Method for controlling a continuous granulating and drying process, and apparatus and system therefor

Abstract

The present invention relates to a method for controlling a continuous granulating and drying process, and to an apparatus and a system therefor, wherein a model takes account of a combination of granulation by the granulator and subsequent drying by the dryer and control parameters or predicted formulation parameters are determined by the model on the basis of state parameters of the apparatus; and / or wherein the model comprises a static part, by which a base value is or has been determined for the relevant control parameter or predicted formulation parameter with the state parameters, and the model comprises a dynamic part, by which the base value is optimized by means of a prediction.

Description

[0001] Method for controlling a continuous granulation and drying process and plant and system therefor

[0002] The present invention relates to a method for controlling a plant and a plant according to the preamble of claim 14 and a system comprising the plant.

[0003] The background of the present invention is primarily the production of drug dosage forms, in particular tablets, capsules, or granules. In principle, however, the invention can also be used in other technical fields, particularly when a formulation is to be produced with a predetermined or predeterminable attribute concerning the particles of the formulation, such as a predetermined particle size distribution and / or moisture content. This is the case, for example, when the formulation is to be tabletted, which is particularly preferred in the pharmaceutical field, but is also possible in principle in connection with cleaning agents, foodstuffs, or the like.

[0004] The present invention particularly relates to a method and a plant for the preferably continuous production of a formulation from a feedstock. The plant is particularly preferably controlled by a proposed method, or the plant is designed, such that a formulation with a predetermined or predeterminable property relating to the particles of the formulation, such as homogeneity or particle distribution, and preferably with a predetermined or predeterminable (relative) humidity, is produced from the feedstock.

[0005] The combination of a granulator and a dryer has proven advantageous for producing the formulation from the feedstock. Thus, the granulator can first produce an intermediate product (granulate) with a specified or predeterminable particle distribution from the feedstock, while the intermediate product can then be conditioned to a specified or predeterminable (relative) humidity using the dryer.

[0006] One difficulty here is that the varying properties of the intermediate product affect the drying process and thus the formulation. Fluidized-bed granulators are generally known. In the solution disclosed therein, a specific particle distribution and conditioning with regard to relative humidity are achieved in the same step. However, solutions based on fluidized-bed granulators have disadvantages regarding the reliable generation of an exact particle size distribution and humidity, and are also only suitable for certain feedstocks.

[0007] In comparison, the use of a granulator followed by the use of a separate dryer has proven more advantageous, as the components can also be used separately and, if necessary, a more precise particle size distribution and relative humidity can be achieved, preferably at least essentially independently of each other. However, the proposed method can also be advantageous for controlling fluidized-bed granulators.

[0008] Combinations of a granulator and a dryer are also generally known, initially in batch mode. In a batch process, a batch of the feedstock undergoes a first manufacturing step, followed by a second manufacturing step for the entire batch before the final product, the formulation, emerges from the manufacturing process.

[0009] In contrast, in a continuous production process, which preferably forms the basis of the present invention, after a start-up phase, feedstock is simultaneously added while previously added feedstock undergoes the production process, and previously added feedstock, which has already been fully subjected to the production process, is removed. Thus, in a continuous process, feedstock is simultaneously added and the result in the form of the formulation is removed.

[0010] The present invention preferably relates to a continuous production of the formulation or a continuous process and the plant therefor or the system comprising the plant, preferably in contrast to a batch process. Advantages of a continuous process are:

[0011] - easy scalability (scaling according to time and throughput possible)

[0012] - small space requirement of the system

[0013] - Shorter downtimes of the plant compared to batch systems - Higher degree of automation possible

[0014] - higher product quality

[0015] Against this background, the object of the present invention is to provide a method as well as a plant and a system by means of which the process for producing the formulation from the starting material can be improved with regard to a reliable and consistent realization of attributes such as, in particular, a particle size distribution and moisture.

[0016] This object is achieved by a method according to claim 1, a plant according to claim 14 or a system according to claim 15. Advantageous further developments are the subject of the dependent claims.

[0017] The present invention relates, on the one hand, to a method for controlling a plant for producing a formulation from a feedstock, wherein the production comprises processing the feedstock with a granulator and drying an intermediate product produced from the feedstock with the granulator by means of a dryer.

[0018] In a first variant of the present invention, control parameters of the plant are determined based on a model. The model takes into account predefined target formulation parameters that represent the desired properties of the formulation produced or to be produced.

[0019] The desired properties of the formulation produced or to be produced are in particular an intended grain property and moisture.

[0020] Variable target formulation parameters can be passed to the model, which the model uses to determine the control parameters. However, the target formulation parameters can be or have been considered alternatively or additionally when creating the model.

[0021] The control parameters are or represent manipulated variables for controlling the system's actuators. State parameters of the system are determined and processed by the model. These are preferably passed to the model, which uses them to determine the control parameters.

[0022] The condition parameters each represent a condition of the system that influences production and are preferably sensor values.

[0023] Furthermore, input material parameters are taken into account by the model. In particular, variable input material parameters are passed to the model, which the model uses to determine the control parameters. However, the input material parameters can alternatively or additionally be or have been taken into account at least partially in the creation of the model.

[0024] The feedstock parameters represent an attribute of the feedstock, in particular a moisture and / or a grain property.

[0025] In one aspect, the dryer is coupled to the granulator in such a way that the intermediate product is automatically conveyed from the granulator to the dryer without interruption. The proposed model takes into account the continuous combination of granulation with the granulator and subsequent drying with the dryer.

[0026] In a second aspect that can be combined with the first, the model has a static part with which a base value is determined for the respective control parameter, and a dynamic part with which the base value is optimized by means of a prediction.

[0027] In a second variant of the present invention, actual formulation parameters are predicted based on a (the same or a different) model.

[0028] The model takes into account the control parameters that are specified or can be specified in this aspect. Preferably, variably specified control parameters are passed to the model, which the model uses to predict the actual formulation parameters. However, alternatively or additionally, the control parameters can be or have been at least partially considered in the creation of the model. The actual formulation parameters represent actual properties of the formulation produced or to be produced, in particular an actual grain size property and moisture content.

[0029] As in the first variant, the state parameters are determined and processed by the model and the input material parameters are taken into account by the model.

[0030] Furthermore, as in the previous variant, in one aspect the dryer is coupled to the granulator in such a way that the intermediate product is automatically conveyed from the granulator to the dryer without interruption. The proposed model takes into account the continuous combination of granulation with the granulator and subsequent drying with the dryer.

[0031] In a second aspect that can be combined with the first, the model in the second variant also has a static part with which a base value is determined for the respective control parameter, and a dynamic part with which the base value is optimized by means of a prediction.

[0032] In a proposed process, in order to control the plant for producing the formulation from a feedstock or to support the control, feedstock parameters are or are initially specified or taken into account which represent a state of the feedstock, in particular a moisture content and / or a particle property such as a particle size distribution.

[0033] A feedstock within the meaning of the present invention is preferably a granulatable substance, i.e., a substance that can be processed into granules through a granulation process. The feedstock is most preferably a powder or granule whose particle properties can be modified through a granulation process.

[0034] Furthermore, the feedstock is preferably a mixture of substances, i.e. an at least substantially homogeneous mixture of different, preferably solid components. These components can comprise an active ingredient, in particular a pharmacologically or otherwise active substance, a filler and / or a disintegrant. In particular, the feedstock is an at least substantially homogeneous powder mixture. During production of the formulation from the feedstock, material parameters of the feedstock passing through the plant, via its intermediate product to the final product (formulation), are preferably not determined by sampling and analysis away from the plant. Instead, only parameters that can be measured during ongoing, continuous operation are used.

[0035] Either no material parameters are determined at all, or at most, in-line measurable material parameters are determined, for example, measured values ​​from contactless measurement methods, a reflection and / or transmission measurement, particularly with infrared radiation, for example, as an indicator of material moisture. Measurement of particle size distributions is preferably avoided in the manufacturing process, at least for intermediate products.

[0036] For the proposed control, control parameters of the plant can be determined, which are or represent manipulated variables for controlling the plant's actuators. The plant can therefore be controlled using these control parameters. This is preferably achieved by controlling different actuators of the plant with the control parameters so that they influence the input material or intermediate product.

[0037] Alternatively or additionally, actual formulation parameters are predicted to support control. These represent the properties of the manufactured formulation under specified boundary conditions. This can be done based on the specified or specifiable control parameters.

[0038] In other words, actual formulation parameters can be predicted and preferably output, preferably by manually pre-setting or entering the control parameters. This allows a user to compare them—again preferably manually—with target formulation parameters and specify varied control parameters to adjust the predicted actual formulation parameters to the target formulation parameters. The varied control parameters are then preferably used as the basis for controlling the system.

[0039] Preferably, at least one granulator drive and a supply for a desiccant, in particular (conditioned) air, are controlled using the control parameters. Additionally, a feed device for the feedstock, an injection device for liquid during granulation, one or more temperature control devices of the granulator, a conveying device for the desiccant for adjusting a desiccant volume flow, and / or a temperature control device for controlling the temperature of the desiccant can be controlled using the control parameters.

[0040] To control the system or to support the control, system status parameters, in particular one or more sensor values, are determined, each of which preferably represents a system status that influences the process for producing the formulation from the feedstock. These include, in particular, temperatures and / or pressures or pressure differences and / or torques and / or volume flows. However, other parameters or sensor values ​​that describe a system status are also conceivable.

[0041] However, the state parameters preferably do not describe, or at least not directly, any material properties of the feedstock or of an intermediate product (granulate) or final product (formulation) formed from it.

[0042] In any case, it is preferred that no particle size distribution, no size, no shape, no density, and / or no active ingredient content of the starting material or the intermediate product formed therefrom is / are determined in the continuous granulation and drying process. In this respect, the present invention pursues a completely different approach compared to the prior art. Properties of the starting material, however, can be determined in advance, and properties of the final product, i.e., the formulation, can be determined for verification after completion and / or for the creation of a model.

[0043] Target formulation parameters are or are preferably specified for control purposes, which represent the desired properties of the formulation produced or to be produced, in particular one or more physical properties such as a particle size, particle size distribution, particle shape or density and / or a moisture content of the formulation.

[0044] Furthermore, for modeling and / or verification, actual formulation parameters are determined, preferably by characterizing the formulation, which represent the actual properties of the formulation produced or to be produced, in particular one or more physical properties such as a particle size, particle size distribution, particle shape or density and / or a moisture content of the formulation.

[0045] Finally, it is intended that the control parameters be determined based on the model. The plant can then be controlled during the formulation production process using the control parameters, which are preferably determined by processing the state parameters with the model.

[0046] The (measured) actual formulation parameters are preferably used to derive or define the model. However, the (measured) actual formulation parameters are preferably not used as a basis for the ongoing process of controlling the plant.

[0047] The control parameters are therefore preferably not determined or derived by processing the (measured) actual formulation parameters. Surprisingly, it has been shown that deriving the control parameters from (measured) actual formulation parameters begins too late. If (measured) actual formulation parameters deviate from the target formulation parameters during the ongoing process, considerable waste is already preprogrammed. However, the aim of the invention is to avoid such waste. Rather, it is preferable to control the system (at least substantially) independently of the (measured) actual formulation parameters or to configure the system for this purpose.

[0048] The (measured) actual formulation parameters are preferably used to determine the model or to derive a scheme in the form of the model, whereby the control parameters are derived from the state parameters of the plant during the process.

[0049] The model preferably has a machine-learning-based structure, in particular a neural network.

[0050] As already mentioned, the plant according to a first aspect of the present invention comprises a granulator for processing the feedstock into an intermediate product and a dryer coupled to the granulator such that the intermediate product is automatically conveyed from the granulator to the dryer without interruption. The model considers, and in particular describes, the combination of granulation with the granulator and subsequent drying with the dryer. Based on the plant's condition parameters and preferably the feedstock parameters, the control parameters, in particular base values ​​of the control parameters, can be determined using the model.

[0051] Alternatively or additionally, predicted (i.e., not measured) actual formulation parameters are determined based on the model. This may be a different model than the one used to determine the control parameters. The predicted actual formulation parameters can be output to enable a (manual) comparison with target formulation parameters and, if necessary, a (manual or automatic) adjustment of the control parameters. In this case, the control parameters are preferably specified for the model.

[0052] Optionally, one or more inline measurable properties of the intermediate product, in particular one or more optically measurable parameters on the intermediate product, can be additionally considered. However, measured values ​​whose determination requires sampling, analysis separate from the plant, or interruption of the manufacturing process are preferably avoided.

[0053] It is therefore not excluded that, in addition to the plant's condition parameters, certain (physical) in-line measurable properties of the feedstock and / or the intermediate product and / or the formulation (actual formulation parameters) are determined or used for control or as an input variable of the model, for example a moisture and / or particle size distribution of the formulation 7 or variables determined therewith, in particular insofar as an in-line measurement is possible for the determination without interrupting the continuous granulation and drying.

[0054] Preferably, a measurement of a particle size distribution is provided, if at all or only, for the formulation, but not for the intermediate product. Accordingly, the system can have a sensor for determining an attribute describing a particle of the formulation, such as a particle size distribution, but preferably only at or after the formulation outlet for discharging the formulation after drying of the intermediate product. The sensor provided as part of the system or a system with the system is, in particular, an inline probe with spatial filter anemometry for particle size measurement. However, other principles are also possible here.

[0055] Moisture is preferably determined by the formulation, but can alternatively or additionally also be determined by the intermediate product. One or more sensors can be used for this purpose. In particular, this is an optical sensor, particularly preferably an infrared radiation-based sensor. In this context, sensors based on near-infrared radiation, especially from the NIR-2 spectrum with a wavelength of 860 to 1040 nm, have proven particularly advantageous.

[0056] It has proven to be particularly advantageous, both independently of and synergistically with the modeling of the combination of granulator and dryer, if the model has a static part with which a base value is or is determined for the respective control parameter or predicted actual formulation parameter from the feedstock parameters and the state parameters, and the model also has a dynamic part with which the base value is or can be optimized by means of a prediction.

[0057] The static part of the model can be or become defined by determining model parameters which are or become determined, in particular measured, in a corresponding manner for several stationary (steady-state, at least essentially static) states of the system.

[0058] The static part of the model can be used to determine base values ​​for the control parameters from the state parameters and / or to predict the actual formulation parameters. The static part of the model is, or preferably will be, specified as fixed.

[0059] The dynamic part of the model allows for continuous adjustment of the model's behavior. For example, the behavior of a system changes over time due to wear, material expansion, aging, etc. Instead of directly accounting for such or similar transient effects by changing (in particular, scaling and / or correcting by adding / subtracting correction terms) the values ​​of the control parameters or predicted actual formulation parameters, the proposed model adjustment allows the model to generate corrected control parameters or actual formulation parameters from the state parameters.

[0060] One or more, possibly different, parameters (e.g., the state parameters) are processed by the model. These parameters can be configured to serve as input values ​​for the model, particularly for the static part of the model and / or the dynamic part of the model.

[0061] The model, especially its static part, can comprise an artificial neural network. The parameters can be passed to nodes of an input layer of the artificial neural network, and the artificial neural network can use them to generate values ​​at the nodes of an output layer of the artificial neural network.

[0062] The values ​​generated at the nodes of the neural network of the static part of the model in the output layer can be optimized with the dynamic part of the model, for example by processing taking into account the same and / or other parameters

[0063] A further aspect of the present invention, which can also be implemented independently, relates to a computer program product or computer-readable storage medium, comprising instructions which, when the program is executed by a computer, cause the computer to carry out the method according to one of the aspects described.

[0064] A further aspect of the present invention, which can also be implemented independently, relates to the plant for producing the formulation from the feedstock, wherein the production comprises processing the feedstock with a granulator and drying an intermediate product produced with the granulator by means of a dryer:

[0065] The system includes sensors for detecting system status parameters, each of which represents a state of the system that influences production. The system includes actuators for directly or indirectly influencing the feedstock. Furthermore, the system includes a control device for controlling the actuators with the control parameters, with target formulation parameters being specified or specifiable to the control device, and the model based on which the control parameters can be determined.The granulator is coupled to the dryer in such a way that the intermediate product is automatically conveyed from the granulator to the dryer without interruption, wherein the model takes into account the combination of granulation with the granulator and the continuously subsequent drying with the dryer and the system is designed so that the control device uses the model to determine the control parameters based on the state parameters of the system.

[0066] Alternatively or additionally, the model comprises a static part, wherein the control device is configured to use the static part to determine a base value for the respective control parameter from the input material parameters and the state parameters, and the model comprises a dynamic part, wherein the control device is configured to use the dynamic part to optimize the base value by means of a prediction. The resulting, optimized base value is then preferably used to control the plant, or the plant can be controlled with it.

[0067] A further aspect of the present invention, which can also be implemented independently, relates to a system comprising the proposed plant and a device for forming the feedstock from a plurality of components, preferably powders, preferably by sieving, and / or a device for further processing, preferably tabletting, of the formulation.

[0068] A formulation within the meaning of the present invention is a substance that has passed through the plant and has been modified in terms of its physical properties. The formulation is therefore preferably the product of a combination of granulation and drying. Subsequent further processing of the formulation, which is preferably in the form of granules (dried or preferably conditioned with respect to their (relative) moisture content), for example, by tableting, is not excluded.

[0069] A feedstock within the meaning of the present invention is preferably a substance that is fed into the system or granulator to change its physical properties. The feedstock is preferably an active ingredient-filler mixture. However, this is not mandatory. The feedstock may already be preprocessed, for example, by at least substantially homogeneously mixing a powder with another powder or other substance, one of which may be or contain an active ingredient. An active ingredient is preferably a pharmacologically active substance.

[0070] A granulator within the meaning of the present invention is preferably a device that mechanically processes the feedstock to change its physical properties. Particularly preferably, the granulator converts the feedstock into granules, in particular a coarse (granular) powder. For this purpose, the feedstock can be fed to the granulator as a powder in order to process it into a coarser or finer-grained powder. The intermediate product is preferably a solid.

[0071] The granulator preferably conveys the feedstock while acting on it. The granulator preferably generates pressure and / or friction in the feedstock, preferably with temperature control / heating and / or moisture supply. Particularly preferably, the granulator changes the granularity, grain size, or particle size distribution of the feedstock. In the present invention, the granulate represents an intermediate product that is subsequently further processed.

[0072] The granulator can have an extruder or be formed by an extruder. In particular, it is a screw extruder, preferably a twin-screw extruder, or the granulator has or resembles one. However, other solutions are also conceivable.

[0073] The granulator is preferably a screw granulator, for example a twin-screw granulator. In screw granulators, the feedstock is conveyed and processed by means of a screw shaft rotating around a rotational axis. For this purpose, the screw shaft can have flights of different pitches and / or processing structures. In a twin-screw granulator, two screw shafts are provided, which are arranged in parallel and / or intermeshing and effect conveying and processing. In principle, other concepts can also be used here. Conveying preferably results in extrusion. The screws are therefore or form one or more extruders or effect extrusion of the feedstock.

[0074] The granulator is preferably a granulator for wet granulation. For this purpose, the granulator may have a liquid injection system for moistening the feedstock, after which the moistened feedstock is processed into the intermediate product by the granulator. The granulator may have an opening, in particular a nozzle, or a valve for adding the liquid to the feedstock, in particular for adding water, ethanol, isopropanol, and / or a mixture thereof, for the purpose of (temporarily) increasing the moisture content.

[0075] A dryer within the meaning of the present invention is a device for reducing the (relative) humidity or water content of a substance, in this case for reducing the (relative) humidity or water content of the feedstock processed into an intermediate product using the granulator. The dryer is preferably a device that removes water from the substance / intermediate product. This is preferably a device that brings the substance / intermediate product into contact with a desiccant that removes water from the substance.

[0076] The desiccant is preferably air or another gas with a relative humidity that allows the absorption of water from the substance. The desiccant is preferably temperature-controlled, in particular to a temperature above ambient temperature. The desiccant is therefore, in particular, preconditioned, warm, and / or dried air, also known as process air. In principle, however, it can also be other, particularly inert gases or other, preferably gaseous, desiccants.

[0077] The dryer is preferably a fluidized-bed dryer, also known as a fluidized-bed dryer. A fluidized-bed dryer, as defined in the present invention, is a device that creates a cushion of air or gas (process air) for a substance to be dried, in this case the intermediate product / granulate. In this context, the air or gas represents the preferably gaseous desiccant.

[0078] The preferably gaseous desiccant is fed from the intermediate product / granules into a dryer bed, preferably through a perforated distributor plate. The preferably gaseous desiccant flows through the bed at a velocity such that the particles of the intermediate product / granules are kept in a fluidized state despite their weight. The fluidized particles of the intermediate product / granules form the fluidized bed. They are dried by desiccants, preferably gaseous. Within the fluidized bed, bubbles can form and collapse to promote intensive particle movement. In this state, the solids behave like a free-flowing, boiling liquid. Very high heat and mass transfer rates are the result of the intimate contact between the individual particles and the preferably gaseous desiccant.However, in principle, other dryer concepts can also be used, even if fluidized bed dryers have proven particularly advantageous in the context of the present invention.

[0079] Sensors for detecting state parameters of the system, each of which represents a state of the system that influences production, within the meaning of the present invention are preferably sensors that characterize a state of devices of the system, i.e. are aligned and configured to measure one or more state parameters.

[0080] State parameters in the sense of the present invention may be or represent one or more of the following attributes:

[0081] • Moments and / or speeds of drives, tools and / or conveying equipment (in particular propellers, turbines, screws / extruders) or corresponding parameters such as current consumption and / or speeds

[0082] • Mass flow rates (of desiccant / process gas, feedstock, additive / granulating liquid, intermediate product, exhaust gas, final product / formulation) or corresponding parameters that characterize, for example, positions of valves, flaps, rotor speeds, pressure differences, or the like

[0083] • Temperatures of parts of the plant that are preferably in direct or indirect contact with the feedstock being processed, with the aggregate / granulating liquid or with the desiccant, of tools for processing the feedstock or the intermediate product, of equipment for adding aggregate or introducing the desiccant

[0084] State parameters preferably do not directly characterize any (physical or chemical) property of the feedstock or an intermediate or final product formed from it, or of the formulation. Therefore, in particular, no measured variable of the feedstock or an intermediate or final product (formulation) formed from it that characterizes a chemical composition or granularity is considered a state parameter of the plant.

[0085] Basically, a distinction can be made between two groups of state parameters.

[0086] The first group includes state parameters that are at least essentially independent of the input material, particularly because they are directly specified by an actuator in the system. These are also referred to as predeterminable or "state parameters unaffected by the input material." Examples of these are predeterminable temperatures or speeds. They can be used, in particular, for actuator control.

[0087] A distinction can be made between the "state parameters unaffected by the feedstock" and a second group of state parameters that are influenced by interaction with the feedstock or with the intermediate product formed with it. These are referred to from now on as "state parameters influenced by the feedstock." Examples include a torque that develops depending on the consistency of the feedstock or the (relative) humidity of exhaust air moistened by the drying process.

[0088] One or more state parameters influenced by the feedstock are preferably used as input variables for the model. The state parameters processed by the model or transferred to the model for this purpose and used by the model to determine the control parameters or predict the actual formulation parameters are therefore preferably state parameters influenced by the feedstock.

[0089] The state parameters that are processed by the model or transferred to the model for this purpose and used by the model to determine the control parameters or to predict the actual formulation parameters are preferably at least one state parameter of the granulator that is influenced by the feedstock, more preferably at least one state parameter of the granulator and the dryer that is influenced by the feedstock, in particular at least two state parameters of the granulator that are influenced by the feedstock and at least one, preferably at least two, state parameters of the dryer that are influenced by the feedstock.

[0090] Preferably, one or more of the parameters unaffected by the input material are used to control the corresponding actuator. Alternatively or additionally, one or more state parameters influenced by the input material are used to determine the control parameter(s), preferably using the model. The control parameters can be target specifications for controlling the actuator(s), on the basis of which the actuator(s) are then controlled.

[0091] An in-line measurable property within the meaning of the present invention is preferably a property that can be measured in parallel in an uninterrupted, continuous manufacturing process.

[0092] Actuators for influencing the feedstock within the meaning of the present invention are preferably drives for tools and / or conveying devices. These can be motors, for example, for driving fans, turbines, conveyor belts, and / or screws, or also temperature control devices, heating, or cooling (e.g., a cooling water conveying device) for controlling the temperature of feedstock-carrying housing parts of the granulator. A temperature control device, in particular a heating or cooling device such as a heating register, for controlling the temperature of the process gas can also be an actuator, since this acts indirectly on the feedstock via the temperature of the desiccant / process gas.

[0093] A control device for controlling the actuators within the meaning of the present invention is preferably an electronic component for influencing the operation of the actuators, the control of motor speeds (of the screw drive and / or the fan motor(s)) and / or for controlling or regulating temperature control devices or heaters of the granulator or for the desiccant.

[0094] Control parameters, which are or represent manipulated variables for controlling the actuators, within the meaning of the present invention are preferably values ​​as a specification for the operation of the actuators, in particular one or more predetermined motor speeds (of the screw drive(s) and / or the fan motor(s)) or corresponding current consumptions and / or specifications for the control or regulation of temperature control devices or heaters of the granulator or for the desiccant or the like. Feed material parameters within the meaning of the present invention are preferably parameters which preferably characterize physical properties of the feed material, such as granularity, particle size, particle size distribution, (relative) humidity and / or temperature of the feed material.

[0095] A state of the feedstock within the meaning of the present invention is preferably determined at least by its moisture content and / or a particle size distribution, optionally supplemented by further properties or replaced by corresponding information which allows the moisture content and / or a particle size distribution to be directly or indirectly inferred or derived therefrom.

[0096] Target formulation parameters within the meaning of the present invention are preferably parameters that represent the desired properties of the formulation produced or to be produced, in particular a granularity or particle size distribution and a moisture content, optionally supplemented by further properties or replaced by corresponding information that allows the moisture content and / or a particle size distribution to be directly or indirectly inferred or derived therefrom.

[0097] Actual formulation parameters within the meaning of the present invention are parameters measured on the formulation which represent the properties of the formulation produced or to be produced, in particular a particle size, particle size distribution and a moisture content, optionally supplemented by further properties or replaced by corresponding information which directly or indirectly indicate the moisture content and / or a particle size distribution or are derived therefrom.

[0098] Actual formulation parameters can be predicted as an alternative to measurement, but are then referred to below as predicted actual formulation parameters.

[0099] A model within the meaning of the present invention is preferably an (abstracted) representation that is preferably limited to essential properties. A model here is preferably a model of the plant and represents—preferably mathematically—properties of the plant directly or indirectly through their effects on the feedstock and / or the intermediate product. For this purpose, the model can comprise or be formed by a mathematical description of the granulator, the dryer, and / or their effects on the feedstock and / or the intermediate product. The model preferably makes it possible to determine control parameters or predict formulation parameters based on plant condition parameters and, where applicable, feedstock parameters.

[0100] A coupling of the granulator to the dryer within the meaning of the present invention is preferably a device for transferring the intermediate product into the dryer and can comprise a conveying device for this transfer, such as a chute, a conveyor belt, a screw, or the like. The coupling preferably ensures that the intermediate product is automatically conveyed continuously, without interruption, or "in-line" from the granulator to the dryer. For this purpose, the granulator can also have an outlet that opens directly into the dryer. Here, the coupling is thus effected by the granulator or by the conveying effect of the granulator.

[0101] Base values ​​of the control parameters within the meaning of the present invention are preferably control parameters that serve as basic settings and are preferably determined or specified independently of measured properties of the final product (the formulation). Base values ​​of predicted actual formulation parameters are starting points for the forecast.

[0102] A static part of the model in the sense of the present invention is preferably a part of the model based on empirical values, with which base values ​​can be determined or predicted.

[0103] A dynamic part of the model within the meaning of the present invention is preferably a part of the model that can be changed dynamically depending on state parameters of the plant, i.e. during ongoing operation of the plant during processing of the feedstock into the intermediate and final product, the formulation, in order to adapt the model to any (future) changes in the states of the plant or the process, preferably so that the model preferably represents the behavior of the plant or the process for producing the formulation from the feedstock with sufficient accuracy. The dynamic part of the model preferably makes it possible to optimize the base value(s) by means of a prediction. Past developments can be taken into account for this purpose. Further aspects, advantages and properties of the present invention emerge from the claims and the following descriptions of a preferred exemplary embodiment with reference to the drawing.It shows:.

[0104] Fig. 1 shows a schematic section of the proposed system;

[0105] Fig. 2 shows a simplified, block diagram view of control-relevant components of the system (see Fig. 1).

[0106] Fig. 3 is a simplified schematic view of an artificial neural network;

[0107] Fig. 4 is a schematic diagram of past and predicted time courses;

[0108] Fig. 5 is a schematic diagram of the result of a control with constant control parameters over time;

[0109] Fig. 6 is a schematic diagram of forecasts offset from one another;

[0110] Fig. 7 is a schematic diagram of process properties over time;

[0111] Fig. 8 is a schematic flow diagram;

[0112] Fig. 9 a system embedded in a system.

[0113] In the drawing, the same reference numerals are used for the same or similar parts, whereby the same or similar properties and advantages can be achieved, even if a repeated description is omitted for reasons of clarity.

[0114] Fig. 1 shows a schematic section of a proposed plant 1 for producing a formulation 2 from a feedstock 3. The plant 1 is preferably designed to carry out a process for producing the formulation 2 from the feedstock 3. Feedstock parameters 4 are or will be specified for the feedstock 3, which represent a state of the feedstock 3, in particular a moisture content, composition, and / or a grain property, such as a particle size distribution.

[0115] The system 1 is preferably designed to determine control parameters 5 of the system 1, which are or represent manipulated variables for controlling actuators 6 of the system 1.

[0116] Furthermore, state parameters 7 of the system 1 are preferably determined or can be determined with the system 1, for example via sensors 8, in particular sensor values ​​or variables derived therefrom as state parameters 7, wherein the state parameters 7 each represent a state of the system 1 that influences production.

[0117] The actuators 6 of the plant 1 can be controlled by a control device 9 of the plant 1 in order to control or influence the production process for forming the formulation 3 from the feedstock 2.

[0118] Target formulation parameters 10 can be or become predetermined, which represent the desired properties of the formulation 2 produced or to be produced, in particular a physical property such as a grain property, in particular particle size distribution, and / or a moisture content.

[0119] The target formulation parameters 10 can be stored and / or maintained in a database 10A. The database 10A can be read by the control device 9, or the target formulation parameters 10 can be retrieved by the control device 9 from the database 10A and used for control purposes.

[0120] Actual formulation parameters 11 can be provided or measured that represent actual properties of the formulation 3 produced or to be produced, in particular a physical property such as a grain property such as the particle size distribution and / or moisture content. Actual formulation parameters 11 are or preferably comprise information on attributes of the formulation 2 that are measured in a separate analysis process, in particular one that is separate from the system 1, preferably not in-line or not in real time / delayed. The system 1, in particular the control device 9, is preferably designed to determine the control parameters 5 based on a model 12.

[0121] The proposed model 12 can alternatively or additionally be used independently of the control device 9 or without direct influence or preferably automatic control of the actuators 6, preferably for a forecast and / or output of forecast (expected under given boundary conditions) actual formulation parameters 11 of the formulation 2. For this purpose, one or more properties of the formulation 2 can be forecast using the model 12 from one or more state parameters 7 and control parameters 5 and, preferably, feedstock parameters 4.

[0122] In one aspect of the present invention, the plant 1 comprises a granulator 13 for processing the feedstock 2 into an intermediate product 14 and a dryer 15 for the intermediate product 14.

[0123] The dryer 15 is coupled to the granulator 13 such that the intermediate product 14 is conveyed automatically and / or without interruption from the granulator 13 into the dryer 15. Granulation and drying preferably form a joint, continuous process.

[0124] A simplified, block diagram view of control-relevant components of system 1 is shown in Fig. 2.

[0125] The control device 9 is preferably designed to control the system 1 or the combination of granulator 13 and dryer 15. The control is preferably based on the model 12, which describes the behavior of the system 1 or the combination of granulator 13 and dryer 15 and, based on at least one or more state parameters 7, enables the determination of control parameters 5 for controlling actuators 6 of the system 1 or the combination of granulator 13 and dryer 15. It is particularly preferred that the model 12 takes into account the combination of granulation with the granulator 13 and the subsequent drying with the dryer 15.

[0126] The model 12 is preferably used to determine the control parameters 5, in particular initially base values ​​of the control parameters 5, preferably based on the state parameters 7 of the plant 1 and, further preferably, based on the feedstock parameters 4.

[0127] For this purpose, the model 12 can have a static part 12A with which a base value is or is determined for the respective control parameter 5 taking into account the input material parameters 4 and based on the state parameters 7, and the model 12 can have a dynamic part 12B with which the (respective) base value can be adapted, preferably optimized by means of a prediction.

[0128] In other words, the static part 12A of the model 12, preferably formed by machine learning, is used to determine base values ​​(basic settings) of the control parameters 5, which can then be adjusted and finalized by means of the dynamic part 12B of the model 12 to ultimately be used to control the actuators 6.

[0129] The dynamic part 12B of the model 12 preferably fine-tunes the base values ​​or basic settings determined by the static part 12A of the model 12. The base values ​​of the control parameters 5 determined by the static part 12A are preferably independent of transient behavior, i.e., the behavior during runtime, for example, under the influence of environmental influences, tolerance changes, wear of the system 1 or the manufacturing process of the formulation 2.

[0130] In contrast, the adjustment values ​​for the control parameters 5, or correspondingly adjusted settings or control parameters 5, determined with the dynamic part 12B are those that take transient or runtime effects into account, for example, via one or more forecasts. This is preferably done by taking past developments into account, in particular by means of time-series forecasting.

[0131] For this purpose, adjustment values ​​for the base values ​​of the control parameters 5 or correspondingly adjusted settings or control parameters 5 can be determined, which preferably take into account a comparison of predicted actual formulation parameters 11 with the target formulation parameters 10 and, by dynamically adjusting the base values ​​of the control parameters 5, bring the actual formulation parameters 11 closer to the target formulation parameters 10, in particular based on current state parameters 7 at runtime. An advantageous feature of this approach is that the base values ​​and adjustments can be determined using different methods. Alternatively or additionally, it is provided that at least the base values ​​are determined using a Kl or a machine learning method, preferably differently than the adjustments.Furthermore, it is particularly preferred that the determination of the base values ​​also depends on the state parameters 7, i.e. both base values ​​and adjustments (correction terms) for determining adjusted or corrected base values ​​are variable.

[0132] It is therefore possible and preferred that the base values ​​depend on the state parameters 7, but are independent of the historical and predicted development of the state parameters 7. On the other hand, both the base values ​​of the control parameters 5 determined with the static part 12A of the model 12 can be changed depending on the (current) state parameters 7, as can correction terms or adjustments determined with the dynamic part 12B of the model 12, which adjust or correct the base values.

[0133] The static part 12A of the model 12 differs from the dynamic part 12B of the model 12 preferably in that with the static part 12A the base values ​​are determined or can be determined without predictions of the influence of changes in control parameters 5 and / or past changes in the state parameters 7, while the dynamic part 12B takes into account an influence of changes in control parameters 5, future changes in state parameters 7 and / or forecasts resulting from past changes in control parameters 5 or state parameters 7, more preferably taking previous changes into account.

[0134] In summary, the static part 12A of model 12 preferably differs from the dynamic part 12B of model 12 in that they implement different methods to determine base values ​​for the control parameters 5 in the case of the static part 12A and correction terms for the base values ​​of the control parameters 5 or corrected / adjusted base values ​​as final control parameters 5 with or from the state parameters 7 and, preferably, taking into account the feedstock parameters 4. The state parameters 7 considered by model 12 as input variables are preferably state parameters 7 influenced by feedstock 3. In this case, model 12 preferably uses at least one state parameter 7 each of the granulator 13 and the dryer 15 influenced by feedstock 3.

[0135] Optionally, model 12 additionally considers as input a (exclusively) in-line measurable formulation property parameter 7K, which describes a physical or chemical property of formulation 2 that is measurable in the uninterrupted, continuous process.

[0136] Alternatively or additionally, the model 12 considers as an input variable an (exclusively) in-line measurable intermediate product property parameter 7L, which describes a physical or chemical property of the intermediate product 14 that is measurable in the uninterrupted, continuously running process. The measurement can be carried out using an intermediate product property sensor 8L, in particular a (NIR) sensor for determining an indicator for a moisture content (a water content) of the intermediate product 14.

[0137] However, the consideration of other parameters is not excluded. Furthermore, the definition of Model 12 can consider parameters that cannot be measured in-line, while Model 12 preferably does not require non-in-line measurable parameters as input variables during operation.

[0138] The static part 12A of the model 12 can implement a K1 or machine learning method, while the dynamic part 12B of the model 12 preferably implements a different, preferably non-K1 or machine learning-based method or no neural network 32. However, it is not excluded that both different methods are K1 or machine learning-based, or that neither is.

[0139] It has surprisingly been found that the control of plant 1 is particularly reliable and accurate, meaning that the actual formulation parameters 11 are particularly close to the target formulation parameters 10 when the previously described aspects are combined. In summary, it is therefore surprisingly particularly advantageous to: control the combination of granulator 13 and dryer 15 or to enable the control, and / or

[0140] • with direct coupling of the plant components without external or non-in-line analysis of the intermediate product 14, and / or

[0141] • excluding any use of analysis results of the intermediate product 14 that are not measured or measurable in-line - i.e. not in the continuously running process or on the continuously moving intermediate product 14 - and / or

[0142] • that base values ​​of the control parameters 5 are determined / can be determined with the static part 12 A of the model 12 using a first, preferably machine-learning-based method, and / or

[0143] • that the base values ​​of the control parameters 5 are determined / can be determined without taking into account the influence of changes in the control parameters 5 changing state parameters 7 in the determination, and / or

[0144] • that dynamic components of the control parameters 5 are determined or can be determined with the dynamic part 12B of the model 12; or if the base values ​​are corrected with the dynamic components or the system 1 is designed to do so, and / or

[0145] • that the dynamic components of the control parameters 5 are determined or can be determined with the dynamic part 12B of the model 12 by means of a different, preferably time-series forecasting or non-machine learning based method than the method used in the static part 12A of the model 12 or supported thereby, and / or

[0146] • that the dynamic part 12B of the model makes and takes into account predictions of the influence of changing state parameters 7 with changes in control parameters 5 or the system 1 is designed for this purpose, and / or • that the base values ​​for the control parameters 5 are corrected or adjusted with the dynamic components or correction terms are determined / can be determined for this purpose or the system 1 is designed for this purpose, and / or

[0147] • that the final control parameters 5 are the base values ​​adjusted with the dynamic components and that the system 1 is controlled with these control parameters 5.

[0148] As shown in Fig. 1, the plant 1 can have several different actuators 6 to act directly or indirectly on the feedstock 3, whereby the intermediate product 14 and ultimately the formulation 2 are formed.

[0149] A control parameter 5 corresponds to the respective actuator 6, which can be a control variable for the actuator 6 or correspond to it in order to control the behavior of the actuator 6.

[0150] The actuators 6 of the granulator 13 that can be controlled via corresponding control parameters 5 include one or more of the following actuators 6: a feed conveyor drive 6A that can be controlled by means of a feed conveyor drive control parameter 5A, preferably for metering or conveying the feedstock 3 to the granulator 13, and / or a granulator drive 6B that can be controlled by means of a granulator drive control parameter 5B, preferably for driving the granulating unit 17 or one or more screws, and / or an injection device 6C that can be controlled by means of an injection device control parameter 5C, preferably for injecting granulating liquid 13B, and / or a granulator temperature control device 6D that can be controlled by means of a granulator temperature control device control parameter 5D, preferably for temperature control (heating and / or cooling) of parts of the granulator that come into contact with the feedstock 3 13.

[0151] The actuators 6 of the dryer 15 that can be controlled via corresponding control parameters 5 include one or more of the following actuators 6: a supply air conveyor drive 6E, preferably a fan, for supplying air 27 to the dryer 15, and / or a supply air temperature control device 6F, preferably for temperature-regulating the air 27, that can be controlled by means of a supply air temperature control device control parameter 5F, and / or an exhaust air conveyor drive 6G, preferably for removing air 27 after the drying process, that can be controlled by means of an exhaust air conveyor drive control parameter 5G, and / or a fluidized bed drive 6H, preferably for rotating a carousel of the dryer 15, that can be controlled by means of a fluidized bed drive control parameter 5H.

[0152] The sensors 8 for measuring corresponding state parameters 7 of the granulator 13 include one or more of the following sensors 8: a feeder sensor 8A for measuring a feeder parameter 7A that characterizes the feed, in particular a throughput and / or a speed of the feed - preferably a feed rate or dosage that can be represented in [kg / h]; and / or a granulator sensor 8B for measuring a granulator parameter 7B that characterizes a functional property of the granulator 13, in particular a torque and / or a speed of its drive 6B, and / or a throughput - preferably a feed rate of the granulator that can be represented in [kg / h] or a corresponding speed [rpm] of the extruder(s) / screw(s);and / or an injection sensor 8C for measuring an injection device parameter 7C that characterizes an addition of liquid 13B, in particular water and / or alcohol, to the feedstock 3 in the region of the granulator 13, in particular a throughput and / or a quantitative ratio compared to the feedstock - preferably a spray rate that can be represented in [g / min]; and / or one or more granulator temperature sensors 8D for measuring one or more granulator temperatures 7D, in particular temperatures of the granulator 13 or (indirectly) of the feedstock 3 at different positions along a transport path for the feedstock 3 through the granulator 13 - preferably in [°C] or corresponding.;

[0153] The sensors 8 for measuring corresponding state parameters 7 of the dryer 15 include one or more of the following sensors 8: a supply air conveyor sensor 8E for measuring a supply air conveyor parameter 7E, which characterizes a supply air conveyance of the dryer 15, in particular a (mass) throughput, a (corresponding) pressure difference and / or a speed of the supplied air 27, preferably in [m A3 / h]; and / or a supply air sensor 8F for measuring a supply air parameter 7F that characterizes a property of the supplied air 27, in particular a temperature - preferably representable in [°C] - and / or a humidity of the supplied air 27, preferably representable in %, wt. %, or [°C] for the dew point; and / or an exhaust air sensor 8G for measuring an exhaust air parameter 7G that characterizes an exhaust air conveyance of the dryer 15, in particular a (mass) throughput, a (corresponding) pressure difference, a temperature of the air 27 discharged from the dryer 15 and / or a humidity thereof; and / or a fluidized bed sensor 8H for measuring a fluidized bed parameter 7H, which preferably characterizes a drying-relevant, variable property relating to the fluidized bed of the dryer 15, in particular a speed such as a carousel speed of a carousel of the dryer - preferably representable in [rpm].

[0154] Alternatively or in addition to the use of sensors 8, properties of actuators 6 or control parameters 5 can also be used to determine or derive one or more of the aforementioned or corresponding variables. These can then be used as the basis for a control system.

[0155] However, when determining the state parameters 7 influenced by the input material 3, measurement using sensors 8 is mandatory. These state parameters 7 are also used as unit values ​​for the model 12.Optionally provided as part of the system 1, alternatively or additionally externally, are a formulation temperature sensor 8I for measuring the formulation temperature 7I, which characterizes a temperature of the formulation 2; and / or a formulation outlet quantity sensor 8J for measuring a formulation outlet quantity parameter 7J, which characterizes a property of the production and / or of the system 1 with regard to the production of the formulation 2, in particular an outlet quantity of the formulation 2; and / or a formulation property sensor 8K, which characterizes an in-line measurable formulation property parameter 7K of the formulation 2, in particular a particle size, particle size distribution, granularity, and / or moisture, particularly preferably a particle size or size distribution (XD10, XD50, XD90) and / or a residual moisture content and / or a drying loss of the intermediate product 14 during processing into the formulation 2.

[0156] The aforementioned control parameters 5, actuators 6, state parameters 7 and sensors 8 are particularly preferred examples with regard to the embodiment of a particularly preferred combination of a granulator 13 - preferably a (twin-)screw granulator 13 - with a dryer 15 - particularly preferably a fluidized bed dryer 15.

[0157] It is understood that for other granulators 13 and / or dryers 15, other control parameters 5, actuators 6, state parameters 7, and sensors 8 are possible alternatively or additionally. The invention is therefore preferably not limited to the aforementioned control parameters 5, actuators 6, state parameters 7, and sensors 8. Furthermore, it is not necessary that all control parameters 5, actuators 6, state parameters 7, and sensors 8 be implemented or used below. Different selections are possible here.

[0158] The production of formulation 2 using system 1 is explained in more detail below using the exemplary embodiment shown in Fig. 1. It is understood that, although the invention can in principle be implemented particularly preferably and advantageously using system 1 described below, it is not limited thereto. In particular, it is not necessary that all of the described components of system 1 or the steps performed therewith are or will be implemented.

[0159] Thus, it is conceivable to use the invention in cases where other granulator and / or dryer technologies are used. Thus, a granulator type other than a twin-screw granulator 13 can be used. Alternatively or additionally, a dryer type other than a fluidized-bed dryer 15 can be used. Notwithstanding this, the invention has proven particularly advantageous in this context.

[0160] In the embodiment according to Fig. 1, the plant 1 comprises the granulator 13 for processing the feedstock 3 into the intermediate product 14 and the dryer 15 for forming the formulation 2 from the intermediate product 14.

[0161] The granulator 13 has a feed conveyor 16 for feeding and / or dosing the feed material 3. The feed conveyor 16 can have a particularly funnel-shaped storage container 16A that holds the feed material 3. Furthermore, the feed conveyor 16 can have a feed device 16B that feeds the feed material 3 at a specific feed rate (amount per unit time) to a granulation unit 17 of the granulator 13 coupled to the feed conveyor 16.

[0162] The feed conveyor 16 can have a drive 6A for the feed device 16B, in particular a motor for driving a screw 16C. However, other feed principles than a screw 16C are also possible, such as a conveyor belt. As already explained above, the drive 6A is preferably controllable by means of the feed parameter 7A. The feed rate can preferably be adjusted using the feed conveyor drive control parameter 5A.

[0163] The feed conveyor 16 preferably has one or more sensors 8A for measuring the throughput and / or a speed - preferably corresponding thereto.

[0164] The feed conveyor 16 is preferably followed by a granulating unit 17 of the granulator 13. The granulating unit 17 is designed to change the grain size or particle size distribution of the feedstock 3. To this end, the granulating unit 17 can physically act on the feedstock 3, in particular by kneading and / or rolling. In the illustrated example, the granulating unit 17 has at least one screw 18, preferably a twin screw. The screws 18 of the twin screw preferably mesh and transport the feedstock 3 while physically acting on it to change the grain size or particle size distribution of the feedstock 3.

[0165] The granulation unit 17 or the screw(s) 18 may have different processing zones 19. In particular, different screw pitches and / or surfaces may be provided to achieve the desired processing of the feedstock 3.

[0166] The granulating unit 17 can have the granulator drive 6B to drive the granulating unit 17, in particular the screw(s) 18. As already explained above, the granulator drive 6B is controllable via the granulator drive control parameter 5B, preferably with regard to throughput and / or speed, in particular the rotational speed of the screw(s) 18.

[0167] The granulator 13 preferably has the granulator sensor 8B, with which a parameter representing the granulation speed and / or processing intensity can be measured, in particular a speed, a throughput or - very particularly preferably - a torque (of the granulating unit 17 / the screw(s) 18).

[0168] The granulator 13 may have an injection device 6C for injecting liquid 13B for mixing with or admixing with the feedstock 3. This may be a sprayer, but alternatively also a dropper or generally a device for adding a liquid substance.

[0169] The injection device 6C is preferably arranged in the region of an inlet 13A or in the first half or in the first third of the transport path formed by the granulating unit 17 for the feedstock 3 between the inlet 13A and an intermediate product outlet 21 for the feedstock 3 of the granulator 13 processed by the granulator 13.

[0170] The granulator 13 preferably has at least one, but preferably several granulator temperature sensors 8D for measuring the temperature 7D of the feedstock 3 or a corresponding temperature 7D of the granulator 15 or its housing 22 at a wide variety of positions.

[0171] In the illustrated example, more than three and / or fewer than ten granulator temperature sensors 8D are provided for measuring the temperature 7D of the feedstock 3 or a corresponding temperature 7D of the granulator 15. The temperature sensors 8D are preferably distributed (at least substantially equidistantly) along the transport path for the feedstock 3 in the granulator 15 or granulating unit 17.

[0172] A coupling device 20 preferably enables a continuous and / or uninterrupted transfer of the intermediate product 14 from the intermediate product outlet 21, which may be formed by the housing 22 of the granulator 13, to an intermediate product inlet 24 of the dryer 15, preferably formed by a housing 23 of the dryer 15.

[0173] Preferably, it is not intended that the intermediate product 14 be stopped for intermediate storage, at least not for more than one, two, or five minutes. Thus, in particular, it is not intended to carry out or wait for a sampling and analysis of the intermediate product 14 that is separate from the system or delays the production process. Rather, within the meaning of the present invention, it is preferred to feed the intermediate product 14 to the dryer 14 at least substantially uninterruptedly and / or continuously.

[0174] One or more of the granulator temperature sensors 8D can be provided to measure a temperature 7D of the intermediate product 14 or a temperature 7D corresponding thereto, in particular of the granulator 15 in the region of the coupling device 20.

[0175] The dryer 15 preferably has a fluidized bed 25. Furthermore, the dryer 15 preferably has an air inlet 26 for the intake of (process) air 27, an optional diffuser 28 for uniformly feeding the fluidized bed 25 with the air 27, and an air outlet 29 for discharging the air 27 after passing through the fluidized bed 25. In the region of or upstream of the air outlet 29, the dryer 15 optionally has a separator 30, such as a cyclone separator or filter, for capturing particles of the intermediate product 14 or the formulation 2 from the air 27. Finally, the dryer 15 preferably has a formulation outlet 31 for discharging the formulation 2, i.e., the intermediate product 14 dried by the dryer 15.The dryer 15 dries the intermediate product 14 entering through the intermediate product inlet 24 with the air 27 and then discharges the formulation 2 formed thereby through the formulation outlet 31, preferably after passing through the fluidized bed 25.

[0176] In order to supply the air 27 to the dryer 15, in particular to the fluidized bed 25, the dryer 15 preferably has an air supply conveyor 15A with an air supply conveyor drive 6E. This can be a fan or, in general, a device for transferring and / or compressing air 27.

[0177] The dryer 15 can have the supply air conveyor sensor 8E, with which a throughput of the air 27 or a corresponding variable can be measured. In particular, the supply air conveyor sensor 8E is a pressure sensor for measuring the air pressure on the discharge side or the side of the supply air conveyor drive 6E facing the fluidized bed 25, or a differential pressure sensor for determining a differential pressure across the supply air conveyor 15A. Alternatively or additionally, the supply air conveyor sensor 8E can be or have a variable associated with the supply air conveyor drive 6E, such as a rotational speed (fan speed), a current consumption, a torque, or the like.

[0178] The supply air conveyor 15A or its supply air conveyor drive 6E can be controlled by means of the supply air conveyor drive control parameter 5E, in particular with regard to a throughput, a speed (of the fan), a power consumption, and / or a pressure or differential pressure. The differential pressure can be a pressure across the supply air conveyor 15A, but alternatively or additionally also across the fluidized bed 25 or the like.

[0179] The air 27 is preferably tempered, in particular heated, before being fed to the fluidized bed 25 or another drying device of the dryer 15. For this purpose, the dryer 15 preferably has the supply air temperature control device 6F, preferably a heating register. The supply air temperature control device 6F can be controlled by the supply air temperature control device control parameter 5F or can be designed to do so. The temperature and / or (relative) humidity of the conditioned air 27 brought or to be brought into contact with the intermediate product 14 for the purpose of drying is preferably measured. For this purpose, the dryer 15 can have the supply air sensor 8F, which measures the temperature and, alternatively or additionally, a (relative) humidity of the air 27 as the supply air parameter 7F or can be designed to do so.

[0180] The supply air sensor 8F can be provided between the fluidized bed 25 and the supply air temperature control device 6F or the supply air conveyor 15A or measure the properties of the air 27.

[0181] After the air 27 has been brought into contact with the intermediate product 14 for drying, the air 27 is discharged through the dryer 15. This preferably occurs via the air outlet 29, which is distinct from a formulation outlet 31 for discharging the formulation 3.

[0182] The air 27 can be conveyed from the air outlet 29 by means of the exhaust air conveyor 15B, in particular a (second) fan. For this purpose, the exhaust air conveyor drive 6G can be provided, which effects the conveying or drives the exhaust air conveyor 15B.

[0183] The exhaust air conveyor 15B can be controlled or regulated using the exhaust air conveyor drive control parameter 5G. In particular, it is provided that the exhaust air conveyor 15B is controlled such that no air 27 escapes through the intermediate product outlet 21 and the formulation inlet 31. For this purpose, the flow rate of the exhaust air conveyor 15B can equal or exceed the flow rate of the supply air conveyor 15A.

[0184] An exhaust air conveyor sensor 8G can measure a throughput, a speed, a temperature and / or humidity of the air 27 discharged from the dryer 15 after the drying process and / or a corresponding variable such as a speed of the exhaust air conveyor drive 6G or impeller of the fan as exhaust air parameter 7G.

[0185] The fluidized bed 25 can have or form a processing zone, particularly on the side facing away from the air inlet 26. The fluidized bed 25 or a structure delimiting it in the direction of the air inlet 26, such as a sieve or perforated plate or carousel, can be driven, particularly set in motion. For this purpose, the fluidized bed drive 6H can be provided, which can be controlled with the fluidized bed drive control parameter 5H.

[0186] The fluidized bed sensor 8H can measure the fluidized bed parameter 7H, which can characterize a property of the fluidized bed 25 such as a movement of the carousel.

[0187] After drying the intermediate product 14 with the dryer 15, the processed feedstock 3 is output as formulation 2.

[0188] During or after the formulation 2, one or more actual formulation parameters 11 can be determined, i.e., parameters that describe physical or chemical properties of the produced formulation 2. This can be done in-line, i.e., in an uninterrupted process. Alternatively or additionally, actual formulation parameters 11 can also be determined subsequently by means of laboratory testing. Actual formulation parameters 11 that cannot be measured in-line are preferably used as the basis for model 12, so that the modeling takes into account actual formulation parameters 11 that cannot be measured in-line. However, actual formulation parameters 11 that cannot be measured in-line are not used directly for the control of system 1 or as input variables for control system 9 or model 12.

[0189] The system 1 can have one or more sensors 8 for the inline characterization of properties of the formulation 2. These include one or more of the formulation temperature sensor 8I for measuring a formulation temperature 7I, the formulation outlet quantity sensor 8J for measuring a formulation parameter 7J that describes the outlet quantity or throughput (mass flow) of formulation, and / or the formulation property sensor 8K, which measures one or more properties of the formulation 2 and outputs them as formulation property parameter 7K, preferably the (relative) humidity or residual moisture and / or the drying loss.

[0190] Formulation parameters 10, 11 preferably comprise at least one parameter characterizing particles of formulation 2, such as a particle size or particle size distribution (XD10, XD50, and / or XD90) or a corresponding parameter. Alternatively or additionally, formulation parameters 10, 11 preferably comprise a (relative) humidity or residual moisture content and / or a loss on drying (LoD) and / or a corresponding parameter.

[0191] The supplementary formulation parameters 10, 11 can be determined as required by the formulation temperature sensor 8I, the formulation outlet quantity sensor 8J and / or the formulation property sensor 8K or as formulation temperature 7I, formulation outlet quantity parameter 7J and / or formulation property parameter 7K.

[0192] In principle, corresponding measured variables, such as the particle size or particle size distribution (XD10, XD50 and / or XD90) of formulation 2, can be used to form model 12. However, it is not mandatory or in every case intended to determine corresponding variables during ongoing operation of plant 1 or to feed them into its control system.

[0193] Optionally, but preferably, an in-line measurable property, in particular moisture, of the intermediate product 14 can be determined as an intermediate product property parameter 8L by means of an intermediate product (humidity) sensor 8L. This can, if provided, also be taken into account in the model 12, in particular used as an input variable, or (additionally) be used as the basis for the control of the system 1.

[0194] Actual formulation parameters 11 can be measured downstream of system 1. This will be discussed in more detail below. However, it should be noted at this point that parameters characterizing the shape, size, or distribution of particles in formulation 2 are also preferably measured in-line.

[0195] On the other hand, a particle characterizing measurement of the intermediate product 14 is preferably avoided.

[0196] In connection with the embodiment according to Fig. 1, various sensors 8 for determining state parameters 7 of the system 1 have been described. However, it is not mandatory that all sensors 8 are provided or state parameters 7 are used. Preferably, at least two or at least three state parameters 7 or

[0197] Sensors 8 are used for the granulator 13 and the dryer 15 respectively.

[0198] The control parameters 5 which are particularly preferably dynamically adjustable for the system control, the selection of which for the further aspects of the invention can represent an independent idea of ​​the invention, include:

[0199] On the granulator 13 side: the feeder drive control parameter 5A, preferably characterizing a dosage of the feed material 3; and / or the granulator drive control parameter 5B, preferably characterizing an extruder speed of an extruder of the granulator 13, which can form a granulating unit 17 of the granulator 13, or a (other) variable corresponding to a conveying or processing speed of the granulator 13; and / or the injection device control parameter 5C, in particular a spray rate characterizing the amount of liquid 13B supplied per unit time.

[0200] On the dryer 15 side: the fluidized bed drive control parameter 5H, preferably characterizing a speed of a carousel of the dryer 15; the supply air conveyor drive control parameter 5E, preferably representing an inlet-side supply air flow, and / or the supply air temperature control device control parameter 5F, in particular representing the temperature of the inlet-side inflow of the air 27.

[0201] The particularly preferred parameters 3, 7 as input variables for the system control or the model, the selection of which for the further aspects of the invention can represent an independent aspect of the invention, include:

[0202] On the granulator 13 side: the granulator parameter 7B, in particular characterizing an extruder torque of the extruder of the granulator 13 or another parameter of the granulator drive 6B that depends on the consistency and / or feed rate of the feedstock 3 being processed; and / or the one or more granulator temperatures 7D, in particular characterizing one or more temperatures of the feedstock 3 being processed in the granulator 13 at preferably different positions along a material flow of the feedstock 3 in the granulator 13 or temperatures corresponding thereto; and / or the granulator temperature 7D or the intermediate product property parameter 7L, which is the temperature of the intermediate product or corresponds thereto.

[0203] On the dryer 15 side: the exhaust air parameter(s) 7G, in particular characterizing the exhaust air temperature and / or (relative) humidity and / or throughput (for example, represented by a pressure difference) of the exhaust air 27; and / or the formulation outlet quantity parameter 7J, in particular characterizing the outlet quantity and / or a pressure difference in connection with the output of the formulation, for example via a filter or sieve.

[0204] On the feedstock 3 side: the feedstock parameter 4, preferably characterizing a moisture and / or granularity of the feedstock 3.

[0205] Optionally, the following can be additionally taken into account for the control of plant 1 on the formulation 2 side: the formulation property parameter 7K, preferably characterizing a particle size distribution, in particular D10, D50 and / or D90, and / or a loss on drying.

[0206] The use of the aforementioned parameters 4, 5, 7 can be supplemented by one or more of the parameters 4, 5, 7 discussed previously and subsequently. Fig. 3 shows a simplified, schematic view of an artificial neural network 32 for determining control parameters 5 for controlling the system 1 or the actuators 6 thereof.

[0207] According to one aspect of the present invention, the model 12, in particular the static part 12A of the model 12, is formed by the artificial neural network 32 or comprises the artificial neural network 32. An example of the structure of the artificial neural network 32 is shown in Fig. 3.

[0208] The artificial neural network 32 has, in an input layer 33, nodes 36 in the form of input nodes which correspond to one or more of the feedstock parameters 4, to one or more of the state parameters 7 of the granulator 13, to one or more of the state parameters 7 of the dryer 15 and / or to one or more of the target formulation parameters 10.

[0209] Preferably, the artificial neural network 32 has nodes 36 in one or more hidden layers 34, via which the input layer 23 can be linked to an output layer 35.

[0210] The artificial neural network 32 may have nodes 36 in the form of output nodes in an output layer 35, which correspond to one or more of the control parameters 5.

[0211] The nodes 36 of different layers 33, 34, 35 can be connected to each other by edges 37. The nodes 36 can form a graph by means of the edges 37. The nodes 36 and / or edges 37 preferably have weights 38 that specify the connections of the nodes 36 that can be represented by the edges 37.

[0212] The artificial neural network 32 is or is preferably trained with data sets that consist of different combinations of the input material parameters 4, state parameters 7 and control parameters 5 as well as actual formulation parameters 11 that are adjusted under these specifications.

[0213] The training data sets each represent a steady state of the plant 1 in which the actual formulation parameters 11 and state parameters 7 have assumed an at least essentially static value based on constant control parameters 5 and feedstock parameters 4.

[0214] The artificial neural network 32 is or is preferably trained by supplying the input nodes 36 with at least one, preferably several, state parameters 7 of the granulator 13, preferably influenced by the feedstock 3, and at least one, preferably several, corresponding state parameters 7 of the dryer 15, preferably influenced by the feedstock 3, of the respective training data set.

[0215] The input nodes (nodes 36 in the input layer 33) are or are preferably each further supplied with one or more corresponding actual formulation parameters 11 and with one or more corresponding input material parameters 4 of the respective training data set.

[0216] By specifying the parameters 4, 5, 7, 11 of a training data set in the input layer 33, control parameters 5 (values ​​of the neural network 32) result at the output nodes (node ​​36 of the output layer 35), from which errors are preferably determined by comparison with the control parameters 5 of the respective training data set, and the errors are reduced or compensated by adjusting the weights 38 of the artificial neural network 32, preferably by means of backpropagation and / or successively.

[0217] In order to ultimately determine control parameters 5 by means of the neural network 32, target formulation parameters 10 together with further current parameters 4, 7 are specified to the artificial neural network 32 (at the input layer 33) instead of actual formulation parameters 11, whereupon control parameters 5 for the control of the system 1 (at the output layer 35) result, on the basis of which the system 1 can be controlled or is (automatically) controlled.

[0218] The parameters 4, 5, 7, 10, for which nodes 36 are provided, are preferably at least:

[0219] Preferably in input layer 33:

[0220] At least one node 36 to at least one corresponding feedstock parameter 4, preferably the moisture content of the feedstock 3 and / or a property of the particles of the feedstock 3, in particular its particle size distribution; and / or

[0221] • A node 36 to the granulator parameter 7B, in particular the screw or extruder torque of the granulator 13; and / or

[0222] • One node 36 to the granulator temperature 7D, in particular several nodes 36 to several granulator temperatures 7D; and / or

[0223] • A node 36 to an exhaust air parameter 7G, in particular to the temperature and humidity of the air 27 and / or the pressure difference of the air 27 across the separator 30;

[0224] Preferably in output layer 35:

[0225] • A node 36 to a granulator drive control parameter 5B; and / or

[0226] • A node 36 to an injector control parameter 5C; and / or

[0227] • A node 36 to a fluidized bed drive control parameter 5H; and / or

[0228] • A node 36 to a supply air conveyor drive control parameter 5E; and / or a node 36 to a supply air temperature control device control parameter 5F.

[0229] Optionally, nodes for one or more of the following state parameters 7 are provided in the input layer 34:

[0230] • A node 36 to an injector parameter 7C;

[0231] • A node 36 to a supply air parameter 7F;

[0232] • A node 36 to a supply air conveyor parameter 7E;

[0233] • A node 36 to a fluidized bed parameter 7H;

[0234] • A node 36 to a formulation temperature 7I;

[0235] • A node 36 to a formulation outlet quantity parameter 7J; and / or

[0236] • A node 36 to a formulation property parameter 7K, and / or to one or more of the following control parameters 5 in the output layer 35 is provided:

[0237] A node 36 to a feeder drive control parameter 5A;

[0238] A node 36 to a granulator tempering device control parameter

[0239] 5D; and / or a node 36 to an exhaust conveyor drive control parameter 5G.

[0240] Thus, it is preferred that the state parameters 7, to which a node 36 corresponds in each case, include the granulator parameter 7B, in particular a granulating unit torque of the granulator 13, one or more of the granulator temperatures 7D at different positions along a transport path of the granulator 13 for the feedstock 3, the formulation temperature 7I, in particular a temperature of the formulation 2 at the formulation outlet 31 of the dryer 15, a formulation property parameter 7K, in particular the humidity, in particular relative humidity, of the formulation 2 at the formulation outlet 31 of the dryer 15, and / or an exhaust air parameter 7G, in particular characterizing a pressure loss across a filter, here by way of example (cyclone) separator 30, of the dryer 15.

[0241] The artificial neural network 32 is preferably trained to generate control parameters 5 from the parameters 4, 7, 10 fed into the nodes 36 of the input layer, with which control parameters 5 the system 1 or the combination of granulator 13 and dryer 15 can be controlled. For this purpose, the control parameters 5 generated by the artificial neural network 32 are preferably, but not necessarily, optimized using the dynamic model 12B before they become the basis for controlling the system 1.

[0242] With the model 12, the control parameters 5 are preferably determined only or primarily on the basis of the feedstock parameter(s) 4, state parameter 7 and target formulation parameter 10.

[0243] It remains the case that, in this case, non-in-line measurable properties of the intermediate product 14 are preferably ignored. Preferably, only the temperature and / or humidity of the intermediate product 14 are taken into account.

[0244] The model 12 preferably takes into account future effects of changes in state parameters 7 on the actual formulation parameters 11, preferably by means of predictions. For this purpose, the dynamic part of the model 12 is preferably designed to take into account changes in the actual formulation parameters 11 caused by long-term effects. The controller 9 preferably takes into account future effects of changes in the control parameters 5 on the actual formulation parameters 11 and / or state parameters 7, preferably by means of predictions. In particular, the dynamic part of the model 12 is designed to pre-compensate for changes in the actual formulation parameters 11 caused by long-term effects.

[0245] For this purpose, the model 12 can base the determination or adjustment of the control parameters 5 on predictions about future developments of the actual formulation parameters 11. For this purpose, the model 12 can have, in addition to the static part 12A, with which a base value is or is determined for the respective control parameter 5 from the state parameters 7 and, preferably, the input material parameters 4 and the target formulation parameter(s) 10, the dynamic part 12B, with which the base value is optimized by means of a prediction.

[0246] With the dynamic part 12B of the model 12, a change in actual formulation parameters 11 when the current state parameters 7 change can be predicted based on input material parameters 4 and current state parameters 7, and based on this forecast, the base values ​​for the control parameters 5 can be adjusted and the plant 1 can be controlled with the adjusted control parameters 5.

[0247] The model 12, in particular the dynamic part 12B of the model 12, is thus designed to predict the long-term effects of changes in the control parameters 5 on the actual formulation parameters 11 and / or state parameters 7. The control system 9 is thus configured by the model 12 to control the system 1 or the combination of granulator 13 and dryer 15 while compensating for the long-term effects. Thus, surprisingly, and despite the direct coupling of granulator 13 and dryer 15, production can be achieved while maintaining small / permissible deviations of the actual formulation parameters 11 from the target formulation parameters 10.

[0248] If only a prediction of properties of the formulation 2 is desired or realized with the model 12, the model 12 or the artificial neural network 32 can be constructed differently than in the case of a preferably fully automatic control by means of the model 12, namely preferably in such a way that properties characterizing the formulation 2, in particular one or more predicted actual formulation parameters 11, can be determined and / or output.

[0249] In this case, the model 12 preferably determines or predicts one or more actual formulation parameters 11 only or primarily on the basis of the input material parameter(s) 4, state parameter 7 and specified control parameter 5.

[0250] With the dynamic part 12B of the model 12, a change in actual formulation parameters 11 can be predicted, preferably based on input material parameters 4, current state parameters 7 and / or control parameters 5 or their base values, taking into account a change in the state parameters 7 associated with a change in the control parameters 5, and based on this forecast, the resulting actual formulation parameters 11 can be predicted and, preferably, output.

[0251] The model 12, in particular the dynamic part 12B of the model 12, is thus preferably designed to predict long-term effects of changes in the state parameters 7 on the actual formulation parameters 11. This surprisingly provides the user with an indicator for selecting suitable control parameters 5, despite the direct coupling of the granulator 13 and dryer 15 with the predicted actual formulation parameters 11.

[0252] Regardless of whether control parameters 5 or predicted actual formulation parameters 11 are used by the model 12, non-in-line measurable properties of the intermediate 14 are preferentially ignored, for example, a physical property of the intermediate 14 that characterizes particles of the intermediate 14. In particular, all properties of the intermediate 14 are ignored, except for the temperature and humidity of the intermediate 14.

[0253] Model 12 therefore preferentially considers only those parameters of the feedstock 2 or the intermediate product 14 being processed that can be measured in-line, i.e., do not require sampling and analysis separate from the plant.

[0254] As already mentioned, the model 12 can alternatively be configured to predict the actual formulation parameter 11. In this case, the model 12 or artificial neural network 32 is configured differently than the artificial neural network 32 for determining the control parameters 5.

[0255] The parameters 4, 5, 7, 10, for which nodes 36 are provided, are in the case of determining predicted actual formulation parameters 11 with the model 12 preferably at least:

[0256] Preferably in input layer 33:

[0257] • A node 36 to at least one corresponding feedstock parameter 4, preferably the moisture content of the feedstock 3 and / or a property of the particles of the feedstock 3, in particular its particle size distribution; and / or

[0258] • A node 36 to a granulator parameter 7B, in particular a screw or extruder torque of the granulator 13, and / or

[0259] • One node 36 to the granulator temperature 7D, in particular several nodes 36 to several granulator temperatures 7D; and / or

[0260] • A node 36 to an exhaust air parameter 7G, in particular to the temperature and humidity of the air 27 and / or the pressure difference of the air 27 across the separator 30; and / or

[0261] • A node 36 to a granulator drive control parameter 5B; and / or

[0262] • A node 36 to an injector control parameter 5C; and / or

[0263] • A node 36 to a fluidized bed drive control parameter 5H; and / or

[0264] • A node 36 to a supply air conveyor drive control parameter 5E; and / or

[0265] • A node 36 to a supply air temperature control device control parameter 5F.

[0266] Preferably in output layer 35:

[0267] • one or more nodes 36 to (each) a (predicted) actual formulation parameter 11 .

[0268] Optionally, nodes 36 are provided for one or more of the following state parameters 7 in the input layer 34:

[0269] • A node 36 to an injector parameter 7C and / or

[0270] • A node 36 to a supply air parameter 7F and / or

[0271] • A node 36 to an air supply conveyor parameter 7E and / or • A node 36 to a fluidized bed parameter 7H and / or

[0272] • A node 36 to a formulation temperature 7I and / or

[0273] • A node 36 to a formulation parameter 7J and / or

[0274] • A node 36 to a formulation property parameter 7K and / or

[0275] • A node 36 to a feeder drive control parameter 5A and / or

[0276] • A node 36 to a granulator tempering device control parameter 5D and / or

[0277] • A node 36 to an exhaust air conveyor drive control parameter 5G.

[0278] In this case, the artificial neural network 32 can be trained with corresponding or the same data sets as the artificial neural network 32 for determining the control parameters 5.

[0279] Fig. 4 shows a schematic diagram of past and predicted trends of one or more parameters 5, 7, 10 and 11 .

[0280] To the left of the Y-axis representing a value of a parameter 5, 7, 11 is, as indicated by an arrow for the past P, the past development of the one or more parameters 5, 7, 11 as well as the past course of a reference trajectory 39 and a measured parameter 7, 11. In addition, as indicated by an arrow for the future F, the future development of these in a forecast horizon 40 is shown.

[0281] According to the proposal, one or more of the parameters 5, 7, 10 can be adjusted at discrete times tk+ P changed or a change is predicted. In the example, the times tk+ Pspaced apart by a sampling time Δt. In principle, however, it is not absolutely necessary for the sampling time Δt to be constant, although this is possible, and the sampling time Δt can be chosen to be short, so that the course of the parameter(s) 5, 7, 11 can be at least essentially continuous.

[0282] As can be seen in Fig. 4, it is possible that the forecast of the parameter(s) 5, 7, 11 can exhibit different, both rising and falling, courses with the aim of approximating a measured value such as the state parameter 7 and / or the actual formulation parameter 11 of the reference trajectory 39. Ultimately, the course according to Fig. 4 represents a possible system behavior, which takes into account an advantageously optimized control of the system 1 or the manufacturing process by means of the model 12 and the development of various variables taken into account both in the past and in the future.

[0283] Fig. 5 shows a schematic diagram of the result of a control with constant control parameters 5 over time. In a predetermined, preferably fixed time window 41, one or more state parameters 7 result due to the at least substantially constant control parameter(s) 5. The state parameter(s) 7 preferably approach asymptotically a fixed value, taking into account quality dynamics 42, which can be measured at discrete points in time.

[0284] Based on one or a combination of the constant control parameters 5 and the resulting state parameter(s) 7, the static part of the model 12A can be determined. In particular, a machine-learning-based model 12A of the static system behavior can be generated on this basis. As previously explained, this can be an artificial neural network 32, but other machine-learning-based methods are also conceivable.

[0285] Fig. 6 shows a schematic diagram of temporally offset forecasts over time t. At each time point tk+ PFuture measured values ​​K depend on previous developments and current control parameters 5 or state parameters 7. Such a regression can be solved by means of a time series forecast, as indicated in Fig. 6. Shown are courses of the measured values ​​K AT different relative times, which are offset from one another by a time difference or sampling time Δt and by means of which it is indicated that these are forecasts of the time series.

[0286] Specifically, it is possible and preferred that predictions made using the model 12 are updated at regular intervals Δt. As soon as the state of the plant 1 has changed and thus one or more state parameters 7 deviate from previous values, a changed or adjusted forecast can be generated - preferably using the model 12 - in particular of one or more control parameters 5, state parameters 7 and / or actual formulation parameters 11. Fig. 7 shows a schematic diagram of process properties over time. The basic idea is to combine the static behavior, as explained, for example, with the dynamic behavior and predicted developments taking past courses into account, so that, as shown in Fig.7, a process is started with static base values, preferably determined by the static part 12A of the model 12, and then during production, by repeated, iterative optimization, in particular by means of a time forecasting approach, the desired attributes, in particular actual formulation parameters 11, can be achieved in a short time and subsequently at least substantially maintained.

[0287] In principle, it is therefore possible to determine the control parameters 5 based on the static part 12A of the model 12 at the start of the continuous processing of the feedstock 3 to the formulation 2 and to easily adjust the control parameters 5 during ongoing operation of the plant 1 using the dynamic part 12B of the model 12. This is indicated in the time periods in Fig. 7 in the range in which the sampling times Δt are entered.

[0288] It is therefore possible that plant 1 initially reaches a quasi-steady state before the readjustment is activated by means of the dynamic part 12B of model 12 and then preferably begins by means of time-series forecasting at intervals of the sampling time Δt. In principle, however, other control strategies are also possible.

[0289] The proposed system 1 can be used particularly advantageously in a system 45 for producing tablets. An expanded process for this purpose is shown in Fig. 8 using a schematic flow diagram. Components for processing the feedstock 3 and / or for post-processing the formulation 2, such as a preferably pneumatic conveying system 46, can be provided in the system 45, which adds additional functions to the system 1, as explained in more detail below with reference to Fig. 9.

[0290] Fig. 9 shows a plant 1 embedded in a system 45. The preparation of the feedstock 3 can be arranged upstream of the plant 1. In the illustrated example according to Fig. 9 and referring to the method according to Fig. 8, the feedstock 3 is produced from components 47 of the feedstock 3 by sieving and / or mixing in a preparation step 48; in this case, by way of example and also overall, preferably a powder mixture of the components 47.

[0291] In a granulation step 49, which preferably takes place continuously with the subsequent drying step 50, the feedstock 3 is then processed into the intermediate product 14 by means of the granulator 13, preferably with the addition of liquid 13B, also called granulation liquid. In the continued continuous process, the intermediate product 14 is then dried by means of the dryer 15, ultimately producing formulation 2. Regarding the continuous processing of the feedstock 3 into formulation 2, please refer to the previous sections.

[0292] Formulation 2 can then be further processed. In one or more post-processing steps 51, formulation 2 can be sieved, for example in a post-processing device 52, in particular for particle selection, to form a post-processed (in particular sieved) formulation 53. Alternatively or additionally, formulation 2 or post-processed formulation 53 is mixed with additives such as disintegrants and / or binders 56 in a mixing process 55 to form a final mixture 54.

[0293] Finally, a dosage form 58, in particular one or more tablets, can be produced from the formulation 2 or the post-processed formulation 53 or the final mixture 54 by means of a tabletting process 57, in particular a compression process.

[0294] The proposed, preferably pharmaceutical process focuses on the continuous process steps of producing solid oral dosage forms, as exemplified by a schematic flow diagram in Fig. 8 and subsequently explained using the system 45 from Fig. 9.

[0295] After an optional, preferably batchwise, preparation of a homogeneous powder premix of components 47 as feedstock 3, preferably by sieving and / or mixing the powders forming feedstock 3 as components 47, the intermediate product 14, preferably wet granules (wet or moist granules), is formed in a continuous granulation step using the granulator 13, preferably a so-called twin-screw granulator (TSG). For this purpose, liquid 13B can be added to the feedstock 3 in the granulator 13 or during the granulation process. This can be done using the injection device 6C.

[0296] The intermediate product 14 is preferably transferred in a continuous product stream directly into the continuously operating dryer 15, in this case a fluidized-bed dryer. After the two continuous process steps of granulation and drying, formulation 2 is produced, preferably a dry granulate (dried wet granulate).

[0297] Formulation 2 is optionally and preferably subsequently sieved (to form the post-processed formulation 53) and / or mixed with an extragranular phase (an additive / disintegrant and / or binder 56) to obtain the final mixture 54. This final mixture 54 (final mixture) preferably forms the starting material for a tabletting process 57, is used for tabletting, or the system 45 is configured for this purpose.

[0298] Preferably, the two continuous process steps of granulation and drying are provided as Plant 1. However, the system preferably combines Plant 1 into a total of at least three, in particular four, different process units or production steps.

[0299] The feeding and twin-screw wet granulation process units are responsible for the continuous granulation process. The following two process units—continuous fluidized bed drying and pneumatic conveying system—are responsible for the continuous drying process.

[0300] A total of - preferably six - control parameters 5 (main input variables) are defined for the control / control device 9 of the continuous granulating and drying plant 1 or are used for the control.

[0301] The at least two, preferably at least three control parameters 5 or main input variables for the granulator 13 are or preferably include the dosing quantity [kg / h] (of the feedstock 3), the extruder speed [rpm] (of the granulator 13) and / or the spray rate [g / min] (of the injection device 6C).

[0302] For the dryer 15, the at least two, preferably at least three control parameters 5 or main input variables are or include the supply air flow [m 3 / h] (into the dryer 15), the supply air temperature [°C] (of the supply air into the dryer 15) and / or the carousel speed [rpm] (of the dryer 15 or a carousel thereof).

[0303] Since a total of preferably at least six main input parameters or control parameters 5 are defined for the system 1 or are used for control, it is easy to understand that understanding the relationships, steering and controlling such a multifactorial system 1 can be challenging.

[0304] A schematic view of the material and data flow of the continuous granulating and drying plant 1 or the system 45 formed thereby is shown in Fig. 8 and 9.

[0305] With regard to material flow, it is preferred that the system 45 operates according to a bin-to-bin approach or is designed to do so. This means that a first container feeds a preferably homogeneous powder premix as feedstock 3 into the continuous line (consisting of granulator 13 and dryer 15) in order to obtain the dried granules as formulation 2 in a second container after the two continuously running process steps—preferably twin-screw wet granulation and fluidized-bed drying.

[0306] The powder premix as feedstock 3 is preferably processed in batches by system 45. Further processing of the dry granulate (Formulation 2) is preferably also carried out in batches, as shown by way of example in Fig. 9.

[0307] In contrast to fully continuous production (from raw material to finished tablets), this bin-to-bin process offers greater flexibility, as the System 45 itself can be used modularly. One advantage, therefore, is embedding the continuous granulation and drying process into a bin-to-bin process in order to achieve the aforementioned advantages while maintaining high flexibility.

[0308] Preferably, three types of data are considered for controlling the system 1 or the system 45. The first is control parameters 5 and preferably includes the six relevant control parameters 5, as already described above.

[0309] The second type of data are state parameters 7 relating to process states (e.g. temperatures, pressure losses, torques), which can be measured via several sensors 8 throughout the system 1, preferably in an online and / or real-time mode.

[0310] The third data type preferably comprises one or more feedstock parameters 4 and / or actual formulation parameters 11. These critical material attributes (of the feedstock 3, the intermediate 14 and / or the formulation 2) are preferably measured as in-process control, preferably separately from the plant 1, discontinuously and / or on the basis of samples and therefore form the third data type with a time delay.

[0311] Preferably, a complete data set with all three data types forms the basis for the creation of the model(s) 12 used in the context of the invention.

[0312] As significant surprising advantages of the continuous granulation and drying plant 1 achieved by the invention, two relevant aspects can be highlighted in particular, even if these are not mandatory:

[0313] The multifactorial interplay of - for example, six - main input variables (control parameters 5 and state parameters 7 that can be directly influenced by control parameters 5 or are unaffected by the feedstock 3) and the resulting output variables is to be understood. The resulting state parameters 7 that are influenced by the feedstock (for example, 10 states; measured online and in real time) as well as the material attributes (preferably attributes characterizing the feedstock 3 and the formulation 2, described with, for example, a total of four material parameters or actual formulation parameters 11; measured offline and with a time delay) can be defined as output variables.

[0314] Furthermore, the continuous granulation and drying system 1 preferably combines various (continuous) process steps. Therefore, the process parameters or control parameters 5 of one process unit also influence the process states or state parameters 7 of the subsequent units and, ultimately, the material properties or actual formulation parameters 11.

[0315] Considering these two aspects, it becomes clear that optimal manual control of the process is challenging and unlikely to produce a good result. Therefore, a major advantage of the preferred machine-learning-based method, which provides predictions and control adjustments to this continuous process, becomes clearly apparent.

[0316] In summary, the following aspects can be implemented individually or in various possible combinations in Annex 1: Integrated continuous process: not a completely continuous process (from the active ingredient and excipients to the final product), but rather the replacement of at least two traditional batch processes by one continuous process step; combination of batch processes with continuous steps = integrated continuous process

[0317] - Use of a flat-bottom dosing device, which enables a very precise and easily controllable mass flow, as feed conveyor 16;

[0318] - Use of a twin-screw wet granulator as granulator 13;

[0319] - The use of a continuous fluid bed dryer as dryer 15 enables dual functionality; the fluid bed dryer can be converted into a fluid bed granulator. This modular and more variable system enables dual use of the main equipment, thereby increasing plant productivity, efficiently utilizing the plant's footprint, and reducing downtime.

[0320] - use of a special continuous dryer 15 with a slowly rotating carousel that divides the large fluidized bed chamber into several small chambers, thus avoiding the formation of partial batches;

[0321] - Reduction of potential mass buildup of wet granules (Intermediate 14) through shorter transport routes and by avoiding valves for wet granules; inline data generation from NIR probes and / or particle size measurement system (Intermediate Sensor 8L / Formulation Property Sensor 8K) for measuring material properties of wet granules (Intermediate 14) and / or dry granules (Formulation 2).

[0322] The invention further relates to a computer program product or computer-readable storage medium comprising instructions which, when the program is executed by a computer, cause the computer to carry out the proposed method / steps of the method or parts thereof.

[0323] Individual aspects of the present invention can be implemented separately from one another, but also in different combinations. In particular, aspects described in connection with the system 45 can be combined or can be advantageous in combination with the aspects previously described in connection with the combination of the granulator 13 and the dryer 15.

[0324] List of reference symbols:

[0325] 1 Appendix 7C Injector parameters

[0326] 2 Formulation 7D Granulator temperature

[0327] 3 Input material 40 7E Supply air conveyor parameters

[0328] 4 Input material parameters 7F Supply air parameters

[0329] 5 Control parameters 7G Exhaust air parameters

[0330] 5A Feeder drive control parameters 7H Fluid bed parameters 7I Formulation temperature

[0331] 5B Granulator drive control parameters 45 7J Formulation outlet flow meter parameters

[0332] 5C Injector control parameters 7K Formulation property parameters

[0333] 5D Granulator temperature control device 7L Intermediate product property control parameters so Parameters

[0334] 5E Supply air conveyor drive control parameter 8A Feed conveyor sensor

[0335] 5F Supply air temperature control device- 8B Granulator sensor

[0336] Control parameter 8C injection sensor

[0337] 5G Exhaust air conveyor drive control parameter(s) 55 8D Granulator temperature sensor(s)

[0338] 5H Fluid bed drive control parameters 8E Supply air conveyor sensor 8F Supply air sensor

[0339] 6 Actuator 8G Exhaust Air Conveyor Sensor

[0340] 6A Feeder drive 60 8H Fluid bed sensor

[0341] 6B Granulator drive 8I Formulation temperature sensor

[0342] 6C injection device sor

[0343] 6D Granulator temperature control device8J Formulation outlet quantity sensor

[0344] 6E Supply air conveyor drive 65 8K Formulation properties

[0345] 6F Supply air temperature control sensor

[0346] 6G exhaust conveyor drive 8L intermediate product sensor

[0347] 6H Fluidized bed drive 9 Control device

[0348] 7 state parameters 10 target formulation parameters

[0349] [7] State parameter value 70 10A database

[0350] 7A Feeder parameters 11 Actual formulation parameters

[0351] 7B Granulator parameters 12 Model 12A static part 36 nodes

[0352] 12B dynamic part 37 edge

[0353] 13 Granulator 35 38 Weight

[0354] 13A Granulator inlet 39 Reference trajectory

[0355] 13B Granulating liquid 40 Forecast horizon

[0356] 14 Intermediate product 41 Fixed time window

[0357] 15 Dryers 42 Quality Dynamics

[0358] 15A Supply air conveyor 40 43 Planning dimension

[0359] 15B Exhaust air conveyor 44 Process state

[0360] 16 feeder 45 system

[0361] 16A Storage container 46 Conveyor system

[0362] 16B Feeding device 47 components

[0363] 16C Snail 45 48 Preparation step

[0364] 17 Granulation plant 49 Granulation step

[0365] 18 screw 50 drying step

[0366] 19 Processing zone 51 Post-processing step

[0367] 20 Coupling device 52 Post-processing device

[0368] 21 intermediate product outlet so 53 post-processed formulation

[0369] 22 Housing 54 final mixture

[0370] 23 Housing 55 Mixing process

[0371] 24 Intermediate product inlet 56 Disintegrating and / or binding agent

[0372] 25 Fluidized bed 57 Tableting process

[0373] 26 Air inlet 55 58 Dosage form

[0374] 27 Air

[0375] 28 Diffuser t Time

[0376] 29 Air outlet tk+p time

[0377] 30 separator At sampling time

[0378] 31 Formulation Outlet 60 P Past

[0379] 32 artificial neural network F future

[0380] 33 Input Layer Relative Time

[0381] 34 Hidden-Layer K measured value

[0382] 35 output layers

Claims

Patent claims:

1. Method for controlling a plant (1) for producing a formulation (2) from a feedstock (3), wherein the production comprises processing the feedstock (3) with a granulator (13) and drying an intermediate product (14) produced from the feedstock (3) with the granulator (13) by means of a dryer (15), wherein based on a model (12): (a) control parameters (5) of the plant (1) are determined, which are or represent manipulated variables for controlling actuators (6) of the plant (1), wherein the model (12) takes into account predetermined target formulation parameters (10) which represent desired properties of the formulation (2) produced or to be produced, in particular an intended grain property and moisture, or (b) actual formulation parameters (11) are predicted, which represent actual properties of the formulation (2) produced or to be produced, in particular an actual grain size property and moisture, wherein the model (12) takes into account predetermined or predeterminable control parameters (5) which are or represent manipulated variables for controlling actuators (6) of the plant (1), wherein state parameters (7), in particular sensor values, of the plant (1), which each represent a state of the plant (1) influencing the production, are determined and processed by the model (12), and wherein feedstock parameters (4), which represent an attribute of the feedstock (3), in particular a moisture content and / or a grain size property, are taken into account by the model (12), and wherein the dryer (15) is coupled to the granulator (13) in such a way,that the intermediate product (14) is automatically conveyed without interruption from the granulator (13) into the dryer (15) and the model (12) takes into account the continuously running combination of granulation with the granulator (13) and subsequent drying with the dryer (15), and / or wherein the model (12) has a static part (12A) with which a base value is determined for the respective control parameter (5) or predicted actual formulation parameter (11), and wherein the model (12) has a dynamic part (12B) with which the base value is optimized by means of a prediction.

2. Method according to claim 1, characterized in that the model (12), in particular the static part (12A) of the model (12), is or comprises an artificial neural network (32), wherein the artificial neural network (32) comprises nodes (36) in an input layer (33) which correspond to at least one state parameter (7) of the granulator (13), and to at least one state parameter (7) of the dryer (15) and, preferably, to the feedstock parameters (4).

3. The method according to claim 1 or 2, characterized in that the artificial neural network (32) has nodes (36) in an output layer (35) that correspond to the control parameters (5), while the artificial neural network (32) has nodes (36) in the input layer (33) that correspond to the target formulation parameters (10); or that the artificial neural network (32) has nodes (36) in the input layer (33) that correspond to the control parameters (5), while the artificial neural network (32) has nodes (36) in the output layer (35) that correspond to the predicted actual formulation parameters (11).

4. Method according to claim 2 or 3, characterized in that the artificial neural network (32) is or is trained with different training data sets, each of which consists of combinations of the feedstock parameters (4), state parameters (7) and control parameters (5) as well as actual formulation parameters (11) which are adjusted under the specification of these.

5. The method according to claim 4, characterized in that the training data sets each represent a stationary state of the plant (1) in which the actual formulation parameters (11) and state parameters (7) have assumed an at least substantially static value based on constant control parameters (5) and feedstock parameters (4).

6. The method according to claim 5, characterized in that the artificial neural network (32) is or is trained in that the nodes (36) are subjected to at least one state parameter (7) of the granulator (13) and at least one corresponding state parameter (7) of the dryer (15) of the respective training data set and, preferably, to corresponding feedstock parameters (4) of the respective training data set.

7. The method according to claim 6, characterized in that the nodes (36) are further each supplied with corresponding control parameters (5) or actual formulation parameters (11), insofar as nodes (36) are provided for this purpose in the input layer (33).

8. The method according to claim 6 or 7, characterized in that values ​​are obtained at the nodes (36) of the output layer (35) for the respective control parameters (5) or predicted actual formulation parameters (11), and errors are determined by comparing these values ​​with the corresponding control parameters (5) or actual formulation parameters (11) of the respective training data set, and wherein the errors are reduced by adapting weights of the artificial neural network (32), preferably by means of backpropagation and / or successively from training data set to training data set.

9. Method according to one of the preceding claims, characterized in that the control parameters (5) comprise a control parameter (5B, 5C, 5D) for controlling the granulator (13) and at least one control parameter (5E, 5F, 5G, 5H) for controlling the dryer (15).

10. Method according to one of the preceding claims, characterized in that the state parameters (7) have at least one parameter (7B, 7C, 7D) which describes an operating state of the granulator (13) and one parameter (7E, 7F, 7G, 7H) which describes an operating state of the dryer (15).

11. Method according to one of the preceding claims, characterized in that the model (12) (a) the control parameters (5) are only based on the feedstock parameters (4), state parameters (7) and actual formulation parameters (11) or (b) the predicted actual formulation parameters (11) are determined only on the basis of the feedstock parameters (4), state parameters (7) and control parameters (5), preferably wherein non-in-line measurable properties of the intermediate product (14), in particular particle-related properties of the intermediate product (14), are disregarded.

12. Method according to one of the preceding claims, characterized in that the model (12), in particular the dynamic part (12B) of the model (12), takes into account long-term effects of changes in the control parameters (5) on the actual formulation parameters (11) and / or state parameters (7); and / or that a change in state parameters (7) is forecast using the dynamic part (12B) of the model (12), and based on this forecast, the base values ​​for the control parameters (5) are adjusted, and the system (1) is controlled using the control parameters (5) optimized in this way.

13. A computer program product or computer-readable storage medium comprising instructions which, when executed by a computer, cause the computer to carry out the method according to any one of the preceding claims.

14. Plant (1) for producing a formulation (2) from a feedstock (3), wherein the production comprises processing the feedstock (3) with a granulator (13) and drying an intermediate product (14) produced with the granulator (13) by means of a dryer (15), comprising: Sensors for detecting condition parameters (7) of the system (1), each representing a condition of the system (1) influencing production, Actuators (6) of the plant (1) for directly or indirectly acting on the feedstock (3), a control device (9) for controlling the actuators (6) with control parameters (5) which are or represent manipulated variables for controlling the actuators (6), wherein the control device (9) is predefined or can be predefined with target formulation parameters (10) which represent the desired properties of the formulation (2) produced or to be produced, in particular a physical property characterizing particles and a moisture content, and a model (12) based on which the control parameters (5) can be determined and / or actual formulation parameters can be predicted, characterized in that the granulator (13) is coupled to the dryer (15) in such a way that the intermediate product (14) is automatically conveyed without interruption from the granulator (13) into the dryer (15), wherein the model (12) represents the combination of Granulation with the granulator (13) and the continuously subsequent drying with the dryer (15) are taken into account, and the system (1) is designed such that the control device (9) uses the model (12) to determine the control parameters (5) and / or to predict actual formulation parameters (11) based on the state parameters (7) of the system (1); and / or that the model (12) has a static part (12A), wherein the control device (9) is designed to determine a base value with the static part (12A) for the respective control parameter (5) or actual formulation parameter (11) using the state parameters (7), and wherein the model (12) has a dynamic Part (12B), wherein the control device (9) is designed to optimize the base value with the dynamic part (12B) by means of a prediction.

15. System (45) comprising a plant (1) according to claim 14 and a device for forming the feedstock (3) from several components (47), preferably Powders, preferably by sieving, and / or a device for further processing (51, 55, 57), preferably tabletting, of the formulation (2).

Images