Method for producing ice cream with solid components

Abstract

In a method of producing ice cream containing inclusions, ice cream flow (A1) is produced by a mixer (11) for mixing inclusions (A2) into the ice cream, and a conveying device (5) is operated to provide inclusions to the mixer (11) at a target feed rate. A control system operates to generate a control signal (C1) for setting the conveying device speed based on a target feed signal and a feedback signal. The feedback signal represents an estimated feed rate of inclusions to the conveying device and is determined based on a weight signal (S1) and a speed signal (S2). Input data obtained from the weight and speed signals are provided to a state observer, such as a Kalman filter, which generates the feedback signal based on a system model relating an instantaneous feed rate of inclusions to an instantaneous speed of the conveyor device.

Description

Technical Field

[0001] This disclosure generally relates to the production of frozen ice cream, and in particular to a technique for adding solid components to an ice cream stream during the production process. Background Technology

[0002] Frozen ice cream containing solid ingredients is a popular product. Solid ingredients may include, for example, fruit pieces, chocolate, or nuts.

[0003] Ideally, consumers should experience one or at least some solid components in every bite or spoonful. Therefore, it is desirable to ensure that the solid components are evenly distributed throughout the ice cream.

[0004] In the production of ice cream containing solid components using known equipment, a continuous flow of mixture is cooled to -4°C to -6°C at the outlet of a flow-through freezer. The mixture typically has approximately 100% expansion, meaning that air is mixed in a volume corresponding to the volume of the liquid mixture before the flow-through freezer inlet, and this air is uniformly distributed in the mixture in the form of small bubbles. After the flow-through freezer outlet, the solid components are mixed into the ice cream stream by a mixing device (e.g., a vane pump), which provides a uniform relationship between the ice cream and the solid components. The ice cream stream can then pass through another mixing device configured to further mix the solid components with the ice cream. The ice cream is then made into a final product, which is packaged in a packaging station and passed through a freezing device that rapidly cools it to a storage temperature typically -12°C to -18°C to achieve product stability, and then the product is stored for further distribution.

[0005] The feeding of solid components into the mixing unit is carried out by a conveyor system. A key factor in achieving a uniform distribution of solid components in ice cream is operating the conveyor system to feed the solid components into the mixing unit at a uniform and well-controlled rate. If the mixing unit receives solid components at an uneven rate, the solid components are likely to be unevenly distributed in the ice cream even with the use of a second mixing unit. Operating the conveyor system to provide a uniform and well-controlled feed rate can be very challenging. The conveyor system may need to handle solid components with different properties such as size, density, and softness. The conveyor system may need to achieve a wide range of feed rates for solid components into the mixing unit. Solid components may exhibit varying degrees of mutual adhesion and adhesion to the conveyor system. In addition, the conveyor system may be subjected to mechanical shocks during operation, causing significant deviations in the feed rate of solid components. Furthermore, the conveyor system may require mechanical adjustments to handle different solid components. Mechanical adjustments make the conveyor system more difficult to operate to achieve a consistent feed rate for solid components. Summary of the Invention

[0006] Its purpose is to at least partially overcome one or more limitations of the existing technology.

[0007] One such objective is to provide a technology for producing ice cream with a uniform distribution of solid components.

[0008] Another objective is to provide such a robust technology.

[0009] Another objective is to provide a technology that is easily adaptable to different delivery systems and solid components with different properties.

[0010] One or more of these objectives, and other objectives that may arise from the following description, are achieved, at least in part, by a method for producing ice cream containing one or more solid components as described in the independent claim, the embodiments of which are defined by the dependent claims.

[0011] A first aspect of this disclosure is a method for producing ice cream containing one or more solid components. The method includes: generating an ice cream stream via a mixer configured to mix one or more solid components into the ice cream; and operating a conveyor in a feeding system to supply one or more solid components to the mixer at a target feed rate given by a target feed signal. The operation of the conveyor includes: obtaining a weight signal representing the weight of one or more solid components in the feeding system; obtaining a speed signal representing the speed of the conveyor; and determining a feedback signal based on the weight signal and the speed signal, the feedback signal representing an estimated feed rate of the conveyor to the one or more solid components. The determination includes providing input data obtained from the weight signal and the speed signal to a state observer, which generates the feedback signal based on a system model that correlates the instantaneous feed rate of the one or more solid components with the instantaneous speed of the conveyor. The method further includes: operating a control system according to the target feed signal and the feedback signal to generate a control signal for setting the speed of the conveyor.

[0012] Stable and robust operation of the conveyor is achieved by using a state observer to generate feedback signals for the control system. This state observer receives input data from weight and velocity signals and operates according to a system model that correlates the instantaneous feed rate of one or more solid components with the instantaneous velocity of the conveyor. One reason for this is that a state observer with a properly configured system model can inherently determine a statistically optimal solution at any given time. The stable and robust operation of the conveyor, in turn, contributes to obtaining ice cream with precise amounts of solid components, uniformly distributed throughout the ice cream. This state operator is also inherently resistant to measurement noise prevalent in industrial facilities. High levels of measurement noise can be inherent in the feeding of solid components. For example, solid components may vibrate as they enter the conveyor, such as passing through a hopper. Furthermore, the conveyor itself, such as a screw conveyor, may vibrate during its operation. All these vibrations can manifest as measurement noise in the weight signal. Moreover, as described herein, the feedback signal generated by the state observer responds quickly and accurately to changes in the weight signal. This rapid response facilitates the task of designing a control system for precisely controlling the conveyor.

[0013] Various embodiments of the first aspect are defined below. These embodiments provide at least some of the aforementioned technical effects and advantages, as well as additional technical effects and advantages that are readily understood by those skilled in the art, for example, from the following detailed description.

[0014] In some implementations, the instantaneous feed rate in the system model is a linear function of the instantaneous speed of the conveying equipment.

[0015] In some implementations, at least one coefficient of the system model of the state observer is updated based on feedback signals and velocity signals.

[0016] In some implementations, the method further includes: determining an instantaneous feed rate of one or more solid components based on a weight signal, and determining an instantaneous speed of the conveying device based on a speed signal, wherein the instantaneous feed rate and instantaneous speed are included in input data provided to a state observer.

[0017] In some implementations, the state observer repeatedly performs predictions at the corresponding current time step to predict the state vector and estimated feed rate for the next time step based on the instantaneous speed of the conveyor, and performs updates to adjust the state vector and estimated feed rate based on the instantaneous feed rate.

[0018] In some implementations, the method further includes pausing updates to the operation status observer upon receiving an alarm signal indicating the start of a refill operation to add a batch of one or more solid components to the feed system.

[0019] In some implementations, the method further includes replacing the state vector of the state observer with a reference state vector after receiving an alarm signal.

[0020] In some implementations, the reference state vector is a previous state vector used by the state observer at an earlier time step, wherein the earlier time step is earlier than the start of the refill operation by a predefined number of data samples, wherein each data sample corresponds to an instance of the input data obtained from the weight signal and the velocity signal and provided to the state observer.

[0021] In some implementations, the method further includes: maintaining a set of recently used state vectors in memory and obtaining a reference state vector from the set of recently used state vectors in memory.

[0022] In some implementations, the method further includes: processing at least one of a weight signal or a feedback signal to detect the start of a refill operation, and generating an alarm signal after the detection.

[0023] In some implementations, the start of a refill operation is detected when the weight increase of one or more solid components in the feeding system per unit time, as indicated by the weight signal, exceeds a weight increase threshold.

[0024] In some implementations, the start of a refill operation is detected when the estimated feed rate given by the feedback signal is lower than a feed rate threshold.

[0025] In some implementations, the method further includes: determining a time series of the differences between the correlation values ​​of the estimated feed rate predicted by the forecast and the correlation values ​​of the estimated feed rate adjusted by the update, and evaluating these differences to detect the start of a refill operation.

[0026] In some implementations, the start of a refill operation is detected when the dispersion measure of these differences exceeds a dispersion threshold.

[0027] In some implementations, the method further includes obtaining the input data by downsampling at least one of the weight signal or the velocity signal.

[0028] In some implementations, downsampling is performed to generate the input data at given time intervals based on the speed of the transmission device.

[0029] In some implementations, the conveying device includes a spiral conveyor, and the input data is generated synchronously with the instantaneous rotational speed of the spiral conveyor.

[0030] In some implementations, the state observer includes a discrete-time state-space Kalman filter.

[0031] Other objects, aspects, features, and implementation schemes, as well as technical advantages, will become apparent from the following detailed description and accompanying drawings. Attached Figure Description

[0032] Figure 1 This is a schematic diagram of an example apparatus for producing ice cream with solid components.

[0033] Figure 2 It is an operation Figure 1 A flowchart of an example method of the device in the diagram.

[0034] Figure 3 It is used for Figure 1 A block diagram illustrating an example arrangement of feedback control for the transmission equipment in a device.

[0035] Figures 4-7 This is a flowchart of an example method for operating a transmission device.

[0036] Figure 8A It is used for Figure 3 A block diagram of an example system for generating feedback signals by a control system. Figure 8B It is by Figure 8A The flowchart shows the method executed by the preprocessing unit of the subsystem.

[0037] Figure 9 The diagram shows the response of the conveyor to a step change in speed for two different feedback signal generation techniques (top plot), as well as the measured speed of the conveyor (middle plot) and the measured weight of the feeding system (bottom plot).

[0038] Figures 10A-10B The diagram illustrates the target and feedback signals (top figure) during closed-loop operation of the control system for two different feedback signal generation techniques, as well as the measured speed of the conveyor (middle figure) and the measured weight of the feeding system (bottom figure).

[0039] Figures 11A-11B Corresponding to Figures 10A-10B , and indicates the data generated in different tests.

[0040] Figure 12 yes Figure 11B An enlarged view of the curve graph in the image. Detailed Implementation

[0041] The embodiments will now be described more fully below with reference to the accompanying drawings, which illustrate some, but not all, of the embodiments. In fact, the subject matter of this disclosure can be implemented 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 meet applicable legal requirements.

[0042] For the sake of brevity and / or clarity, well-known functions or constructions may not be described in detail. Unless otherwise defined, all terms used herein (including technical and scientific terms) have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0043] The same reference numerals refer to the same elements throughout the text.

[0044] Before describing the implementation plan in more detail, some definitions will be given.

[0045] As used herein, "solid components" refers to any non-fluid component that can be added to ice cream in batches during the production process. These components will appear essentially as inclusions in the finished ice cream, distinguishing it from the surrounding ice cream by taste, appearance, and / or consistency. Examples of solid components include fruits, chocolate, nougat, nuts, cookies, cakes, brownies, hard candy, fudge, etc.

[0046] As used herein, a "state observer," in its usual sense, refers to a device, module, or unit that provides an estimate of the internal state of a given real system based on measurements of its inputs and outputs. State observers are typically implemented by computers and are also known as "state estimators." State observers can be defined in discrete-time or continuous-time. Examples of state observers include Kalman filters, particle filters, extended state observers, etc.

[0047] As used herein, “Kalman filter” in its general sense refers to a class of state observers. Kalman filters can be defined in discrete-time or continuous-time. As used herein, the term “Kalman filter” includes the original optimal state estimator for linear systems with Gaussian measurements or process noise, as well as its variants such as the Extended Kalman Filter (EKF), Unscented Kalman Filter (UKF), Square Root Unscented Kalman Filter (SR-UKF), Integrated Kalman Filter (EnKF), etc.

[0048] Figure 1This is a schematic diagram of an example apparatus 1 for producing ice cream with fillings. Apparatus 1 includes a feeding system 5' for forming solid components into ice cream fillings. The feeding system includes a hopper 2 that defines a housing for receiving and holding the solid components via a feed opening 2A. A stirrer 3 is arranged within the hopper 2 and configured to stir the solid components to facilitate mixing within the hopper and prevent clogging. The stirrer 3 is driven by a drive unit 4, such as an electric motor, to rotate within the hopper 2. A conveying device 5 is arranged below the hopper 2 to receive the solid components and provide a well-controlled flow of solid components for mixing into the ice cream. The hopper 2 is placed on or attached to the conveying device 5. In the example shown, the conveying device 5 is a spiral conveyor that includes an auger 6, which is driven by a drive unit 7, such as an electric motor, to rotate. Other conveying devices are contemplated, such as vibrating conveyors, belt conveyors, drag conveyors, bucket conveyors, etc.

[0049] A weight sensor WSA 8 is arranged to measure the weight of the feed system 5' and any solid components contained therein. WSA 8 is configured to output a sensor signal (“weight signal”) S1 indicating the weight or mass measured by WSA 8. It will be appreciated that a change in the measured weight corresponds to a change in the weight of the solid components, and thus to a change in the feed rate of the solid components exiting the feed system 5'. WSA 8 may include one or more weighing units distributed along the conveyor, or a single integral weighing scale. WSA 8 is standard equipment in the ice cream making industry and is well known to those skilled in the art.

[0050] The apparatus 1 also includes a funnel 9 for receiving solid components from the feed system 5', and a feed pipe 10 for receiving ice cream from an upstream freezing system (not shown). A mixer 11 is arranged to receive the ice cream stream via the feed pipe 10 and the solid component stream via the funnel 9, and is operable to mix or blend the solid components into the ice cream. In the example shown, the mixer 11 is a vane pump. Alternatively, the mixer 11 can be an impeller pump, a piston pump assembly, an inline mixer, etc. The mixer 11 is driven by a drive unit 12, such as an electric motor. A delivery pipe 13 is arranged to receive the mixture of ice cream and solid components from the mixer 11 and to deliver the mixture to another mixing device 14, which is operable to mitigate the formation of solid component pockets in the ice cream. In the example shown, the mixing device 14 is an inline mixer with a rotating element. The rotating element is driven by a drive unit 15, such as an electric motor. Alternatively, a static inline mixer can be used. The outlet pipe 16 is arranged to receive the processed mixture of ice cream and solid components from the mixing unit 14 and to transport the processed mixture for further processing into the final product.

[0051] During the operation of device 1, solid component A1 is supplied to hopper 2 in batches. The supply of solid component A1 can be repeated automatically or manually during the operation of device 1 to refill hopper 2. Conveying device 5 is operated to feed solid component A1 to mixer 11. Simultaneously, ice cream A2 is fed to mixer 11 via feed pipe 10. The processed mixture A3 of ice cream A2 and solid component A1 is output as an ice cream stream containing fillings from output pipe 16. Depending on the configuration of device 1, the flow in output pipe 16 can be continuous or intermittent.

[0052] Figure 1 It also includes a control device 20 configured to operate the control device 1. The control device 20 may or may not be part of the device 1. The control device 20 can be implemented by hardware or a combination of software and hardware. In the example shown, the control device 20 includes processor circuitry 21 and computer memory 22. The processor circuitry 21 may include, for example, one or more of the following: a CPU (“Central Processing Unit”), a DSP (“Digital Signal Processor”), a microprocessor, a microcontroller, an ASIC (“Application-Specific Integrated Circuit”), a combination of discrete analog and / or digital components, or some other programmable logic device such as an FPGA (“Field-Programmable Gate Array”). In a non-limiting example, the control device 20 is a PLC. A control program including computer instructions may be stored in memory 22 and executed by processor circuitry 21 to perform the methods and processes described herein. The control program may be provided to the control device 20 on a computer-readable medium, which may be a tangible (non-transient) product (e.g., magnetic media, optical disc, read-only memory, flash memory, etc.) or a propagating signal.

[0053] The control device is configured to receive input signals denoted as S1, S2, ..., Sn, and to generate and output control signals denoted as C1, C2, ..., Cm for operating device 1. In the example shown, the control signals include signal C1 for operating drive unit 7, signal C2 for operating drive unit 4, signal C3 for operating drive unit 12, and signal C4 for operating drive unit 15. Figure 1 In this example, the input signal includes a weight signal S1 from WSA 8. The input signal may also include a speed signal S2 from speed sensor 17, which can be associated with the conveying device 5 for sensing the speed of the drive unit 7. In the example shown, speed sensor 17 is integrated into the drive unit 7. In the example given below, speed signal S2 specifies the speed as a percentage of the maximum speed of the conveying device 5, but the speed can be given in any suitable unit. Speed ​​sensor 17 can be an electromechanical encoder, photoelectric encoder, potentiometer, accelerometer, etc.

[0054] although Figure 1 Although not shown, control device 20 may also be connected to a user interface (UI) device, which may be configured to receive operator input for configuring control device 20 and / or generate feedback to the operator. For example, UI devices may include one or more of a keyboard, control buttons, a touchscreen, a microphone, a display, indicator lights, a speaker, an alarm, etc.

[0055] In a non-limiting example, the feed rate of the solid component is in the range of 15-1200 liters per hour, depending on the conveying device 5 and the configuration of the solid component. The density of the solid component can be in the range of 200-1400 g per liter. It is understood that control equipment 20 needs to be configured to operate the conveying device 5 accurately and reliably within these wide ranges.

[0056] Figure 2 This is a flowchart of an exemplary overall method for operating device 1. Method 200 can be performed by... Figure 1 The control device 20 performs the operation. Step 201 involves generating a continuous flow of ice cream via the feed pipe 10 through the mixer 11. As mentioned above, the ice cream flow can be generated by an upstream freezing system (not shown). In step 202, the conveyor device 5 is operated to supply solid components to the mixer 11. In the example shown, step 202 involves operating the conveyor device 5 to supply solid components to the mixer 11 at a target feed rate given by the target signal TS. The target feed rate is set by the control device 20 or a separate control device to achieve the required amount of contents per unit of ice cream in the final product. Therefore, the target signal TS can be set based on the flow rate of the ice cream entering or exiting the mixer 11.

[0057] Figure 3 It is used according to Figure 2 Step 202 of the diagram shows an example installation of the operating device 1, particularly the control system 30 for operating the conveying device 5. The control system 30 is configured to generate a control signal C1 for setting the speed of the conveying device 5. In the example of a screw conveyor, the control signal C1 defines the rotational speed of the auger 6. As a result of the control signal C1, the conveying device 5 will produce an actual feed rate AD of solid components to the mixer 11. The actual feed rate is estimated using a weight signal S1 from the WSA 8. As mentioned above, signal S1 represents the instantaneous weight or load measured by the WSA 8. However, signal S1 is affected by an integral effect inherent to device 1, represented by the integration unit 32. Signal S1 also includes unknown interference (represented as interference signal DS) and measurement noise (represented as disturbance signal NS). The merging of DS and NS is conceptually determined by… Figure 3 The corresponding addition units 33 and 34 in the diagram represent this.

[0058] The control system 30 is configured to operate on a target signal TS and a feedback signal FS to generate a control signal C1. The feedback signal FS contains an estimate of the instantaneous feed rate of the solid component to the mixer 11. The control system 30 can implement any suitable control algorithm for generating the control signal C1, such as P, PI, PD, or PID well known in the art. In the example shown, the control system 30 includes a subtraction unit 31A configured to generate an error signal ES representing the instantaneous difference between TS and FS, and a controller 31B configured to implement the aforementioned control algorithm.

[0059] The processing module 35 generates a feedback signal FS based on the weight signal S1 from WSA 8 and a signal S2 representing the instantaneous speed of the conveying device 5. As described above, signal S2 can originate from the speed sensor 17 in device 1. Alternatively, signal S2 can be given by or obtained from control signal C1. In some embodiments, the processing module 35 obtains signal values ​​from S1 and S2 at regular sampling intervals. In a non-limiting example, the sampling interval is in the range of 0.01-1 second. In a particular non-limiting example, the sampling interval is 0.1 seconds.

[0060] The control system 30 and the processing module 35 can be implemented by the control device 20, for example by software routines executed by the processor circuit 21.

[0061] The inventive effort presented herein focuses on processing module 35, aiming to provide a feedback signal FS that represents the actual feed rate AD as accurately as possible in both time and amplitude. It should be understood that the performance of control system 30 is highly dependent on the quality of the feedback signal FS. For example, the dynamic performance of the feedback signal FS will affect the responsiveness of device 1 to changes in ice cream production rate. If the response time of the feedback signal FS is long, there is a risk that the final product may contain insufficient or inconsistent amounts of inclusions, and the final product may need to be discarded.

[0062] Before describing the implementation scheme of processing module 35 according to the implementation scheme, a reference implementation will be given. The following will refer to... Figures 9-12 The performance of the implementation scheme is compared with that of the reference implementation. In the reference implementation, processing module 35 is configured to apply one or more low-pass filters to the weight signal S1 to reduce measurement noise (see [reference]). Figure 3(NS in the original text). Then, the current value of signal S1 is subtracted from the previous value in signal S1 to estimate the instantaneous feed rate. This is repeated in consecutive time steps to generate a primary feedback signal, which is further low-pass filtered to generate a secondary feedback signal, which forms the feedback signal FS provided to the control system 30. The reference implementation exhibits poor dynamic behavior because there is a noticeable delay before changes in AD are displayed in FS. Furthermore, the reference implementation is susceptible to disturbances that may occur during operation of device 1, such as when hopper 2 is refilled with a batch of solid components, or when someone climbs onto device 1 or otherwise mechanically impacts the feed system 5' (see [link to relevant documentation]). Figure 3 The DS in the middle is very sensitive.

[0063] Figure 4 It is based on Figure 2 A flowchart of an example method 210 for operating the feeding system 5', particularly the conveying device 5, according to the embodiment of step 202. Method 210 includes steps 211-213 performed by the processing module 35 and step 214 performed by the control system 30. In step 211, a weight signal S1 is obtained. In step 212, a speed signal S2 is obtained. In step 213, a feedback signal FS is determined based on S1 and S2 to represent the estimated feed rate of the solid component of the conveying device 5 over time. As shown, step 213 may include steps 213A-213C, which are repeated at consecutive time steps. In step 213A, input data is determined according to each of S1 and S2. In step 213B, the input data is provided to a Kalman filter. The Kalman filter is configured to generate the feedback signal FS based on a system model that correlates the instantaneous feed rate of the solid component with the instantaneous speed of the conveying device 5. In step 213C, the Kalman filter is operated to determine the signal value of the feedback signal FS at the current time step.

[0064] In step 214 (which is also repeated in consecutive time steps), the operation control system 30 operates, for example, as referred to above. Figure 3 The control signal C1 for setting the speed of the transmission device 5 is generated based on the target signal TS and the feedback FS.

[0065] By using a Kalman filter in step 213C instead of the low-pass filter of the reference implementation, a feedback signal FS is generated to more accurately represent the actual feed rate AD. In other words, the responsiveness of the feedback signal FS is improved. This, in turn, improves the operation of the control system 30.

[0066] The primitive form of the Kalman filter (KF) is the optimal estimator of the system state with respect to the mean square error of the state (MSE). KF estimates the unknown process by combining a system model of the unknown process with measurements of the unknown process. KF then recursively calculates the optimal Kalman gain to minimize the MSE of these states. The ratio between the covariance of the measurement noise and the covariance of the model noise determines whether KF maximizes the weights of the measurements or the model. If the covariance of the model noise is small, the system model is considered reliable, and the state estimation will be primarily based on the system model. Conversely, if the covariance of the measurement noise is small, the estimation will be primarily based on the measurements.

[0067] To define the operation of a standard Kalman filter, a discrete-time state-space format can be used: in k It is a time index. x ( k ) is the state vector. u ( k ) is the input. w ( k ) is model noise. v ( k () is the measurement of noise. y ( k ) is the observable output. F It is the state transition matrix. 𝐺 It is a control-input matrix. H It is the observation matrix.

[0068] The operation of a Kalman filter can be divided into a prediction (or projection) phase and an update phase.

[0069] In the prediction phase, the Kalman filter is based on previous samples and before using information from the measurements. a priori Predict the current value: in, It is given Time under each sample k State estimates, It is given Time under each sample State estimates, It is given Time under each sample k The output estimate, It is given Time under each sample k The covariance estimate, It is given Time under each sample The covariance estimate, It is the model noise covariance.

[0070] In the update phase, all estimates from the prediction phase are updated to determine the posterior estimates, meaning these values ​​now use information from the measurements: in e ( k (This is time) k News y ( k) For time k The measured values ​​taken, S ( k (This is time) k The new information covariance, R It measures the noise covariance. Kf ( k (This is time) k Kalman gain, It is given k Each sample in time k Updated state estimate, Indicates a given k The updated covariance estimate of each sample at time k. To implement the Kalman filter, it is necessary to select the system model and... Q and R value. Q and R Values ​​are typically chosen based on experience or through testing. In the context of generating the feedback signal FS, the system model defines the feed rate of the conveyor as a function of one or more variables. In the example given in this paper, the system model defines the feed rate as a linear function of the conveyor speed. This results in a system model in state-space format: in, It is a time index k The estimated feed rate is as follows. It is a time index k The speed difference below, It is a time index k The expected feed rate is as follows. It can be noted that... This is because it is assumed that the previous feed rate does not affect the current feed rate. Because the output is the same as the state. Control-input matrix. G It is a single coefficient that can be specific to each solid component and each configuration of the conveying equipment. Therefore, if mechanical adjustments are made to the conveying equipment, such as changing the auger 6, then... G This may change. Therefore, in some embodiments, different configurations are predefined for device 1 based on the solid components and the configuration of the conveying device 5.G Values. For example, each ice cream recipe that device 1 is to implement can be specified. G Values. For example, before starting ice cream production, the operator can input predefined values ​​into control device 20. G Value. Alternatively, control device 20 can automatically, for example, export from a database based on recipe data. G value.

[0071] It should be understood that The speed signal S2 determines the feed rate. Additionally, the weight signal S1 is used to determine the measured feed rate. This is used for the update phase. In some implementations, the measured feed rate is determined by the differential weight signal S1. For example, It can be by Given, optionally by k and The time difference between them is normalized, where and This refers to the signal value in the weight signal S1. Here, the time index... k This is the current time step and time index. This is the previous time step. It can be noted that... Determining the weight signal may involve averaging the signal values ​​in the weight signal S1. This averaging can be performed by downsampling the weight signal S1 and then processing the signal to determine the weight. Similarly, the velocity signal S2 can be downsampled and then processed to determine... See below for reference. Figures 8A-8B Provide an example of downsampling.

[0072] It should also be understood that, in addition to the speed of the transmission equipment, the system model may depend on one or more other variables. These other variables can be included in the system model as linear or nonlinear terms. It is also conceivable that the system model be defined as having a nonlinear dependence on the speed of the transmission equipment. However, it is currently believed that a linear system model improves the performance of processing module 35.

[0073] For a Kalman filter defined according to equations 11-12 above, such as Figure 4 As determined in step 213A, the input data for the Kalman filter is thus the velocity value. u (Instantaneous speed) and measured feed rate y (Instantaneous feed rate), speed value u It is obtained from the speed signal S2, and is a measured feed rate. y It is obtained from the weight signal S1.

[0074] It should also be noted that the above discussion applies equally to any Kalman filter variant (EKF, UKF, etc.). Furthermore, it applies to other state observers. Other state observers operate in a manner similar to the Kalman filter to determine one or more states of the system based on a system model. Therefore, even though this description is given in reference to the original Kalman filter, it applies equally to Kalman filter variants and general state observers.

[0075] Figure 5 This is a flowchart of example method 220 executed by processing module 35, which includes a Kalman filter 40 (…). Figure 3 (e.g., defined by equations 11-12). The method includes steps 221 and 222, which are repeated at consecutive time steps, for example as... Figure 4 A portion of step 213C is executed. In other words, steps 221 and 222 are executed consecutively over time. Step 221 is the prediction step, corresponding to the prediction phase of the Kalman filter described above. In step 221, based on the input data... u The given instantaneous velocity prediction of the estimated feed rate for the next time step of the conveyor (see [reference]). The input data and the state vector (step 213B). Typically, step 221 is performed according to equations 3-5. Step 222 is the update step, corresponding to the update phase described above. In step 222, based on the input data... y The measured feed rate is used to adjust the estimated feed rate from step 221 (see [reference]). The estimated feed rate, adjusted accordingly, is then output as the signal value in the feedback signal FS. (Step 213B) This is followed by the state vector. Typically, step 222 is performed according to Equations 6-10.

[0076] like Figure 5 As shown, method 220 may include step 223 of updating one or more coefficients of the system model based on the feedback signal FS and the velocity signal S2. In the example of equations 11-12, step 223 determines the coefficients. G Memory coefficient G Indicates the estimated feed rate ( The linear relationship between () and velocity (u(k)) G The ratio of a first value derived from the feedback signal FS to a second value derived from the velocity signal S2 can be calculated. The first and second values ​​can be taken at any time interval. In some implementations, step 223 is performed continuously at every time step except during disturbances (e.g., refill operations).

[0077] In the specific example of step 223, the calculation is performed at each time step relative to the previous time step. GThe instantaneous values ​​are obtained and stored in the FIFO memory. Thus, the FIFO memory stores M of the latest values. G Instantaneous values. The values ​​used in the system model at each time step. G The value is determined based on M instantaneous values ​​in the FIFO memory, such as the average, median, etc.

[0078] Step 223 provides a significant technological advancement because it relaxes the need to input predefined G values ​​for different operating states of device 1 (solid composition, conveying equipment configuration, etc.). It also mitigates the risk that processing module 25 may use incorrect G values, thereby reducing the accuracy of the feedback signal FS. Instead, through step 223, processing module 25 is operable to autonomously adjust G values ​​during operation. G Adjust to the correct value. At startup, the processing module 25 can adjust the default value. G Perform the operation.

[0079] exist Figure 5 In the example, method 200 also includes step 224 of storing the state vector in a FIFO (First-In-First-Out) memory. The FIFO memory can be memory 22 ( Figure 1 It is part of a FIFO memory and can be configured to store multiple state vectors. The number N of state vectors in the FIFO memory can be set to correspond to a predefined time. See below for reference. Figure 7 Figure 8 illustrates the use of the FIFO memory.

[0080] Affecting weight signal S1 (see Figure 3 Interference from DS (in the hopper 2) may interfere with the operation of conveyor 5. Such interference occurs when hopper 2 is refilled with another batch of solid components. Figure 1 During the operation of device 1, refilling operations can be performed very frequently, therefore it is desirable to mitigate the impact of refilling operations on the control of conveying device 5. This is achieved through... Figure 6 Example method 230 is implemented as follows, which includes step 231 detecting the start of a refill operation based on weight signal S1 and / or feedback signal FS, and step 232 generating an alarm signal AS whenever a refill operation is detected in step 231. The alarm signal can trigger the control device to take one or more actions to mitigate the effect of the refill operation on the feedback signal FS, for example, as shown below. Figure 7 The method 230 can be performed by the control device 20, for example, as part of the processing device 25. Figure 6 Example sub-steps 231A-231C are shown, which can be performed individually or in any combination as part of step 231 to detect the start of the refill operation.

[0081] In step 231A, the weight signal S1 is processed to detect a weight increase per unit time above a weight increase threshold T1. The basic principle of step 231A is that the load on WSA 8 may increase as solid components are filled into hopper 2. This weight increase can be derived from the weight signal over a period of time, set to be detectable within a reasonable time after the refilling operation begins. A potential drawback of step 231A is that it may take some time to detect the refilling operation performed by slowly pouring solid components into hopper 2.

[0082] In step 231B, the feedback signal FS is processed to detect estimated feed rates below the feed rate threshold T2. The basic principle of step 231B is a Kalman filter and the resulting feedback signal FS responding to changes in weight measured by the WSA 8. Therefore, even a slow refill operation can cause the estimated feed rate to decrease, or even become negative.

[0083] In step 231C, the estimated feed rate predicted in step 221 and adjusted in step 222 is determined and evaluated according to the detection criteria (see [reference]). Multiple differences (residuals) between the relevant values. The adjustment in step 222 is based on the measured feed rate (see...). The control is activated in response to changes in the weight signal S1 from WSA 8. Step 231C has been found to be highly responsive to the initiation of the filling operation. In some embodiments, the detection criterion in step 231C involves detecting whether a deviation measure exceeds a deviation threshold T3. The basic principle of using a deviation measure is that, in the absence of disturbance, multiple differences are expected to follow a Gaussian distribution with known deviations and a zero mean, and it has been found that refilling operations significantly increase the degree of deviation. Therefore, the deviation threshold can be set based on the known degree of deviation. Any deviation measure can be used, such as standard deviation, variance, coefficient of variation, sum of differences, energy, power, sum of absolute deviations from the mean, average of absolute differences from the mean, etc. In a non-limiting example, in the absence of disturbance, the deviation threshold is set in the range of 2-6 (e.g., 4.5 times) of the standard deviation of the Gaussian distribution.

[0084] In some implementations, two or more sub-steps 231A-231C are executed, and the start of a refill operation is detected by logically combining the results of the respective sub-steps. In one example, a logical OR is used so that a refill operation is detected if at least one sub-step indicates a refill operation. Using a logical OR between sub-step 231A and at least one of sub-steps 231B, 231C mitigates the aforementioned disadvantage of sub-step 231A. Alternatively or additionally, a logical AND can be used to combine two or more sub-steps, so that a refill operation is detected only if all such combined sub-steps indicate a refill operation.

[0085] It is conceivable that method 230 could be modified to detect interference not caused by the refill operation.

[0086] Figure 7 This is a flowchart of an example method 240 for operation processing module 35 to mitigate the impact of refill operation on the operation of device 1. In step 241, the Kalman filter is operated to pause update step 222 during the refill operation. Step 241 can be triggered, for example, by an alarm signal AS (see...). Figure 6 (Step 232 in the previous step). As a result of step 241, the Kalman filter will continue to output the feedback signal FS, but the signal value will be given by the estimated feed rate predicted in step 221, rather than by the feed rate adjusted in step 222. By pausing the update step 222, the weight signal S1 has no effect on the feedback signal FS. In other words, the operation of the Kalman filter is shielded or isolated from the effects of the weight signal S1 and thus from the refill operation.

[0087] It should also be noted that the state covariance increases each time prediction step 221 is performed, and decreases each time update step 222 is performed. Therefore, the state covariance will increase during step 241. As a result, when update step 222 is finally resumed, its impact on the estimated feed rate is greater than that of prediction step 221. In other words, the Kalman filter automatically assigns a greater weight to the measured feed rate (given by S1) than to the model prediction based on instantaneous velocity (given by S2). This is advantageous because the feedback signal FS will quickly adapt to the weight measurement.

[0088] It can be noted that the pause in step 241 may not be applicable to all types of state controllers, because for some state controllers, the prediction and update phases are inseparable.

[0089] The applicant has discovered that the feedback signal FS may be negatively affected if there is a delay in detecting a refill operation. During this delay, the Kalman filter performs both prediction step 221 and update step 222, subjecting the state vector to the refill operation. Step 243 may be included to mitigate the effects of this detection delay. In step 243, the state vector is replaced by a reference state vector xr, which is preferably unaffected by the refill operation. Step 243 results in a “reset” of the Kalman filter. Once a refill operation is detected, the state vector can be replaced with the reference vector. In some embodiments, the reference state vector is based on one or more previous instances of the state vector in the Kalman filter. By configuring the reference state vector to represent one or more previous instances, step 243 will cause the Kalman filter to reset to a previous operating state. This is also referred to herein as “rollback”. Step 242 is an example of a technique for implementing rollback. In step 242, the reference state vector is based on the state vector currently stored in the FIFO memory, such as Figure 5 As shown in step 224, the FIFO memory stores a set of recently used state vectors. In step 242, at least one state vector is selected based on a rollback time set to exceed the detection delay. Thus, at least one state vector is unaffected by the refill operation. It is recognized that the FIFO memory needs to contain enough state vectors to accommodate the rollback time. The reference state vector can be given by a single state vector in the FIFO memory or by multiple state vectors (e.g., as their average).

[0090] The rollback time can, but does not necessarily, be defined in clock time units. In a variant, the rollback time is given as a predefined number of data samples, where each data sample corresponds to an instance of the input data obtained in step 213A and provided to the Kalman filter in step 213B. Referring to Equation 1-12, this corresponds to the rollback time indexed by time. k The number of steps is given.

[0091] It should be understood that the pause in step 241 will eventually terminate. The termination point can be determined by analogy with step 231 and its sub-steps 231A-231C, by detecting that the refill operation no longer has an effect. Alternatively, the pause in step 241 can have a predefined duration. It is also conceivable that the pause in step 241 has a minimum duration, which can be given based on the time unit or the number of samples, to avoid updating step 222 by starting and closing too quickly.

[0092] It can be noted that method 240 is equally applicable if the refill operation is predictable, for example, if it is performed automatically. Method 240 is also applicable to reducing the impact of any other disturbances besides detectable or predictable refill operations.

[0093] Figure 8A This is a block diagram of an example processing module 35. Processing module 35 includes a Kalman filter 40, which is arranged to receive input data. The time series, and generate including The feedback signal FS of the time series of values. y The value is generated by the weight signal S1. u The value is generated from the velocity signal S2. In the example shown, the first preprocessing unit 41 is arranged to downsample the input signals S1, S2 to reduce measurement noise. In a non-limiting example, unit 41 performs downsampling by calculating the median of the M most recent data samples in the corresponding signals S1, S2, where M is in the range of 5-20. For example, if the sampling interval of S1, S2 is 0.1 seconds and M is 10, then one input data point will be generated per second. Sampling. In one variation, only one of the signals S1 and S2, for example, signal S1, is downsampled. Depending on the format of the signal value in the weight signal S1, a second preprocessing unit 42 can be installed to generate the desired format. y Value. For example, unit 42 can be configured to differentially process the input signal, for example, by subtracting a previous weight sample from the current weight sample, and optionally by scaling the difference between the resulting current and previous weight samples over time.

[0094] In the example shown, processing module 35 is also configured to execute Figures 6-7 Methods 230 and 240. The refill detector 43 is arranged to perform method 230 based on at least one of the weight signal S1 (step 231A), the feedback signal FS (step 231B), or the residual rs from the Kalman filter (step 231C).

[0095] Filter controller 44 is configured to receive an alarm signal AS from refill detector 43 and execute method 240 in response to the alarm signal. Filter controller 44 is connected for communication with FIFO memory 50. As part of step 241, filter controller 40 generates a pause signal ss for Kalman filter 40, which causes the update step in Kalman filter 40 to pause. As shown, according to steps 242-243, filter controller 40 also retrieves one or more state vectors from FIFO memory 50, generates a reference state vector xr, and replaces the current state vector with the reference state vector xr.

[0096] Figure 8B It is an operation Figure 8A The flowchart illustrates an example method 250 of the first preprocessing unit 41 (optionally together with the second preprocessing unit 42). In step 252, the weight signal S1 and the velocity signal S2 are downsampled, respectively. Downsampling can be performed to generate pairs of corresponding signal values ​​from S1 and S2. In step 253, the input data... The sampled signal is output to Kalman filter 40. In some implementations, step 253 involves calculating based on the downsampled signal value from S1. y For example, as described in reference unit 42 above. The time interval between consecutive samples of the input data corresponds to the downsampling period used in step 252.

[0097] In some implementations, the downsampling in step 252 is given with a fixed downsampling period, which can be given as a fixed time interval or a fixed number of data samples in the corresponding signals S1, S2. In other implementations, as shown in step 251, the downsampling period is dynamically set according to the speed signal S2, such that the downsampling period decreases as the speed increases. In a specific example, the downsampling period is set to correspond to one or more complete rotations of the auger 6 in the conveyor device 5. Thus, samples of input data are generated synchronously with the instantaneous rotational speed of the auger 6. It has been found that this improves the feedback signal FS by reducing the impact of vibrations originating from the rotation of the auger 6.

[0098] Include Figures 9-12 This is used to illustrate the practicality of the above-mentioned technology. Figure 9 It is based on method 210-240 ( Figures 4-7 A comparison between the processing module 35 for the operation and the aforementioned reference implementation. The figure above depicts the response to... Figure 1 The step change in the speed of the transmission device 5 is achieved from feedback signal 301 from the processing module and feedback signal 302 from the reference implementation. The middle figure depicts the speed 303 of the transmission device, thus corresponding to speed signal S1. As shown, a step change is performed at 96 seconds. The bottom figure depicts the weight 304 measured by WSA, thus corresponding to weight signal S1. As shown, signal 301 responds significantly faster to the speed change. Furthermore, as indicated by signal 304, a refill operation RO is performed between 405 and 414 seconds. This is achieved using method 240 ( Figure 7 Signal 301 is largely unaffected by the refill operation, while signal 302 exhibits significant instability during and after the refill operation.

[0099] Figure 10ATest results for feedback control based on feedback signals from a reference implementation are shown. In the top figure, signal 401 is the target signal (TS), and signal 402 is the feedback signal (FS). The middle figure shows the speed 403 of the transmission device given by the speed signal S2. It should be understood that the speed is controlled by the control system 30 ( Figure 3 The control signal C1 generated is used for control. The background diagram shows the measured weight 404 given by the weight signal S1. Figure 10B The results show the results based on methods 210-240 ( Figures 4-7 The feedback control of the processing module 35 is based on the corresponding test results of the feedback signal. Therefore, signal 501 is the target signal (TS), signal 502 is the feedback signal (FS), signal 503 is the speed signal (S2), and signal 504 is the weight signal (S1). This is achieved through comparison. Figures 10A-10B As can be seen from the top diagram, using processing module 35 results in more consistent and stable operation. Figure 10B The settling time of signal 502 in the signal is approximately 4 seconds, while Figure 10A The settling time for signal 402 is approximately 55 seconds. From Figure 10A It can also be seen that the feedback control using the reference implementation is highly sensitive to disturbances occurring at times 518 seconds, 694 seconds, and 842 seconds.

[0100] Figures 11A-11B Corresponding to Figures 10A-10B And it shows the results of another comparative test. Figure 11A Signals 601-604 in the diagram correspond to... Figure 10A Signals 401-404 in the middle, Figure 11B Signals 701-704 in the diagram correspond to... Figure 10B Signals 501-504 in the dataset. Figure 11B The settling time of the 702 signal is approximately 4 seconds, while Figure 11A The settling time for the 602 signal is approximately 119 seconds. Figure 11A During the process, a refill operation (RO) was performed at time 840 seconds, causing a significant interruption in the feedback signal, such as... Figure 11A Signal 602 is shown in the top diagram. Figure 11B In the middle, a refill operation (RO) was performed at time 231 seconds, resulting in... Figure 11B Signal 702 in the top image shows almost no visible change.

[0101] Figure 12 yes Figure 11BThe enlarged view shows a time span of 230-250 seconds, which includes the refill operation RO. Throughout the shown time span, the target feed rate (701 in the above figure) remains at 43.7 g / s. The refill operation begins at 231.0 seconds, as shown at point 705. The refill detector detects the refill operation at 231.2 seconds, as shown at point 706. Therefore, the detection delay is 0.2 seconds. During this period, the Kalman filter is updated using invalid weight data, which causes the signal value in the feedback signal (702 in the above figure) to drop from 43.7 g / s to 37.6 g / s. In the example shown, when the start of the refill operation is detected, the operation filter controller 44 ( Figure 3 The Kalman filter update is paused during the pause period SP. After the pause period, the feedback signal (702 in the figure above) needs some time to reach the target value again.

[0102] exist Figure 12 In the test, processing module 35 does not include the rollback function in steps 242-243. If the rollback function were implemented, the feedback signal (702 in the figure above) would drop from 43.7 g / s to 37.6 g / s, and then immediately recover to 43.7 g / s after the refill operation is detected to have started. Therefore, the effect of the refill operation on the feedback signal would only last for the detection delay, which is 0.2 seconds in this case.

[0103] While the subject matter of this disclosure has been described in conjunction with embodiments currently considered to be the most practical, it should be understood that the subject matter of this disclosure is not limited to the disclosed embodiments, but rather is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims. Furthermore, although operations are depicted in a specific order in the drawings, this should not be construed as requiring such operations to be performed in the specific order or sequence shown, or requiring all of the shown operations to achieve the desired result.

Claims

1. A method for producing ice cream (A3) containing one or more solid components (A2), the method comprising: An ice cream (A1) stream is generated by a mixer (11) configured to mix one or more solid components (A2) into the ice cream (A1), and a conveying device (5) of a feeding system (5') is operated (202) to supply the one or more solid components (A2) to the mixer (11) at a target feed rate given by a target feed signal (TS). The operation (202) of the transmission device includes: Obtain (211) a weight signal (S1) representing the weight of one or more solid components in the feeding system (5'), Obtain (212) a speed signal (S2) representing the speed of the conveying device (5), and determine (213) a feedback signal (FS) based on the weight signal (S1) and the speed signal (S2), the feedback signal (FS) representing the estimated feed rate of the conveying device (5) to the one or more solid components (A2), the determination (213) including: The input data (y, u) obtained from the weight signal (S1) and the speed signal (S2) are provided (213B) to the state observer (40), which generates (213C) the feedback signal (FS) based on a system model that correlates the instantaneous feed rate of the one or more solid components (A2) with the instantaneous speed of the conveying device (5), and The control system (30, 31) operates (214) based on the target feed signal (TS) and the feedback signal (FS) to generate a control signal (C1) for setting the speed of the conveying device (5).

2. The method according to claim 1, wherein in the system model, the instantaneous feeding rate is a linear function of the instantaneous speed of the conveying device (5).

3. The method according to claim 1 or 2, wherein at least one coefficient of the system model of the state observer (40) is updated (223) based on the feedback signal (FS) and the velocity signal (S2).

4. The method according to any one of the preceding claims, further comprising: The instantaneous feed rate (y) of the one or more solid components (A2) is determined (213A) based on the weight signal (S1), and the instantaneous speed (u) of the conveying device (5) is determined (213A) based on the speed signal (S2), wherein the instantaneous feed rate and the instantaneous speed are included in the input data provided to the state observer (40).

5. The method according to claim 4, wherein the state observer (40) repeatedly performs prediction (221) at the corresponding current time step to predict the state vector and estimated feed rate for the next time step based on the instantaneous speed (u) of the conveying device (5), and performs update (222) to adjust the state vector and the estimated feed rate based on the instantaneous feed rate (y).

6. The method of claim 5, further comprising: Upon receiving an alarm signal instructing the feed system (5') to begin a refill operation of adding a batch of one or more of the solid components (A2), the status observer (40) is operated (241) to pause the update (222).

7. The method of claim 6, further comprising: Upon receiving the alarm signal, the state vector of the state observer (40) is replaced (243) with the reference state vector.

8. The method according to claim 7, wherein the reference state vector is a previous state vector used by the state observer (40) at an earlier time step, wherein the earlier time step is a predetermined number of data samples earlier than the start of the refill operation, wherein each data sample corresponds to an instance of the input data (y, u) obtained from the weight signal (S1) and the velocity signal (S2) and provided to the state observer (40).

9. The method according to claim 6 or 7, further comprising: A set of recently used state vectors is maintained (224) in the memory (50), and the reference state vector is obtained (242) from the set of recently used state vectors in the memory (50).

10. The method according to any one of claims 6 to 9, further comprising: Process (231A, 231B) at least one of the weight signal (S1) or the feedback signal (FS) to detect the start of the refill operation, and generate (232) the alarm signal when the start is detected.

11. The method of claim 10, wherein the start of the refilling operation is detected when the increase in weight of the one or more solid components per unit time in the feed system (5') given by the weight signal (S1) exceeds a weight increase threshold.

12. The method of claim 10 or 11, wherein the start of the refilling operation is detected when the estimated feed rate given by the feedback signal (FS) is lower than the feed rate threshold.

13. The method according to any one of claims 6 to 9, further comprising: Determine (231C) the time series of the difference between the correlation value of the estimated feed rate predicted by the prediction (221) and the correlation value of the estimated feed rate adjusted by the update (222), and evaluate (231C) the difference to detect the start of the refill operation.

14. The method of claim 13, wherein the start of the refill operation is detected when the deviation metric of the difference exceeds a deviation threshold.

15. The method according to any one of the preceding claims, further comprising: The input data (y, u) is obtained (250) by downsampling (252) at least one of the weight signal (S1) or the speed signal (S2).

16. The method of claim 15, wherein the downsampling is performed to generate the input data (y, u) based on the time interval given by the speed of the transmission device (5).

17. The method according to claim 15 or 16, wherein the conveying device (5) includes a spiral conveyor (6), and wherein the input data (y, u) is generated synchronously with the instantaneous rotational speed of the spiral conveyor (5).

18. The method according to any of the preceding claims, wherein the state observer (40) comprises a discrete-time state-space Kalman filter.

Images