A Method for Optimizing and Simulating the Steaming Process Parameters of Zongzi Leaves Based on Digital Twins

By deploying a temperature and pressure sensor array in the rice dumpling leaf steaming equipment, constructing a three-dimensional fluid twin, solving the steam flow field characteristics and correcting the heat flux density, and iteratively optimizing the nozzle control parameters, the problem of uneven local heating caused by nonlinear disturbances in the steam flow field was solved, and uniform heating and high-precision simulation of the rice dumpling leaf steaming process were achieved.

CN123069958APending Publication Date: 2026-11-10SUIPING LINGXIN AGRI SCI & TECHNTOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611231858.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-14
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

In existing technologies, the assumption that the steam flow field inside the equipment is ideally uniformly distributed leads to a discrepancy between the heating state of materials in the digital twin space and the physical reality, resulting in uneven heating in certain areas. This fails to effectively address the issues of overcooking due to heat accumulation in some areas and undercooking due to insufficient energy input in others.

Method used

By deploying a temperature and pressure sensor array to collect multidimensional temperature and pressure time-series data in real time, a three-dimensional fluid twin is constructed. The Navier-Stokes equations are solved to obtain the characteristics of the steam velocity vector field and vortex distribution. The vortex attenuation deviation in the flow field is extracted and converted into a heat flux density correction coefficient, which is then injected into the thermodynamic twin model. The nozzle control parameters are iteratively optimized to eliminate the local temperature distortion corresponding to the dead zone in the flow field.

Benefits of technology

It effectively solves the problems of local temperature distortion and heat simulation distortion caused by dead zones in the flow field, realizes dynamic optimization of steam parameters, ensures uniform heating of materials, and improves the simulation accuracy and actual execution effect of the rice dumpling cooking process.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123069958A_ABST
    Figure CN123069958A_ABST
Patent Text Reader

Abstract

This invention relates to the field of computer-aided design technology and discloses a method for optimizing and simulating the steaming process parameters of zongzi leaves based on digital twins. This method collects multi-dimensional temperature and pressure time-series data in physical space using an array of temperature and pressure sensors. This data is then used as initial boundary conditions input into a three-dimensional fluid twin containing the topology of the equipment's internal cavity and nozzles. The Navier-Stokes equations are solved to obtain the steam velocity vector field and vortex distribution characteristics. The measured temperature and pressure distribution field in physical space is spatially aligned and differentially calculated with the virtual flow field solution results. The vortex attenuation deviation is extracted and converted into a heat flux density correction coefficient, which is then injected into the zongzi leaf thermodynamic twin model as a dynamic boundary compensation term. This scheme breaks the assumption of constant boundary conditions, making the thermodynamic solution respond to physical airflow disturbances and solving the problem of localized heating simulation distortion caused by dead zones in the flow field.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of computer-aided design technology, and discloses a method for optimizing and simulating the steaming process parameters of zongzi leaves based on digital twins. Background Technology

[0002] Current digital twin-based food cooking process simulation technology typically employs the construction of a three-dimensional geometric model encompassing the internal geometry of the equipment and the topology of the steam nozzles, upon which a thermodynamic simulation environment is established. The closest conventional approach generally assumes an ideally uniform steam flow field within the equipment, neglecting the nonlinear disturbances to airflow caused by the layout of baffles, material supports, and nozzles within the physical equipment. This conventional approach directly applies constant temperature and pressure boundary conditions to the surface of the material to be cooked, uses the heat conduction equation to solve for the temperature field changes within the material, and uses a set target temperature curve as a benchmark to perform inverse iterative calculations on the global steam supply parameters. Finally, the optimized process parameters—either constant temperature or a single heating rate—are sent to the steam regulating valve group of the physical equipment for execution.

[0003] The core technical problem with the aforementioned existing solutions lies in the fact that, because the steam flow field inside the equipment is set to an ideal, uniform distribution, the material heating state calculated in the digital twin space is seriously inconsistent with the physical reality of vortex dead zones. When dead zones in the flow field are generated inside the physical equipment due to structural interference, the global isothermal parameters obtained and issued based on constant boundary conditions cannot replenish energy to the local dead zones. This results in the process parameters output based on distorted simulation results failing to address the problem of uneven local heating during actual execution, leading to the coexistence of overcooked materials due to heat accumulation and undercooked materials due to insufficient energy input. Summary of the Invention

[0004] The purpose of this invention is to provide a method for optimizing and simulating the steaming process parameters of zongzi leaves based on digital twins, which can effectively solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows:

[0006] The method for optimizing and simulating the process parameters of zongzi leaf steaming based on digital twins includes: real-time acquisition of multi-dimensional temperature and pressure time-series data in the physical space by using an array of temperature and pressure sensors deployed on the inner wall of the physical steaming equipment and around the steam nozzles.

[0007] A three-dimensional fluid twin containing the internal geometry of the equipment cavity and the topology of the steam nozzle is constructed. Using the multidimensional temperature and pressure time series data as the initial boundary conditions, the Navier-Stokes equations are solved to obtain the steam velocity vector field and vortex distribution characteristics in the virtual flow field.

[0008] Spatial alignment and differential calculation are performed between the measured temperature and pressure distribution field in physical space and the solution results of the virtual flow field to extract the flow field vortex attenuation deviation.

[0009] The flow field vortex attenuation deviation is converted into a heat flux density correction coefficient, and the heat flux density correction coefficient is injected as a dynamic boundary compensation term into the zongzi leaf thermodynamic twin model located inside the equipment geometric model.

[0010] In the twin space, the local temperature distortion corresponding to the dead zone of the flow field is eliminated as a constraint condition. The flow distribution timing and opening duty cycle of each steam nozzle are iteratively optimized, and the optimized nozzle control parameter matrix is ​​sent to the steam regulating valve group of the physical equipment.

[0011] Preferably, the real-time acquisition of multi-dimensional temperature and pressure time-series data in the physical space by means of an array of temperature and pressure sensors arranged on the inner wall of the physical cooking equipment and around the steam nozzle includes: arranging temperature sensors on the inner wall of the physical cooking equipment according to the intersection nodes of a three-dimensional grid, and arranging pressure sensors at equal intervals in the circumferential direction of the steam nozzle outlet section.

[0012] A unified high-frequency clock synchronization signal is configured for the temperature sensor and the pressure sensor. The rising edge of the high-frequency clock synchronization signal is used as the trigger reference to perform synchronous analog-to-digital conversion on the temperature analog signal output by the temperature sensor and the pressure analog signal output by the pressure sensor, thereby generating multi-dimensional temperature and pressure time series data carrying an absolute timestamp.

[0013] The high-frequency glitches caused by steam flow pulsation in the multidimensional temperature and pressure time series data are removed by a finite impulse response filter, and the multidimensional temperature and pressure time series data is output after smoothing.

[0014] Preferably, the construction of a three-dimensional fluid twin including the internal geometry of the equipment cavity and the topology of the steam nozzle, using the multidimensional temperature and pressure time series data as the initial boundary conditions, and solving the Navier-Stokes equations to obtain the steam velocity vector field and vortex distribution characteristics in the virtual flow field, includes: using an unstructured hexahedral mesh to mesh the internal geometry of the equipment cavity and the topology of the steam nozzle, and performing local mesh refinement in the interface area between the steam nozzle outlet region and the inner wall.

[0015] The pressure data in the smoothed multidimensional temperature and pressure time series data is mapped to the inlet dynamic pressure boundary condition of the three-dimensional fluid twin, and the temperature data is mapped to the inlet enthalpy boundary condition of the energy equation.

[0016] The filtered Navier-Stokes equations are solved transiently using a large eddy simulation turbulence model. The viscous dissipation rate of steam vortices is calculated using a subgrid-scale model. The output includes the steam velocity vector field and vortex distribution characteristics, which contain the correspondence between three-dimensional spatial coordinates.

[0017] Preferably, the step of spatially aligning and differentially calculating the measured temperature and pressure distribution field in physical space with the solution results of the virtual flow field to extract the flow field vortex attenuation deviation includes: using the unstructured hexahedral mesh nodes in the three-dimensional fluid twin as the reference coordinates, and employing a radial basis function interpolation algorithm to project the measured temperature and pressure distribution field in physical space from the sensor physical coordinates onto the reference coordinates to generate the interpolated measured physical field;

[0018] The interpolated measured physical field is subtracted node by node from the flow field parameters of the corresponding grid nodes in the virtual flow field solution results to generate a spatial difference residual matrix.

[0019] Extract the set of spatial grid nodes from the spatial difference residual matrix where the velocity gradient change rate is lower than the preset turbulence activation threshold and the temperature change rate is higher than the preset thermal accumulation threshold, and define the mean of the difference residuals corresponding to the set of spatial grid nodes as the flow field vortex attenuation deviation.

[0020] Preferably, the step of converting the flow field vortex attenuation deviation into a heat flux density correction coefficient and injecting the heat flux density correction coefficient as a dynamic boundary compensation term into the zongzi leaf thermodynamic twin model located inside the equipment geometric model includes: establishing a mapping function between the flow field vortex attenuation deviation and the attenuation ratio of the steam convection heat transfer coefficient, substituting the flow field vortex attenuation deviation corresponding to the spatial grid node set into the mapping function, and calculating the heat flux density correction coefficient corresponding to each spatial grid node;

[0021] The fluid boundary layer mesh of the outer surface of the wrapping leaf is extracted inside the device geometric model, and the heat flux density correction coefficients corresponding to each spatial mesh node are distributed to the corresponding nodes of the fluid boundary layer mesh according to the spatial distance weight.

[0022] The fluid boundary layer grid node data carrying the heat flux density correction coefficient is used as the third type of boundary condition and superimposed on the surface energy conservation equation of the thermodynamic twin model of the zongzi leaf.

[0023] Preferably, the step of iteratively optimizing the flow distribution timing and duty cycle of each steam nozzle in the twin space with the constraint of eliminating the local temperature distortion corresponding to the dead zone of the flow field, and sending the optimized nozzle control parameter matrix to the steam regulating valve group of the physical equipment includes: constructing an optimization model with the objective function of minimizing the difference between the maximum local temperature difference and the average local temperature difference on the surface of the thermodynamic twin model of the zongzi leaf, and with the constraint that the time interval between two adjacent opening states of a single steam nozzle is greater than the preset mechanical response delay time;

[0024] A non-dominated sorting genetic algorithm with an elite retention strategy is used to solve the Pareto front solution set by performing crossover and mutation operations on the initial flow distribution time sequence and initial duty cycle of each steam nozzle as individuals in the population.

[0025] The solution with the fewest nozzle state switching times is selected from the Pareto front solution set as the nozzle control parameter matrix, and the nozzle control parameter matrix is ​​converted into a control level signal and sent to the steam regulating valve group.

[0026] Preferably, configuring a unified high-frequency clock synchronization signal for the temperature sensor and the pressure sensor includes: deploying a field-programmable gate array in the control cabinet of the physical cooking equipment, and generating a high-frequency clock synchronization signal through the field-programmable gate array;

[0027] The hardware trace delay time of the high-frequency clock synchronization signal transmitted from the field programmable gate array to each temperature sensor and each pressure sensor is measured, and the hardware trace delay time is written into the data register inside each sensor.

[0028] When the rising edge of the high-frequency clock synchronization signal arrives, each sensor reads the sensor measurement value at the current moment and compensates the hardware trace delay time into the count value of the absolute timestamp to generate multidimensional temperature and pressure timing data after transmission delay correction.

[0029] Preferably, the step of using a large eddy simulation turbulence model to perform transient solution of the filtered Navier-Stokes equations and calculating the viscous dissipation rate of steam vortices through a subgrid-scale model includes: using a dynamic Smagorinsky-Lilly model as the subgrid-scale model, and calculating the velocity-strain rate tensor after dual filtering by grid-scale filtering and test filtering by setting a test filter in the steam nozzle outlet region of the physical cooking equipment.

[0030] Based on the aforementioned velocity-strain rate tensor, the model coefficients are dynamically solved using the German fixed-point iterative method, and the solved model coefficients are confined to a non-negative interval that does not cause numerical divergence.

[0031] Substituting the model coefficients within the defined range into the eddy viscosity expression, the subgrid eddy viscosity coefficient, which dynamically varies with spatial and temporal coordinates, is calculated. The viscous dissipation rate of the steam vortex within the three-dimensional fluid twin is then calculated using the subgrid eddy viscosity coefficient.

[0032] Preferably, the step of projecting the measured temperature and pressure distribution field of the physical space from the sensor physical coordinates to the reference coordinates using the radial basis function interpolation algorithm includes: selecting a multivariate quadratic function as the basis function of the radial basis function interpolation algorithm;

[0033] Calculate the shortest spatial distance between the physical coordinates of each sensor, and use the preset scaling factor of the shortest spatial distance as the initial shape parameter of the basis function;

[0034] With the goal of minimizing the sum of squared interpolation residuals of the interpolated measured physical field at non-sensor physical coordinate positions, the golden section method is used to perform a one-dimensional search and iteration on the initial shape parameters in a preset shape parameter space, and the shape parameter corresponding to the convergence of the sum of squared interpolation residuals is determined as the target shape parameter.

[0035] An interpolation matrix is ​​constructed using the target shape parameters, and the interpolation matrix is ​​solved to obtain the interpolated measured physical field data of each node on the reference coordinates.

[0036] Preferably, the step of distributing the heat flux density correction coefficients corresponding to each spatial grid node to the corresponding nodes of the fluid boundary layer grid according to the spatial distance weight includes: extracting vapor streamlines from each spatial grid node to the corresponding node of the fluid boundary layer grid in the three-dimensional fluid twin;

[0037] Calculate the curve integral path length along the steam streamline, and use the curve integral path length as the streamline distance weight to replace the straight-line spatial distance weight;

[0038] Obtain steam density distribution data within the area through which the steam streamline passes, and multiply the steam density distribution data with the streamline distance weight to generate a comprehensive transfer weight;

[0039] The heat flux density correction coefficient corresponding to the spatial grid node is multiplied by the comprehensive transfer weight and accumulated to the initial heat flux density value of the corresponding node of the fluid boundary layer grid to complete the injection of the dynamic boundary compensation term.

[0040] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0041] 1. This invention acquires multidimensional temperature and pressure time-series data in physical space by deploying a sensor array. This data is then used as initial boundary conditions to solve the Navier-Stokes equations in a three-dimensional fluid twin, obtaining the vapor velocity vector field and vortex distribution characteristics. By spatially aligning and differentially calculating the measured temperature and pressure distribution field in physical space with the virtual flow field solution, the vortex attenuation deviation is extracted and converted into a heat flux density correction coefficient, which is then injected into the thermodynamic twin model as a dynamic boundary compensation term. This technique introduces the actual fluid topology deviation inside the equipment into the simulation solution process, breaking the assumption of constant temperature and pressure boundary conditions in existing technologies. This allows the thermodynamic solution logic to respond to nonlinear airflow disturbances in physical space, solving the problems of local temperature distortion and localized heating simulation distortion caused by dead zones in the flow field.

[0042] 2. A high-frequency clock synchronization signal is generated using a field-programmable gate array (FPGA), and the absolute timestamp of the acquired data is corrected by combining the hardware trace delay time. This eliminates the interference of differences in the physical path of sensor signal transmission on the timing alignment accuracy, providing a consistent data source for the subsequent spatial difference calculation of the physical field and the virtual flow field. A dynamic subgrid-scale model combined with the German fixed-point iterative method is used to solve the Navier-Stokes equations, limiting the model coefficients to the non-negative interval and suppressing numerical divergence under high velocity gradients in the steam nozzle outlet region. A radial basis function interpolation algorithm based on the golden section method to determine the target shape parameters is used to reduce the sum of squared interpolation residuals when projecting physical coordinates to the grid reference coordinates. A comprehensive transfer weight is constructed by introducing the curve integral path length along the steam streamline and the steam density distribution data, ensuring that the distribution process of the heat flux density correction coefficient to the fluid boundary layer grid closely matches the actual physical path of heat transfer. Attached Figure Description

[0043] Figure 1 This is the main flowchart of the present invention;

[0044] Figure 2 This is a flowchart of the multidimensional temperature and pressure time-series data acquisition sub-process of the present invention;

[0045] Figure 3 This is a flowchart of the three-dimensional fluid twin construction and flow field solution of the present invention;

[0046] Figure 4 This is a flowchart of the sub-process for extracting the flow field vortex attenuation deviation in this invention;

[0047] Figure 5 This is a flowchart of the heat flux density correction coefficient injection process of the present invention;

[0048] Figure 6 This is a flowchart of the iterative optimization and parameter distribution process of this invention. Detailed Implementation

[0049] Please refer to the attached document. Figure 1 This embodiment provides a method for real-time acquisition of multi-dimensional temperature and pressure time-series data within a physical space by using a temperature and pressure sensor array deployed on the inner wall of the physical cooking equipment and around the steam nozzles. The inner cavity of the physical cooking equipment is a closed cylindrical cavity structure, and the steam nozzles are arranged in an array along the axial and circumferential directions of the cavity. The temperature and pressure sensor array includes two types of sensing units: temperature sensors and pressure sensors.

[0050] Please refer to the attached document. Figure 2Temperature sensors are deployed at the intersections of a three-dimensional grid on the inner wall of the physical steam cooking equipment. The axial and circumferential node spacing of the three-dimensional grid is set according to the cavity size and steam flow characteristics. The spatial positions of adjacent nodes cover the entire flow field monitoring area inside the cavity. Pressure sensors are deployed circumferentially along the outlet cross-section of each steam nozzle, with equal spacing along the circumference of the outlet cross-section, to monitor the changes in steam dynamic pressure at the steam nozzle outlet. The acquisition period of the temperature and pressure sensor array is matched with the transient change characteristics of the steam flow field. The acquired raw data includes the spatial coordinate information of the corresponding sensing unit, the measured value of the physical quantity, and the acquisition time information. After preliminary processing, multidimensional temperature and pressure time-series data is generated.

[0051] Please refer to the attached document. Figure 3 A three-dimensional fluid twin was constructed, incorporating the internal geometry of the equipment and the topology of the steam nozzles. Using multidimensional temperature and pressure time-series data as initial boundary conditions, the Navier-Stokes equations were solved to obtain the steam velocity vector field and vortex distribution characteristics in the virtual flow field. The geometric model of the three-dimensional fluid twin maintained topological consistency with the physical structure of the physical cooking equipment. Through three-dimensional reverse modeling, the geometric features of the equipment's internal cavity, such as the cavity contour, inner wall protrusions, material supports, and the installation position and outlet diameter of the steam nozzles, were obtained, generating a three-dimensional computational domain model that conforms to the actual physical boundaries.

[0052] A three-dimensional computational domain model is meshed to generate a mesh model suitable for fluid computation. The acquired multidimensional temperature and pressure time-series data are mapped to the initial boundary conditions for fluid computation, where pressure data corresponds to the inlet pressure boundary of the flow field calculation, and temperature data corresponds to the inlet temperature boundary of the energy equation. Based on the Navier-Stokes equations for compressible fluids, transient numerical solutions are performed on the steam flow process within the three-dimensional computational domain. The viscosity, compressibility, and thermal conductivity of steam are considered during the calculation. The solution yields the steam velocity vector, pressure value, and temperature value corresponding to each mesh node in the computational domain at each time step, as well as the vortex distribution characteristics calculated based on the velocity vector field. The vortex distribution characteristics are characterized by the modulus of vorticity and its spatial distribution range.

[0053] Please refer to the attached document. Figure 4This method spatially aligns and differentially calculates the measured temperature and pressure distribution field in physical space with the virtual flow field solution, extracting the vortex attenuation deviation in the flow field. The measured temperature and pressure distribution field in physical space is constructed by combining all temperature and pressure measurements corresponding to the same time stamp in multidimensional temperature and pressure time series data with the spatial coordinates of the corresponding sensors. The spatial alignment process uses the grid node coordinates of the three-dimensional fluid twin as a reference, interpolating discrete sensor measurements to the grid nodes corresponding to the reference coordinates to generate measured physical field data with the same spatial coordinate system as the virtual flow field solution. The differential calculation process performs node-by-node subtraction between the measured physical field data and the virtual flow field solution at the same spatial coordinates and time stamp, obtaining the flow field parameter residual values ​​of the corresponding grid nodes. The residual values ​​of all grid nodes constitute the spatial difference residual matrix.

[0054] A set of grid nodes meeting preset conditions is selected from the spatial difference residual matrix. The flow field regions corresponding to these grid nodes have velocity gradient change rates below a preset turbulence activation threshold and temperature change rates above a preset thermal accumulation threshold. These regions correspond to vortex attenuation regions and dead zones in the physical flow field. The mean of the residual values ​​corresponding to the selected set of grid nodes is defined as the vortex attenuation deviation, which characterizes the difference in vortex dissipation characteristics between the virtual flow field solution and the actual physical flow field.

[0055] Please refer to the attached document. Figure 5 The flow field vortex attenuation deviation is converted into a heat flux density correction coefficient, which is then injected as a dynamic boundary compensation term into a leaf-shaped thermodynamic twin model located within the device's geometric model. A mapping relationship is pre-established between the flow field vortex attenuation deviation and the attenuation ratio of the steam convection heat transfer coefficient. This mapping relationship is obtained by fitting convective heat transfer experimental data and reflects the decrease in convective heat transfer capacity caused by vortex attenuation. Substituting the flow field vortex attenuation deviation into this mapping relationship, the heat flux density correction coefficient for the corresponding spatial grid node is calculated. The numerical range of the heat flux density correction coefficient covers the entire range of convective heat transfer capacity attenuation and is used to correct the overestimated convective heat transfer in the virtual flow field.

[0056] The thermodynamic twin model of zongzi leaves is constructed based on the physical properties and heat conduction laws of zongzi leaves. It includes thermodynamic parameters such as the geometric structure, thermal conductivity, specific heat capacity, and density of the leaves, and can solve for the temperature field distribution inside and on the surface of the leaves. In the computational domain of the three-dimensional fluid twin, the fluid boundary layer mesh covering the outer surface of the zongzi leaves is extracted. This boundary layer mesh corresponds to the region where steam and the leaf surface undergo convective heat transfer. The heat flux density correction coefficients corresponding to each spatial mesh node are distributed to the corresponding nodes of the fluid boundary layer mesh according to their spatial correspondence, generating boundary condition data with correction coefficients. This boundary condition data is used as a dynamic boundary compensation term and superimposed on the surface energy conservation equation of the zongzi leaf thermodynamic twin model, enabling the solution process of the zongzi leaf thermodynamic twin model to respond to changes in convective heat transfer capacity caused by vortex attenuation in the physical flow field.

[0057] In the twin space, with the elimination of local temperature distortion corresponding to the flow field dead zone as a constraint, the flow distribution timing and duty cycle of each steam nozzle are iteratively optimized. The optimized nozzle control parameter matrix is ​​then sent to the steam regulating valve group of the physical equipment. A parameter optimization model is constructed in the twin space. The input of the optimization model is the flow distribution timing and duty cycle of each steam nozzle, and the output is the temperature field distribution on the surface of the zongzi leaf calculated by the thermodynamic twin model of the zongzi leaf. The constraint of the optimization model is to eliminate the local temperature distortion corresponding to the flow field dead zone, that is, the difference between the temperature value of the zongzi leaf surface region corresponding to the flow field dead zone and the target cooking temperature is within a preset allowable range.

[0058] The optimization process adjusts the flow distribution timing and duty cycle of each steam nozzle using a preset iteration step size. After each parameter adjustment, the flow field solution of the 3D fluid twin, deviation extraction, heat flux density correction coefficient calculation, and temperature field solution of the zongzi leaf thermodynamic twin model are re-executed to obtain the temperature field distribution results after parameter adjustment. When the iteration process meets the convergence condition, the iteration stops, and the current steam nozzle flow distribution timing and duty cycle parameters are output, generating a nozzle control parameter matrix. The nozzle control parameter matrix contains the control timing, opening value, and opening duration information of each steam regulating valve. This matrix is ​​converted into control signals that can be recognized by the physical equipment and sent to the steam regulating valve group of the physical cooking equipment. The steam regulating valve group then executes the opening and closing and opening adjustment actions of the corresponding valves according to the control signals.

[0059] In this embodiment, in order to clarify the composition and corresponding physical meaning of the multidimensional temperature and pressure time series data, a multidimensional temperature and pressure time series data field correspondence table as shown in Table 1 is constructed.

[0060] Table 1. Correspondence between fields and physical meanings of multidimensional temperature and pressure time series data.

[0061] Field Name Data types Physical meaning Data source Absolute timestamp 64-bit unsigned integer The absolute time reference for data acquisition, in microseconds. Synchronous acquisition clock counter Sensor spatial coordinates X 32-bit floating-point The X-axis coordinate of the sensor installation location in the device coordinate system, in meters. Equipment 3D geometric model calibration data Sensor spatial coordinates Y 32-bit floating-point The sensor's installation location is the Y-axis coordinate of the device coordinate system, in meters. Equipment 3D geometric model calibration data Sensor spatial coordinates Z 32-bit floating-point The Z-axis coordinate of the sensor installation location in the device coordinate system, in meters. Equipment 3D geometric model calibration data Temperature measurement value 32-bit floating-point The corresponding medium temperature value acquired by the sensor at the time of data collection, in Kelvin. Analog-to-digital conversion results of temperature sensing unit Pressure measurement value 32-bit floating-point The static pressure value of the medium acquired by the sensor at the corresponding acquisition time, in Pascals. Analog-to-digital conversion results of pressure sensing unit Sensor type identification 8-bit unsigned integer Distinguishing between temperature sensors and pressure sensors: type identification Sensor hardware address encoding

[0062] The fields in Table 1 constitute the smallest data unit of the multidimensional temperature and pressure time series data. All data units under the same absolute timestamp constitute the physically measured temperature and pressure distribution field at that moment, providing a data source with a unified format for subsequent flow field solutions and spatial difference calculations. The numerical range and precision of each field match the transient change characteristics and spatial distribution characteristics of the steam flow field, ensuring that the collected data can fully reflect the temperature and pressure change process of the physical flow field.

[0063] The implementation of this embodiment drives the solution process of the virtual flow field through measured temperature and pressure data in physical space. The deviation between the physical flow field and the virtual flow field is transformed into a dynamic boundary compensation term and injected into the thermodynamic twin model of the zongzi leaf. This enables the simulation solution process to fit the actual flow and heat transfer characteristics inside the physical device, corrects the simulation distortion problem caused by the assumption of an ideal uniform flow field, and provides a simulation model basis that conforms to physical reality for the parameter optimization of steam nozzles.

[0064] In a preferred embodiment, temperature sensors are arranged at three-dimensional grid intersections on the inner wall of the physical cooking equipment, and pressure sensors are evenly spaced along the circumferential direction of the steam nozzle outlet cross-section. The temperature sensors are arranged in a three-dimensional grid along the axial, circumferential, and radial directions of the equipment cavity. Axial nodes cover the entire length of the cavity, circumferential nodes are evenly distributed along the circumference, and radial nodes correspond to different radial heights on the inner wall, ensuring that the temperature monitoring range covers the temperature field changes throughout the entire cavity. The number of pressure sensors matches the outlet cross-sectional size of the steam nozzles, and they are evenly spaced along the circumferential direction of each nozzle outlet cross-section. Adjacent pressure sensors have the same central angle, enabling complete monitoring of the steam pressure distribution along the circumferential direction of the nozzle outlet and capturing the non-uniform airflow characteristics at the nozzle outlet.

[0065] A unified high-frequency clock synchronization signal is configured for both temperature and pressure sensors. Using the rising edge of this signal as the trigger reference, synchronous analog-to-digital conversion is performed on the analog temperature and pressure signals output by the temperature and pressure sensors, generating multi-dimensional temperature and pressure time-series data with absolute timestamps. The frequency of the high-frequency clock synchronization signal matches the sensor's analog-to-digital conversion rate, ensuring a complete signal acquisition and conversion is completed within each clock cycle. All analog-to-digital conversion trigger terminals of the temperature and pressure sensors are connected to the same high-frequency clock synchronization signal, using the rising edge as a unified conversion trigger reference to eliminate sampling start-time discrepancies between different sensors. After analog-to-digital conversion, an absolute timestamp of the current moment is appended to each converted digital data point. The count of the absolute timestamp is synchronized with the cycle number of the high-frequency clock synchronization signal, ensuring that the data acquired by different sensors have a unified time reference.

[0066] A finite impulse response (FIR) filter is used to remove high-frequency glitches caused by steam flow pulsations from multidimensional temperature and pressure time-series data, resulting in smoothed multidimensional temperature and pressure time-series data. The FIR filter is a linear phase filter with a finite-length unit impulse response, capable of filtering high-frequency noise while maintaining the signal's phase characteristics without distortion, thus preventing time base shifts in the time-series data. The filter's passband and stopband cutoff frequencies are set based on the normal frequency variation of the steam flow field and the frequency of high-frequency noise caused by airflow pulsations. The passband covers the effective frequency variation range of the steam flow field, and the stopband corresponds to the frequency range of high-frequency glitches caused by airflow pulsations. Temperature and pressure measurements from the acquired multidimensional temperature and pressure time-series data are input into the FIR filter. After filtering, the smoothed measurements are output, replacing the corresponding fields in the original data to generate smoothed multidimensional temperature and pressure time-series data.

[0067] In this embodiment, the difference equation of the finite impulse response filter is defined by the following formula:

[0068]

[0069] in, For the filter number The output value at each sampling time, i.e. the smoothed temperature and pressure measurement value; The order of the filter; For the filter's first The unit impulse response coefficient; For the filter number The input value at each sampling time, i.e. the original temperature and pressure measurement value collected.

[0070] Furthermore, a field-programmable gate array (FPGA) is deployed in the control cabinet of the physical cooking equipment to generate a high-frequency clock synchronization signal. The FPGA integrates a phase-locked loop (PLL) module, which multiplies the input reference clock signal to generate a stable high-frequency clock synchronization signal. The frequency jitter of the high-frequency clock synchronization signal is within a preset allowable range, ensuring the accuracy of the sampling trigger timing. The input and output ports of the FPGA are connected to the trigger terminals of each temperature and pressure sensor via hardware wiring, transmitting the generated high-frequency clock synchronization signal to each sensor.

[0071] The hardware trace delay time of the high-frequency clock synchronization signal transmitted from the field-programmable gate array (FPGA) to each temperature sensor and each pressure sensor is measured, and this delay time is written into the data register inside each sensor. The hardware trace delay time is obtained using time-domain reflectometry (TD-RS). During the measurement, the FPGA sends a pulse signal to the corresponding sensor's hardware trace, and the transmission time of the pulse signal and the reception time of the reflected signal are recorded. Half of the time difference between these two moments is the transmission delay time of that hardware trace. The delay time is measured for each hardware trace connected to a sensor, and this value is written into the corresponding sensor's internal non-volatile data register for subsequent timestamp correction.

[0072] When the rising edge of the high-frequency clock synchronization signal arrives, each sensor reads its current measurement value and compensates for the hardware trace delay time in the absolute timestamp count, generating multidimensional temperature and pressure timing data corrected for transmission delay. The timestamp counter inside the sensor is synchronized with the high-frequency clock synchronization signal. When the rising edge of the high-frequency clock synchronization signal reaches the sensor's trigger terminal, the sensor latches the current analog signal measurement value and simultaneously reads the current timestamp count and the hardware trace delay time stored in the data register. The hardware trace delay time is subtracted from the timestamp count to obtain the corrected absolute timestamp. The corrected absolute timestamp corresponds to the actual physical time of signal acquisition, eliminating transmission delay deviations caused by differences in hardware trace lengths and ensuring that all sensor data has a strictly unified time reference.

[0073] In this embodiment, the transmission delay correction process for absolute timestamps is defined by the following formula:

[0074]

[0075] in, For the first The calibrated absolute timestamps corresponding to each sensor; For the first The raw timestamp count value read when each sensor latches a measurement value; For the first Hardware trace delay time for each sensor.

[0076] In this embodiment, to clarify the coefficient configuration and corresponding filtering characteristics of the finite impulse response filter, a table showing the correspondence between filter coefficient configuration and filtering characteristics as shown in Table 2 is constructed.

[0077] Table 2. Correspondence between coefficient configuration and filtering characteristics of finite impulse response filters.

[0078] Filter order passband cutoff frequency stopband cutoff frequency Maximum passband attenuation Minimum stopband attenuation Window function type Applicable data types 32nd order 50Hz 200Hz 0.5dB 60dB Hanning Window Temperature measurement time series data 64th level 100Hz 500Hz 0.3dB 80dB Blackman window Pressure measurement time series data 16th order 20Hz 100Hz 1.0dB 40dB rectangular window Low-frequency temperature and pressure trend data

[0079] The filter configurations in Table 2 correspond to different types of acquired data. Temperature measurement data has a relatively slow rate of change, so a lower passband cutoff frequency and a 32nd-order filter are used to filter out high-frequency noise while maintaining the trend characteristics of the temperature signal. Pressure measurement data is greatly affected by steam flow pulsation and has richer high-frequency noise components, so a higher-order filter and a steeper transition band are used. Low-frequency temperature and pressure trend data are used as initial boundary conditions for flow field solutions, and low-order filters are used to achieve fast smoothing. All filter configurations maintain linear phase characteristics to avoid phase distortion of time-series data caused by filtering, ensuring the consistency of the time base for subsequent transient flow field solutions.

[0080] This embodiment achieves synchronous data acquisition from multiple sensors through a unified high-frequency clock synchronization signal. Combined with the measurement and compensation of hardware trace delay time, it eliminates sampling time deviations between different sensors, providing a data source with a strictly consistent time reference for the differential calculation of the physical field and the virtual flow field. A linear-phase finite impulse response filter removes high-frequency pulsating noise from the acquired data, improving data smoothness while maintaining signal phase characteristics and preventing noise signals from interfering with subsequent flow field solution processes.

[0081] In another preferred embodiment, an unstructured hexahedral mesh is used to mesh the internal geometry of the equipment cavity and the topology of the steam nozzle, with local mesh refinement in the boundary region between the steam nozzle outlet area and the inner wall. The unstructured hexahedral mesh has good geometric adaptability, able to conform to the complex boundary structure of the equipment cavity and the outlet topology of the steam nozzle, avoiding the mesh distortion problem that occurs with structured meshes on complex curved surfaces. During mesh generation, the computational domain is decomposed into four sub-regions: the steam nozzle region, the main flow field region, the near-wall region, and the leaf-shaped surface boundary layer region. Different mesh sizes are set for the flow characteristics of different sub-regions.

[0082] The steam nozzle outlet region is characterized by high velocity gradients and high turbulence intensity. Flow separation and vortex generation occur at the inner wall interface. Mesh refinement is applied locally to these two regions, and the refined mesh size captures the small-scale flow structure and transient characteristics within these areas. The flow in the main flow field is relatively stable, so a relatively large mesh size is used to reduce computational complexity while maintaining accuracy. The mesh size for the boundary layer region on the bamboo leaf surface is matched to the boundary layer thickness of the steam flow, ensuring accurate calculation of the velocity and temperature gradients within the boundary layer and providing accurate flow field data for subsequent convective heat transfer calculations.

[0083] The pressure data in the smoothed multidimensional temperature and pressure time series data is mapped to the inlet dynamic pressure boundary condition of the 3D fluid twin, and the temperature data is mapped to the inlet enthalpy boundary condition of the energy equation. The flow field calculation inlet of the 3D fluid twin corresponds to the steam nozzle outlet section of the physical device, with each steam nozzle outlet section being an independent inlet boundary. For each inlet boundary, the smoothed pressure data collected by the pressure sensors deployed at the corresponding steam nozzle outlet is mapped to the dynamic pressure boundary condition of that inlet boundary. The dynamic pressure boundary condition is dynamically updated with the time step, keeping pace with the measured pressure changes in the physical space. Based on the smoothed temperature data and combined with the thermodynamic characteristics of the steam, the steam enthalpy at the corresponding temperature is calculated, and this enthalpy is mapped to the inlet enthalpy boundary condition of the energy equation. The inlet enthalpy is also dynamically updated with the time step, keeping pace with the measured temperature changes in the physical space. Except for the inlet boundary, the inner wall of the equipment cavity is set as a no-slip wall boundary condition. The wall temperature is dynamically updated using measured data from temperature sensors installed on the inner wall. The surface of the zongzi leaf is set as a fluid-structure interaction wall boundary condition for subsequent data interaction with the zongzi leaf thermodynamic twin model.

[0084] A large eddy simulation (LES) turbulence model is used to transiently solve the filtered Navier-Stokes equations. The viscous dissipation rate of steam vortices is calculated using a subgrid-scale model, outputting a steam velocity vector field and vortex distribution characteristics that include the correspondence between three-dimensional spatial coordinates. The LES turbulence model decomposes the vortex structure in the flow field into large-scale solvable vortices and small-scale subgrid-scale vortices through a filtering function. Large-scale vortices are solved directly using the Navier-Stokes equations, while the influence of small-scale vortices on large-scale vortices is modeled using a subgrid-scale model, avoiding the extremely high requirements for grid resolution and computational resources imposed by direct numerical simulation.

[0085] The filtering process employs a grid-scale box filter function to filter the Navier-Stokes equations for compressible fluids, yielding filtered continuity, momentum, and energy equations. Transient solutions are then performed based on these filtered governing equations. During the solution process, subgrid-scale stresses and heat fluxes induced by small-scale vortices are calculated using a subgrid-scale model, thereby calculating the viscous dissipation rate of the steam vortices. This viscous dissipation rate is then incorporated into the solution of the momentum and energy equations. After each time step, the system outputs steam velocity vectors, pressure, temperature, vorticity, and viscous dissipation rate data corresponding to all grid nodes within the computational domain. This generates a steam velocity vector field and vortex distribution characteristics containing three-dimensional spatial coordinate correspondences. The vortex distribution characteristics are characterized by the spatial distribution of vorticity and the numerical values ​​of the viscous dissipation rate.

[0086] In this embodiment, the filtered Navier-Stokes continuity equation and momentum equation for compressible fluids are defined by the following formulas:

[0087]

[0088]

[0089] in, The density of the filtered steam; For time; , The filtered velocity components are averaged using the Favre mass-weighted method. , These are spatial coordinate components; The filtered steam pressure; The molecular dynamic viscosity of steam; This is the filtered velocity-strain rate tensor; For subgrid stress tensor, it characterizes the effect of small-scale vortices on large-scale flow.

[0090] Furthermore, the filtered velocity-strain rate tensor is defined by the following formula:

[0091]

[0092] The physical meaning of each parameter is consistent with the aforementioned formula, with no redefinition.

[0093] A dynamic Smagorinsky-Lilly model was employed as the subgrid-scale model. A test filter was placed in the steam nozzle outlet region of the physical cooking equipment to calculate the velocity-strain rate tensor after dual filtering using both grid-scale and test filters. The test filter's filtering scale is twice that of the grid-scale filter, enabling the extraction of larger-scale flow structure information and providing a data foundation for the dynamic solution of model coefficients. The solved velocity field was then filtered using both grid-scale and test filters to obtain dual-filtered velocity components. The corresponding velocity-strain rate tensor was calculated based on these dual-filtered velocity components for subsequent model coefficient calculation.

[0094] Based on the velocity-strain rate tensor, the German fixed-point iterative method is used to dynamically solve for the model coefficients, confining the obtained model coefficients within a non-negative interval that does not cause numerical divergence. The German fixed-point iterative method constructs fixed-point equations to iteratively solve for the Smagorinsky model coefficients, avoiding the dependence on spatial and temporal averaging found in traditional dynamic models. This allows for local dynamic adjustment of the model coefficients, better adapting to the local variations in the flow field. During the iteration process, the model coefficients at the current step size are used as initial values ​​and substituted into the fixed-point equations for iterative calculation. The iteration stops when the residual value of the iteration result is less than a preset convergence threshold, and the current model coefficients are output. To prevent negative values ​​in the obtained model coefficients from causing numerical divergence, the lower limit of the model coefficients is set to 0, and the upper limit is set to a preset maximum allowable value, confining the iteratively solved model coefficients within this non-negative interval and ensuring the stability of the numerical solution.

[0095] Substituting the model coefficients within a defined interval into the eddy viscosity expression, the subgrid eddy viscosity coefficient, dynamically varying with spatial and temporal coordinates, is calculated. This subgrid eddy viscosity coefficient is then used to calculate the viscous dissipation rate of steam vortices within a three-dimensional fluid twin. The subgrid stress tensor is modeled using the eddy viscosity assumption, correlating the anisotropic component of the subgrid stress with the filtered velocity-strain rate tensor. The subgrid eddy viscosity coefficient characterizes the viscous dissipation effect of small-scale vortices. Substituting the dynamically solved model coefficients into the eddy viscosity expression, the subgrid eddy viscosity coefficient corresponding to each grid node and time step is calculated. This subgrid eddy viscosity coefficient dynamically adjusts with the spatial structure and temporal changes of the flow field, accurately reflecting the dissipation characteristics of small-scale vortices in different regions and at different times. Based on the subgrid eddy viscosity coefficient and the velocity-strain rate tensor, the viscous dissipation rate of steam vortices is calculated. This viscous dissipation rate characterizes the rate at which the mechanical energy of small-scale vortices is converted into thermal energy, providing accurate dissipation term data for subsequent energy equation solutions and convective heat transfer calculations.

[0096] In this embodiment, the formula for calculating the subgrid eddy viscosity coefficient is defined as follows:

[0097]

[0098] in, The subgrid eddy viscosity coefficient; These are the coefficients of the Smagorinsky model obtained through dynamic solution; For grid filtering scale; The magnitude of the filtered velocity-strain rate tensor is defined as follows: .

[0099] In this embodiment, to clarify the grid division rules and encryption requirements for different regions of the computing domain, a computing domain grid division region encryption rule table as shown in Table 3 is constructed.

[0100] Table 3. Computation Domain Grid Partitioning and Densification Rules

[0101] Computational domain subregion Basic grid size Encryption multiplier Grid type Core encryption control indicators Steam nozzle outlet area Basic grid size 8x Encryption Unstructured hexahedron Velocity gradient capture accuracy at nozzle exit Inner wall junction area Basic grid size 4x encryption Unstructured hexahedron Near-wall flow separation and vortex structure resolution Boundary layer area on the surface of zongzi leaves Basic grid size 6x encryption Boundary layer structured hexahedron The y+ value within the boundary layer is in the range of 1-5. Mainstream areas Basic grid size Unencrypted Unstructured hexahedron Mesh orthogonality greater than 0.3, aspect ratio less than 5

[0102] The mesh refinement rules in Table 3 are set according to the flow characteristics of different sub-regions. The high velocity gradient in the steam nozzle outlet region requires a higher mesh resolution, so an 8x refinement is used. The flow separation and vortex generation in the inner wall boundary region need to capture the medium-scale flow structure, so a 4x refinement is used. The boundary layer region on the surface of the bamboo leaf needs to accurately solve the velocity and temperature gradient within the boundary layer, so a 6x refinement is used and the y+ value of the boundary layer mesh is controlled to ensure the accuracy of the convective heat transfer calculation. The flow in the main flow field region is relatively stable, so the basic mesh size is used, while controlling the orthogonality and aspect ratio of the mesh to ensure the stability and accuracy of the numerical solution.

[0103] This embodiment improves the mesh resolution of key flow regions while controlling computational load through unstructured hexahedral mesh partitioning and local refinement, providing a high-quality mesh foundation for accurate flow field solutions. By employing a dynamic Smagorinsky-Lilly subgrid-scale model combined with the German fixed-point iteration method, the model coefficients are confined to the non-negative interval, suppressing numerical divergence in high-velocity gradient regions and improving the accuracy of the vapor vortex viscous dissipation rate solution. This allows the virtual flow field solution to better reflect the actual flow characteristics in physical space.

[0104] In another preferred embodiment, using the unstructured hexahedral mesh nodes in the three-dimensional fluid twin as reference coordinates, a radial basis function interpolation algorithm is employed to project the measured temperature and pressure distribution field in physical space from the sensor's physical coordinates onto the reference coordinates, generating the interpolated measured physical field. The radial basis function interpolation algorithm is a meshless interpolation method that can perform high-precision interpolation of physical quantity values ​​at any reference coordinate point based on discrete sensor measurement point data. It is not limited by the mesh topology and can be well adapted to the reference coordinate system of unstructured hexahedral meshes. The interpolation process uses the sensor's physical coordinates as discrete data points, the corresponding temperature and pressure measurements as discrete data values, and the mesh node coordinates of the three-dimensional fluid twin as interpolation target points. By constructing a radial basis function interpolation matrix, the temperature and pressure values ​​corresponding to all interpolation target points are solved, generating the interpolated measured physical field with the same spatial coordinate system as the virtual flow field solution.

[0105] Multivariate quadratic functions are chosen as the basis functions for the radial basis function interpolation algorithm. Multivariate quadratic functions possess good numerical stability and interpolation accuracy, and can adapt to arbitrary discrete point distributions in three-dimensional space, avoiding numerical singularities that occur with some basis functions when the discrete point distribution is uneven. The shortest spatial distance between the physical coordinates of each sensor is calculated, and a preset scaling factor of the shortest spatial distance is used as the initial shape parameter of the basis function. The initial shape parameter is set to match the sensor deployment density, ensuring that the support range of the basis function can cover the spatial region between adjacent sensors.

[0106] The objective is to minimize the sum of squared interpolation residuals of the interpolated measured physical field at non-sensor physical coordinate positions. The golden section method is used to perform a one-dimensional search and iteration on the initial shape parameter within a preset shape parameter space. The upper and lower limits of the preset shape parameter space are set according to the maximum and minimum spacing of the sensor deployment, ensuring that the iteration process covers the range of optimal shape parameter values. In each iteration, the current shape parameter is substituted into the radial basis function to construct the interpolation matrix and solve for the interpolation result. The corresponding sum of squared interpolation residuals is calculated. When the change in the sum of squared interpolation residuals is less than a preset convergence threshold, the iteration stops, and the current shape parameter is determined as the target shape parameter. An interpolation matrix is ​​constructed using the target shape parameter, and solving the interpolation matrix yields the interpolated measured physical field data for each node on the reference coordinates. The interpolated measured physical field data includes the temperature and pressure values ​​corresponding to each reference coordinate node, perfectly matching the spatial coordinates and data dimensions of the virtual flow field solution.

[0107] In this embodiment, the formula for calculating radial basis function interpolation is defined as follows:

[0108]

[0109] in, For the target interpolation point The interpolation result at that point, i.e., the temperature and pressure values; This represents the total number of discrete data points from the sensor. For the first Interpolation weight coefficients corresponding to each discrete data point; These are multivariate quadratic radial basis functions; For the target interpolation point With the discrete data points from a sensor The Euclidean distance between them.

[0110] Furthermore, the expression for the multivariate quadratic radial basis function is defined as follows:

[0111]

[0112] in, The Euclidean distance between the two points; These are the target shape parameters obtained through optimization using the golden section method.

[0113] The flow field parameters at corresponding grid nodes in the interpolated measured physical field and the virtual flow field solution are subtracted node by node to generate a spatial difference residual matrix. The flow field parameters include the temperature and pressure values ​​of the corresponding grid nodes. For each reference coordinate grid node, the difference between the interpolated measured physical field temperature value and the temperature value in the virtual flow field solution, as well as the difference between the measured pressure value and the pressure value in the virtual flow field, are calculated to obtain the temperature and pressure residual values ​​for that node. The temperature and pressure residual values ​​of all grid nodes constitute the spatial difference residual matrix. The dimension of the spatial difference residual matrix matches the total number of grid nodes in the 3D fluid twin, and each matrix element corresponds to a unique spatial grid node, containing the temperature and pressure residual information for that node.

[0114] In this embodiment, the formula for calculating the elements of the spatial difference residual matrix is ​​defined as follows:

[0115]

[0116]

[0117] in, For the first Temperature residual values ​​of each grid node; For the first The interpolated measured temperature value of each grid node; For the first Temperature values ​​are obtained by solving the virtual flow field at each grid node. For the first Pressure residual values ​​of each grid node; For the first The interpolated actual pressure value of each grid node; For the first The pressure value is obtained by solving the virtual flow field of each grid node.

[0118] The spatial grid node set is extracted from the spatial difference residual matrix, where the velocity gradient rate of change is lower than a preset turbulence activation threshold and the temperature rate of change is higher than a preset heat accumulation threshold. The mean of the difference residual corresponding to the spatial grid node set is defined as the vortex attenuation deviation of the flow field. The velocity gradient rate of change is calculated based on the velocity vector field obtained from the virtual flow field solution, representing the turbulence intensity in the region where the corresponding grid node is located. Regions where the velocity gradient rate of change is lower than the preset turbulence activation threshold indicate weak turbulence or laminar flow, corresponding to the dead zone of the flow field after vortex attenuation. The temperature rate of change is calculated based on the temperature distribution of the interpolated measured physical field, representing the degree of temperature accumulation in the corresponding region. Regions where the temperature rate of change is higher than the preset heat accumulation threshold indicate temperature distortion caused by heat accumulation. All grid nodes are screened, and grid nodes that simultaneously meet the above two conditions are extracted to form the spatial grid node set. Calculate the mean of the temperature residual and pressure residual values ​​corresponding to all grid nodes in the set, and define the mean as the flow field vortex attenuation deviation. The flow field vortex attenuation deviation characterizes the simulation deviation of the changes in flow and heat transfer characteristics caused by vortex attenuation during the virtual flow field solution process.

[0119] In this embodiment, the formula for calculating the flow field vortex attenuation deviation is defined as follows:

[0120]

[0121] in, This represents the deviation in flow field vortex attenuation. This represents the total number of nodes within the spatial grid node set. The weighting coefficient for the temperature residual; The weighting coefficient for the pressure residual; , The temperature and pressure residuals for the corresponding grid nodes are defined in the same way as the aforementioned formulas.

[0122] A mapping function is established between the vortex attenuation deviation in the flow field and the attenuation ratio of the steam convection heat transfer coefficient. Substituting the vortex attenuation deviation corresponding to each set of spatial grid nodes into the mapping function, the heat flux density correction coefficient for each spatial grid node is calculated. The mapping function is obtained by fitting convective heat transfer experimental data. The independent variable is the vortex attenuation deviation in the flow field, and the dependent variable is the attenuation ratio of the convective heat transfer coefficient. The attenuation ratio is the ratio of the actual convective heat transfer coefficient to the convective heat transfer coefficient under the assumption of an ideal uniform flow field, with a value ranging from 0 to 1. The smaller the value, the more severe the attenuation of convective heat transfer capacity. Substituting the vortex attenuation deviation corresponding to each spatial grid node into the mapping function, the attenuation ratio of the convective heat transfer coefficient corresponding to that node is calculated. This attenuation ratio is defined as the heat flux density correction coefficient. The heat flux density correction coefficient is used to correct the overestimated convective heat transfer in the virtual flow field, making the simulated convective heat flux density closely match the actual heat transfer capacity.

[0123] In this embodiment, the mapping function for the heat flux density correction coefficient is defined as follows:

[0124]

[0125] in, For the first Heat flux density correction factor for each spatial grid node; For the first The flow field vortex attenuation deviation corresponding to each spatial grid node; The proportionality coefficient is obtained by fitting experimental data, and it is a positive constant.

[0126] The fluid boundary layer mesh of the outer surface of the wrapping leaf is extracted within the device's geometric model. The heat flux density correction coefficients corresponding to each spatial mesh node are then assigned to the corresponding nodes of the fluid boundary layer mesh according to spatial distance weights. The fluid boundary layer mesh represents the fluid mesh near the leaf surface, corresponding to the region where steam and the leaf surface undergo convective heat transfer. The flow and heat transfer characteristics within this region directly determine the heat flux density input to the leaf surface. Steam streamlines pointing from each spatial mesh node to the corresponding node of the fluid boundary layer mesh are extracted from the 3D fluid twin. These steam streamlines represent the trajectories of steam particles in the flow field, and the direction of the streamlines represents the actual physical path of steam flow and heat transfer.

[0127] The curve integral path length along the steam streamline is calculated and used as the streamline distance weight instead of the straight-line spatial distance weight, avoiding the limitation of the straight-line distance weight in reflecting the actual flow path. Steam density distribution data within the streamline's path region is acquired, and this data is multiplied by the streamline distance weight to generate a comprehensive transfer weight. Steam density is directly related to the steam's heat-carrying capacity; higher density regions have stronger heat transfer capabilities. The heat flux density correction coefficient corresponding to the spatial grid node is multiplied by the comprehensive transfer weight and accumulated into the initial heat flux density value of the corresponding node in the fluid boundary layer grid, completing the injection of the dynamic boundary compensation term.

[0128] In this embodiment, the formula for calculating the comprehensive transfer weight is defined as follows:

[0129]

[0130] in, For the first The spatial grid node to the first The overall transfer weight passed by each fluid boundary layer mesh node; To connect the first The node and the first The average steam density of the region through which the steam streamline passes at each node; The integral path length along the steam streamline is denoted as .

[0131] The fluid boundary layer mesh node data carrying a heat flux density correction coefficient is used as a third type of boundary condition and superimposed on the surface energy conservation equation of the zongzi leaf thermodynamic twin model. This third type of boundary condition is a convective heat transfer boundary condition. Given the convective heat transfer coefficient between the fluid and the solid wall and the mainstream fluid temperature, the surface energy conservation equation of the zongzi leaf thermodynamic twin model is constructed based on the laws of heat conduction and energy conservation, describing the balance between convective heat transfer on the zongzi leaf surface and internal heat conduction. By using the convective heat transfer coefficient carrying the heat flux density correction coefficient and fluid temperature data as the third type of boundary condition and superimposing them on the surface energy conservation equation, the solution process of the zongzi leaf thermodynamic twin model can respond to changes in convective heat transfer capacity caused by vortex attenuation in the physical flow field, correcting the distortion of heat transfer boundary conditions caused by the assumption of an ideal homogeneous flow field.

[0132] In the twin space, the local temperature distortion corresponding to the dead zone in the flow field is eliminated as a constraint. The flow distribution timing and duty cycle of each steam nozzle are iteratively optimized, and the optimized nozzle control parameter matrix is ​​sent to the steam regulating valve group of the physical equipment. An optimization model is constructed with the objective function of minimizing the difference between the maximum local temperature difference and the average local temperature difference on the surface of the zongzi leaf thermodynamic twin model, and the constraint that the time interval between two adjacent opening states of a single steam nozzle is greater than the preset mechanical response delay time. The optimization objective of the objective function is the uniformity of temperature distribution on the surface of the zongzi leaf. The smaller the difference between the maximum local temperature difference and the average local temperature difference, the more uniform the temperature distribution on the surface of the zongzi leaf and the lower the degree of local temperature distortion.

[0133] The constraints consider the mechanical response characteristics of the steam control valve. The time interval between two consecutive opening states of a single steam nozzle must be greater than the mechanical response delay time of the valve to avoid mechanical wear and control failure caused by frequent valve switching. The design variables of the optimization model are the flow distribution sequence and the opening duty cycle of each steam nozzle. The flow distribution sequence is the flow setpoint of each steam nozzle at different time steps, and the opening duty cycle is the ratio of the valve opening time to the total cycle time within a single control cycle.

[0134] In this embodiment, the objective function of the optimization model is defined as follows:

[0135]

[0136] in, The objective function value; The maximum surface temperature of the zongzi leaf calculated using a thermodynamic twin model; The average surface temperature of the zongzi leaf is calculated using a thermodynamic twin model of zongzi leaves.

[0137] Please refer to the attached document. Figure 6 A non-dominated sorting genetic algorithm with an elitist retention strategy is employed. The initial flow allocation sequence and initial duty cycle of each steam nozzle are used as the population individuals for crossover and mutation operations to solve the optimization model and obtain the Pareto front solution set. This algorithm can handle multi-objective optimization problems. It stratifies the population through non-dominated sorting, maintains population diversity through crowding calculation, and the elitist retention strategy directly preserves the best individuals from the current generation to the next, avoiding the loss of optimal solutions and improving the convergence speed and stability of the optimization process.

[0138] The initial population size is set based on the number of steam nozzles and the length of the control sequence. Each individual in the population corresponds to a complete set of steam nozzle flow distribution timing and duty cycle parameters. Selection, crossover, and mutation operations are performed on the individuals to generate a new generation of the population. After each generation, the parameters corresponding to the individuals are input into a 3D fluid twin and a thermodynamic twin model of a bamboo leaf to calculate the corresponding objective function value and constraint satisfaction. The individuals in the population are then non-dominated and sorted. Iteration stops when the number of iterations reaches the preset maximum, or when the change in the objective function value of the optimal individual in the population is less than a preset convergence threshold. The non-dominated solution of the current population is then output, forming the Pareto front solution set.

[0139] The solution with the fewest nozzle state switching times is selected from the Pareto front solution set as the nozzle control parameter matrix. This matrix is ​​then converted into control level signals and sent to the steam regulating valve group. Each solution in the Pareto front solution set is a non-dominated solution that satisfies the constraints, corresponding to different optimization objective equilibrium results. Selecting the solution with the fewest nozzle state switching times reduces the operating frequency of the steam regulating valve, decreases valve mechanical wear, and extends equipment life. The selected solution includes the flow distribution timing and duty cycle parameters of all steam nozzles. These parameters are arranged according to the steam regulating valve address and control timing to generate the nozzle control parameter matrix. The values ​​in the nozzle control parameter matrix are converted into control level signals recognizable by the steam regulating valve. The amplitude and pulse width of the control level signal correspond to the valve opening degree and opening duration. The control level signal is sent to the steam regulating valve group of the physical cooking equipment via fieldbus. The steam regulating valve group then executes the corresponding valve opening and closing and opening degree adjustment actions according to the control signal, realizing parameter optimization and closed-loop control of the zongzi leaf cooking process.

[0140] In this embodiment, in order to clarify the physical meaning of the constraints and corresponding parameters of the optimization model, a table of constraints and parameters of the optimization model as shown in Table 4 is constructed.

[0141] Table 4. Constraints and Parameter Definitions for the Optimization Model

[0142] Constraint Types Parameter name Physical meaning of parameters Parameter value range Constraint Decision Rules Mechanical property constraints Minimum time interval for valve status switching The minimum time interval between two consecutive opening / closing actions of a single steam nozzle ≥Preset mechanical response delay time If the value is less than the threshold, it is determined to be an infeasible solution. Flow boundary constraints Maximum flow rate per nozzle Maximum permissible flow rate of a single steam nozzle 0 to rated maximum flow If the solution is outside the specified range, it is considered an infeasible solution. Temperature constraint Highest surface temperature of zongzi leaves Maximum temperature at any point on the surface of the bamboo leaf ≤ Upper limit of target cooking temperature If the solution exceeds the upper limit, it is considered an infeasible solution. Temperature constraint Minimum temperature on the surface of zongzi leaves Minimum temperature at any point on the surface of a zongzi leaf ≥Target cooking temperature lower limit If the solution is below the lower limit, it is considered an infeasible solution. Flow field constraints Maximum range of dead zone in flow field The spatial volume ratio corresponding to the dead zone of the flow field ≤Preset maximum allowed percentage If the proportion exceeds the limit, it is determined to be an infeasible solution.

[0143] The constraints in Table 4 cover four dimensions: valve mechanical characteristics, flow boundary, cooking temperature requirements, and flow field characteristics. This ensures that the optimized parameters not only achieve uniform temperature distribution on the surface of the bamboo leaves but also meet the mechanical operating limitations of the equipment and the temperature requirements of the cooking process, avoiding the generation of unexecutable parameter solutions. Each constraint has a clearly defined judgment rule. During the optimization iteration process, the constraints are checked for each individual in the population. Individuals that do not meet the constraints are judged as infeasible solutions and do not enter the subsequent non-dominated sorting process, ensuring that all solutions in the Pareto front solution set obtained through optimization are executable and feasible solutions.

[0144] This embodiment achieves high-precision projection of discrete sensor data onto grid reference coordinates through a radial basis function interpolation algorithm based on the golden section method to optimize shape parameters. This reduces the sum of squared interpolation residuals and improves the accuracy of spatial difference calculation. By constructing a comprehensive transfer weight using the curve integral path length along the steam streamline and the steam density distribution, the allocation process of the heat flux density correction coefficient closely matches the actual physical path of heat transfer, improving the accuracy of dynamic boundary compensation. A non-dominated sorting genetic algorithm with an elite retention strategy is used to solve the multi-objective optimization model. Under the premise of satisfying the mechanical characteristics constraints of the equipment, global optimization of the steam nozzle flow distribution timing and duty cycle is achieved, eliminating local temperature distortion corresponding to dead zones in the flow field and realizing uniform temperature control during the steaming process of zongzi leaves.

Claims

1. A method for optimizing and simulating the steaming process parameters of zongzi leaves based on digital twins, characterized in that, include: By deploying an array of temperature and pressure sensors on the inner wall of the physical cooking equipment and around the steam nozzles, multi-dimensional temperature and pressure time-series data in the physical space are collected in real time. A three-dimensional fluid twin containing the internal geometry of the equipment cavity and the topology of the steam nozzle is constructed. Using the multidimensional temperature and pressure time series data as the initial boundary conditions, the Navier-Stokes equations are solved to obtain the steam velocity vector field and vortex distribution characteristics in the virtual flow field. Spatial alignment and differential calculation are performed between the measured temperature and pressure distribution field in physical space and the solution results of the virtual flow field to extract the flow field vortex attenuation deviation. The flow field vortex attenuation deviation is converted into a heat flux density correction coefficient, and the heat flux density correction coefficient is injected as a dynamic boundary compensation term into the zongzi leaf thermodynamic twin model located inside the equipment geometric model. In the twin space, the local temperature distortion corresponding to the dead zone of the flow field is eliminated as a constraint condition. The flow distribution timing and opening duty cycle of each steam nozzle are iteratively optimized, and the optimized nozzle control parameter matrix is ​​sent to the steam regulating valve group of the physical equipment.

2. The method for optimizing and simulating the steaming process parameters of zongzi leaves based on digital twins according to claim 1, characterized in that, The method of collecting multi-dimensional temperature and pressure time-series data in the physical space in real time by deploying an array of temperature and pressure sensors on the inner wall of the physical cooking equipment and around the steam nozzle includes: deploying temperature sensors on the inner wall of the physical cooking equipment according to the intersection nodes of a three-dimensional grid, and deploying pressure sensors at equal intervals in the circumferential direction of the steam nozzle outlet section. A unified high-frequency clock synchronization signal is configured for the temperature sensor and the pressure sensor. The rising edge of the high-frequency clock synchronization signal is used as the trigger reference to perform synchronous analog-to-digital conversion on the temperature analog signal output by the temperature sensor and the pressure analog signal output by the pressure sensor, thereby generating multi-dimensional temperature and pressure time series data carrying an absolute timestamp. The high-frequency glitches caused by steam flow pulsation in the multidimensional temperature and pressure time series data are removed by a finite impulse response filter, and the multidimensional temperature and pressure time series data is output after smoothing.

3. The method for optimizing and simulating the steaming process parameters of zongzi leaves based on digital twins according to claim 2, characterized in that, The construction of a three-dimensional fluid twin including the internal geometry of the equipment cavity and the topology of the steam nozzle, using the multidimensional temperature and pressure time series data as the initial boundary conditions, and solving the Navier-Stokes equations to obtain the steam velocity vector field and vortex distribution characteristics in the virtual flow field, includes: using an unstructured hexahedral mesh to mesh the internal geometry of the equipment cavity and the topology of the steam nozzle, and performing local mesh refinement in the interface area between the steam nozzle outlet region and the inner wall. The pressure data in the smoothed multidimensional temperature and pressure time series data is mapped to the inlet dynamic pressure boundary condition of the three-dimensional fluid twin, and the temperature data is mapped to the inlet enthalpy boundary condition of the energy equation. The filtered Navier-Stokes equations are solved transiently using a large eddy simulation turbulence model. The viscous dissipation rate of steam vortices is calculated using a subgrid-scale model. The output includes the steam velocity vector field and vortex distribution characteristics, which contain the correspondence between three-dimensional spatial coordinates.

4. The method for optimizing and simulating the steaming process parameters of zongzi leaves based on digital twins according to claim 3, characterized in that, The step of spatially aligning and differentially calculating the measured temperature and pressure distribution field in physical space with the solution results of the virtual flow field, and extracting the flow field vortex attenuation deviation, includes: using the unstructured hexahedral mesh nodes in the three-dimensional fluid twin as the reference coordinates, and using the radial basis function interpolation algorithm to project the measured temperature and pressure distribution field in physical space from the sensor physical coordinates to the reference coordinates, thereby generating the interpolated measured physical field; The interpolated measured physical field is subtracted node by node from the flow field parameters of the corresponding grid nodes in the virtual flow field solution result to generate a spatial difference residual matrix. Extract the set of spatial grid nodes from the spatial difference residual matrix where the velocity gradient change rate is lower than the preset turbulence activation threshold and the temperature change rate is higher than the preset thermal accumulation threshold, and define the mean of the difference residuals corresponding to the set of spatial grid nodes as the flow field vortex attenuation deviation.

5. The method for optimizing and simulating the steaming process parameters of zongzi leaves based on digital twins according to claim 4, characterized in that, The step of converting the flow field vortex attenuation deviation into a heat flux density correction coefficient and injecting the heat flux density correction coefficient as a dynamic boundary compensation term into the zongzi leaf thermodynamic twin model located inside the equipment geometric model includes: establishing a mapping function between the flow field vortex attenuation deviation and the attenuation ratio of the steam convection heat transfer coefficient; substituting the flow field vortex attenuation deviation corresponding to the set of spatial grid nodes into the mapping function; and calculating the heat flux density correction coefficient corresponding to each spatial grid node. The fluid boundary layer mesh of the outer surface of the wrapping leaf is extracted inside the device geometric model, and the heat flux density correction coefficients corresponding to each spatial mesh node are distributed to the corresponding nodes of the fluid boundary layer mesh according to the spatial distance weight. The fluid boundary layer grid node data carrying the heat flux density correction coefficient is used as the third type of boundary condition and superimposed on the surface energy conservation equation of the thermodynamic twin model of the zongzi leaf.

6. The method for optimizing and simulating the steaming process parameters of zongzi leaves based on digital twins according to claim 5, characterized in that, The process of eliminating the local temperature distortion corresponding to the dead zone of the flow field in the twin space as a constraint condition, iteratively optimizing the flow distribution timing and opening duty cycle of each steam nozzle, and sending the optimized nozzle control parameter matrix to the steam regulating valve group of the physical equipment includes: constructing an optimization model with the objective function of minimizing the difference between the maximum local temperature difference and the average local temperature difference on the surface of the zongzi thermodynamic twin model, and with the constraint condition that the time interval between two adjacent opening states of a single steam nozzle is greater than the preset mechanical response delay time; A non-dominated sorting genetic algorithm with an elite retention strategy is used to solve the Pareto front solution set by performing crossover and mutation operations on the initial flow distribution time sequence and initial duty cycle of each steam nozzle as individuals in the population. The solution with the fewest nozzle state switching times is selected from the Pareto front solution set as the nozzle control parameter matrix, and the nozzle control parameter matrix is ​​converted into a control level signal and sent to the steam regulating valve group.

7. The method for optimizing and simulating the steaming process parameters of zongzi leaves based on digital twins according to claim 2, characterized in that, The provision of a unified high-frequency clock synchronization signal for the temperature sensor and the pressure sensor includes: deploying a field-programmable gate array in the control cabinet of the physical cooking equipment, and generating a high-frequency clock synchronization signal through the field-programmable gate array; The hardware trace delay time of the high-frequency clock synchronization signal transmitted from the field programmable gate array to each temperature sensor and each pressure sensor is measured, and the hardware trace delay time is written into the data register inside each sensor. When the rising edge of the high-frequency clock synchronization signal arrives, each sensor reads the sensor measurement value at the current moment and compensates the hardware trace delay time into the count value of the absolute timestamp to generate multidimensional temperature and pressure timing data after transmission delay correction.

8. The method for optimizing and simulating the steaming process parameters of zongzi leaves based on digital twins according to claim 3, characterized in that, The method of using a large eddy simulation turbulence model to solve the filtered Navier-Stokes equations transiently and calculating the viscous dissipation rate of steam vortices through a subgrid-scale model includes: using a dynamic Smagorinsky-Lilly model as the subgrid-scale model, and calculating the velocity-strain rate tensor after dual filtering by grid-scale filtering and test filtering by setting a test filter in the steam nozzle outlet area of ​​the physical cooking equipment. Based on the aforementioned velocity-strain rate tensor, the model coefficients are dynamically solved using the German fixed-point iterative method, and the solved model coefficients are confined to a non-negative interval that does not cause numerical divergence. Substituting the model coefficients within the defined range into the eddy viscosity expression, the subgrid eddy viscosity coefficient, which dynamically varies with spatial and temporal coordinates, is calculated. The viscous dissipation rate of the steam vortex within the three-dimensional fluid twin is then calculated using the subgrid eddy viscosity coefficient.

9. The method for optimizing and simulating the steaming process parameters of zongzi leaves based on digital twins according to claim 4, characterized in that, The step of projecting the measured temperature and pressure distribution field of the physical space from the sensor physical coordinates to the reference coordinates using the radial basis function interpolation algorithm includes: selecting a multivariate quadratic function as the basis function of the radial basis function interpolation algorithm; Calculate the shortest spatial distance between the physical coordinates of each sensor, and use the preset scaling factor of the shortest spatial distance as the initial shape parameter of the basis function; With the goal of minimizing the sum of squared interpolation residuals of the interpolated measured physical field at non-sensor physical coordinate positions, the golden section method is used to perform a one-dimensional search and iteration on the initial shape parameters in a preset shape parameter space, and the shape parameter corresponding to the convergence of the sum of squared interpolation residuals is determined as the target shape parameter. An interpolation matrix is ​​constructed using the target shape parameters, and the interpolation matrix is ​​solved to obtain the interpolated measured physical field data of each node on the reference coordinates.

10. The method for optimizing and simulating the steaming process parameters of zongzi leaves based on digital twins according to claim 5, characterized in that, The step of distributing the heat flux density correction coefficients corresponding to each spatial grid node to the corresponding nodes of the fluid boundary layer grid according to the spatial distance weight includes: extracting vapor streamlines from each spatial grid node to the corresponding node of the fluid boundary layer grid in the three-dimensional fluid twin; Calculate the curve integral path length along the steam streamline, and use the curve integral path length as the streamline distance weight to replace the straight-line spatial distance weight; Obtain steam density distribution data within the steam streamline path area, and multiply the steam density distribution data with the streamline distance weight to generate a comprehensive transfer weight; The heat flux density correction coefficient corresponding to the spatial grid node is multiplied by the comprehensive transfer weight and accumulated to the initial heat flux density value of the corresponding node of the fluid boundary layer grid to complete the injection of the dynamic boundary compensation term.