Configuration of control device in food product production plant

Abstract

A method is implemented on a computer device that is configured for closed-loop control of control devices of subsystems in a food product production plant. The computer device obtains (401) definition data DD indicating a task performed by the subsystem in the production of the food product and a device in the subsystem for performing the task; deriving (402) a candidate process model, CPM, for the subsystem based on the DD; obtaining (405) measurement data MD generated by the subsystem while running in accordance with the test sequence TS; estimating (406) a constant parameter of a differential equation in the CPM based on TS and MD; defining (407) a final process model APM of the subsystem based on the differential equation and the constant parameter; and running (408) a tuning algorithm on the APM to determine control parameters of the control device (12) for closed-loop control.

Description

Technical Field

[0001] The present disclosure relates generally to food product production and, more particularly, to techniques for configuring a control device for closed-loop control of food product production. Background Art

[0002] The main requirement of food production is to create a product that can be packaged, marketed and subsequently consumed. This involves developing a mixture of ingredients that can be efficiently and economically combined to make it palatable. Products should be consistent in appearance and taste and should be produced in a way that makes them easy or convenient to consume, such as being portioned or packaged appropriately.

[0003] Food manufacturing requirements must prioritize safety and cleanliness to minimize the risk of foodborne illness. Potentially hazardous ingredients must be stored and handled at safe temperatures, either by refrigerating them or heating the ingredients to a temperature sufficient to kill pathogens and then rapidly cooling the resulting ingredient combination to prevent further bacterial growth. Additionally, production equipment must be kept clean.

[0004] For example, the food product may be packaged liquid food such as beverages, dairy products, sauces, oils, creams, custards, soups, etc.

[0005] Food production plants have specific and unique requirements. Specifically, plants need to be configured to produce food products that not only meet hygiene requirements but also meet consumers' subjective expectations (such as taste, appearance, flavor, etc.). In addition, the equipment in the plant can be reconfigured to produce different food products from a variety of ingredients. In addition, the properties of ingredients may vary over time or from one plant to another. For example, the content of raw milk may vary over time and from one region to another.

[0006] Food production plants typically include a large number of components operated by multiple control devices. Given the specific and unique requirements of food production plants, as described above, configuring the respective control devices to perform their respective tasks and ensuring that the control devices operate in coordination to produce food products as required is a daunting task.

[0007] When setting up or modifying a food production plant, the current practice is for commissioning engineers to manually tune the respective control devices until their performance is acceptable. This tuning is done by trial and error and is highly dependent on the commissioning engineer’s experience. The results are often less than perfect and can result in inefficient plant operation. In a food production plant, a large number of different components are used and / or combined in a variety of different ways. This means that commissioning engineers will encounter a variety of control scenarios. As a result, configuring control devices in a food production plant can be a time-consuming task even for experienced commissioning engineers. Summary of the invention

[0008] It is an object to at least partially overcome one or more limitations of the prior art.

[0009] One of the aims is to simplify the configuration of control equipment in food production plants.

[0010] Another object is to provide a technique to improve the accuracy and consistency of control equipment used for closed-loop control in food production plants.

[0011] One or more of these objects, as well as further objects that may emerge from the following description, are at least partly achieved by a computer-implemented method, a computer-readable medium and a computer device configuring a control device according to the independent claims, embodiments of which are defined by the dependent claims.

[0012] A first aspect of the present invention is a computer-implemented method configured for closed-loop control of a control device in a plant for the production of food products. The method comprises: obtaining definition data of a subsystem, the subsystem being included in the plant and being operated by the control device, wherein the definition data indicates a task to be performed by the subsystem in the production of the food product and a combination of components included in the subsystem for performing the task; deriving a candidate process model of the subsystem based on the definition data, the candidate process model comprising one or more differential equations representing the task to be performed by the subsystem in the production of the food product by using the combination of components; obtaining measurement data generated by the subsystem when operating according to a test sequence; estimating a set of constant parameters of one or more differential equations based on the test sequence and the measurement data; defining a final process model of the subsystem based on the one or more differential equations and the set of constant parameters; and running a tuning algorithm on the final process model to determine one or more control parameters of the control device for the closed-loop control.

[0013] A first aspect provides a computer-implemented technique for configuring a control device for closed-loop control of a subsystem in a food production plant. The technique simplifies the configuration of the control device by automatically or semi-automatically generating a digital version or "digital twin" of the subsystem and its performance of a specified task in food product production. By generating a digital twin, the tuning of the control device can be performed at least partially in the digital domain, thereby reducing the need for manual labor. Therefore, in the method of the first aspect, a tuning algorithm is run on the digital twin to determine the control parameters installed in the control device. The digital twin provides a simulation environment that enables systematic, time-efficient and well-controlled tuning of the control device, thereby potentially improving the performance of the closed-loop control, such as in terms of accuracy and consistency, compared to the traditional method of manually tuning the control device based on engineering experience, which may be performed when the control device is operated to control the physical subsystem in the plant. Therefore, the technology of the first aspect will improve the ability of the plant to meet specific requirements related to food production.

[0014] A second aspect of the invention is a computer readable medium comprising computer instructions which, when executed on a processing circuit, cause the processing circuit to perform the method of the first aspect or any embodiment thereof.

[0015] A third aspect of the invention is a computer device comprising a processing circuit configured to perform the method of the first aspect or any embodiment thereof, and a signal interface for acquiring definition data and measurement data.

[0016] Other objects, aspects and embodiments as well as features and advantages will become apparent from the following detailed description and the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1A is a flow chart of an example factory for producing food products. Figure 1B is a schematic diagram of a production plant divided into separately controlled subsystems.

[0018] Figure 2 It is a functional block diagram of an apparatus for closed-loop control of a subsystem in a production plant.

[0019] Figure 3 is a block diagram of a computer device used to configure a control device to perform closed-loop control of a subsystem.

[0020] Figure 4 is a flow chart of an example method of configuring a control device.

[0021] Figure 5 yes Figure 3 Functional block diagram of the computer device in .

[0022] Figure 6 Example subsystems and variables associated with level control in the subsystems are shown. DETAILED DESCRIPTION

[0023] Embodiments will be described more fully hereinafter with reference to the accompanying drawings, in which some, but not all, embodiments are shown. Indeed, the subject matter of the present disclosure may be embodied in many different forms and should not be construed as limited to the embodiments described herein; rather, these embodiments are provided so that this disclosure may satisfy applicable legal requirements.

[0024] Like reference numerals refer to like elements throughout.

[0025] For brevity and / or clarity, well-known functions or constructions may not be described in detail. Unless otherwise defined, all terms (including technical and scientific terms) used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present invention belongs.

[0026] As used herein, "food product" refers to any nutrient or combination of nutrients that can be eaten or drunk by humans or animals. Examples of food products include liquid food products (e.g., beverages, dairy products, sauces, oils, creams, custards, etc.), and solid food products (e.g., meat, pasta, cereals, flour, etc.), and composite food products comprising liquid and solid ingredients.

[0027] As used herein, "closed-loop control" (also referred to as "feedback control") is used in its ordinary sense to refer to a control procedure that controls a process based on feedback from the process. Thus, the process is automatically adjusted based on the feedback to achieve the desired output (target value) of the process.

[0028] As used herein, "open-loop control" (also called "non-feedback control") is used in its ordinary sense to refer to a control procedure in which a process is controlled independently of its output (ie, without feedback from the process).

[0029] As used herein, "and / or" includes any and all combinations of elements, a "set" of elements means one or more elements are provided, and a "plurality" of elements means at least two elements are provided.

[0030] As mentioned in the background section, food product production differs from other types of production in many ways, such as the need to comply with food safety requirements, meet consumer subjective expectations, handle variations in ingredient properties, etc. A food production plant is a complex industrial facility that includes a variety of components used to process ingredients into food products. The same equipment is often used to produce different types of food products, so different components may need to be combined in different ways for different production lines. It should be understood that whenever a new production line is set up by combining different components, the control system of the production line needs to be carefully adjusted to the combination of components in the production line based on the products to be produced and the ingredients to be processed. Therefore, process control configuration is a common and time-consuming task in food production plants.

[0031] To provide a general illustration of the complexity of food production plants, see Figure 1A , which shows an example production line 1 ′ for producing a food product P in the form of a long-shelf stirred yogurt package. The production line 1 ′ comprises a number of different components that work together to produce the product P. Without going into detail Figure 1A By way of example, it can be appreciated that a food production plant may include individual components (e.g., tanks, pipe manifolds, valves, stirring devices, mixing devices, heaters, heat exchangers, pumps, flow controllers, vacuum chambers, etc.), as well as integrated machines configured for specific purposes (e.g., filling machines, packaging machines, homogenizers, freezers, separators, degassers, etc.). In addition, different variants of the corresponding components may be installed in the plant, such as tanks of different specifications, pumps of different capacities and types, heat exchangers of different capacities and types, etc. It can also be appreciated that the same component may operate on different fluids with different characteristics, depending on where it is installed in the production line 1′.

[0032] In order to facilitate production control, the production line 1' is usually divided into subsystems, which are operated by respective control devices. Figure 1A A plurality of such control devices 12 are shown, which are configured to collectively operate the production line 1' to produce the food product P. Each control device 12 provides one or more control signals to its corresponding subsystem 10. Figure 1A In the diagram, control signals from the control device 12 to n different subsystems in the production line 1' are represented by C1-Cn.

[0033] The respective control devices 12 may be modular units, such as programmable logic controllers (PLCs), microcontrollers, single board computers, programmable logic replays (PLRs) or any other specific or general purpose computing devices. In one variant, at least a subset of the control devices 12 are implemented on one and the same computing device.

[0034] It will be appreciated that the control device 12 may in turn be operated by one or more higher level control devices, for example to synchronize the operation of different subsystems. Alternatively, a group of control devices 12 may be communicatively coupled to each other to synchronize their operation as desired.

[0035] Figure 1B is a schematic block diagram of an example production plant 1, which includes a plurality of subsystems 10, which can be combined to form a production line 1 ', such as Figure 1A As shown. Each subsystem 10 is operated by a control device 12. The subsystem 10 includes one or more controllable or actuatable components 11A (one is shown) and one or more feedback components 11B (one is shown), as well as other components 11. The subsystem 10 is configured to perform a task by using its components. The task corresponds to a processing operation in the production of food products. The corresponding control device 12 operates its subsystem 10 through closed-loop control, by providing one or more control signals to the controllable component 11A based on one or more feedback signals from the feedback component 11B. Therefore, in the context of the subsystem 10, the corresponding controllable component 11A can be represented as an actuator, and the corresponding feedback device 11B can be represented as a sensor or measuring device. In a food production plant, the controllable component 11A can be a valve, a pump, a flow controller, a heat exchanger or a heater, although other components are also possible, such as stirring equipment, mixing equipment, integrated machines, etc.

[0036] Figure 2 is a schematic block diagram of an example control device 12, which is arranged to operate the subsystem 10 through closed-loop control. As previously described, the subsystem 10 performs processing operations (tasks). Figure 2 In , the processing operation is represented by a transfer function G, which is well known in control theory. The transfer function G operates on one or more controlled variables (represented by u) and provides one or more observed variables (represented by y). Figure 1B In the context of , each controlled variable u is provided by a control signal to the controllable component 11A, and each observed variable y is given by a feedback signal from the feedback component 11B.

[0037] In a food production plant, the observed variable could represent flow rate, temperature, pressure, or level, although other physical properties such as concentration, viscosity, etc. are possible.

[0038] exist Figure 2, the control device 12 includes a controller 121 having a control algorithm, here represented by a transfer function F. The controller 121 is configured to calculate a current value of a corresponding controllable variable u based on a current value of a corresponding error variable (represented by e). A difference calculator 122 is configured to calculate the current value of the corresponding error variable e as the difference between the current value of the corresponding reference variable (represented by r) and the current value of the corresponding observation variable y from the subsystem 10. The reference variable r defines a target value or set point for the observation variable y. The control device 12 includes a target generator 123, which is configured to generate and provide a signal representing the reference variable r to the difference calculator 122. Figure 2 1 is a conventional representation of a control device 12 configured for closed-loop control. The controller 121 may include any type of closed-loop control algorithm, such as proportional (P), proportional differential (PD), or proportional integral differential (PID). In order to achieve the desired dynamic behavior of the closed-loop control, one or more control parameters of the control algorithm (F) in the controller 121 need to be adjusted according to the processing operation (G) of the subsystem 10. This adjustment is performed iteratively and is generally referred to as "tuning". In the example of the PID algorithm, the desired dynamics can be obtained by adjusting three control parameters or "gains": proportional gain, integral gain, and differential gain. In the following, the values ​​of one or more control parameters of the control device 12 are represented as control parameter data CPD.

[0039] In some embodiments, the control device 12 is also operable to perform open loop control by using a different control algorithm in the controller 121 and by configuring the difference calculator 122 to provide the controller 121 with a reference variable r instead of the error variable e.

[0040] The present disclosure relates to a technique for simplifying the time-consuming task of tuning control equipment 12 in a food production plant 1. The technique is based on the insight that tuning can be simplified by providing a "digital twin" of the process operation of the subsystem 10 to be controlled. A digital twin is a virtual representation that serves as a real-time digital counterpart to a physical process. Therefore, determining the digital twin is equivalent to determining an operational process model of the subsystem 10. Figure 2 In the example of , such a process model will define the transfer function G of the subsystem 10. When the process model is available, the transfer function F of the controller 121 can be determined by using computer equipment, such as by trial and error tuning by an operator or by using existing algorithms for so-called automatic tuning, automated tuning or auto-tuning. Examples of automatic tuning algorithms include, but are not limited to, Ziegler-Nichols tuning, Tuning, Nyquist tuning, lambda tuning, and internal model control (IMC) tuning. Further examples of automatic tuning algorithms are given in the following articles: et al., “Automatic tuning and adaptation for PID controllers-a survey”, published in IFAC Proceedings Volumes, Vol. 25, No. 14, pp. 371-376 (1992), and Berner et al., “An Experimental Comparison of PID Autotuners”, published in Control Engineering Practice, Vol. 73, pp. 124-133 (2018), which are incorporated herein by reference.

[0041] Thus, by providing a digital twin, tuning of the control device 12 may be performed in a computer device separate from the physical subsystem 10 .

[0042] The digital twin, i.e. the process model, is also determined using computer equipment, which may or may not be the same as the equipment used for tuning, based on measurement data acquired from the subsystem 10. Figure 4 Give more details.

[0043] Figure 3 is a schematic diagram of a machine 30 configured to determine a process model of a subsystem 10 and to determine control parameter data CPD of a control device 12 for controlling the operation of the subsystem 10. Thus, the machine 30 corresponds to the computer device described above. The machine 30 includes a processing circuit 31, such as a central processing unit (CPU), a graphics processing unit (GPU), a digital signal processor (DSP), a microcontroller, one or more application-specific integrated circuits (ASICs), a field programmable gate array (FPGA), or any combination thereof. The machine 30 also includes a system memory 32, which may include computer memory in the form of volatile and / or non-volatile memory, such as read-only memory (ROM), random access memory (RAM), or flash memory. The memory 32 may store computer instructions (e.g., software or program code) to enable the machine 30 to perform any of the methods discussed herein. The instructions may be provided to the machine 30 via a computer-readable medium, which may be a tangible (non-transitory) product (e.g., magnetic media, optical media, read-only memory, flash memory, digital tape, etc.) or a propagated signal.

[0044] The machine 30 also includes an I / O interface 33 for connecting to a user interface (UI) system 34 to enable user interaction. Typically, the UI system 34 includes a presentation device configured to present output data from the machine 30 to a user and an input device configured to allow a user to enter input data into the machine 30. For example, the UI system 34 may include one or more of a keyboard, a keypad, a computer mouse, control buttons, a printer, a microphone, a display device, an indicator light, a speaker, a touch screen, a camera, a voice control system, a gesture control system, a USB port, etc.

[0045] exist Figure 3 In the example of , the input data includes definition data (DD) and measurement data (MD), and the output data includes test sequence (TS) and control parameter data (CPD). Figure 4 Describe the input and output data in detail. Figure 3 As shown, MD is generated by feedback component 11B of subsystem 10. Specifically, MD is generated in response to subsystem 10 operating according to TS, such as by control device 12. In the example shown, a local interface (LI) device 13 is provided to enable data to be input into control device 12. LI device 13 can be integrated into control device 12 or can be a separate unit that is connected to control device 12 by wire or wirelessly for communication. A user can transmit TS to control device 12 via LI device 13. Figure 3 In the example of , MD is obtained directly from component 11B, although it can also be output to the user via LI device 13.

[0046] As shown, the machine 30 may also include a signal interface 33′ for wired or wireless communication with the control device 12 via any suitable protocol. For example, the machine 30 may be configured to output instructions on the interface 33′ to cause the control device 12 to execute the TS and / or receive the MD from the control device 12 via the interface 33′. Alternatively or additionally, the machine 30 may transmit the CPD to the control device 12 via the interface 33′, so that the control device 12 configures its controller 121 accordingly. The signal interface 33′ may be combined with the I / O interface 33 into a single physical interface.

[0047] Figure 4 4 is a flow chart of an example method 400 for configuring a control device for closed-loop control in a food production plant. The method 400 may be performed by the machine 30 and will be referred to as Figure 3 The method 400 allows a user to interact with the machine 30 to enable it to determine the above-mentioned digital twin of the subsystem 10. The method 400 also allows a user to perform or initiate a tuning operation to determine the control parameter data ( Figure 3 in CPD). Figure 4Optional steps are indicated by dashed lines.

[0048] In step 401, the definition data of the subsystem 10 to be controlled is obtained ( Figure 3 DD in the definition data). The definition data indicates the task to be performed by the subsystem 10 and the related equipment in the subsystem 10 that participates in performing the task, i.e., a combination of components. The definition data can take any form, and the tasks and related equipment can be defined at any level of detail. In some embodiments, the tasks, i.e., the input and output of the processing operation in the subsystem 10, are defined by including controllable variables u and observation variables y in the definition data. The components in the subsystem 10 can be defined by categories, subcategories, performance, models, serial numbers, etc. according to the implementation scheme. A category can indicate the type of component, such as a pump, a valve, a heater, a heat exchanger, etc. Subcategories can further distinguish components of the same type. For example, for a pump, a subcategory can distinguish between a gear pump, an impeller pump, a centrifugal pump, etc. Performance can be defined in different ways for different components. For a pump, performance can indicate pumping capacity.

[0049] Step 401 may involve a user manually inputting definition data for the respective component. It is also contemplated that the user provides definition data by selecting a component from a library of predefined components of the plant, or even by selecting from predefined component groups, where each group may correspond to an actual or potential subsystem in the plant. For example, the user selection may be made in a graphical user interface presented on the UI system 34.

[0050] In step 402, at least one candidate process model (CPM) of the subsystem 10 is derived based on the definition data. A CPM is a dynamic model that includes one or more differential equations ([DE]) that represent tasks to be performed using associated devices in the subsystem 10. In some embodiments, the CPM is predefined and stored in a database accessible to the machine 30. The database can associate different identifiers given by the definition data with one or more CPMs. Therefore, step 402 can include determining one or more identifiers based on the definition data, and searching the database by using the identifier to retrieve one or more CPMs stored in the database in association with the identifier. The identifier can be included in the definition data, for example as a model identifier, a serial number, or any other unique or semi-unique identifier. Alternatively, the identifier can be generated algorithmically based on the content of the definition data. In an alternative embodiment, the CPM is generated on demand, for example by running a machine learning (ML)-based algorithm on the definition data or a portion thereof. The ML algorithm can be trained based on a large number of DD and CPM pairs.

[0051] A CPM may include ordinary differential equations (ODEs) and / or partial differential equations (PDEs). The corresponding differential equations may be linear or nonlinear and of arbitrary order. Typically, but not necessarily, at least one differential equation of a CPM has time as an independent variable and includes time derivatives of arbitrary order.

[0052] In some embodiments, the one or more differential equations of the CPM include the controllable variables u and the observed variables y of the subsystem, and a set of constant parameters or coefficients. Each constant parameter does not change with time. In the CPM derived in step 402, the value of each constant parameter is unknown.

[0053] The CPM can be a so-called custom model, also called a mechanical model, which is defined according to basic physical principles, where the constant parameters represent system parameters that have a physical interpretation at least in principle. However, in order to make method 400 more generally applicable and easy to implement, the CPM can be a so-called black box model, also called a ready-made model. Typically, the constant parameters of a black box model have no direct physical interpretation, but are used to describe the properties of the input-output relationship of a subsystem. There are a large number of black box models available. These are standard models that are known empirically to handle a variety of different system dynamics. Examples of linear black box model types include Box-Jenkins (BJ), Output Error (OE), ARMAX, and ARX. All of these black box model types contain a set of structural parameters that define the black box model. Therefore, different black box models can be obtained for different structural parameter values. The set of structural parameters can define the dynamics of the model, for example, in terms of its order, delay values, etc. For example, examples of linear and nonlinear black-box models can be found in Ljung, L.'s article "Black-box models from input-output measurements", published in the 18th IEEE Instrumentation and Measurement Technology Conference, Vol. 1, pp. 138-146 (2001), and the references cited therein.

[0054] In some embodiments, step 402 derives a single CPM for subsystem 10 and provides the single CPM for subsequent processing (see steps 406-407 below). This can increase the speed and processing efficiency of method 400.

[0055] In other embodiments, step 402 derives multiple CPMs for the subsystem 10, which are processed and evaluated by subsequent steps to select the best CPM for the subsystem 10. This can improve the accuracy of the resulting digital twin, thereby improving the performance of the tuned control device 12.

[0056] In step 405, the method obtains measurement data ( Figure 3 MD in ). Figure 3 As explained, MD is responsive to the subsystem 10 following the test sequence ( Figure 3 MD is generated by operating the TS in the controllable variable u(t). MD thus represents how the subsystem 10 reacts to the stimulus. In some embodiments, TS defines the time change of the controllable variable u(t), and MD includes the resulting response of the observed variable y(t). For example, TS can define a step change of the corresponding controllable variable u, so that MD represents the resulting step change of the observed variable y. Alternatively or additionally, TS can define oscillations, positive and negative changes, ramp changes, etc.

[0057] like Figure 4 As shown, the TS may be derived in step 403 and output in step 404 for application by the subsystem 10 to generate the MD. The TS may be derived in step 402 by analogy with the CPM. In one example, the TS is derived from a database based on one or more identifiers derived from the definition data. In another example, the TS is derived by running an ML algorithm on the definition data or a portion thereof.

[0058] The test sequence may be output in different formats in step 404. In some embodiments, the TS is presented to the user on a presentation device of the UI system 34. Here, the TS may be given by characteristic parameter values ​​defining u(t), such as amplitude of change, rise time, fall time, oscillation frequency, oscillation amplitude, etc. The user may then Figure 3 The LI device 13 in the TS manually inputs the characteristic parameter value into the control device 12. In some embodiments, the TS Figure 3 The TS is transmitted electronically to the control device 12 via the signal interface 33′ in the embodiment. Here, the TS may be given by any form of control data that causes the control device 12 to operate the subsystem 10 according to the TS. In some embodiments, the TS is transmitted to an auxiliary control device, such as a handheld device, via the signal interface 33′, which is then connected to the subsystem 10 to operate the subsystem 10 according to the TS.

[0059] It should be noted that steps 403-404 are optional. The TS may be provided to the user and applied to the subsystem in other ways. For example, the TS may be standardized for all subsystems, groups of subsystems, or a single subsystem. The user may determine the TS through a printed manual or a separate digital lookup system, and then configure the control device 12 or an auxiliary control device to operate the subsystem 10 according to the TS.

[0060] The measurement data is generated to represent the processing operation of the subsystem 10 (see Figure 3G in ). Therefore, MD should be generated without closed-loop control of subsystem 10. Therefore, when subsystem 10 operates according to TS, control device 12 is configured to switch to open-loop control. Similarly, auxiliary control device (if used) is configured to perform open-loop control.

[0061] Step 405 may acquire the measurement data in a variety of ways. In some embodiments, MD is manually input by a user through an input device in the UI system 34, for example, in the form of characteristic parameter values ​​defining y(t). Alternatively, MD may be stored on a storage medium (e.g., a USB memory) and manually transferred to the machine 30. In some embodiments, MD is manually input by a user through an input device in the UI system 34, for example, in the form of characteristic parameter values ​​defining y(t). Figure 3 The signal interface 33' in the machine 30 is electronically transmitted to the control device 12. In some embodiments, the auxiliary control device is connected to the machine 30 at the signal interface 33' and operates to transmit MD.

[0062] Step 405 may pre-process the measurement data, for example by applying one or more filters, before providing it for use in step 406 .

[0063] In step 406, the values ​​of the unknown constant parameters of the CPM derived in step 402 are estimated based on the test sequence and the measured data. Specifically, the values ​​of the constant parameters are estimated so that when the differential equations of the CPM are configured using these values, the CPM reproduces the MD as closely as possible when operating on the TS. In other words, a set of constant parameters is determined so that the CPM reproduces the dynamic behavior represented by the combination of the TS and the MD. Step 406 can use any suitable algorithm to fit the parameterized model to the data. Such an algorithm is represented herein as a fitting algorithm (FA). Examples of fitting algorithms include, but are not limited to, regression-based algorithms, statistical algorithms, and iterative algorithms. Fitting algorithms can generally be viewed as outputting a parameter vector (PV) containing estimated values ​​of the unknown constant parameters. Therefore, in some embodiments, step 406 includes running a fitting algorithm on the CPM to determine the PV given the TS and the MD. As previously described, the fitting algorithm can be iterative. Such a fitting algorithm can be configured to modify the PV until the CPM is considered to generate the MD from the TS when configured by the PV. The optimal PV is then output by the fitting algorithm.

[0064] As described above, step 402 may derive multiple CPMs. In such embodiments, step 406 may process each CPM from step 402 to determine a corresponding PV including an estimate of a constant parameter. Depending on the implementation, step 406 may provide each combination of PV and CPM for further processing or remove PV and CPM combinations that cannot generate an MD from the TS with sufficient accuracy.

[0065] In step 407, a final or actual process model (APM) is defined for the subsystem 10 based on one or more PV and CPM combinations from step 406. The APM is the digital twin described above. If step 406 provides a PV for a single CPM, the APM can be defined by configuring the CPM according to the value of the PV, i.e., by inserting the estimated values ​​of the constant parameters into the differential equations of the CPM.

[0066] If step 406 provides multiple combinations of PVs and CPMs, step 407 may include a step 407A of selecting one of the CPMs based on the selection criteria, and a step 407B of defining the APM based on the selected CPM and its associated PV. For example, the selection in step 407B may be performed by operating the corresponding CPM configured by its associated PV according to the TS to generate virtual or "synthetic" measurement data. The synthetic measurement data is then compared to the MD to generate a performance score representing the similarity between the synthetic measurement data and the MD. One of the CPMs is then selected based on the performance score of the corresponding CPM. In a variant, step 407B is performed as part of step 406, and the performance score is generated by a fitting algorithm.

[0067] In some embodiments of step 405, the incoming measurement data is divided into two disjoint subsets, with the first subset designated for use in step 406 to estimate the PV and the second subset designated for use in step 407B to generate the performance score. This can improve the relevance of the performance score.

[0068] In step 408, a tuning algorithm (TA) is run on the APM to determine the controller 121 ( Figure 3 ) in a closed-loop control algorithm. When APM is available, step 408 is straight-forward and any conventional tuning algorithm may be used. For example, step 408 may implement a tuning algorithm that allows a user to tune by trial-and-error or a tuning algorithm for automatic tuning.

[0069] In step 409, control parameter data (CPD) including the control parameter values ​​determined in step 408 is output for installation in the control device 12. The CPD may be transmitted similarly to the TS to the control device 12. For example, the CPD may be manually input via the LI device 13, or electronically transmitted via the signal interface 33'.

[0070] Figure 5 is a schematic block diagram of a structure for implementing method 400, for example, in machine 30 ( Figure 3 )middle. Figure 5 Each block in may be implemented by a processing circuit or a combination of a processing circuit and a computer instruction.

[0071] The first box 51 is configured to perform steps 401 and 402. In the example shown, the user inputs DD to the box 51, which is configured to process DD to determine the first identifier ID1. The box 51 is also configured to retrieve one or more CPMs from the database 40A by using ID1. The database 40A may be located in the internal memory 32 of the machine 30 or in an external memory.

[0072] The second block 52 is configured to perform steps 403 and 404. In the example shown, block 52 receives DD from block 51. Block 52 is configured to process DD to determine a second identifier ID2. Block 52 is also configured to retrieve TS from database 40B by using ID2. TS is then output by block 52. Database 40B may be located in the internal memory 32 of machine 30 or in an external memory. In a variant, databases 40A, 40B may be merged into a single database, where CPM and TS may be retrieved using a single identifier.

[0073] The third block 53 is configured to perform steps 405 and 406 by using the MD and one or more CPMs derived by block 51. As shown, block 53 includes a fitting algorithm (FA) and is configured to run the fitting algorithm on the MD and one or more CPMs to estimate the PV of the corresponding CPM.

[0074] The fourth block 54 is configured to perform step 407 to define an APM using one or more pairs of CPM and PV provided by block 53. As indicated by the dashed arrow, block 54 may be further configured to update the data record associated with ID1 in database 40A based on the result of step 406. For example, if a CPM results in a low performance score, the CPM may be deleted from the data record. Conversely, one or more CPMs may be added to the data record based on the performance scores of different CPMs. Thus, by block 54, Figure 5 The structure in implements a learning function that can improve performance over time. The learning function can also be implemented by block 54.

[0075] The fifth block 55 is configured to perform steps 408-409 by using the APM from block 54. As shown, block 55 includes a tuning algorithm (TA) and is configured to run the tuning algorithm on the APM to determine the CPD. Block 55 is configured to output the CPD thus determined for installation in the control device.

[0076] Figure 6 A simple example is provided to present subsystem 10 and related tasks. Figure 6In Figure 1, the subsystem 10 includes a tank 60, an inlet pipe 61, an outlet pipe 62, a valve 63, and a pressure sensor 66, which are components of the subsystem 10. The inlet pipe 61 is connected to a fluid port 64 in the side wall of the tank 60, while the outlet pipe 62 is connected to a fluid port 65 in the bottom wall of the tank 60. The valve 63 is arranged in the inlet pipe 61. The task of the subsystem 10 is to maintain the fluid 67 in the tank 60 at a given level (h). The level of the fluid in the tank is affected by the inflow of the fluid (Q) through the inlet pipe 61. in ) and the fluid outflow (Q out ) is affected. The outflow rate depends on the pressure head (p t Those skilled in the art will recognize that a custom model can be created to accomplish the task of maintaining a given fluid level (h) in tank 60, where the observable variable is the pressure head (p t ), the controllable variable is the inflow (Q in ). This custom model includes a set of constant parameters that are unknown and can be determined based on the physical characteristics of the subsystem 10 and its components. By including the custom model as a CPM in method 400, the custom model can be generated according to step 403 based on the measurement data (p t ) determines the set of constant parameters. The test sequence defines the inflow (Q in ) over time. Those skilled in the art also understand that, using basic knowledge of the dynamics of the subsystem 10, one or more black box models can be defined for maintaining a given fluid level (h) in the tank 60. In a first example, the black box model is defined as an ARX model with structural parameters na=1 and nb=1. In a second example, the black box model is defined as an ARMAX model with structural parameters na=1, nb=1, and nc=0. In a third example, the black box model is defined as a BJ model with structural parameters nb=1, nc=0, nd=1, and nf=1. In a fourth example, the black box model is defined as an OE model with structural parameters nb=1 and nf=1. These models also include a structural parameter nk, which defines a delay and can be set to different values, resulting in different CPMs in the context of method 400. Like the custom model, each black box model includes a set of constant parameters that are unknown and can be determined in accordance with step 403.

Claims

1. A computer-implemented method configured for closed-loop control of a control device (12) in a plant (1) for the production of a food product (P), the method comprising: Acquiring (401) definition data (DD) of a subsystem (10), said subsystem (10) being included in said plant and being operated by said control device (12), wherein said definition data (DD) indicates a task to be performed by said subsystem (10) in said production of said food product (P) and a combination of components (11A, 11B, 11) included in said subsystem (10) for performing said task; deriving (402) a candidate process model (CPM) of the subsystem (10) based on the definition data (DD), the candidate process model comprising one or more differential equations representing tasks to be performed by the subsystem (10) in the production of the food product (P) by using the combination of the components (11A, 11B, 11); Acquiring (405) measurement data (MD) generated by the subsystem (10) when operating according to the test sequence (TS); estimating (406) a set of constant parameters of the one or more differential equations based on the test sequence (TS) and the measurement data (MD); defining (407) a final process model (APM) of the subsystem (10) based on the one or more differential equations and the set of constant parameters; as well as A tuning algorithm is run (408) on the final process model (APM) to determine one or more control parameters (CPD) of the control device (12) for the closed-loop control.

2. The method according to claim 1, wherein the task is defined by one or more controllable variables (u) of the subsystem (10) and one or more observed variables (y) of the subsystem (10).

3. The method according to claim 2, wherein the one or more differential equations include the one or more controllable variables (u), the one or more observation variables (y) and the set of constant parameters.

4. The method according to claim 2 or 3, wherein the one or more observed variables (y) include at least one of flow rate, temperature, pressure or liquid level.

5. A method according to claim 2, 3 or 4, wherein the one or more controllable variables (u) comprise a control signal for at least one of a valve, a pump, a flow controller, a heat exchanger or a heater.

6. Method according to any one of claims 2 to 5, wherein the test sequence (TS) defines a variation of the one or more controllable variables (u).

7. The computer-implemented method according to any of the preceding claims, wherein said deriving (402) a candidate process model (CPM) comprises: The candidate process model (CPM) is retrieved by searching a database (40A) storing a plurality of predefined candidate process models based on one or more identifiers (ID1) given by the definition data (DD).

8. The computer-implemented method of any preceding claim, further comprising: The test sequence is presented (404) on a presentation device (34), or the control device (12) is caused to operate the subsystem (10) through open loop control to execute the test sequence (TS).

9. The method according to claim 8, further comprising: The test sequence is derived (403) based on the definition data (DD).

10. The computer-implemented method according to any of the preceding claims, wherein the candidate process model (CPM) is a black-box model.

11. The computer-implemented method of any one of the preceding claims, wherein said estimating (406) said set of constant parameters comprises: Given the test sequence (TS) and the measurement data (MD), a fitting algorithm (FA) is run on the candidate process model (CPM) to determine a parameter vector (PV) containing estimated values ​​for the set of constant parameters.

12. A computer implemented method according to claim 11, wherein the fitting algorithm (FA) is configured to modify the parameter vector (PV) until the candidate process model (CPM) when configured by the parameter vector (PV) is considered to produce the measurement data (MD) from the test sequence (TS), and then output the parameter vector (PV).

13. The computer-implemented method of claim 11 or 12, wherein said deriving (402) comprises deriving a plurality of candidate process models (CPM) of said subsystem (10) based on said definition data (DD), wherein said estimating (406) comprises running said fitting algorithm (FA) on each of said candidate process models (CPM) given said test sequence (TS) and said measurement data (MD) to determine a corresponding parameter vector (PV), and wherein said defining (407) comprises: One of the candidate process models (CPM) is selected (407A) and the final process model (APM) is defined (407B) based on the one or more differential equations and parameter vectors (PV) thereof of the thus selected candidate process model (CPM).

14. The computer-implemented method of claim 13, wherein the selecting (407A) comprises: running a corresponding candidate process model (CPM) configured by its parameter vector (PV) according to the test sequence (TS) to generate synthetic measurement data; A performance score is determined based on the synthetic measurement data and the measurement data (MD); and one of the candidate process models (CPM) is selected based on the performance score of the corresponding candidate process model (CPM).

15. A computer-implemented method according to any of the preceding claims, wherein the one or more control parameters (CPD) are configured to cause the control device (12) to perform closed-loop control on the one or more controllable variables (u) to achieve a target value (r) of the one or more observation variables (y).

16. A computer readable medium comprising computer instructions which, when executed on a processing circuit (31), cause the processing circuit (31) to perform the method of any one of claims 1-15.

17. A computer device (30) comprising a processing circuit (31) configured to perform the method of any one of claims 1 to 15, and a signal interface (33) for acquiring the definition data (DD) and the measurement data (MD).

Images