Beverage sterile cold filling line self-adaptive PID closed-loop control system and control method thereof

Abstract

The invention relates to the technical field of automatic control, and discloses a beverage sterile cold filling line adaptive PID closed-loop control system and a control method thereof, and the system comprises a process event perception and prediction module which is used for analyzing a production instruction into a process event containing future working condition change information; the event-based adaptive identification strategy module is used for dynamically adjusting a forgetting factor of an identification algorithm according to a process event and improving the identification sensitivity during working condition switching; the dynamic coupling factor online identification module is used for identifying a dynamic coupling factor matrix of the system in real time; and a predictive feedforward-feedback compound controller. According to the method, the problems of large overshoot and long stabilization time caused by parameter time varying and multivariable strong coupling in dynamic processes such as formula switching of a production line are solved, and the adaptive capacity, robustness and control precision of a control system are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of automation control technology, in particular to an adaptive PID closed-loop control system for a beverage aseptic cold filling line and a control method thereof. Background Art

[0002] Aseptic cold-filling lines for beverages are a critical component of the modern food industry. Their production processes require a high degree of continuity and stringent control of process parameters. A single production line often needs to frequently switch between different products based on market demand, such as switching from low-viscosity tea beverages to high-viscosity, pulp-containing juices. This process requires precise control of multiple key process variables, such as the material temperature during ultra-high temperature (UHT) sterilization and the liquid level in the aseptic buffer tank, to ensure product quality and production safety.

[0003] However, controlling these production processes presents significant technical challenges. First, complex multivariable coupling exists within the system. For example, adjusting the feed pump speed to control liquid level simultaneously alters the material's residence time in the heat exchanger, significantly affecting the sterilization temperature. Conversely, adjusting the steam valve to control temperature indirectly affects the liquid level due to changes in the fluid's physical properties. Furthermore, when switching between product types, material physical properties such as viscosity, density, and specific heat capacity change, causing the dynamic characteristics of the entire control object to change accordingly, creating a time-varying system.

[0004] The PID (Proportional-Integral-Derivative) controller widely used in current industrial settings is insufficiently suited to such complex processes. Traditional control strategies typically configure a separate PID controller for each controlled variable. This decentralized approach fundamentally ignores the inter-variable coupling, often leading to interference between control loops and a loss of focus, resulting in slow system response and oscillation. Furthermore, once tuned, PID controller parameters typically remain fixed and effective only for specific operating conditions (such as the production of a specific product). Once the production recipe changes and the system's dynamic characteristics change, the original PID parameters become suboptimal, often leading to a significant decline in control quality, significant temperature overshoot, and long settling times. This inherently "static" and "passive" control mode fails to anticipate and proactively adapt to changing operating conditions. This results in a large number of substandard products during the dynamic transition between production lines, significantly impacting the overall operational efficiency and economic benefits of the production line. Therefore, there is an urgent need for advanced control methods that can proactively adapt to changing operating conditions and effectively handle multi-variable coupling. Summary of the Invention

[0005] The technical problem solved by the present application is that the control parameters of the existing beverage aseptic cold filling line control system are difficult to dynamically adjust when facing production condition switching, material property changes and the like, and the system adaptability is insufficient because the coupling effects between multiple process variables cannot be effectively handled.

[0006] To solve the above technical problems, the present application provides a beverage aseptic cold filling line adaptive PID closed-loop control system and a control method thereof.

[0007] The present application provides a beverage aseptic cold filling line adaptive PID closed-loop control system in the first aspect, comprising: A process event perception and prediction module is configured to receive production instructions of an upper-layer manufacturing execution system and parse the production instructions into process events containing event information; A dynamic coupling factor online identification module is configured to use a recursive algorithm containing a forgetting factor to online identify a dynamic coupling factor matrix representing the coupling characteristics of multiple variables of the system according to real-time input and output data of the production line; An event-based adaptive identification strategy module is configured to dynamically adjust the value of the forgetting factor in the dynamic coupling factor online identification module before or during the occurrence of the process event according to the process event; A predictive feed-forward-feedback compound controller is configured to generate a compound control signal for controlling the beverage aseptic cold filling line according to the process event and the dynamic coupling factor matrix.

[0008] Preferably, the event-based adaptive identification strategy module is specifically configured to: Set the forgetting factor to a first value during and within a preset time window before the occurrence of the process event; Set the forgetting factor to a second value during a stable production stage without the occurrence of the process event; Wherein, the first value is less than the second value.

[0009] Preferably, the compound control signal generated by the predictive feed-forward-feedback compound controller includes a feed-forward control component. The predictive feed-forward-feedback compound controller is specifically configured to: Determine a target output change amount according to the event information contained in the process event, and calculate the feed-forward control component based on the inverse or pseudo-inverse operation of the dynamic coupling factor matrix to actively compensate for the disturbance caused by the process event.

[0010] In a specific embodiment, the predictive feed-forward-feedback compound controller is configured to calculate the feed-forward control component The formula is: ; wherein, is an estimate of the dynamic coupling factor matrix at discrete time is a pseudo-inverse of the estimate of the dynamic coupling factor matrix, is the target output variation.

[0011] Preferably, the composite control signal further comprises a decoupling feedback control component; The predictive feedforward-feedback composite controller is further configured to: construct a dynamic decoupling matrix based on the dynamic coupling factor matrix, and perform decoupling operation on the outputs of the PID controller group using the dynamic decoupling matrix to obtain the decoupling feedback control component, so as to eliminate the coupling effect between the controlled variables.

[0012] In one specific embodiment, the dynamic decoupling matrix is calculated according to the following formula: ; wherein, is an operation of taking diagonal elements to form a diagonal matrix, is a diagonal element of the estimate of the dynamic coupling factor matrix .

[0013] Preferably, the recursive algorithm adopted by the dynamic coupling factor online identification module is recursive least squares. For the kth output of the system, the update step of the parameter estimation vector includes: 1. Calculate the prediction error : ; wherein, is a regression vector composed of control inputs, is the parameter estimation vector at the previous time.

[0014] 2. Calculate the gain vector : ; wherein, is the forgetting factor, is the covariance matrix at the previous time.

[0015] 3. Update the parameter estimation vector : ; 4. Update the covariance matrix : ; ​wherein, is an identity matrix.

[0016] Preferably, the process event comprises at least one of the following information: an event identifier, an event parameter related to the event, and a timestamp or a triggering condition of the event.

[0017] The second aspect of the present application provides an adaptive PID closed-loop control method applied to a beverage aseptic cold filling line, comprising the following steps: a) receiving and parsing production instructions from an upper-layer manufacturing execution system to obtain a process event containing event information; b) dynamically adjusting a value of a forgetting factor for an online identification algorithm according to the process event; c) obtaining a dynamic coupling factor matrix representing system multivariate coupling characteristics by using a recursive algorithm for online identification based on the forgetting factor adjusted in step b); d) generating a composite control signal according to the process event and the dynamic coupling factor matrix; and e) outputting the composite control signal to an actuator of the production line to realize closed-loop control.

[0018] The present application provides a beverage aseptic cold filling line adaptive PID closed-loop control system and a control method thereof. The present application has the following beneficial effects: 1. The present application establishes an association between a process event and a forgetting factor in an online identification algorithm, so that the control system can actively adjust the update strategy of its mathematical model according to foreseeable operating condition changes. During operating condition switching, the system uses a smaller forgetting factor value to accelerate the tracking of changes in system dynamic characteristics; under stable operating conditions, a larger forgetting factor value is used to increase the stability of the identified model, thereby improving the accuracy and timeliness of the system model; 2. The present application calculates a feedforward control component based on a process event and a dynamically identified dynamic coupling factor matrix, which can pre-compensate for foreseeable system disturbances caused by production operating condition switching and the like. This active compensation method can exert control action before the disturbance significantly affects the process variable, thereby effectively reducing the fluctuation amplitude of the process variable during operating condition switching; 3. The present application uses a dynamically identified dynamic coupling factor matrix to construct a dynamic decoupling matrix and generate a decoupling feedback control component, which can effectively reduce the mutual coupling effect between control loops in a multivariate control system. This makes the PID feedback adjustment for a single process variable more independent and accurate, simplifies the controller parameter tuning, and improves the overall control performance of the system. BRIEF DESCRIPTION OF DRAWINGS

[0019] Figure 1A functional module structure diagram of a beverage aseptic cold filling line adaptive PID closed-loop control system according to an embodiment of the present application; Figure 2 A flow chart of a beverage aseptic cold filling line adaptive PID closed-loop control method according to an embodiment of the present application; Figure 3 A comparison chart of control effects of the method of the present application and the traditional PID method in response to the aforementioned working condition switching.

[0020] Among them, 10, process event awareness and prediction module; 20, event-based adaptive identification strategy module; 30, dynamic coupling factor online identification module; 40, predictive feed-forward-feedback composite controller. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical scheme and advantages of the present application clearer, specific embodiments of the present application will be described in detail below with reference to the drawings. It should be noted that the descriptive embodiments here are only used to explain the present application, and do not constitute a limitation on the protection scope of the present application.

[0022] Referring to the drawings Figure 1 , Figure 1 A functional module structure diagram of a beverage aseptic cold filling line adaptive PID closed-loop control system according to an embodiment of the present application. The beverage aseptic cold filling line adaptive PID closed-loop control system provided by the present application can include: a process event awareness and prediction module 10, an event-based adaptive identification strategy module 20, a dynamic coupling factor online identification module 30, and a predictive feed-forward-feedback composite controller 40.

[0023] The system is established on a multi-input multi-output (MIMO) discrete-time process model, which describes the dynamic process of the beverage aseptic cold filling line near a specific working point, and its mathematical expression is: ; Among them, is a discrete-time step; is a process output vector of the system, whose elements are process variables measured by sensors on the production line, such as sterilization temperature, tank liquid level, etc.; is a control input vector of the system, whose elements are control signals applied to each actuator, such as valve opening, heating power, etc.; is a dynamic coupling factor matrix to be identified, whose elements represent the influence coefficient of a specific control input on a specific process output; is an error vector composed of unmodeled dynamics and measurement noise of the system.

[0024] a process event perception and prediction module 10, which is connected with a higher-level manufacturing execution system (MES) or SCADA system to receive production instructions containing production condition adjustment information, and whose output is connected with an event-based adaptive identification strategy module 20 and a predictive feed-forward-feedback composite controller 40 to transmit structured process events generated after analysis;

[0025] the event-based adaptive identification strategy module 20, which is connected with the process event perception and prediction module 10 to receive process events, and whose output is connected with a dynamic coupling factor online identification module 30 to transmit a forgetting factor value dynamically calculated according to the process events;

[0026] the dynamic coupling factor online identification module 30, which is connected with the event-based adaptive identification strategy module 20 and sensors and actuators in the production field to receive a forgetting factor, a process output vector and a control input vector , and whose output is connected with the predictive feed-forward-feedback composite controller 40 to transmit a dynamically identified dynamic coupling factor matrix .

[0027] the predictive feed-forward-feedback composite controller 40, which is connected with the process event perception and prediction module 10 and the dynamic coupling factor online identification module 30 to receive process events and a dynamic coupling factor matrix , and simultaneously receives an externally given set value vector and a process output vector fed back by a sensor, and whose output is connected with an actuator in the production line to output a final composite control signal.

[0028] The working principle of the system is as follows: the process event perception and prediction module 10 converts high-level production instructions into internal process events and distributes them to the module 20 and the module 40. The module 20 adjusts the identification algorithm parameters (forgetting factor) of the module 30 according to the event. The module 30 calculates the current dynamic coupling factor matrix using the adjusted parameters and real-time production data, and provides it to the module 40. Finally, the module 40 synthesizes the process event, the dynamic coupling factor matrix and the real-time control deviation to generate and output a composite control signal to the actuator, thereby completing the closed-loop control of the production process.

[0029] The process event perception and prediction module 10 is responsible for translating macro production instructions from the upper layer manufacturing execution system (MES) into structured process events executable within the system. The module receives production instructions through a pre-defined communication interface, and the data format of the instructions can be either XML (Extensible Markup Language) messages or JSON (JavaScript Object Notation) strings.

[0030] Upon receiving a production instruction, the process event perception and prediction module 10 parses it and extracts key information related to the control strategy. For example, a production instruction for changing product recipe can include parameters such as target viscosity, target density, target sterilization temperature, etc.; a production instruction for adjusting production rate can include a new target filling flow rate.

[0031] After parsing, the module 10 encapsulates the extracted information into a standardized data structure, i.e., a process event. In a specific embodiment, the process event object contains the following fields: Event identifier : An encoding used to uniquely identify the type of event, for example, 0x01 represents a "recipe switching event" and 0x02 represents a "production rate adjustment event".

[0032] Event parameters : One or more quantitative parameters directly related to the event, stored in the form of key-value pairs. For a "recipe switching event", the event parameters can be {target_viscosity: 1.5, target_temperature: 135.0}; for a "production rate adjustment event", the event parameters can be {target_flowrate: 2000.0}.

[0033] Timestamp or trigger condition : The time point at which the event is expected to take effect or the logical condition that needs to be met for the event to take effect. This field can be an absolute timestamp indicating the exact time of event occurrence; or a logical expression, for example, when the liquid level of a certain upstream tank is below a certain threshold, the event is triggered.

[0034] After generating the structured process event, the process event perception and prediction module 10 sends the process event object to the event-based adaptive recognition strategy module 20 and the predictive feed-forward-feedback composite controller 40 through the system's internal communication bus or function call mechanism, so that they can perform subsequent calculations and decision-making.

[0035] An event-based adaptive identification strategy module 20, which dynamically provides the forgetting factor values for the recursive algorithm of the dynamic coupling factor online identification module 30 according to the process events received from the process event perception and prediction module 10. The core of this dynamic adjustment mechanism is to make the data weighting strategy of the online identification algorithm able to respond to known production regime changes in advance, switching between stable and fast tracking states.

[0036] In one specific embodiment, the event-based adaptive identification strategy module 20 is internally configured with the following key parameters: The first value, denoted as , is used for the phase when the system dynamic characteristics change significantly. This value is a number less than 1, and its value range can be [0.90, 0.98]. The purpose of choosing this value is to reduce the weight of historical data in the recursive algorithm, i.e., to shorten the “memory length” of the algorithm, thereby increasing the sensitivity of the identification algorithm to recent data changes, enabling it to quickly track changes in system dynamic characteristics during regime switching.

[0037] The second value, denoted as , is used for the phase when the system is in stable production. This value is a number very close to 1, and its value range can be [0.99, 1.0). The purpose of choosing this value is to increase the weight of historical data in the recursive algorithm, thereby enhancing the anti-high-frequency measurement noise capability of the identification algorithm, ensuring that the parameters of the dynamic coupling factor matrix identified under stable conditions have higher stability.

[0038] The preset pre-position time is used to define the time window for switching to the first value in advance before the event officially occurs. The setting of this parameter is based on the physical response delay of the production line, ensuring that the update rate of the identification algorithm has been increased before the physical change actually occurs.

[0039] The preset post-position time is used to define the duration for maintaining the first value after the event occurs. The setting of this parameter is based on the typical settling time of the production line after a regime switch, ensuring that the identification algorithm can capture the entire dynamic transition process.

[0040] When the event-based adaptive identification strategy module 20 receives a process event, it determines the output value of the forgetting factor according to the timestamp or trigger condition in the event.

[0041] If is an exact timestamp, the module performs the judgment at each discrete control time : if the current time In the time interval If the module 20 forgets the factor The output value is set to the first value ; Otherwise, set it to the second value .

[0042] like For example, a logical trigger condition can be a composite event, such as "when the liquid level of upstream tank A is lower than 5%" and "the status of downstream pump B is started". Once the logical condition is met, the module 20 immediately sets the forgetting factor to The output value is switched to the first value , and keep the value equal to the preset post-time The duration of the value, and then automatically switches back to the second value .

[0043] Finally, this module 20 at each moment Calculated forgetting factor It is transmitted to the dynamic coupling factor online identification module 30 as one of the core parameters for the recursive calculation in the current control cycle, directly affecting the update of its covariance matrix.

[0044] Finally, the forgetting factor calculated by this module 20 is It is transmitted to the dynamic coupling factor online identification module 30 as one of the core parameters for performing recursive calculation in the current control cycle.

[0045] The core function of the dynamic coupling factor online identification module 30 is to use the recursive least squares (RLS) algorithm with a variable forgetting factor to perform online parameter identification on the MIMO process model of the beverage aseptic cold filling line to obtain the estimated value of the dynamic coupling factor matrix in real time. .

[0046] To explain the implementation of this module in detail, a , and the two process outputs , The system is taken as an example, but the present invention is not limited to this dimension. In this example, the process output vector and the control input vector The specific form is: ; The corresponding dynamic coupling factor matrix is a 2x2 matrix whose elements Characterizes the Input pair The influence coefficient of each output: ; To apply the RLS algorithm, this module first logically decomposes the MIMO system into two parallel multiple-input single-output (MISO) subsystems. The first subsystem describes all inputs to the first output. The second subsystem describes the effect of all inputs on the second output The scalar equations of the two subsystems are: ; ; In order to perform recursive calculation in a unified vector form, a regression vector is defined and a parameter vector to be identified In this embodiment, the regression vector is composed of all control inputs of the system:

[0047] ;

[0048] No. The parameter vector of the subsystem By the dynamic coupling factor matrix No. Row elements consist of: ; Therefore, the above two MISO subsystems can be uniformly expressed as , or written in the predicted form .

[0049] At each discrete control moment The dynamic coupling factor online identification module 30 receives the forgetting factor from the event-based adaptive identification strategy module 20 , and for each MISO subsystem (i.e. and ) performs the following recursive update steps in parallel to compute the parameter estimate vector : 1. Calculate the prediction error , the error is Actual process output The difference between the output predicted by the model based on the parameters at the previous moment: ; in, For the The parameter vector at time estimated value.

[0050] 2. Compute gain vector g(k) g(k) which is used to adjust the step size and direction of parameter update: ; where, is the covariance matrix of the algorithm at time .

[0051] 3. Update parameter estimate vector by adding the correction term computed from the prediction error and gain vector to the estimate at the previous time step: ; 4. Update covariance matrix : ; where, is an identity matrix of the same dimension as the covariance matrix.

[0052] At system initialization, all parameter estimate vectors can be set to zero vectors, and the covariance matrix can be set to , where is a large positive number (e.g. 1000), is the identity matrix.

[0053] After the parameter estimate vectors and of all subsystems are computed, the dynamic coupling factor online identification module 30 transposes these two column vectors to row vectors, combines them in order, to form the complete estimate of the dynamic coupling factor matrix at the current time step : ; This matrix is then passed to the predictive feedforward-feedback composite controller 40.

[0054] The predictive feedforward-feedback composite controller 40 functions to receive and synthesize information from other modules to generate and output a final composite control signal . This composite control signal is composed of the linear superposition of a predictive feedforward control component and a dynamic decoupling feedback control component.

[0055] The computation of the predictive feedforward control component is triggered by a process event received from the process event perception and prediction module 10.

[0056] When the predictive feedforward-feedback composite controller 40 receives a process event, it first computes the event parameters determining a target output change vector needed to counteract the disturbance introduced by the event . Continuing with the 2x2 system example, if a process event indicates a product recipe change that requires the target value of the first process variable (e.g., sterilization temperature) to be increased by 5 units, while the target value of the second process variable (e.g., tank level) remains unchanged, the target output change vector can be specifically represented as .

[0057] Subsequently, the predictive feedforward-feedback composite controller 40 obtains the latest dynamic coupling factor matrix estimate provided by the dynamic coupling factor online identification module 30 and computes the feedforward control component by solving the pseudo-inverse of the matrix ; where is the Moore-Penrose pseudo-inverse of the matrix . The pseudo-inverse operation is used to ensure that a solution with the smallest norm is obtained even when is a non-square or singular matrix. During periods when no process event occurs, the target output change vector is a zero vector, and the feedforward control component is also a zero vector.

[0058] The computed is a 2x1 vector whose elements are the feedforward compensation values for the two control inputs.

[0059] The calculation of the dynamic decoupling feedback control component is used to track and correct the process output in real time. The implementation process includes the following steps: The predictive feedforward-feedback composite controller 40 computes the error vector between the system setpoint vector and the sensor-measured process output vector : ; The error vector is input to a parallel PID controller bank. In a 2x2 system, the controller bank consists of two independent digital PID controllers. Each PID controller corresponds to an error component At time , the th PID controller calculates a preliminary control output value based on its corresponding error . In a specific embodiment, the calculation uses a positional digital PID algorithm whose formula is: ; wherein: , , are the pre-set proportional, integral and derivative gain coefficients for the first control loop, respectively.

[0060] is the integral accumulation term, which is calculated as .

[0061] is the error value at the last control time.

[0062] The two independent PID controller outputs constitute a 2x1 preliminary control intention vector .

[0063] To eliminate the coupling effect between control loops, the predictive feedforward-feedback compound controller 40 utilizes the dynamic coupling factor matrix estimate to construct a dynamic decoupling matrix .

[0064] For a system with equal number of inputs and outputs, the dynamic decoupling matrix is calculated as: ; wherein, is the inverse matrix of , is the operation of taking the diagonal elements of a matrix to form a diagonal matrix, and are the diagonal elements of . The role of the dynamic decoupling matrix is to pre-compensate the control intention vector .

[0065] For a 2x2 system, the specific calculation formula of the decoupling matrix is: ; wherein, is the determinant of .

[0066] Finally, by left-multiplying the dynamic decoupling matrix to the preliminary control intention vector , a 2x1 decoupled feedback control component is obtained: ; This component is the actual control signal for feedback correction after decoupling processing.

[0067] Finally, the predictive feed-forward-feedback compound controller 40 adds the two parts of 2x1 control components to get the final 2x1 compound control signal , and outputs its two elements to the corresponding actuators on the beverage aseptic cold filling line respectively.

[0068] Referring to the drawings Figure 2 , Figure 2 is a flow chart of a beverage aseptic cold filling line adaptive PID closed-loop control method according to an embodiment of the present application. The control method provided by the present application is applied to the control system described above, and the method sequentially performs the following steps in each discrete control period . Step S201, acquisition and analysis of process events.

[0069] The process event perception and prediction module 10 receives and analyzes the production instructions from the upper-layer manufacturing execution system (MES) through its preset communication interface. The module extracts and encapsulates the key information (such as the type of working condition switching, target parameter value, and effective timestamp) in the instructions into a standardized process event data object. Subsequently, the process event object is distributed to the event-based adaptive identification strategy module 20 and the predictive feed-forward-feedback compound controller 40.

[0070] Step S202, adaptive adjustment of identification strategy.

[0071] The event-based adaptive identification strategy module 20 receives the process event. The module determines whether the current time kk is within the preset time window defined by the timestamp of the process event . If it is within the window, the module 20 outputs a first value as the forgetting factor of the current period; if it is not within the window or there is no valid process event, a second value is output as the forgetting factor . The forgetting factor is transmitted to the dynamic coupling factor online identification module 30.

[0072] Step S203, online identification of system model.

[0073] The dynamic coupling factor online identification module 30 obtains the process output vector of the current period from the sensors in the production field, and obtains the control input vector of the previous period from the controller record. Combined with the forgetting factor , the module performs one iteration of the RLS algorithm inside. This iteration contains the calculation of the prediction error, the update of the gain vector, the update of the parameter estimates of each subsystem and the update of the covariance matrix. After the iteration is finished, all the updated parameter estimates are recombined to get the latest estimate of the dynamic coupling factor matrix for the current period, which is then sent to the predictive feed-forward-feedback composite controller 40.

[0074] Step S204, generation of the composite control signal.

[0075] The predictive feed-forward-feedback composite controller 40 receives the latest estimate of the dynamic coupling factor matrix and performs the following calculations: Calculation of the dynamic decoupling feedback control component: the module first calculates the error vector between the setpoint vector and the process output vector . This error vector is input to a parallel set of PID controllers to generate the preliminary control intention vector . Meanwhile, the module constructs the dynamic decoupling matrix . Finally, the dynamic decoupling feedback control component is calculated by matrix multiplication . Calculation of the predictive feed-forward control component: the module checks the process events received from step S201. If the current time is within the effective period of an event, the module calculates the target output change vector

[0076] according to the event parameters, and calculates the predictive feed-forward control component by pseudo-inverse operation . If there is no effective process event, this component is a zero vector.

[0077] Combination of the control signals: the two control components are added vectorially to get the final composite control signal .

[0078] Step S205, execution of the control and loop.

[0079] The composite control signal generated in step S204 is output to the corresponding actuator on the production line, such as a control valve, a variable frequency pump or a heater, to complete the control action for the current control period. The system then enters the next control period and repeats steps S201 to S205.

[0080] Embodiment: To further illustrate the application of the technical solutions of the present application in actual industrial scenarios and the technical effects, a specific application example is given below.​

[0081] Referring to the drawings Figure 1 and the accompanying Figure 2 , the application scenario of this embodiment is a beverage aseptic cold filling production line, and the specific working condition is to perform a production formula switching from "low viscosity product A" to "high viscosity, fruit pulp containing product B". The switching process mainly involves two key process variables and two corresponding control variables, constituting a 2x2 MIMO system: Process output vector : including the outlet material temperature of the ultra-high temperature instantaneous sterilization machine (UHT) and the liquid level height of the sterile buffer tank .

[0082] Control input vector : including the steam regulating valve opening of the UHT heating section and the feed pump rotating speed of the conveying material to the sterile tank .

[0083] In this system, the steam valve opening mainly affects the sterilization temperature , but the change of temperature will cause the change of fluid density and fluidity, thereby indirectly affecting the liquid level . At the same time, the feed pump rotating speed mainly affects the liquid level , but the change of flow rate will directly and significantly change the residence time of the material in the UHT heat exchanger, thereby strongly affecting the sterilization temperature . This two-way coupling is a technical problem that is difficult to handle by traditional control methods.

[0084] The specific execution process of the control method of this embodiment is as follows: Initial stable stage: The production line is stably producing product A. The process event perception and prediction module 10 has no effective event. The event-based adaptive identification strategy module 20 outputs the second value (0.995, for example) as the forgetting factor. The dynamic coupling factor online identification module 30 continuously runs, and the dynamic coupling factor matrix identified by the dynamic coupling factor online identification module 30 converges to a stable value, which reflects the physical properties of product A. The feedforward component of the compound controller 40 at this time is zero, and mainly relies on the dynamic decoupling feedback control component to maintain the set values of temperature and liquid level.

[0085] Event triggering and control preparation stage: 1. Step one: the upper MES system issues an instruction to require switching at time The production recipe is switched to Product B and the sterilization temperature setpoint is raised from 95°C to 105°C. The process event awareness and prediction module 10 receives and parses the instruction, generates a process event object containing the new setpoint and a timestamp, and distributes it.

[0086] 2. Step two: At the current time (t, for example, 30 seconds in advance), the event-based adaptive recognition strategy module 20 switches the output of the forgetting factor from to a first numerical value (e.g., 0.92). This increases the sensitivity of the recognition algorithm, preparing it to capture the upcoming system dynamic change.

[0087] Process switch and adaptive control phase: 3. Step three: At time , the production line begins to switch materials. Product B, which is highly viscous, enters the system, causing a dramatic change in the system's physical dynamics. Because the forgetting factor has been set to a small value , the dynamic coupling factor online identification module 30's RLS algorithm is able to quickly track this change, the elements of which begin to update dramatically to reflect the dynamics of Product B (e.g., the absolute value of increases significantly, indicating that the pump speed's effect on the temperature has increased).

[0088] 4. Step four: At time , the predictive feedforward-feedback composite controller 40 performs the following operations simultaneously: Feedforward control: The module calculates the target output change vector 95,0 ] T = 10,0 T using the real-time updated . Using the real-time updated , the feedforward control component is immediately calculated. This component directly acts on the steam valve and the feed pump, actively and proactively compensating for the disturbance caused by the setpoint step and the change in material properties.

[0089] Decoupled feedback control: At the same time, the PID controller group calculates the error using the new temperature setpoint, generating a preliminary control intention . This intention is compensated by the real-time dynamic decoupling matrix (which is also calculated based on the latest ), generating the feedback component .

[0090] Performance and effects: 5. Step five: final composite control signal is applied to the actuator. Since the feedforward component pre-compensates most of the disturbances, and the feedback component eliminates the coupling effects between loops through dynamic decoupling, the sterilization temperature can quickly and smoothly rise to 105℃, with the overshoot being controlled within a very small range. At the same time, the buffer tank liquid level fluctuates slightly throughout the switching process, basically maintaining the set value.

[0091] Through the above method, compared with the traditional PID control scheme, the overshoot of the temperature during the formula switching process can be reduced from 10-15% to within 2%, and the system's settling time can be shortened from 5-8 minutes to 1-2 minutes, significantly reducing the production of unqualified products during the switching process, and improving the overall operation efficiency of the production line.

[0092] Referring to the accompanying Figure 3 , the figure intuitively compares the control effects of the method of the present application and the traditional PID method in response to the aforementioned working condition switching in a graphical manner. The accompanying Figure 3 contains two subgraphs, and the horizontal coordinates are both time.

[0093] In the upper half of the sterilization temperature response comparison graph, when the temperature set value undergoes a step change at the 60th second, the solid line curve representing the method of the present application exhibits the characteristics of fast response, minimal overshoot, and stability at the new set value in a short time. In contrast, the dot-dash line curve representing the traditional PID method exhibits significant overshoot and subsequent sustained oscillation, and it takes a longer time to reach stability. This proves the significant advantage of the method of the present application in set point tracking performance.

[0094] In the lower half of the buffer tank liquid level response comparison graph, the suppression effect of the coupling effect is clearly demonstrated. When the method of the present application is used, the solid line curve shows that the buffer tank liquid level basically remains near the set value throughout the process, with slight fluctuations. When the traditional PID method is used, the dot-dash line curve shows that the liquid level has a sharp instantaneous drop due to the coupling effect, deviating significantly from its set value. This clearly demonstrates the ability of the present application to effectively suppress the adverse coupling between control loops through the dynamic decoupling matrix.

[0095] In summary, the present application, through the combination of predictive feedforward and adaptive dynamic decoupling, can achieve faster, more stable, and more accurate control than traditional control methods in complex working condition switching scenarios, thereby effectively improving production efficiency and reducing material loss.

[0096] While embodiments of the application have been shown and described, it is to be understood that the embodiments described are merely exemplary of the principles and application of the present application. Numerous modifications and adaptions can be effected without departing from the spirit and scope of the present application, which is not limited to the exact construction and arrangement described. It is intended, therefore, to cover all modifications and adaptions that fall within the scope of the claims and their equivalents.

Claims

1. Adaptive PID closed-loop control system for aseptic cold filling line of beverage, characterized by: include: The process event perception and prediction module is used to receive production instructions from the upper-level manufacturing execution system and parse them into process events containing event information; The dynamic coupling factor online identification module is configured to use a recursive algorithm including a forgetting factor to identify the dynamic coupling factor matrix representing the multivariable coupling characteristics of the system based on the real-time input and output data of the production line. An event-based adaptive identification strategy module, configured to dynamically adjust the value of the forgetting factor in the dynamic coupling factor online identification module according to the process event before or during the occurrence of the event; A predictive feedforward-feedback composite controller is used to generate a composite control signal for controlling the beverage aseptic cold filling line according to the process event and the dynamic coupling factor matrix.

2. The adaptive PID closed-loop control system for the aseptic cold filling line of beverage according to claim 1 is characterized in that: The event-based adaptive identification strategy module is specifically used to: setting the forgetting factor to a first value during and within a preset time window before the process event occurs; During a stable production phase where no process event occurs, setting the forgetting factor to a second value; The first value is smaller than the second value.

3. The adaptive PID closed-loop control system for aseptic cold filling line of beverage according to claim 1 is characterized in that: The composite control signal generated by the predictive feedforward-feedback composite controller includes a feedforward control component; The predictive forward-feedback composite controller is specifically used for: The target output variation is determined according to the event information contained in the process event, and the feedforward control component is calculated based on the inverse or pseudo-inverse operation of the dynamic coupling factor matrix to actively compensate for the disturbance caused by the process event.

4. The adaptive PID closed-loop control system for the aseptic cold filling line of beverage according to claim 3 is characterized in that: The composite control signal also includes a decoupling feedback control component; The predictive feedforward-feedback composite controller is further specifically used for: A dynamic decoupling matrix is ​​constructed based on the dynamic coupling factor matrix, and the dynamic decoupling matrix is ​​used to perform a decoupling operation on the output of the PID controller group to obtain the decoupling feedback control component to eliminate the coupling influence between the controlled variables.

5. The adaptive PID closed-loop control system for the beverage aseptic cold filling line according to claim 1 is characterized in that: The recursive algorithm used by the dynamic coupling factor online identification module is the recursive least squares method, which is specifically used in each control cycle: Calculate the prediction error; Calculating a gain vector based on the forgetting factor; Updating a parameter estimation vector using the gain vector and the prediction error; The covariance matrix is ​​updated using the gain vector.

6. The adaptive PID closed-loop control system for the aseptic cold filling line of beverage according to claim 1 is characterized in that: The process event includes at least one of the following information: An event identifier, event parameters associated with the event, and a timestamp or triggering condition for the event.

7. The adaptive PID closed-loop control system for the aseptic cold filling line of beverage according to claim 3 is characterized in that: The predictive feedforward-feedback composite controller is used to calculate the feedforward control component, including: The target output variation is determined according to event parameters of the process event, and the feedforward control component is calculated by solving the pseudo-inverse of the dynamic coupling factor matrix.

8. The adaptive PID closed-loop control system for the aseptic cold filling line of beverage according to claim 4 is characterized in that: The decoupling feedback control component is obtained by multiplying the dynamic decoupling matrix and the output vector of the PID controller group.

9. The adaptive PID closed-loop control system for the aseptic cold filling line of beverage according to claim 8 is characterized in that: The dynamic decoupling matrix is ​​obtained by multiplying the inverse matrix of the dynamic coupling factor matrix by a diagonal matrix composed of the diagonal elements of the dynamic coupling factor matrix.

10. An adaptive PID closed-loop control method for a beverage aseptic cold filling line, based on a system according to any one of claims 1 to 9, characterized in that: The following steps are involved: Receive and parse production instructions from the upper-level manufacturing execution system to obtain process events containing event information; dynamically adjusting a value of a forgetting factor used in an online identification algorithm based on the process event; Based on the adjusted forgetting factor, a recursive algorithm is used to online identify a dynamic coupling factor matrix that characterizes the multivariable coupling characteristics of the system; generating a composite control signal according to the process event and the dynamic coupling factor matrix; The composite control signal is output to the actuator of the production line to achieve closed-loop control.

Images