Digital twin-driven model prediction control method for food baking equipment

Abstract

The invention relates to the technical field of baking equipment, in particular to a food baking equipment digital twinning driven model prediction control method, which comprises the following steps: constructing a structured variable based on equipment side operation information and food side initial information, the structured variable being used for representing a digital twinning operation state; selecting a food center temperature state, a food overall water content state and a boundary condition of a current period from the structured variables, and sequentially assembling into a state description vector; calculating a constraint tightening coefficient of the state description vector, and calculating a dynamic security constraint based on the constraint tightening coefficient; the target center temperature and the target water content level are introduced, the temperature deviation and the water content deviation of the current period are calculated, the temperature deviation, the water content deviation, the power constraint interval endpoint and the air supply constraint interval endpoint are combined, a control instruction is output through a control function, and the control instruction is written into an execution mechanism setting channel of the food baking equipment.

Description

A Model Predictive Control Method Driven by Digital Twin for Food Baking Equipment Technical Field

[0001] This invention relates to the field of baking equipment technology, and more particularly to a model predictive control method for food baking equipment driven by digital twins. Background Technology

[0002] Food baking equipment is widely used in the industrial and semi-industrial production of bread, pastries, biscuits, and other foods. Its core objective is to achieve stable control over the internal quality of food under limited time, energy consumption, and equipment capacity. The baking process is essentially a complex dynamic process involving heat transfer, moisture migration, and the gradual shaping of the food structure. The evolution of the internal temperature and moisture content of the food directly determines the taste, appearance, and consistency of the final product. However, in actual equipment operation, only external operating parameters such as oven temperature, heating power setting, and air supply status can usually be reliably obtained. The internal state of the food is difficult to measure directly online, resulting in a natural gap in the control system's understanding of key process states.

[0003] To compensate for this deficiency, existing technologies generally employ preset baking curves or adjustment strategies based on empirical rules. However, these methods heavily rely on human experience and have limited adaptability to batch variations in raw materials, changes in loading conditions, and operational disturbances, making it difficult to meet the requirements of modern food manufacturing for consistent quality and process stability. With the increasing adoption of model predictive control (MMC) in industrial control, its ability to predict future states and optimize control inputs offers new insights for complex process control. However, it still faces practical constraints in food baking scenarios. On the one hand, MMC relies heavily on initial state values ​​and the predictive model, while the internal state of food cannot be directly obtained, leading to significant uncertainty in the prediction results. On the other hand, the food baking process is inherently irreversible; once excessive dehydration, overheating, or structural deviations occur, they are difficult to correct. This makes it impossible for the control system to explore feasible control boundaries through trial and error in real equipment. Furthermore, for food safety and equipment reliability considerations, actual equipment operation is unlikely to withstand abnormal conditions such as overheating and sudden power surges, causing control strategies to often rely on static or conservative safety thresholds, further limiting control performance.

[0004] Therefore, existing technologies in the control of food baking equipment generally suffer from problems such as difficulty in obtaining internal states, difficulty in dynamically determining control boundaries, and difficulty in engineering the implementation of predictive control. There is an urgent need for a technical solution that can provide a continuous state basis and executable safety boundaries for predictive control without increasing complex sensing conditions or relying on trial and error with real equipment. Summary of the Invention

[0005] To address the above problems, this invention provides a model predictive control method for food baking equipment driven by a digital twin.

[0006] To achieve the above objectives, the technical solution adopted by this invention is: a model predictive control method for digital twin-driven food baking equipment, comprising: constructing structured variables based on equipment-side operating information and food-side initial information, wherein the structured variables are used to represent the operating state of the digital twin; selecting the food center temperature state, the overall moisture content of the food, and the boundary conditions of the current cycle from the structured variables, and assembling them sequentially into a state description vector; calculating the constraint tightening coefficient of the state description vector, calculating dynamic safety constraints based on the constraint tightening coefficient, wherein the dynamic safety constraints are used to adjust the allowable window temperature of the next cycle, and the dynamic safety constraints include a power constraint interval and an air supply constraint interval; introducing the target center temperature and the target moisture content level, and calculating the temperature deviation and moisture content deviation of the current cycle, combining the temperature deviation, moisture content deviation, power constraint interval endpoints, and air supply constraint interval endpoints, outputting control commands through a control function, and writing the control commands into the actuator setting channel of the food baking equipment.

[0007] Preferably, the structured variables include a subset of the initial internal state of the food and a subset of the equipment boundary starting points. The subset of the initial internal state of the food includes at least the initial center temperature and the initial moisture content. The subset of the equipment boundary starting points includes at least the furnace cavity representative temperature, the heating power setting, and the air supply setting.

[0008] Preferably, before selecting the food core temperature state, the overall moisture content of the food, and the boundary conditions of the current cycle from the structured variables, the method further includes: the digital twin reads the refresh values ​​of the furnace cavity representative temperature, heating power setting, and air supply setting in the register within the same cycle, while retaining the initial fields of the structured variables; and updating the structured variables in a preset order through the heat transfer model and the moisture migration model.

[0009] Preferably, the parameters of the constraint function include the food's core temperature state, the food's overall moisture content state, the oven cavity representative temperature, the oven cavity representative temperature value, the heating power setting value, and the air supply setting value.

[0010] Preferably, the parameters of the control function include an adjustable gain, which is used to convert the temperature deviation and moisture content deviation into the adjustment range of the actuator.

[0011] Preferably, the furnace cavity representative temperature is calculated by averaging the readings of the temperature probes of each furnace cavity at the same time.

[0012] Preferably, the temperature state at the center of the food is obtained by the temperature value at the center grid node of the heat transfer model, and the overall moisture content of the food is obtained by the volume-weighted average of the moisture content of the twin grid.

[0013] Preferably, the boundary conditions of the current cycle include the furnace cavity representative temperature, heating power setting, and air supply setting for the current cycle.

[0014] Preferably, the power constraint interval and the air supply constraint interval are stored as lower limit and upper limit fields, respectively.

[0015] Preferably, the target center temperature and target moisture content are obtained through the target card of the current process segment. The target center temperature and target moisture content are written into the controller recipe area when the batch is issued and updated when the process segment is switched.

[0016] The beneficial effects of this invention are as follows: This application constructs a digital twin operating state starting from equipment operating parameters and initial food process parameters. Within the digital twin, the evolution of the internal heat and moisture state of the food is advanced at a rhythm consistent with the actual control cycle. State variables directly related to control decisions are extracted from this evolution, forming a complete baking state description that can be directly used by model predictive control. Furthermore, the method combines the internal state of the food with the current operating conditions of the equipment to generate dynamic safety constraints that continuously change with the food state, ensuring that the control boundary matches the baking stage and the food's tolerance. Model predictive control performs control calculations under the combined effect of the aforementioned state description and safety constraints, and directly applies the resulting control commands to the actuators of the food baking equipment, enabling the equipment operating state to continuously evolve towards the process objective while ensuring safety and controllable quality risks. The construction of the digital twin operating state, state advancement and extraction, generation of state-driven dynamic constraints, and execution of constrained predictive control are organically integrated. Attached Figure Description

[0017] Figure 1 is a flowchart of the model predictive control method driven by digital twin for food baking equipment in a specific embodiment of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Referring to Figure 1, this application proposes a model predictive control method driven by a digital twin for food baking equipment, including: S101: Constructing structured variables based on equipment-side operating information and food-side initial information. These structured variables represent the operating state of the digital twin. Specifically, this step assembles the equipment-side operating information and food-side initial information of the "current oven" into an operating starting point that can be directly loaded into the digital twin, binding the digital twin to the real baking task from the outset. Equipment-side information comes from the process and setpoint quantities of the control system, while food-side information comes from batch records and loading confirmation records in the recipe / process sheet. Taking a common industrial hot air baking oven as an example, multiple temperature probes are typically arranged inside the oven cavity. The controller simultaneously maintains variables such as the power setting, fan speed setting, or damper opening setting for heating execution; these variables can be read from the controller's communication register. Initial food process parameters are usually given by the recipe sheet and written into the controller or host computer during production preparation, such as the batch loading quantity, product size category, and initial moisture content; these variables can be entered through the HMI or sent from the host computer to the controller's recipe area. The operating state of a digital twin is expressed as "initial values ​​that can be directly used for subsequent deductions." That is, the initial state inside the food and the starting point of the external boundary applied by the equipment are placed in the same state object, and after loading, it can start to advance over time.

[0020] The operational state of a digital twin is represented by a structured variable. This indicates that it includes a subset of the initial internal state of the food and a subset of the device boundary starting points. The subset of the food's internal state contains at least one initial center temperature. With an initial water level The subset of equipment boundary starting points must contain at least one furnace cavity representing a specific temperature. A heating power setting With an air supply setting .in The average of the readings obtained by multiple temperature probes in the furnace cavity at the same time is used as the boundary temperature input for the digital twin at the initial time. and Read the controller's current setpoint register directly; Read the initial moisture content field of the formula / process record for this batch; The values ​​are provided by environmental records or process settings before food loading and are written into the controller's formula area as batch start conditions during loading. These "initial values" exist as numerical fields in the control system and are directly written into the initial state buffer of the digital twin after being read. The heat transfer and moisture migration calculation module of the digital twin reads the initial values ​​from this buffer and then begins operation.

[0021] Furnace cavity represents temperature The calculation uses the average of the readings of each probe at the same time to ensure that differences in probe arrangement do not cause the digital twin boundary input to be biased towards a certain local area. The specific calculation formula is as follows:

[0022] in, Indicates the first The readings of each furnace cavity temperature probe at the initial sampling moment are written into the temperature acquisition register by the controller acquisition module and then read. This indicates the number of temperature probes participating in the averaging process, and the value is taken from the equipment configuration parameter table (written into the controller configuration area by the equipment factory configuration or maintenance calibration). The furnace cavity representative temperature value used for the initial boundary of the digital twin is written into the initial state buffer of the digital twin and used as the starting point of the boundary input in subsequent simulations.

[0023] In obtaining , , , and Afterwards, the operational status of the digital twin. Assembled using "internal initial values ​​+ boundary initial values", forming a starting point for the direct loading of the digital twin, where the digital twin's operating state... The specific expression is:

[0024] in, It represents the operational status of the digital twin and serves as a unified input object for subsequent simulations and control calculations; This indicates the initial temperature of the food's core. The data is obtained by writing the batch process settings or environmental records before feeding into the formula area. This indicates the initial moisture content of the food, and the data is obtained by writing the recipe / process sheet field or batch record into the recipe area. This indicates that the furnace cavity represents the starting temperature, which is calculated by the previous formula and written into the digital twin buffer. This indicates the starting point for setting the heating power; the data is read from the controller's heating setting register. This indicates the start point for the air supply setting; the data is read from the controller's fan / damper setting register. After assembly, The running status area of ​​the digital twin is written, and the heat transfer module of the digital twin reads it. , , As the boundary starting point, the internal modules of the food are read. , This serves as an internal initial value, thus entering a running state consistent with the current baking task.

[0025] S102: Select the food's core temperature state, overall moisture content, and current cycle boundary conditions from the structured variables, and assemble them sequentially into a state description vector. Specifically, step S102 revolves around the digital twin operating state obtained in step one. Expanding this process, we arrive at the twin's internal state corresponding to the current control moment, and organize the state variables directly related to model predictive control into a complete baking state description that can be directly used as state input. In engineering, the digital twin typically operates synchronously with the food baking equipment's control system within the same control cycle: the controller refreshes the heating power setting and air supply setting in each cycle, and after sampling the oven cavity temperature probe, writes these values ​​into the process variable register; the digital twin reads these registers in the same cycle, simultaneously reading and holding them. The batch initial field carried in (e.g.) , (etc.) serves as the starting point for internal evolution. Subsequently, the digital twin performs a fixed-sequence calculation of "heat-driven → moisture-driven" within this cycle: the heat transfer module updates the internal temperature field of the food under the drive of the furnace cavity boundary conditions, and the moisture migration module updates the internal water content state field of the food under the drive of the updated temperature field. To ensure that subsequent predictive control can stably use the state input, this step does not output the entire temperature / water content field as is, but instead extracts representative state variables from the discrete grid within the twin: the food's center temperature state. The temperature value is directly read from the heat transfer module at the center grid node of the food; these three values ​​represent the furnace cavity temperature in the current cycle of the controller, which is the boundary condition of the equipment. Heating power setting With air supply settings Composition, in which The calculation method is the same as in step one. Consistent, but corresponding to the probe sampling results of the current control cycle, they come from the temperature acquisition register, heating setting register, and fan / damper setting register, respectively, and are aligned with the twin cycle. Overall moisture content of the food. The water content of the twin mesh is obtained by volume-weighted averaging. The weights are determined by the cell volume allocation during mesh generation and written into the twin configuration area. The weights are directly read and used at runtime, so that the water content summary under different size categories or different mesh densities still maintains a consistent physical meaning.

[0026] Overall moisture content of food Calculated using a grid volume-weighted average:

[0027] in, This indicates the overall moisture content of the food at the current control moment, calculated by the digital twin moisture migration module after completing the current cycle. Indicates the internal structure of the digital twin. The water content of each grid cell at the current virtual moment is directly read after being written into the grid state buffer by the water migration module; Indicates the first The volume weight of each grid cell is determined by the food size category and grid division rules and written into the twin configuration area during twin initialization; This indicates the number of grid cells, with values ​​derived from the twin grid configuration table and determined according to the food size category.

[0028] In obtaining and And read the current period boundary conditions. , , Then, a complete description of the baking process. They are assembled in a fixed order into a state vector for model predictive control.

[0029] in, This represents the complete baked state description used by the model predictive control at the current control moment, which serves as the input for the next step of generating dynamic safety constraints; This indicates the temperature status at the center of the food. The data is output by the digital twin heat transfer module at the central grid node and written into the temperature status buffer. This indicates the overall moisture content of the food, calculated using the previous formula and written into the moisture content summary field; This indicates the temperature of the furnace cavity. The data is sampled by the furnace cavity temperature probe, provided in the controller's temperature acquisition register, and read in the current cycle. This indicates the heating power setting, with data provided by the controller's heating setting register and read in the current cycle; This indicates the air supply setting; the data is provided by the controller's fan / damper setting register and read in the current cycle. Through the above-described process of advancement and assembly, the output of this step... Simultaneously carrying the internal evolution results of the food and the conditions applied by the equipment, subsequent constraint generation and predictive control calculations are carried out around the state benchmark at the same moment.

[0030] S103: Calculate the constraint tightening coefficient of the state description vector, and calculate the dynamic safety constraint based on the constraint tightening coefficient. The dynamic safety constraint is used to adjust the allowable window temperature for the next cycle. The dynamic safety constraint includes a power constraint range and an air supply constraint range. Specifically, step S103 uses the complete baking state description output in step two. As the sole input, dynamic safety constraints that can be directly used by model predictive control are generated within each control cycle. In the actual operation of food baking equipment, the "safe adjustment range" of the same heating and air supply actuators varies depending on the internal state of the food: when the overall moisture content of the food is high, the internal buffer against thermal shock is stronger, allowing for faster power and air supply adjustments to keep up with the process rhythm; when the temperature at the center of the food rises and approaches the setting range, surface dehydration and structural solidification amplify the quality risks brought about by rapid adjustments, and at this time, a gentler adjustment boundary is more suitable. The internal state of food ( , ) and the current operating conditions of the equipment ( , , By binding at the same time, this step transcribes "the food's ability to withstand external forces at this moment" into "how the actuator is allowed to move next", so that the subsequent model predictive control solution naturally falls within the feasible range.

[0031] During engineering implementation, the controller can provide the current status in each cycle. , , ,and and The output comes from the digital twin after progressing within the same cycle; therefore, this step directly reads the data within the same cycle. The five components are calculated, and a constraint calculation is performed. To ensure that the constraints change smoothly with the state and are insensitive to local fluctuations, this step first generates a "constraint tightening coefficient" through the constraint function. The larger the value, the greater the allowable adjustment range; the smaller the value, the tighter the boundary. The calculation treats the food's moisture content as a positive contribution to the "adjustable space," the food's core temperature as a negative contribution to the "near-settling risk," and the oven cavity representative temperature and current heating / air supply settings as correction terms for the "intensity of external influences." A limiting method is used in the calculation to ensure... It always falls within the range that can be directly used for boundary scaling. The coefficient values ​​are determined by process engineers during the equipment commissioning phase in conjunction with digital twin offline testing and are fixed in the controller parameter area. During operation, it is executed according to a fixed formula, which facilitates consistent risk response under different batches and different load capacities. The specific constraint function expression is:

[0032] in, This represents the constraint tightening coefficient, the value of which is calculated in this step within the current control cycle and used to generate dynamic safety constraints; This represents the overall moisture content of the food obtained in step two, derived from the summary results of the internal grid state obtained by the digital twin moisture migration module. This indicates the food center temperature state obtained in step two, which comes from the output of the digital twin heat transfer module at the center grid node; This indicates the furnace cavity temperature, which is sampled from the furnace cavity temperature probe and provided by the controller in the current cycle. This indicates the heating power setting for the current cycle, describing the complete baking process. The data is read from the controller's heating setting register; This indicates the air supply setting for the current cycle, from The data is read from the controller's fan / damper setting register. The constant coefficients in the formula are determined during equipment commissioning and stored in the controller parameter area, directly participating in calculations during runtime; limiting calculations... and This is implemented by the controller or bypass computing unit using conventional comparison instructions.

[0033] get Then, this step will The function applies to "window width that can be adjusted in the next cycle", setting the window relative to the current actuator. , Binding is used to create a dynamic boundary for the model's predictive control output. To allow the boundary to automatically tighten as the actuator approaches saturation, this step uses... and The adjustable window is resized so that the adjustable range naturally decreases as the setting approaches its upper limit; similarly, the window is relatively wider when the setting is lower, facilitating rapid power increases or airflow boosts to keep up with process trajectories. Dynamic safety constraints. Output in structured interval format, where the power constraint interval and the air supply constraint interval are stored as lower and upper limit fields, respectively, for direct use as endpoints of the feasible region in subsequent control calculations. These endpoints will change in each control cycle. renew.

[0034]

[0035] in, This represents the dynamic safety constraints output in this step, which will serve as constraint inputs for the next step of model predictive control solution. This represents the heating power setpoint variable to be solved in the next control cycle of the model predictive control. Its specific value will be determined by control calculation in step four and then executed by the controller. This indicates the air supply settings to be solved in the next control cycle. The solution must fall within the given range and be executed by the controller. and These are the heating power setting and air supply setting for the current control cycle, respectively, derived from the controller setting register and already included. middle; This is the constraint tightening coefficient calculated using the previous formula; the interval endpoints are calculated by the controller or bypass calculation unit according to a fixed formula and then written into the constraint structure. In practice, and Typically, the values ​​are stored and computed in the controller as a proportion of the actuator's rated capacity. Therefore, interval calculations can be performed directly within the controller's numerical domain and maintain the same numerical system as the solution variables for subsequent model predictive control.

[0036] The output of this step is dynamic safety constraints. Its content is determined within the current control cycle by The five components together determine the feasible domain constraint, which is then directly used as a constraint by the model predictive control in the next step. The constraint automatically adjusts with the evolution of the food's internal moisture content and core temperature, making the control sufficiently responsive in the early stages of baking and more robust in the near-setting and dehydration sensitive stages. This solidifies the "irreversible and overshoot-prone" risks in the food baking scenario into a control boundary in a calculable and executable manner.

[0037] S104: Introduce the target center temperature and target moisture content level, and calculate the temperature deviation and moisture content deviation for the current cycle. Combining the temperature deviation, moisture content deviation, power constraint interval endpoints, and air supply constraint interval endpoints, output control commands through a control function, and write the control commands into the actuator setting channel of the food baking equipment. Specifically, step S104 involves writing the complete baking state description output in step two within each control cycle. The dynamic safety constraints output in step three Simultaneously, it is used for control calculations, and the calculated control commands are written into the actuator setting channel of the food baking equipment, so that the equipment operates with the new heating and air supply settings in the next control cycle. When entering this step... A unified expression has been given for the current internal state of the food and the current operating conditions of the equipment (including...). , And the current , , ), The settings window for the next cycle has been provided (including...) , , , (Four endpoints). The controller side simultaneously saves the target card for the current process segment. This includes the target center temperature. With target water level These two target values ​​are written into the controller's recipe area when the recipe / process sheet is issued in batches and updated when the process segment is switched. The calculation in this step adopts the order of "first generating candidate settings according to the current deviation, and then clamping with a safety window", so that the output instructions can both move towards the target and be implemented in the register within the executable range of the current cycle.

[0038] The generation of candidate settings depends on only two types of deviations: temperature deviation. With moisture content deviation Temperature deviation reflects the difference between the internal thermal state of the food and the target, while moisture content deviation reflects the difference between the dehydration progress and the target. In baking scenarios, these two types of deviations jointly determine whether the heating and air supply in the next cycle need to be "stronger" or "gentler." This step uses two adjustable gain... and The deviation is converted into an increment in the actuator, corresponding to the adjustment level of heating power and air supply setting, respectively. These two gains are determined and fixed in the controller parameter area during equipment commissioning and are used according to a fixed formula during operation. To avoid frequent jittering of the actuator near the boundaries, candidate settings are... The given interval endpoints are clamped. The clamping operation is implemented using the controller's standard comparison and selection instructions. The clamping result is directly written to the setting register and executed by the heating module and the fan / damper module in the next control cycle; for example... Write to the heating power setting register. Write the fan speed setting or damper opening setting register. After the register is refreshed, the underlying driver outputs the latest setting.

[0039] The control function is calculated in the following form for this period:

[0040] in, This indicates the heating power setting instruction for the next control cycle, which is then written to the controller's heating setting register after calculation. This indicates the air supply setting instruction for the next control cycle, which is then written into the controller's fan / damper setting register after calculation. and This indicates that the heating power setting and air supply setting for the current control cycle are both from... Read; and This indicates the core temperature and overall moisture content of the food during the current control period, both from... Read; and This indicates the target center temperature and target moisture content of the current process section, derived from the target card in the controller's formulation area. ; , The endpoints of the power allowable interval generated in step three. , The four endpoints are the air supply allowable interval endpoints generated in step three, and are derived from this step. Read directly from the structure; and The gain written to the controller parameter area during the equipment commissioning phase is used to convert the "deviation composite amount" into the adjustment range of the actuator; and Used to clamp candidate settings within a safe window, this is a commonly used comparison and selection operation in controllers.

[0041] and The setup was completed using the same method as the baking equipment debugging process: during debugging, a standard baking of the target product category was selected, the controller operated according to a fixed control cycle, and small step tests were performed on the heating power setting and air supply setting respectively. The output of the digital twin in subsequent cycles was recorded. and Trend of change; while maintaining Under the premise of unchanged constraints, gradually increase Until The rate of change should meet the process cycle time without exhibiting obvious periodic repetition, and then be adjusted in the same way. make The rate of change meets the dewatering rhythm of the process section. After debugging, and The parameters are written to the controller parameter area and directly called during runtime, thus maintaining the determinism and consistency of the control calculation process. The calculated parameters... As the output of this step, it is written into the setting register in the current cycle so that it takes effect in the next cycle; after the actuator runs according to the new settings in the next cycle, the new device sampling and setting values ​​enter the subsequent cycle, and the new digital twin drives the formation of new... And a new one is generated from step three. This creates a continuous control execution process.

[0042] The above embodiments are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A model predictive control method for food baking equipment driven by a digital twin, characterized in that, include: Structured variables are constructed based on equipment-side operating information and food-side initial information. These structured variables represent the operating state of the digital twin. The food's core temperature state, overall moisture content, and current cycle boundary conditions are selected from these structured variables and assembled sequentially into a state description vector. A constraint tightening coefficient is calculated using a constraint function, and a dynamic safety constraint is calculated based on this coefficient. This dynamic safety constraint is used to adjust the allowable window temperature for the next cycle and includes a power constraint interval and an air supply constraint interval. A target core temperature and target moisture content level are introduced, and the temperature deviation and moisture content deviation for the current cycle are calculated. Combining the temperature deviation, moisture content deviation, power constraint interval endpoints, and air supply constraint interval endpoints, a control command is output through a control function and written into the actuator setting channel of the food baking equipment.

2. The model predictive control method for digital twin-driven food baking equipment according to claim 1, characterized in that, The structured variables include a subset of the initial internal state of the food and a subset of the equipment boundary starting points. The subset of the initial internal state of the food includes at least the initial center temperature and the initial moisture content. The subset of the equipment boundary starting points includes at least the furnace cavity representative temperature, the heating power setting, and the air supply setting.

3. The model predictive control method for digital twin-driven food baking equipment according to claim 1, characterized in that, Before selecting the food core temperature state, the overall moisture content of the food, and the boundary conditions of the current cycle from the structured variables, the process further includes: the digital twin reading the refresh values ​​of the furnace cavity representative temperature, heating power setting, and air supply setting in the register within the same cycle, while retaining the initial fields of the structured variables; and updating the structured variables in a preset order through the heat transfer model and the moisture migration model.

4. The model predictive control method for digital twin-driven food baking equipment according to claim 1, characterized in that, The parameters of the constraint function include the food's core temperature state, the food's overall moisture content state, the oven cavity representative temperature, the oven cavity representative temperature value, the heating power setting value, and the air supply setting value.

5. The model predictive control method for digital twin-driven food baking equipment according to claim 2, characterized in that, The parameters of the control function include an adjustable gain, which is used to convert temperature deviation and moisture content deviation into the adjustment range of the actuator.

6. The model predictive control method for digital twin-driven food baking equipment according to claim 2, characterized in that, The furnace cavity temperature is calculated by averaging the readings of the temperature probes in each furnace cavity at the same time.

7. The model predictive control method for digital twin-driven food baking equipment according to claim 1, characterized in that, The temperature state at the center of the food is obtained by the temperature value at the center grid node of the heat transfer model, and the overall moisture content of the food is obtained by the volume-weighted average of the moisture content of the twin grid.

8. The model predictive control method for digital twin-driven food baking equipment according to claim 1, characterized in that, The boundary conditions for the current cycle include the furnace cavity representative temperature, heating power setting, and air supply setting for the current cycle.

9. The model predictive control method for digital twin-driven food baking equipment according to claim 1, characterized in that, The power constraint range and the air supply constraint range are stored as the lower limit and upper limit fields, respectively.

10. The model predictive control method for digital twin-driven food baking equipment according to claim 1, characterized in that, The target center temperature and target moisture content are obtained through the target card of the current process segment. The target center temperature and target moisture content are written into the controller recipe area when the batch is issued and updated when the process segment is switched.

Images