[0001] The present invention relates to a computing system for simulating a spray dryer for providing nutritional formulations in powder form, such as milk powder. The present invention also relates to a computer implemented method performed by the computing system. The present invention also relates to a spray drying system comprising the computing system and to a production line including the spray drying system.

Background art

[0002] Food in the form of powder products is known and widely consumed. Examples of food powder products include milk powder products, which are manufactured with a combination of ingredients, as they include protein content, fat content , and carbohydrates content, specific minerals and vitamins, and more specifically the exact choice of protein, fat and/or carbohydrate and their content, such as for example whey and/or casein as the protein. Each type of food or milk powder product has a specific recipe with a specific proportion of all the ingredients. These specific ingredients and proportions thereof provide certain characteristics to the powder product. This has an impact on the way the factories producing said powder products are set up.

[0003] One way of manufacturing powder products, such as food or milk powder products, is by using spray drying techniques. A spray drying system allows to produce a dry powder from a liquid or paste, by drying it with hot gas. For each product recipe (ingredients and proportion thereof) a spray dryer or spray drying system needs to have specific process settings. The liquid or paste can be a liquid milk, a liquid milk fractionate or a mixture of milk-based or milk-derived components, or components from a different source, such as plant based components, in a liquid (mostly water). In practice, in order to efficiently produce a food powder product, a number of trials need to be performed, by testing different process settings until reaching the right combination for the desired powder product result. These trials take place first at pilot scale and later on in the factory, and as a result, before a spray drying system can be set up with the right process settings for a specific product recipe, a lot of time and resources are required.

[0004] During actual production, it is common practice that the spray drying system comprises an Advanced Process Control (APC) connected to the spray dryer. The APC is a digital control application which can steer the powder production process by giving instructions to the spray dryer about the process settings. The APC can receive instructions from a computing system. The APC can, based on the input or instructions received, modify process settings to for example reach a target powder property. For example, if a specific powder moisture is desired, in the systems known in the art, in order for the APC to provide the control instructions to the spray dryer to reach the target moisture, the APC needs to have information about the powder moisture being obtained with current process settings, and based on that and on the target moisture, determine which process settings to modify to achieve the target moisture. However, in order to have this information about powder moisture, it is required to operate the spray dryer, obtain a final powder, analyze it by manually performing offline laboratory tests, and obtain the powder moisture information from the laboratory analysis, or to have inline sensors which are able to measure the powder moisture content. When a spray drying system does not include an APC, the operators of the spray dryer need to manually provide the process settings to the spray dryer, and these process settings are also based on laboratory results or sensor information. In any of these cases, when steering the process towards a desired final product property, this control performed only relying on historical data is slow.

[0005] In addition, in order to initialize a spray dryer with the most appropriate process settings for a specific performance, such as to achieve a specific value for a target powder property, many tests have to be done on pilot scale before the spray dryer can start operating in the factory in production mode. This is because, , for different recipes, the process settings will be different, and it is only possible to determine them by testing and analyzing the obtained results with different process settings. In addition, in order to produce one recipe, the process settings for the spray dryer need to go through trial and error until the appropriate process settings for that recipe are achieved, but if that same recipe is to be used in another spray dryer, process setting for the new spray dryer also have to be set, and it is not possible to transfer recipes in an efficient manner.

[0006] The spray drying systems in the art have thus the disadvantages of requiring many tests and much time before having the optimal initialization settings, of not allowing for efficient transfer of recipes between spray drying systems, and of requiring much time after they have been initialized until a fine tuning of the process parameters is achieved to obtain the target powder, thereby causing a waste of time and resources, without achieving an optimal target powder. Summary of the invention

[0007] The present invention aims to overcome at least some of these disadvantages, as it allows, on one hand, to accelerate recipe introduction in a factory and recipe transfer between spray drying systems, and on the other hand, to optimize the process settings while in operation to obtain the target product faster and more accurately. This includes determining the right process settings for scaling up from a pilot plant trial to successful factory production, determining the right process settings for recipe transfer from one factory to another, because all factories are different, and fine tuning the process settings while the spray dryer is in operation via an APC, to faster and more accurately achieve the target powder properties and limit the amount of blockages.

[0008] For the purposes of this document, food powder product is defined as a powdery product containing one or more macronutrient components, such as proteins (including hydrolyzed proteins, oligopeptides or amino acids), fats, and carbohydrates, and optionally one or more other components such as micronutrients, minerals, vitamins, enzymes, and probiotics. In embodiments of the invention, the food powder is a plant based food product in powder form that contains components derived from plant basis such as soy, legumes, such as pea, beans, lentils, nuts, grains or rice. In preferred embodiments the food powder is a milk powder. Milk powder is defined as a powdery product that contains one or more macronutrient components as they are found in animal (including human) milk, such as proteins (including hydrolyzed proteins, oligopeptides, or amino acids), fat, and carbohydrates. The milk powders can additionally contain further components such as micronutrients, minerals, vitamins, enzymes, and probiotics. Milk powders for the purposes of this document also cover products in which part of the components are derived from animal milk together with components from another source, such as from plants, or in other words, the milk powder’s compositions do not have to match the composition of the animal milk from which they are derived. Furthermore, for the purposes of this document, as a spray dryer any spray dryer as known in the art to be suitable for preparing food or milk powder is covered. Such spray dryers include co-current spray dryers. Also in embodiments, the spray dryers contain fluidized bed elements that may be either internal or external to the spray dryer. A combination of internal and external fluidized beds is also covered. The spray dryer can be single stage, two stage or a multistage spray dryer. Also several types of nozzles can be used like for example rotating wheels, pressure-operated nozzles or two-fluid nozzles. In embodiments spray dryers can contain more than one nozzle.

[0009] According to the present invention, a computing system for simulating a spray dryer for providing food powder, preferably milk powder, is provided. The computing system comprises: an input unit configured to receive first information related to dimensions and setup of a spray dryer, second information comprising process-related or powder-related specific information, and third information including at least one of sensor information from a plurality of sensors located on and around the spray dryer while the spray dryer is in operation, and target information of at least one target powder property; a computing unit configured to perform a simulation of a production line where the spray dryer is located based on the first information, the second information and the third information, and obtain, as a result, information about at least one predicted powder property based on the sensor information, or process setting information for the spray dryer based on the target information, and an output unit, configured to output the result so that the result is used to improve the control of the spray dryer.

[0010] The invention thus allows to perform a simulation, that is, a digital twin, of the production line in which a spray dryer is located or will be located, and it allows to determine either process settings for initializing the spray dryer with the most appropriate parameters, in order to achieve a specific powder property target, or information about at least one predicted powder property to be used to operate the spray dryer, or a control unit therefor, while it is in operation. When referring to the simulation of the production line, it should be interpreted as a simulation including the spray dryer (and its characteristics), certain environment information (temperature, humidity, and the like), the sensor information measured from the spray dryer and the surroundings, and optionally other information which can be relevant for the operation of the spray dryer. The present invention is also applicable to a spray-drier per se, and in one embodiment the production line is interpreted as the spray dryer. The scope of operation of the simulation, and the mathematical model which it uses, may however not include the liquid phase preparation, product treatment after the fluidized bed(s), handling of the outlet air stream and the recycle streams other than dry lactose and rework additions to the internal fluidized bed.

[0011] Through the present disclosure, the terms spray drying system and spray dryer may be used interchangeably, although the spray drying system may be considered to also include other elements such as an advanced process control (APC) system. [0012] According to embodiments of the invention, the at least one predicted powder property comprises powder moisture and/or powder flow rate, preferably the at least one target powder property includes a target powder moisture, more preferably the at least one target powder property and the at least one predicted powder property comprise powder moisture. Herein, powder moisture typically refers to the water content of the powder based on its total weight. Thus, in at least some embodiments, the computing system aims at assisting the spray drying system to obtain a powder product with a specific powder moisture. For that, the computing system is able to determine process settings for initializing the spray dryer to achieve a certain powder moisture. In addition or alternatively, once the spray dryer has been initialized and is in operation, the computing system is able to calculate a predicted powder moisture from the sensor information received from the spray dryer, so that this predicted powder moisture can be used to operate the spray dryer. In some embodiments, the predicted powder moisture may be transmitted to the APC connected to the spray dryer, so that the APC uses the predicted powder moisture as soft sensor information to adjust the process settings of the spray dryer in a faster manner than having to rely on actual sensor data and moisture information obtained from lab results. In embodiments where there is no APC connected to the spray dryer, the predicted powder moisture can also be used with (semi-) manual control performed by the operator(s) of the spray dryer, as a soft sensor. In some embodiments, in addition or instead of the powder moisture, powder flow rate may be predicted.

[0013] According to embodiments of the invention, the second information comprising process-related or powder-related specific information includes at least one of sorption isotherm information, equilibrium distance information, glass transition information and spray dryer heat loss information. The specific information comprising information which is product dependent or process dependent can also be called empirical information, which is different from the sensor information. By providing this specific information to the mathematical model, the prediction of the results does not only rely on statistical calculations and is therefore more accurately obtained. The sorption isotherm information indicates an equilibrium between air humidity (moisture) and water content of the product. The equilibrium distance information indicates a deviation from equilibrium between the outlet air and the powder, and it is dependent on the powder and spray dryer system. The glass transition information indicates the gradual transition in amorphous materials from a hard and relatively brittle or glassy state into a viscous or rubbery state as the temperature and moisture content changes.

[0014] According to embodiments of the invention, for performing the simulation, the computing unit is configured to run a mathematical model based on energy and mass balances using the first information, the second information and the third information. Advantages of using empirical information over fully statistical models are, for example; that the model can more easily be transferred from pilot to factory and from factory to factory; and that generally this model has a higher confidence when extrapolating; cause and effect will be better understood by examining the model.

[0015] The output unit comprises a communication unit configured to transmit the information about the at least one predicted powder property to an APC operatively connected to the spray dryer, to be used by the APC as soft sensor information. Soft sensor information is hardware-sensor like information, produced by a software, instead of a hardware sensor, but used by the recipient as if it was hardware sensor information. This allows the computing system to operate in real time in communication with the APC of the spray dryer which is in operation. In the art, as indicated above, if a specific powder moisture is desired, the spray dryer is operated, the obtained powder moisture is analyzed in a lab, and the moisture obtained from the analysis is input to the APC, which compares it with the target moisture and based on that modifies the required process settings for the spray dryer, such as feed flow rate, main air temperature and main air flow rate. By providing the predicted moisture as soft sensor to the APC, the reaction time of the APC is much faster and the target moisture can be achieved faster and more accurately. If there is no APC connected to the spray dryer, the predicted powder moisture can also be used with (semi-) manual control performed by the operator(s) of the spray dryer, as a soft sensor. [0016] The sensor information comprises at least one of inlet air sensor information and fluid feed sensor information. The inlet air sensor information may comprise information about at least one of the flow rate, humidity and temperature of the air flows that enter the main spray dryer, and the internal and/or external fluidized bed(s). The fluid feed sensor information may comprise information about at least one of the flow rate, temperature and density of the fluid feed. The sensor information may be received from the sensors located on or near the spray dryer while in operation, and this information thus allows the computing system to receive real time information from the inputs being fed to the spray dryer. With this information, the computing system is able to obtain a predicted powder moisture and feed it to the APC of the spray dryer so that it can react faster and fine tune the settings of the spray dryer to more accurately and faster reach the target moisture. According to embodiments, the plurality of sensors located on and around the spray dryer comprise at least two sensors, preferably at least ten sensors, and most preferably between ten and one hundred sensors.

[0017] According to embodiments of the present invention, for obtaining the information about the predicted powder moisture, the computing unit is configured to perform the simulation by using the first information, the second information and the third information with operations of energy and mass balances to obtain humidity information of an air outlet of the spray dryer, using the humidity information and the equilibrium distance information to obtain product water activity (content) information, and using the product water activity information and the sorption isotherm information to obtain the predicted powder moisture. In other words, the humidity information and the equilibrium distance information can be used to obtain powder moisture via the powder specific sorption isotherm. The advantage of this approach is that process dependent parameters (such as the heat loss and the equilibrium distance) can be separated from product dependent parameters (such as the sorption isotherm). The separation of parameter dependency results in the transferability of the recipes from pilot to factory and factory to factory.

[0018] According to embodiments of the present invention, for obtaining the information about the predicted powder moisture, the computing unit is configured to perform the simulation by using the first information, the second information, and the third information with operations of energy and mass balances, using the heat loss information, the equilibrium distance information and the sorption isotherm information to obtain the predicted powder moisture and humidity and or temperature of the air leaving the spray dryer system.

[0019] According to embodiments of the present invention, the at least one predicted powder property comprises powder moisture and powder flow rate, and the computing unit is configured to use the first information, the second information and the third information to obtain, as a result of the simulation, at least one of predicted air outlet temperature information and predicted air outlet humidity information of the spray dryer.

[0020] According to embodiments of the present invention, the computing unit is further configured to use the powder flow rate, predicted air outlet temperature information and predicted air outlet humidity information to determine whether an operation point of the spray drying system is within a specific stickiness threshold, by determining a stickiness curve using the predicted powder moisture and the glass transition information, and determining whether the operation point of the spray drying system is within a specific region with respect to the stickiness curve. [0021] At least one purpose of the present invention is to obtain a powder product with a desired powder property with accuracy and in a reduced time. Besides determining the most appropriate process settings for initializing the spray dryer, and providing the predicted powder moisture to the APC so that the APC can react faster to steer the spray dryer process, the computing system of the present invention may further or alternatively obtain as a result at least one of powder flow rate information, predicted air outlet temperature information and predicted air outlet humidity information. In some embodiments, this further or alternative result can be used to monitor that the spray dryer is performing within certain thresholds related to the stickiness of the powder. According to embodiments of the present invention, at least one of the predicted powder flow rate information, predicted air outlet temperature information and predicted air outlet humidity information of the spray dryer may be also used by the APC or by an operator (if the APC is not present in the system) as soft sensor.

[0022] The computing system of the present invention is configured to determine a stickiness curve based on the predicted powder moisture and on the glass transition information, and is thereby able to continuously determine, in real time with the spray dryer operation, on which region with respect to the stickiness curve the spray dryer is operating. The stickiness curve may represent values of powder stickiness, such as stickiness temperature, for different values of the air outlet temperature and air outlet humidity, said values delimiting a safe region for spray drying and a dangerous region with risk of blocking of the spray dryer. Herein, stickiness refers to the stickiness of the powder obtained in the spray-dryer. The stickiness threshold refers to a threshold value above which the powder becomes so sticky that it may block the spray-drier. The skilled person is able to determine the stickiness threshold, e.g. based on the specific spray-drier and/or the specific powder that is produced. [0023] According to embodiments of the present invention, the output unit is further configured to output alert information if the operation point of the spray drying system reaches or surpasses the specific stickiness threshold. [0024] According to embodiments of the present invention, the output unit comprises a storage unit configured to store the process setting information so as to be accessed by an operator of the spray dryer and to be used to provide manual inputs to the spray dryer. Although the computing system of the present invention may be configured to operate in communication with the spray drying system and the APC system when it is in operation in the factory, the computing system according to embodiments of the present invention is also configured to operate offline, that is, not in direct communication with an operating spray drying system. In both configurations, online and offline, the computing system is able to control the spray drying system. When the present disclosure refers to controlling the spray dryer, controlling may refer to simulating the spray dryer (in the production line) and providing control information to the spray dyer directly or indirectly. In some embodiments (online configuration), it may refer to simulating or modelling the spray dryer and production line where it is located, and directly providing the result of the simulation to an APC of the spray dryer, or to the operators controlling the spray dryer. In some embodiments (offline configuration), controlling may refer to simulating or modelling the spray dryer and production line, and storing the result so that the result can then be further processed and used by the operators to determine the process settings to be input to the spray dryer.

[0025] In an offline configuration, the computing system of the present invention can perform the simulation or mathematical simulation of a spray dryer located in a production line. The process settings obtained as a result can be stored, or transmitted to other systems, for further analysis so that operators of the factory in which the spray dryer is to be used can use them as guidance to determine the process settings to initialize the spray dryer. [0026] According to embodiments of the present invention, the process setting information comprises at least one of inlet air information and fluid feed information. The inlet air information may comprise information about at least one of humidity, temperature and amount (flow rate) of the air flows that enter the main spray dryer, and the internal and/or external fluidized bed(s). The fluid feed information may comprise information about at least one of flow rate, temperature and density of the fluid feed.

[0027] According to embodiments of the present invention, for obtaining the process setting information for the spray dryer, performing the simulation comprises performing a mathematical optimization. Optimization operations are able to determine the most optimal output parameters for a specific set of input parameters. The mathematical backbone of the mathematical model is the same for the online and offline configuration. However, the mathematical model may be executed in one direction or in the opposite direction depending on the inputs received and the outputs to be generated. In one implementation, in the online configuration, process settings may be input in the form of sensor information, and predicted moisture information may be obtained, while in the offline configuration a target moisture information may be input and process settings may be obtained.

[0028] According to the present invention, a computer implemented method for simulating a spray dryer for providing food powder is provided, performed by a computing system as described above, the method comprising performing the following steps, such as iteratively performing the following steps: receiving first information related to dimensions and setup of a spray dryer in operation, second information comprising process-related or powder- related specific information, and third information including sensor information from a plurality of sensors located on and around the spray dryer while the spray dryer is in operation, performing a simulation of a production line where the spray dryer is located based on the first, second and third information, and obtaining, as a result, information about at least one predicted powder property; and outputting the information about at least one predicted powder property to be used to operate the spray dryer.

[0029] The present invention thus allows to predict upfront the correct process settings. This is done by mathematically modelling the spray dryer located in a production line of a factory, a technique also referred to as a digital twin (DT) of the production line where the spray dryer is located. In addition, the present invention allows that during production the DT extracts real time data from the factory sensors and uses it to predict the powder moisture, powder stickiness and outlet air conditions. This information may be passed to the Advanced Process Control device (an automated process optimizer) and may enable this device to increase the process control.

[0030] According to the present invention, the computer implemented method may additionally or alternatively comprise iteratively performing the following steps: receiving the first information related to dimensions and setup of a spray dryer, the second information comprising process-related or powder-related specific information, and the third information including target information of at least one target powder property, performing the simulation of the production line where the spray dryer is located based on the first, second and third information, and obtaining, as a result, process setting information for the spray dryer; and storing the process setting information. [0031] According to the present invention, a spray drying system for providing food powder is provided, the spray drying system comprising: a computing system as described above, and a spray dryer comprising or operating with a plurality of sensors. According to embodiments, the spray drying system further comprises an advanced process control, APC, system operatively connected to the spray dryer, and configured to receive, as input, at least one of the information about at least one predicted powder property, and to transmit control instructions to the spray dryer based on the input.

Brief description of the drawings

[0032] The present invention will be discussed in more detail below, with reference to the attached drawings, in which:

[0033] Fig. 1 schematically represents a spray dryer with inputs and outputs;

[0034] Fig. 2 schematically shows a computing system and a spray drying system according to embodiments of the present invention;

[0035] Fig. 3 schematically shows a computing system and a spray drying system according to embodiments of the present invention;

[0036] Fig. 4 shows invention flow charts of an operation of embodiments of the present invention;

[0037] Fig. 5 represents a stickiness region for spray drying according to embodiments of the present invention; [0038] Fig. 6 represents example implementations of the computing system and methods according to embodiments of the present invention.

[0039] Fig. 7 represents a second example of a stickiness region for spray drying according to embodiments of the present invention.

Description of embodiments [0040] Fig. 1 schematically represents a spray dryer 100 with inputs and outputs. As the computing system and methods according to the present invention perform a simulation of the production line comprising the spray dryer, and fluidized bed(s), using a mathematical model, and as in some embodiments the actual sensor information measured from the spray dryer is used as input information for the computing system, Fig. 1 is useful to illustrate an example of the information used in the computing system and methods according to the present invention. The example displays a co-current spray dryer, however the invention is not limited to this type of dryer. The invention could be applied to any spray dryer system used for (food, milk) powder production. The liquid feed with the food or milk concentrate is fed to a spray dryer vessel. Under high pressure the liquid (milky) stream is atomized. At the same time hot air is introduced in the vessel. The heat supplied by the hot air evaporates the water from the droplets, and transforms the droplets to a powder product. The powder falls downwards and is collected in the fluidized beds. Here the powder is gradually cooled down to ambient conditions before packing.

[0041] The inputs to the spray dryer include the product or feed 102 to be dried, which generally comprises a mixture of water and suspended or dissolved solids and or emulsified oils, and a gas such as (hot) air 101. Sensors are located on or near the spray dryer, and can measure certain properties of these inputs. For example, sensors located before the entrance of the spray dryer may detect at least one of a density of the feed (product) flow, a flow rate of the feed flow, a temperature of the feed, a flow rate of the inlet air flow, a humidity of the inlet air flow, and a temperature of the inlet air. The fluidized bed(s) of the spray dryer, which can include an internal static fluidized bed, an external (or vibrating) fluidized bed, or both, also receive additional inputs, comprising gas such as warm air 103. It should be noted that when preparing a food powder in a spray dryer it is possible to start with a liquid or paste of some of the components and add further dry components to the spray dried powder, for example in the fluidized bed. In some embodiments, the lactose component is added as a dry powder. Thus, in some embodiments the fluidized bed(s) may also receive as input dry components, such as dry lactose and slightly out of spec product. In an embodiment it is also possible to recycle slightly out of spec product of an earlier spray dry run as a dry powder to the spray dryer. Sensors located on or in proximity to the entrance of the fluidized bed(s) are able to detect for example at least one of a flow rate of the air flow, a humidity of the air flow, and a temperature of the air.

[0042] At least part of the sensor information shown above is used as input to the computing system and method according to at least some embodiments of the present invention, as will be described in more detail in relation with the next figures.

[0043] As output, the spray dryer produces powder product 106, but also air, result from the inputted air and the evaporated liquid comprising the input product. There may be outlet air from the spray dryer 104 and outlet air from the external fluidized bed 105. Sensors located at the outputs of the spray dryer may also be able to measure at least one of the humidity of the outlet air, the temperature of the outlet air, the flow rate of the powder, and the moisture of the powder. The results obtained from the computing system and method according to at least some embodiments of the present invention can be compared with the sensor measurements provided by the sensors on or around the spray dryer, to perform an evaluation of the accuracy of the results of the computing system and method, and recalibrate the method.

[0044] Fig. 2 schematically shows a computing system and a spray drying system according to embodiments of the present invention. The computing system 200 according to embodiments of the present invention comprises a computing unit configured to perform a simulation, by using a mathematical model, of a production line where a spray dryer is located. The computing system according to embodiments of the present invention is configured to operate in two configurations, either alternatively or simultaneously, which differ from each other in the inputs given to the computing unit and the outputs generated by the computing unit, and in the fact that one configuration is executed online and the other configuration is executed offline, but the mathematical modeling is in the two configurations similar. Throughout the description, the computing unit can be understood as a computer executing a series of instructions which may be stored in a memory of the computing system, such that the computing unit may run an application or program to execute the instructions allowing the computing unit to perform the simulation and obtain the results.

[0045] In a first configuration, represented in Fig. 2, the computing system 200 is configured to operate in communication with a spray dryer 100 in operation. In this configuration, the computing system receives as input first information related to dimensions and setup of the spray dryer, and also second information including product related or process related specific (or empirical) information. The specific information may include at least one of sorption isotherm information, glass transition information, heat loss information, and equilibrium distance information. In addition, the computing unit receives as input third information including sensor information from a plurality of sensors located on and around the spray dryer, that is, live production data coming from the sensors located on or in proximity to the spray dryer. As indicated above with respect to Fig. 1, and although not shown in detail in Fig. 2, sensor information may comprise at least two of inlet air sensor information and feed sensor information, more specifically, information about at least one of a density of the feed (product) flow, a flow rate of the feed flow, a temperature of the feed, a flow rate of the inlet air flow, a humidity of the inlet air flow, and a temperature of the inlet air. In addition, for initializing the operation of the computing system in the first configuration, moisture information from an actual lab sample may be additionally input to the computing system, as starting point, to start calibrating the model. With each new predicted powder moisture from the model (mathematical model), the model parameters are re-adjusted. For example, from the predicted moisture, the APC 300 operatively connected to the spray dryer modifies process parameters of the spray dryer, such as for example increasing a temperature of the inlet air. This will be measured by the sensors which will feed the measurement to the computing system, which will in turn re run the mathematical model to obtain a new predicted moisture.

[0046] In the first configuration, the computing system is configured to perform a simulation of the production line where the spray dryer is located, which may include only the spray dryer, based on a mathematical model of that production line, to obtain, as a result, predicted powder moisture information. The computing system supplies the predicted powder moisture information to the APC system connected to the spray dryer, so that the APC can use the received powder moisture information to optimize the factory performance. This is because the computing system is able to calculate a predicted powder moisture from the sensor information received from the spray dryer, so that this predicted powder moisture can be used by the APC of the spray dryer as soft sensor input information, allowing the APC to modify required process settings of the spray dryer in a faster manner than having to rely on actual sensor data and moisture information obtained from lab results. The APC is in this way able to react faster to process variations and control the spray dryer better. Moreover, the APC is able to optimize the process to reach the target moisture content in a faster and more accurate way. The predicted moisture information may also be displayed on a dashboard in the control room of the spray drying system. [0047] It should be noted that when in the present disclosure APC is used, it can be understood as any control unit or control system which can provide control instructions to the spray dryer.

[0048] In addition, in the first configuration, the computing system may also obtain, as result, outlet air temperature information, outlet air humidity information, and powder flow rate information. These results can be used to obtain stickiness information of the powder. The outlet air temperature and humidity information may correspond to the outlet air 104, to the outlet air 105, or to both. The computing system may use the information about powder flow rate, outlet air temperature and outlet air humidity, together with the predicted moisture information, to monitor the performance of the spray dryer, more specifically for continuously positioning an operation point of the spray dryer with respect to a stickiness curve. That is, besides or as alternative to providing the predicted powder moisture to the APC in operation so that the APC can react faster to steer the spray dryer process, the computing system in the first configuration may further monitor that the spray dryer is performing within certain threshold(s) that are related to the stickiness of the powder, which could potentially block the spray dryer.

[0049] The safe region of a spray dryer operation is dictated by the degree of stickiness of the material to dry (powder). If the stickiness of the material (powder) exceeds a certain threshold, the spray dryer is at risk of blocking by excessive powder build-up in the spray dryer vessel and production must be stopped. That is why monitoring the stickiness of the product is of utmost importance when producing powder with spray drying techniques. However, it is usually difficult to determine, in real time, while the spray dryer is in operation, the stickiness level of the product. In systems of the prior art, there is no actual way of analyzing the stickiness of the powder in real time, and there is therefore a risk that the powder reaches a stickiness threshold that will block the spray dryer. The computing system of the present invention is configured to determine a stickiness curve based on the predicted powder moisture and on glass transition temperature calculations, and is able to determine on which region with respect to the stickiness curve the spray dryer is operating. The stickiness curve may be determined using known equations in the art such as the Couchman Karazs equation and Gordon Taylor model, and specific or empirical information input to the computing system, such as the glass transition information. An example of the determination of the stickiness curve is shown later in relation with Fig. 5. The stickiness curve represents values of the powder stickiness for different values of the air outlet temperature and air outlet humidity, wherein the values of the stickiness (such as a stickiness temperature) defining the curve delimit a safe region for spray drying and a dangerous regions with risk of blocking of the spray dryer.

[0050] According to embodiments of the present invention, information representing the spray dryer operation point with respect to the stickiness curve may be displayed in real time - based on the latest predicted powder moisture - on a screen or dashboard which the operators of the spray drying system can consult. For example, the output unit of the computer system may include a display for displaying the predicted moisture information, predicted outlet air information, predicted outlet air humidity information and an operation point with respect to the stickiness curve, for example in the form of a graph or chart. In addition, the output unit may further output alert information if the operation point of the spray drying system reaches or surpasses a specific stickiness threshold. This alert information can be in the form of a visual or audio signal, and the stickiness threshold may be a threshold that indicates that the operation point is reaching a critical point after which the stickiness level will no longer be acceptable, as there will be risk for blocking the spray dryer. The alert information thus allows the operators to take action to avoid that such critical point is not reached, by manually changing the process settings, such as by changing the water content of the product fed to the spray dryer, or by increasing the inlet air temperature.

[0051] The outlet air temperature information, outlet air humidity information, and powder flow rate information may, besides being used to obtain the stickiness curve, also be transmitted to the APC connected to the spray dryer, or output (displayed) so that the operators of the spray dryer use them to adjust process settings for the spray dryer, in a similar way as the predicted powder moisture information is used. The outlet air humidity information may be predicted using mass and energy balances. It may also be determined by humidity sensors located on or around the spray dryer. The outlet air temperature information may also be predicted using the mass and energy balances, but it may also be determined by temperature sensors located on or around the spray dryer.

[0052] In the second configuration, the computing system 200 may not operate in communication with the sensors around an operating spray dryer and an APC connected to the spray dryer. Instead, or in addition or simultaneously, the computing system may perform an upfront assessment of the factory settings when a new product recipe is to be introduced. For this, in the second configuration, the computing system also receives, as input, first information related to dimensions and setup of the spray dryer, and second information comprising process-related or product-related specific information. The information related to the setup of the spray dryer, in the second configuration, also includes information about the feed flow ranges and air flow ranges between which the production line or the factory is limited to operate. The setup information of the spray dryer thus also includes limit conditions for certain parameters such as feed (product) and air flow. However, sensor information is not input to the computing system. Instead, third information including target information about at least one target powder property, such as target moisture information, is input to the computing system. With the information related to the dimensions and setup of the spray dryer, the target moisture information, and the product or process related specific information (process information such as equilibrium distance information, and product information such as sorption isotherm and product glass transition information), the computing system is able to obtain, by performing a simulation of the production line the spray dryer is located and using a mathematical optimization, the most optimal process setting information for the spray dryer. Most optimal may mean providing stable factory production, avoiding product fouling/stickiness, and/or minimizing the energy utilized per kilogram of powder produced. This results into requiring less trial and error, leading to less waste of product and resources, and leading to less loss of capacity.

[0053] The process setting information comprises at least one of inlet air information and feed information. The inlet air information may comprise information about humidity, temperature and amount (flow rate) of the air flows that enter the main spray dryer, and the internal and/or external fluidized bed(s). The fluid feed information may comprise information about at least one of flow rate, temperature and density of the fluid feed.

[0054] That is, the information obtained as a result in the second configuration can be stored and be further processed so that the operators can use it as guidance to set the initial process settings for the spray dryer.

[0055] Fig. 3 schematically shows a computing system 200 according to embodiments of the present invention. The computing system comprises an input unit 201, a computing unit 202, and an output unit 203. As seen above, depending on the configuration in which the computing system is operating, it may receive different inputs and produce different outputs. However, in both cases, the computing unit performs a simulation of the production line in which the spray dryer is or will be located, using the same mathematical model. The model or digital twin includes the mass and energy balances, and the stickiness curve are common to both configurations. [0056] In the first configuration, at least one powder property, such as the powder moisture is predicted. In order to obtain the information about the predicted powder moisture, the computing unit may be configured to perform a simulation by using the first information about the dimensions and setup of the spray dryer, the second information including product-related or process-related specific information, and the third information including sensor information, with mathematical operations of energy and mass balances to obtain a humidity of an air outlet of the spray dryer. Then, the humidity of the air outlet may be used with the equilibrium distance information, also received as input, to obtain product water activity information, and then the product water activity information may be used with sorption isotherm information, also received as input, to obtain the predicted powder moisture. In the operations of energy and mass balances, the heat loss information is also used.

[0057] The equilibrium distance information, which may comprise one or more equilibrium distance parameters, is input to the computing system to estimate a deviation or distance from equilibrium between the outlet air and the powder moisture. The equilibrium distance parameter may be different for different products. The sorption isotherm information may be obtained from a sorption model used to calculate moisture content known in the art, such as the Guggenheim-Anderson-de Boer, GAB, model or the Brunauer-Emmett-Teller, BET, model, or may be obtained from interpolation of equilibrium data obtained with an offline method.

[0058] In the first configuration, also the outlet air humidity and temperature, and the powder flow rate may be predicted, and may be used to monitor the operation point of the spray dryer with respect to a stickiness curve. The outlet air humidity is obtained from the energy and mass operations as seen above, and the outlet air temperature is predicted also by the mass and energy balances. The outlet air temperature and humidity may also be determined by sensor information.

[0059] In the second configuration, as seen above, the process settings for initializing a spray dryer are calculated. The input information is the first information about dimensions and setup of the spray dryer, the second information including the product or process related specific parameters or information (equilibrium distance information, glass transition temperature information, heat loss information, and sorption isotherm information), and the third information including target information of at least one target powder property, such as target moisture information. For obtaining the process settings, the computing system is configured to find an optimal operation point with respect to the stickiness curve, by solving an optimization problem. In order to obtain the process setting information, the computing unit may be configured to perform a mathematical optimization by using the mathematical model, with the information about the dimensions and setup of the spray dryer, the process or product related specific information and the desired target moisture information. With the optimization, the process settings are determined such that the operation point of the spray dryer (outlet air temperature and outlet air humidity) is within a safe region of the stickiness curve. [0060] It should be noted that, as seen above, the mathematical model requires the use of additional specific information or parameters whose value is not obtained from sensors but which may also be input to the computing system. For example, the equilibrium distance parameter required for obtaining the predicted moisture, a K-value or a deltaT used to obtain the glass transition for determining the stickiness curve (as will be seen later in relation with Fig. 5) can be referred to as specific or empirical parameters, which are also input to the computing system, and which can have values which are product dependent, or are given a value taken from comparable products already produced on the spray dryer. [0061] The output unit may comprise a communication unit configured to transmit the information about the predicted powder moisture to the APC. This allows the computing system to communicate with the APC while the spray dryer is operating, in the first configuration. The output unit may also comprise a storage unit configured to store the process setting information for the spray dryer obtained in the second configuration. This information can then be exported or in any way accessed by operators of the spray dryer, and be used as guidance to apply the process settings to the spray dryer.

[0062] Fig. 4 shows invention flow charts of an operation of embodiments of the present invention;. The computing system is configured to perform method steps. As seen in Fig. 4a, in one embodiment, the computer implemented method comprises the steps of: receiving 401 first information related to dimensions and setup of a spray dryer in operation, second information comprising process-related or powder-related specific information, and third information including at least one of sensor information from a plurality of sensors located on and around the spray dryer while the spray dryer is in operation, performing 402 a simulation of a production line where the spray dryer is located based on the first, second and third information, and obtaining, as a result, information about at least one predicted powder property; and outputting 403 the information about at least one predicted powder property to be used to operate the spray dryer. The specific steps taken to perform the simulation and obtain the results, and the specific steps taken to output the results, have been explained above.

[0063] Additionally or alternatively, the computing system may be configured to perform the method steps described in Fig. 4b, which comprise: receiving 411 the first information related to dimensions and setup of a spray dryer, the second information comprising process-related or powder-related specific information, and third information including target information of at least one target powder property, performing 412 a simulation of a production line where the spray dryer is located based on the first, second, and third information, and obtaining, as a result, process setting information for the spray dryer; and outputting 413 and storing the process setting information.

Example [0064] An example of an implementation of the first configuration will be given below.

With respect to the example, Fig. 5 represents a stickiness region for spray drying according to embodiments of the present invention; and Fig. 6 represents an example implementation of the computing system and methods according to embodiments of the present invention. [0065] In this example a factory producing infant milk powder was used. The computing system focusses on predicting powder moisture and stickiness by simulating the spray dryer with an external fluidized bed (EFB). It should be clear that the computing system could also perform these predictions for a spray dryer with a different setup suitable for producing (milk) powders. The boundaries around the computing system are the concentrate (feed, product) and air flows to the spray dryer and all outlet conditions, including the sensors immediately before and after, as shown in Error! Reference source not found.. When there are multiple feed lines of milk concentrate to the dryer some logic filters can be used to identify the active process line.

[0066] The computing system is configured to execute a mechanistic model that assumes a mixture of water and suspended solid is fed to the spray dryer, where it is atomized. The atomized droplets then dry as they travel through the spray drying chamber. The computing system calculates the equilibrium drying conditions (powder moisture, powder flow rate, outlet air temperature and outlet air humidity) in the spray dryer. Exchange of material and energy between the powder and by-pass of the fluidizing gas, is considered as well. The mechanistic model consists of mass and energy balances. All material and energy flows in this example are based on the schematic of Error! Reference source not found, and are listed in Table 1, except for the heat loss, which is the energy flow from the spray dryer system to the environment. The heat loss is treated as a free parameter and needs to be determined from off line data. The preferred way of working is to receive sensor values and/or lab values for all listed variables. The list of Table 1 can be altered to the requirements per site deployment. In case a site deployment requires one of the output variables to be fixed it can be switched around with one of the input variables per request. [0067] Table 1 : Overview of input and output variables.

Model input Model output

*air streams to the main dryer, internal and external fluid bed

[0068] In order to obtain the predicted moisture content of the powder, the mass and energy balances need to be solved for the spray dryer system. It is known from experience that an equilibrium between the outlet air and the milk powder is never realized, therefore a parameter is introduced to estimate the deviation or distance from equilibrium, the so-called equilibrium distance parameter. Then the product water activity is specified as:

[0069] Aw = RH dryer out + Eq. distance

[0070] where RH dryer out represents the relative air outlet humidity, which is obtained with the mass and energy balances.

[0071] A sorption isotherm model is used with the obtained product water activity to calculate the product moisture content, that is, the predicted powder moisture. The model can be, for example, a BET, or GAB model or interpolation of equilibrium data obtained with an offline method. [0072] In order to obtain the stickiness curve, first a calculation of the mixture glass transition temperature is done via the Couchman Karazs equation. The glass transition temperature is calculated based on a dry material. The moisture dependency is included via the Gordon Taylor model, which includes a glass transition parameter, a K-value. In the present invention, the K-value is fine-tuned with off line data and is specific to the spray dryer and the material to be dried. [0073] The moisture input for the Gordon Taylor model is the predicted product moisture content described above. Finally, the glass transition temperature is transformed to a stickiness temperature with the following equation:

[0074] T stick = T glass + deltaT,

[0075] where deltaT in this equation is also an empirical parameter fine-tuned with offline data, or is obtained from typical values from academic literature, for example a difference of 20K.

[0076] The safe region of operation based on T stick can then the displayed in a graph together with the spray dryer outlet conditions (outlet air temperature and outlet air humidity). Fig. 5 shows the operation in the example for two sets of data from two different pilot plant trials data 1 and data 2, where the same milk powder composition was used. [0077] The materials used in this example are a number of commercially available products. The composition of these products is specified in Table 2.

[0078] Table 2: Macro composition of materials

Recipe code Composition A Composition B Composition C Composition D

* vegetable oil with 2% fish oil

[0079] In Fig. 6, the moisture content prediction as obtained in the example with composition A is compared to the laboratory measurements and to the calculation made by the APC. As seen in Fig. 6, the trial was performed for a duration of 7 hours. The crossed- dotted line represents the values of moisture obtained from actual laboratory measurements. The dotted line represents the moisture values calculated by the APC from sensor information. The solid line represents the moisture values predicted by the computing system (digital twin) according to the present invention. The results show two benefits of the computing system prediction: the moisture prediction of the computing system resembles the lab results more closely than the prediction by the APC system, and the output of the computing system calculation is more frequent than the lab results, allowing to react faster to adjust process settings. A third advantage is the minimization of labor and a fourth advantage is that the operation is less prone to human errors. Similar results are obtained for compositions B, C and D. [0080] Fig. 7 represents a second example of a stickiness region for spray drying according to embodiments of the present invention, wherein a different composition is used compared to the compositions discussed in reference to figures 5 and 6. With this composition, the temperature can be higher (e.g. at or above 100 degrees Centigrade, depending on humidity), without risk of spray dryer blocking. The simulation according the second example yielded a temperature of about 90 degrees Centigrade at a humidity of about 33 g/kg.