoLAB (ICC 121), Chopin Rheo F4 (AACC 89-01.01).
The above list contains less than 90 methods, as most methods measure several analytical parameters.
c/ The diversity of recipes and processes is taken into consideration in level 2 of the system, in which the user is able to define his own recipes and processes and make calculations based on his own data. On the contrary, in level 1, standard recipes and processes of rolls and toast breads are used.
d/ Level 1 is highly easy to use and enables any bakery to start instantly using the system, provided that they get analytical certificates related to the flour batches they process. In level 1, the user can define the specification of the flour(s) he uses with reference values for each analytical parameter. The reference values should be chosen from a typical flour batch that performs well on the production line with standard recipe and production parameters, and delivers baked goods that are well in line with their specification. There is a description provided for each analytical parameter, which contributes to educating the bakery employees (as the flour analysis is nearly always performed in the mill, the bakery employees are not always familiar with the analytical methods).
Once the specification is defined and stored in the system, the user may enter the analytical data of a particular batch of flour he wishes to use. The FLOURplus system then calculates the difference between the reference data and the actual data for each analytical parameter provided. Based on the experimental data from the project, the expectable impact of this difference on the final product is being calculated. Furthermore, again based on the project data, a calculation is made about whether an adaptation of the water dosage and/or of the kneading time would increase the product quality. If so, a qualitative recommendation is given. The product attributes considered in level 1 are for the rolls the specific volume, the ratio height/width, and the cell density; for the breads the specific volume, the cell density, the firmness of the crumb and its elasticity.
e/ Level 2 provides more advanced functionalities. As in level 1, the user needs to define its flour specification. In addition, he has the possibility to define the other ingredients he uses. In the next step, the user needs to define his final products into the system. This implies providing the recipe, the main process parameters, and the specification of the final products. In the same way the users selected flour analytical parameters, he can choose from a variety of typical parameters and attributes of final products which. As far as the recipes and the process parameters are concerned, it is important to understand that the user is not required to provide the full recipe nor the complete process parameters. In fact, only the parameters which are being varied in daily production are relevant. This can be typically the amount of water or of yeast, the kneading time, the proofing time, but also the temperature of the air in the factory or of the ingredients if those vary. A wise selection of the relevant parameters is of upmost importance to allow a good accuracy of the system, as it is crucial to monitor the relevant parameters. Here, the expertise of the user is required. Discussions with potential users, especially at the demonstration session, showed the validity of this approach, as each user is the expert for his production and has the best knowledge about which parameter should be considered.
Once the products are defined, the next step is to enter data about a high number of single productions. The user may start with an empty database and enter the flour analytical data, the process information, possible recipe adaptions he performs as well as the properties of the final baked goods on a day-to-day basis. Or he may choose to import data into the system from past productions, if those data are already available. During the project, data was imported from a pilot user describing the production of 13 SKUs over six years, showing the feasibility of data import from different existing sources. In this step, traceability is a key, i.e. it is necessary to be able to trace which batch of flour was used in manufacturing each particular batch of final product.
Once the FLOURplus system has enough data for a particular product, it becomes possible to use the prediction functionality. This means that, before starting a production, selecting a batch of flour allows to calculate the expectable properties of the final baked goods. Should these be out of spec, a calculation is made about whether a change of recipe and production parameters would lead to better results. If so, appropriate parameters are suggested. The user can also use the prediction algorithm in another way: he can enter the recipe or process adaption that seems meaningful to him, based on his experience, and let the system calculate the expectable outcome. It is important to understand that as the system makes calculation based on past data, it will only propose to make adaptions on parameters which have been varied on the past, and in a range corresponding to past productions. Based on the suggestions made by the system, the user can choose how he will run the production (the decision remains with him). One the final product is obtained, its actual data can be entered into the system.
In this way, the system learns from each production and its predictions become more accurate with time. This approach imitates the learning process of the production employees, but uses the power of computer systems for quantitative modelling of complex relationships and the possibility to take into account all past productions in modelling.
In the current prototype version of the system, data must be entered by hand into the system. An alternative is to copy and paste tables of data from different sources. These tables need to be set in the right format before import. It is a wish of several potential users to have interfaces developed for the automated data transfer from various systems into FLOURplus. Those interfaces were not developed within the project but may be implemented as an additional service to the users. As there is no data standard for flour data, bakery process data and baked good data, no universal interface can be created up to now. It could be the objective of a further R&D project or of a standardization group to developed such a data standard.
f/ The user interface has a multi-language capacity. Four languages were implemented within the project: English, French, German and Spanish, in line with the countries represented by the bakery associations. Starting from English as the reference language, the associations and their members provided accurate translations of the specific technical terms. Any further language can be implemented easily by providing a translation of a pre-defined list of technical terms and interaction phrases. The number of possible languages is not limited. In the system interface, the user can conveniently switch between the languages at any time without having to leave the current view.
The system itself is composed of 3 components: the database, the user interface and the calculation algorithms. The database is composed of a variety of tables. The main ones is a list of the system user, and any data contained in the further tables is clearly related to one specific user, who is the owner of this data. The further tables provide room for the storage of flour analytical data, process data and baked good data.
The user interface provides the functionality described above. For each user, it is possible to define whether he is authorised to use only level 1 or also level 2. In addition, there is a specific administrator profile which allows to define new users and edit the descriptions of parameters in a convenient way.
Finally, the calculation algorithms provide the modelling functionality.
2/ The FLOURplus device for inline monitoring of dough development in the kneading process
It is well-known that different flour batches need different kneading times. This can be shown very well by analytical devices like the Brabender Farinograph and the Chopin Mixolab, which show different kinetics of dough development, which is related to varying gluten and starch properties of the dough. While the FLOURplus system is able to calculate an optimal kneading time based on the flour analytical data and data from past production, the FLOURplus device monitors the kneading process inline when a dough is manufactured in bakery production. Because it can be used for any type of kneader and any type of dough, as long as there is network formation, it is a universal device.
Extensive trials at Backhaus Häussler showed that the sensor is able to indeed indicate the point of maximal network formation, which is usually the point at which the dough has the best properties. The major advantage of in-line measurement is that it is independent of the ingredients and that therefore process control is possible.
3/ Recommendations on optimised flour analytical setups
In the trials performed within the project FLOURplus, each batch of flour was analysed with all the methods stated above, resulting in 90 single analytical values for each flour. Doing this, we wanted to have the most comprehensive analytical approach possible. It is clear that flour mills are not able to perform such a high number of analyses on their flours. Analytical certificates of wheat flours typically include 5 to 10 values.
It is also well understandable that the 90 analytical parameters we considered are not independent from one another. For instance, dough hydration as measured by the Farinograph and the Mixolab is related to the level of damaged starch and the functionality of gluten and pentosans. The hydration itself has an influence on the dough rheology, as a dryer dough will be firmer. The level of protein is in most cased related to the level of gluten, and gluten functionality is related to dough properties. Based on these considerations it is clear that the method considered measure different aspects of the same underlying properties, and the question is to choose an optimal combination of methods, which delivers the highest amount of non-redundant information for the least possible analytical effort.
A principal component analysis of the dataset showed that 4 main flour properties explained 64,8% of the total variance. These properties are (PC1) starch gelatinization properties, (PC2) hydration properties, (PC3) dough resistance at variable water amount, and (PC4) dough strength at fixed water amount.
In a further step, it was checked by means of a PLS regression which analytical parameters allow for the best prediction of the properties of the final baked goods. As could be expected, numerous analytical parameters are significantly correlated with the properties of the baked goods, but none of them is alone a good predictor.
This illustrates the multi-dimensional nature of the investigated relationship, and is why it is so difficult for a bakery employee to infer from a flour analytical certificate how the process and/or the recipe should be adapted. Humans have difficulties to make representation of multi-dimensional relationships. This is where the modelling approach of the project shows its power.
Based on the data analysis made in the project, it was possible to identify analytical methods that are more highly correlated with the baked goods properties, and combination of analytical methods that are more favourable than others, because they measure properties which have only a low correlation with each other.
Five different yeasts in comparison to normal baker’s yeast, used as a control, were investigated for their bread making performance.
• Baker’s yeast - Puratos, Belgium
• s-23 - Fermentis Division of S. I. Lesaffre, France
• T-58 - Fermentis Division of S. I. Lesaffre, France
• us-05 - Fermentis Division of S. I. Lesaffre, France
• wb-06 - Fermentis Division of S. I. Lesaffre, France
• Blanc - Vinoferm, Belgium
Six different sourdoughs were investigated for bread making.
• M - Böcker Sauerteig, Germany
• F - Böcker Sauerteig, Germany
• W200 - Böcker Sauerteig, Germany
• Direct 25 - Böcker Sauerteig, Germany
• Lactobacillus reuteri R29 - culture collection of the Cereal Science laboratory in University College Cork, Ireland.
• Weissella cibaria MG1 - culture collection of the Cereal Science laboratory in University College Cork, Ireland.
Response surface methodology (RSM) was used to evaluate the effect of the independent variables proofing time and temperature on the dependent variables volume, hardness and number of cells.
The optimized process conditions were used to determine the influence of the starter cultures on specific volume, moisture, water activity and colour of the different bread samples. Additionally, shelf-life and crumb structure profile were evaluated using the Texture Profile (over a storage period of five days) analyser and C-Cell image analysis software, along with sensory analysis.
Based on this, recommendations were formulated for the use of the investigated yeast strains and sourdoughs and the expectable outcome in terms of sensory profiles.
Potential Impact:
The potential impact for artisan and industrial bakeries of using the FLOURplus system and the FLOURplus device is primarily an increase in product quality, an increase in productivity and a reduction of food waste. Hence, both the system and the device allow to react pro-actively to variations of flour quality by adapting the recipe and/or the production process. In this way, down times and out-of-spec products can be avoided.
This is essential in a sector that operates with low margins, where any loss of raw materials has a strong impact on profitability. In addition, avoiding down-times is critical both for bakeries producing fresh bread (which has to be delivered every day on time) and frozen bread (because the capacity of the production lines is often saturated and any loss of time cannot be caught up).
Reducing food waste is in addition an important societal challenge, given the increase of the world population and the limited production capacity of agriculture.
Improving the quality of bakery products will increase consumer satisfaction and help reducing food waste in households.
Indirect benefits of the FLOURplus system and the FLOURplus device are:
• Contribute to make the employees in bakery companies more knowledgeable by helping them understanding the impact of flour quality variations on baked good quality, and the related solutions in terms of recipe and process adaptions. This will facilitate the employability of these workforces, especially considering the ongoing transformation of the bakery business to more automation.
• Help bakeries to become less dependent from their flour supplier(s) by allowing them to use different qualities.
• Help bakeries to identify critical parameters in their production process.
• Help bakeries to use the full potential of their production data by connecting raw material, process and product data with one another.
• Help millers to support bakers when quality variations occurs, especially when a strong change in quality occurs as is the case each year with the new harvest.
In a next step, the use of the FLOURplus system can be widened to food products other than baked goods. Hence, it is a general issue in food processing that raw materials have varying properties, which need to be compensated in the transformation steps. The modelling algorithms used in the FLOURplus system can be used independently of the application. However, the database structure and the user interface of FLOURplus are designed to match the bakery process. Therefore, a transposition to other application will need to take into account its specific characteristics. If this is the case, the benefits expected in the baking sector (increase in quality and productivity, reduce of waste) can be achieved in any sector of the food industry.
List of Websites:
www.flourplus.eu
The project FLOURplus had the objective to provide a solution for bakery companies that helps reacting to quality variations of wheat flour.
The solution developed, the so-called "FLOURplus system", is an online IT system that gives recommendations on how to adapt the recipe and the bakery process, based on the analytical data of flour.
The FLOURplus system takes into account the variety of situations encountered in Europe's bakery companies: different degrees of automation, different recipes and processes, different analytical data available, and different skills of the production employees.
There are two levels of services provided by the FLOURplus system: level 1 is easy to use, allows a quick start and provides an interpretation of the analytical certificates delivered by the mills, based on extensive experimental data generated in the project. Level 2 but requires more effort from the user. Hence, the calculations made in level 2 are based on the user's own data. The user can define his own flour, recipes, processes and final products in the system. In daily production, data can be entered into the system, from which correlation models are calculated. When a new batch of a particular product is to be produced, the FLOURplus calculates the expectable properties of the final baked good and proposes recipe and process adjustment as required.
The concept of the FLOURplus system was developed in close cooperation with bakery companies to match their processes and wishes in the best possible way. It was implemented as an easy-to-handle prototype that offers an intuitive interface in 4 languages, a database able to host data both from the project and the users, powerful modelling algorithms and clear recommendations. In level 2, as the system learns from the user's data, its accuracy increases with each single production batch. It is possible to import data from past productions, in order not to start with an empty system.
The experimental work conducted in the project also allowed the formulation of recommendations on optimal flour analytical setups. This will help bakers and millers to challenge and improve their current flour analytical practice.
Next to the FLOURplus system, an inline kneading monitoring instrument, the "FLOURplus device" was developed. This system allows monitoring dough development in real time at stop kneading at the optimal time. The FLOURplus device can be used in combination with the FLOURplus system or independently.
In addition, the project resulted in recommendations on the use of alternative strains of yeasts and lactic acid bacteria in baking, including process specifications, which enables bakery companies producing breads and rolls with a wider diversity or sensory properties.
Numerous companies from various countries, belonging to the core target group of the project team (bakery companies and flour millers) but also to associated businesses (manufacturer of bakery ingredients) from France, Germany, Belgium, Spain and the Netherlands indicated to be highly interested in using the project's results. The expected benefits for them are an increase in product quality, an increase in productivity and a reduction of food waste.
Project Context and Objectives:
The project context includes several aspects. One is digitalization, which offers new perspectives in many facets of daily life but also in industrial production. The vision is that in a connected production, information available in real time allows to make faster and better decisions, ultimately in according to automated decision algorithms. This approach can be adapted in the food industry to the decision-making related to variations of raw material quality.
Hence, the food industry transforms agricultural materials from natural origins, which are subject to quality variations. This is especially true in the case of wheat flour, as many factors have an influence on the flour quality and its functionality: the wheat varieties, the climatic conditions, the culture parameters (including fertilization), the harvesting process, the storage, and the milling process.
To compensate for these variations, flour millers have the possibility to select grain batches and to blend them in an appropriate way. However, this allows only reducing variation, not to fully standardize flour quality. As a result, the quality delivered to bakeries vary from batch to batch. The highest variations are observed when switching from the grain of one season to the next one.
In the mills, analyses are made on each batch of flour and an analytical certificate is delivered to the baker. However, most bakery companies have no method for interpreting the analytical values. In addition, the analytical parameters measured vary from mill to mill and even more from country to country.
It was clearly shown in the first part of the project that almost all bakeries have one recipe and one process setting for each bakery product and that they always use those standard conditions when starting a production. However, due to the variability of the quality of the raw materials and especially of the flour, some problems may appear on the production line (e.g. sticky dough) of with the final product which may be out of spec (wrong size, volume, colour, structure of the crumb, sensory properties. When these problems occur, the production employees have to adapt the recipe or the process. Up to now, this is performed in a non-systematic way, based on the experience of the employees, and with a high risk to fail. The consequence is down-time, food waste and a suboptimal quality.
The project FLOURplus was based on the assumption that the relationship between flour quality as described by the analytical data, process parameters and baked good properties can be modelled. Based on those models, it should be possible to calculate the expectable baked good properties as a function of the analytical data and the process parameters. The pre-requisite is a database that contains enough production data, and appropriate modelling algorithms. In addition, it is necessary to make sure that the most relevant analytical properties and process parameters are measured and taken into consideration.
It is striking that in the food industry and especially in bakery companies, lots of raw material quality data, process data and final product data are monitored, but that these data are in most cases only compared with reference data and not analysed in relation to each other in a modelling approach.
Starting from this situation, it was the ambition of the project FLOURplus to deliver a system which helps bakers making decisions on recipe and process adaptions in a proactive way, based on the flour analytical data, along with recommendations on which analytical data are the most helpful. A further objective was to deliver an inline kneading monitoring device which allows process adjustments based on data generated in real-time from the semi-finished product. Finally, a third objective was to get knowledge on the potential of alternative strains of yeast and lactic acid bacteria to obtain a wider range of sensory properties of breads and rolls.
Reaching these objectives involved:
- Reviewing the wheat flour analytical practice and the bakery process, especially the management of quality and process data across Europe,
- Developing a datable appropriate to host flour quality data, process data, and baked good data provided by the experimental work performed in the project and by future users of the FLOURplus system.
- Developing modelling algorithms capable to take into consideration the fact that each user is willing to use its own combination of parameters for modelling. These algorithms need to be able to make predictions of the baked good quality and appropriate recommendations on recipe and process adjustments, depending on the flour analytical values and based on data from past productions.
- Developing an easy-to-use user interface based on typical processes of bakery companies, allowing a fast decision-making in production.
- Making flour analysis and baking trials with a high number of flour batches and process settings, and characterising the final baked products both by instrumental measurements and sensory analysis, in order to deliver data that feeds the database.
- Developing a dough kneading monitoring sensor.
- Making extensive process trials with selected alternative strains of yeast and lactic acid bacteria and characterising the final baked products both by instrumental measurements and sensory analysis, in order to identify the optimal process settings for each of them and to describe the obtainable sensory properties.
Project Results:
The main S&T results of the project FLOURplus are:
1/ The FLOURplus system for calculation of recipe and process adaptions based on flour analytical data
2/ The FLOURplus device for inline monitoring of dough development in the kneading process
3/ Recommendations on optimised flour analytical setups
4/ Recommendations on the use of alternative strains of yeasts and lactic acid bacteria in baking
This section gives publishable further details on these four main results.
1/ The FLOURplus system
The FLOURplus system was designed to be a practical help in bakery production in order to allow bakers to make fast and reliable decisions before starting processing a particular batch of flour. The following constrains were taken into consideration:
a/ Bakery companies have diverse IT infrastructures. The system should be able to run on most current IT systems without high investment costs.
b/ Analyses performed on flour are very diverse across Europe. It is not realistic to impose a particular analytical setup as a pre-requisite of the use of the FLOURplus system. The system should be customizable to take into account the analytical data available in the companies. Nevertheless, based on the recommendations formulated in the project (see point 3 below), bakers and millers have the possibility to optimize their analytical setup if they wish to do so, which will contribute to improve the prediction accuracy of the FLOURplus system.
c/ Although the main steps of bread and roll production are universal (kneading, cutting of dough pieces, proofing, baking), and the main ingredients are flour, water, yeast and salt, there is a wide diversity of recipes and processes across Europe.
d/ Some bakeries are more artisan-oriented and do not gather extensive quality and production data. These bakeries need a prediction tool that does not require the input of complex data, even at the price of prediction accuracy.
e/ Other bakeries already monitor numerous quality and production data, which are currently stored in diverse IT systems. These bakeries may wish to make use of the full potential of these data to develop prediction models based on their own data. This allows the highest prediction accuracy.
f/ The system needs to operate in the local language of the employees to be accepted.
The following features of the FLOURplus system address the above-mentioned constrains:
a/ The FLOURplus system was developed as an online system which can be operated using just a web browser on a desktop or a mobile device with internet access. The interface developed supports Firefox, Chrome, Edge and Safari. All data required by the system, both from the project and from the users, are stored on a central server. State-of-the art data encryption system may be use in the future commercial use of the system, to make sure that data is securely transferred and stored. All calculations performed by the FLOURplus system are run on the server, which means that no particular computing capacity is necessary from the user's side.
b/ The FLOURplus system currently supports 90 single analytical parameters, which were chosen to reflect the diversity of analytical methods used across Europe. When defining flour specifications in the system, the user can select freely which parameters he desires to use out of them. Only the selected parameters will appear and be taken into consideration in the following steps. The 90 values are based on the following methods and instruments:
Moisture (ICC 110/1), Ash content ICC 104/1, Protein content (ICC 105/2), Sedimentation value (ICC 116/1), Brabender Glutopeak (setting: 25°C, 3000 rpm), Perten Glutomatic (ICC 155), pH and acidity (ICC 145), Falling number (ICC 107/1), Rapid Visco Analyser, Brabender Micro Visco Amylograph, Chopin SDmatic (ICC 172), SRC-CHOPIN (AACC 56-11.2) Mütek Particle Charge Analyser, Brabender Farinograph (ICC 115/1), Chopin Mixolab (ICC 173), Brabender Extensograph (ICC 114/1), Chopin AlveoLAB (ICC 121), Chopin Rheo F4 (AACC 89-01.01).
The above list contains less than 90 methods, as most methods measure several analytical parameters.
c/ The diversity of recipes and processes is taken into consideration in level 2 of the system, in which the user is able to define his own recipes and processes and make calculations based on his own data. On the contrary, in level 1, standard recipes and processes of rolls and toast breads are used.
d/ Level 1 is highly easy to use and enables any bakery to start instantly using the system, provided that they get analytical certificates related to the flour batches they process. In level 1, the user can define the specification of the flour(s) he uses with reference values for each analytical parameter. The reference values should be chosen from a typical flour batch that performs well on the production line with standard recipe and production parameters, and delivers baked goods that are well in line with their specification. There is a description provided for each analytical parameter, which contributes to educating the bakery employees (as the flour analysis is nearly always performed in the mill, the bakery employees are not always familiar with the analytical methods).
Once the specification is defined and stored in the system, the user may enter the analytical data of a particular batch of flour he wishes to use. The FLOURplus system then calculates the difference between the reference data and the actual data for each analytical parameter provided. Based on the experimental data from the project, the expectable impact of this difference on the final product is being calculated. Furthermore, again based on the project data, a calculation is made about whether an adaptation of the water dosage and/or of the kneading time would increase the product quality. If so, a qualitative recommendation is given. The product attributes considered in level 1 are for the rolls the specific volume, the ratio height/width, and the cell density; for the breads the specific volume, the cell density, the firmness of the crumb and its elasticity.
e/ Level 2 provides more advanced functionalities. As in level 1, the user needs to define its flour specification. In addition, he has the possibility to define the other ingredients he uses. In the next step, the user needs to define his final products into the system. This implies providing the recipe, the main process parameters, and the specification of the final products. In the same way the users selected flour analytical parameters, he can choose from a variety of typical parameters and attributes of final products which. As far as the recipes and the process parameters are concerned, it is important to understand that the user is not required to provide the full recipe nor the complete process parameters. In fact, only the parameters which are being varied in daily production are relevant. This can be typically the amount of water or of yeast, the kneading time, the proofing time, but also the temperature of the air in the factory or of the ingredients if those vary. A wise selection of the relevant parameters is of upmost importance to allow a good accuracy of the system, as it is crucial to monitor the relevant parameters. Here, the expertise of the user is required. Discussions with potential users, especially at the demonstration session, showed the validity of this approach, as each user is the expert for his production and has the best knowledge about which parameter should be considered.
Once the products are defined, the next step is to enter data about a high number of single productions. The user may start with an empty database and enter the flour analytical data, the process information, possible recipe adaptions he performs as well as the properties of the final baked goods on a day-to-day basis. Or he may choose to import data into the system from past productions, if those data are already available. During the project, data was imported from a pilot user describing the production of 13 SKUs over six years, showing the feasibility of data import from different existing sources. In this step, traceability is a key, i.e. it is necessary to be able to trace which batch of flour was used in manufacturing each particular batch of final product.
Once the FLOURplus system has enough data for a particular product, it becomes possible to use the prediction functionality. This means that, before starting a production, selecting a batch of flour allows to calculate the expectable properties of the final baked goods. Should these be out of spec, a calculation is made about whether a change of recipe and production parameters would lead to better results. If so, appropriate parameters are suggested. The user can also use the prediction algorithm in another way: he can enter the recipe or process adaption that seems meaningful to him, based on his experience, and let the system calculate the expectable outcome. It is important to understand that as the system makes calculation based on past data, it will only propose to make adaptions on parameters which have been varied on the past, and in a range corresponding to past productions. Based on the suggestions made by the system, the user can choose how he will run the production (the decision remains with him). One the final product is obtained, its actual data can be entered into the system.
In this way, the system learns from each production and its predictions become more accurate with time. This approach imitates the learning process of the production employees, but uses the power of computer systems for quantitative modelling of complex relationships and the possibility to take into account all past productions in modelling.
In the current prototype version of the system, data must be entered by hand into the system. An alternative is to copy and paste tables of data from different sources. These tables need to be set in the right format before import. It is a wish of several potential users to have interfaces developed for the automated data transfer from various systems into FLOURplus. Those interfaces were not developed within the project but may be implemented as an additional service to the users. As there is no data standard for flour data, bakery process data and baked good data, no universal interface can be created up to now. It could be the objective of a further R&D project or of a standardization group to developed such a data standard.
f/ The user interface has a multi-language capacity. Four languages were implemented within the project: English, French, German and Spanish, in line with the countries represented by the bakery associations. Starting from English as the reference language, the associations and their members provided accurate translations of the specific technical terms. Any further language can be implemented easily by providing a translation of a pre-defined list of technical terms and interaction phrases. The number of possible languages is not limited. In the system interface, the user can conveniently switch between the languages at any time without having to leave the current view.
The system itself is composed of 3 components: the database, the user interface and the calculation algorithms. The database is composed of a variety of tables. The main ones is a list of the system user, and any data contained in the further tables is clearly related to one specific user, who is the owner of this data. The further tables provide room for the storage of flour analytical data, process data and baked good data.
The user interface provides the functionality described above. For each user, it is possible to define whether he is authorised to use only level 1 or also level 2. In addition, there is a specific administrator profile which allows to define new users and edit the descriptions of parameters in a convenient way.
Finally, the calculation algorithms provide the modelling functionality.
2/ The FLOURplus device for inline monitoring of dough development in the kneading process
It is well-known that different flour batches need different kneading times. This can be shown very well by analytical devices like the Brabender Farinograph and the Chopin Mixolab, which show different kinetics of dough development, which is related to varying gluten and starch properties of the dough. While the FLOURplus system is able to calculate an optimal kneading time based on the flour analytical data and data from past production, the FLOURplus device monitors the kneading process inline when a dough is manufactured in bakery production. Because it can be used for any type of kneader and any type of dough, as long as there is network formation, it is a universal device.
Extensive trials at Backhaus Häussler showed that the sensor is able to indeed indicate the point of maximal network formation, which is usually the point at which the dough has the best properties. The major advantage of in-line measurement is that it is independent of the ingredients and that therefore process control is possible.
3/ Recommendations on optimised flour analytical setups
In the trials performed within the project FLOURplus, each batch of flour was analysed with all the methods stated above, resulting in 90 single analytical values for each flour. Doing this, we wanted to have the most comprehensive analytical approach possible. It is clear that flour mills are not able to perform such a high number of analyses on their flours. Analytical certificates of wheat flours typically include 5 to 10 values.
It is also well understandable that the 90 analytical parameters we considered are not independent from one another. For instance, dough hydration as measured by the Farinograph and the Mixolab is related to the level of damaged starch and the functionality of gluten and pentosans. The hydration itself has an influence on the dough rheology, as a dryer dough will be firmer. The level of protein is in most cased related to the level of gluten, and gluten functionality is related to dough properties. Based on these considerations it is clear that the method considered measure different aspects of the same underlying properties, and the question is to choose an optimal combination of methods, which delivers the highest amount of non-redundant information for the least possible analytical effort.
A principal component analysis of the dataset showed that 4 main flour properties explained 64,8% of the total variance. These properties are (PC1) starch gelatinization properties, (PC2) hydration properties, (PC3) dough resistance at variable water amount, and (PC4) dough strength at fixed water amount.
In a further step, it was checked by means of a PLS regression which analytical parameters allow for the best prediction of the properties of the final baked goods. As could be expected, numerous analytical parameters are significantly correlated with the properties of the baked goods, but none of them is alone a good predictor.
This illustrates the multi-dimensional nature of the investigated relationship, and is why it is so difficult for a bakery employee to infer from a flour analytical certificate how the process and/or the recipe should be adapted. Humans have difficulties to make representation of multi-dimensional relationships. This is where the modelling approach of the project shows its power.
Based on the data analysis made in the project, it was possible to identify analytical methods that are more highly correlated with the baked goods properties, and combination of analytical methods that are more favourable than others, because they measure properties which have only a low correlation with each other.
Five different yeasts in comparison to normal baker’s yeast, used as a control, were investigated for their bread making performance.
• Baker’s yeast - Puratos, Belgium
• s-23 - Fermentis Division of S. I. Lesaffre, France
• T-58 - Fermentis Division of S. I. Lesaffre, France
• us-05 - Fermentis Division of S. I. Lesaffre, France
• wb-06 - Fermentis Division of S. I. Lesaffre, France
• Blanc - Vinoferm, Belgium
Six different sourdoughs were investigated for bread making.
• M - Böcker Sauerteig, Germany
• F - Böcker Sauerteig, Germany
• W200 - Böcker Sauerteig, Germany
• Direct 25 - Böcker Sauerteig, Germany
• Lactobacillus reuteri R29 - culture collection of the Cereal Science laboratory in University College Cork, Ireland.
• Weissella cibaria MG1 - culture collection of the Cereal Science laboratory in University College Cork, Ireland.
Response surface methodology (RSM) was used to evaluate the effect of the independent variables proofing time and temperature on the dependent variables volume, hardness and number of cells.
The optimized process conditions were used to determine the influence of the starter cultures on specific volume, moisture, water activity and colour of the different bread samples. Additionally, shelf-life and crumb structure profile were evaluated using the Texture Profile (over a storage period of five days) analyser and C-Cell image analysis software, along with sensory analysis.
Based on this, recommendations were formulated for the use of the investigated yeast strains and sourdoughs and the expectable outcome in terms of sensory profiles.
Potential Impact:
The potential impact for artisan and industrial bakeries of using the FLOURplus system and the FLOURplus device is primarily an increase in product quality, an increase in productivity and a reduction of food waste. Hence, both the system and the device allow to react pro-actively to variations of flour quality by adapting the recipe and/or the production process. In this way, down times and out-of-spec products can be avoided.
This is essential in a sector that operates with low margins, where any loss of raw materials has a strong impact on profitability. In addition, avoiding down-times is critical both for bakeries producing fresh bread (which has to be delivered every day on time) and frozen bread (because the capacity of the production lines is often saturated and any loss of time cannot be caught up).
Reducing food waste is in addition an important societal challenge, given the increase of the world population and the limited production capacity of agriculture.
Improving the quality of bakery products will increase consumer satisfaction and help reducing food waste in households.
Indirect benefits of the FLOURplus system and the FLOURplus device are:
• Contribute to make the employees in bakery companies more knowledgeable by helping them understanding the impact of flour quality variations on baked good quality, and the related solutions in terms of recipe and process adaptions. This will facilitate the employability of these workforces, especially considering the ongoing transformation of the bakery business to more automation.
• Help bakeries to become less dependent from their flour supplier(s) by allowing them to use different qualities.
• Help bakeries to identify critical parameters in their production process.
• Help bakeries to use the full potential of their production data by connecting raw material, process and product data with one another.
• Help millers to support bakers when quality variations occurs, especially when a strong change in quality occurs as is the case each year with the new harvest.
In a next step, the use of the FLOURplus system can be widened to food products other than baked goods. Hence, it is a general issue in food processing that raw materials have varying properties, which need to be compensated in the transformation steps. The modelling algorithms used in the FLOURplus system can be used independently of the application. However, the database structure and the user interface of FLOURplus are designed to match the bakery process. Therefore, a transposition to other application will need to take into account its specific characteristics. If this is the case, the benefits expected in the baking sector (increase in quality and productivity, reduce of waste) can be achieved in any sector of the food industry.
List of Websites:
www.flourplus.eu