Ink heat transfer method and device based on agent model, equipment and storage medium

Abstract

The present application relates to inkjet printing technology field, especially ink heat transfer method, device and equipment based on agent model and storage medium, the method is through the standardized feature mapping operator converts engineering parameter into the computable high-dimensional feature vector, so that non-professional personnel can complete parameter optimization without mastering complex professional knowledge; The built-in agent model of the system replaces the traditional high time-consuming real-time numerical solution, greatly compresses the simulation process, and improves the process iteration efficiency; At the same time, by integrating scalar performance index, physical field data matrix and prediction uncertainty result, the flow resistance and ink rheological safety are considered while pursuing heat transfer enhancement, which avoids the process risk caused by single index optimization; The final output of the visual report and the standardized control instruction realize the connection from theoretical simulation to industrial field, provide a technical path for viscosity regulation in OLED inkjet printing, and realize the effective optimization of ink pulsating flow preheating parameters.

Description

Technical Field

[0001] This invention relates to the field of inkjet printing technology, and in particular to an ink heat transfer method, apparatus, device, and storage medium based on a proxy model. Background Technology

[0002] In OLED inkjet printing, ink flows from a storage container through an ink supply line to the printhead and is finally ejected onto the substrate. The degree of matching between its physical state and process parameters directly determines the printing quality. Appropriately reducing the ink viscosity can ensure high-speed and stable ejection, while increasing it can easily cause problems such as nozzle clogging and decreased droplet size control accuracy. Therefore, ink usually needs to be preheated before ejection to adjust its viscosity.

[0003] However, existing preheating methods mainly employ static flow heating methods such as heating the outer wall of the ink storage unit and heating belts along the ink supply pipeline. The heat exchange boundary layer lacks an active intervention mechanism, resulting in low preheating efficiency and difficulty in adapting to the complex rheological characteristics of different inks.

[0004] Pulsating flow, as an active heat transfer enhancement technology, introduces a periodic velocity field into the flow channel, which can effectively disrupt the heat transfer boundary layer and enhance fluid mixing, thereby significantly improving convective heat transfer efficiency. Existing research shows that pulsating flow is highly effective in enhancing the heat transfer of Newtonian fluids. For example, for Newtonian fluids in helical tubes, at pulsating frequencies of 2–10 Hz, the pressure drop increase is only 3%–7%, while the convective heat transfer enhancement can reach up to 39%. For shear-thinning non-Newtonian fluids, the heat transfer enhancement effect of pulsating flow can also reach 30%–35%, and there is a significant nonlinear coupling relationship between pulsation and shear-thinning effects. However, current research and application of pulsating flow-enhanced heat transfer technology mainly focus on the cooling scenarios of Newtonian fluids. For non-Newtonian fluids such as OLED inkjet printing inks, which have complex rheological properties such as shear-thinning and viscoelasticity, the mechanism of pulsating flow-enhanced heat transfer has not yet been systematically studied.

[0005] Furthermore, existing OLED inkjet printing temperature control solutions primarily focus on temperature regulation of the ink cartridge and printhead, neglecting the preheating and parameter optimization of the ink supply pipeline's pulsed flow. Existing technologies involving pulsed flow-enhanced heat transfer mainly target cooling scenarios such as electronic device heat dissipation, failing to consider the impact of the special rheological properties of non-Newtonian fluids on the pulsed flow enhancement effect, and also failing to introduce this technology into the preheating stage of the OLED ink supply pipeline. In addition, existing technologies exhibit complex coupling relationships between the geometric parameters of the preheating pipeline and control parameters such as pulsed frequency, amplitude, and waveform. Parameter selection heavily relies on manual experience, lacking intelligent optimization methods. Simultaneously, the selection of geometric structures, setting of physical property parameters, and simulation calculations lack an integrated encapsulation and calling system, resulting in low model reuse efficiency and making it difficult for non-professionals to participate in design iterations.

[0006] It is evident that existing technologies still need improvement and enhancement. Summary of the Invention

[0007] To overcome the shortcomings of existing technologies, the present invention aims to provide an ink heat transfer method based on a proxy model, which provides a technical path for viscosity control in OLED inkjet printing and achieves effective optimization of ink pulsating flow preheating parameters.

[0008] The first aspect of this invention provides an ink heat transfer method based on a surrogate model, comprising: pre-constructing a heat transfer optimization system, the heat transfer optimization system including a data support layer, a model resource layer, a parameter calling layer, and an optimization decision layer interconnected via interfaces; obtaining engineering parameters through the parameter calling layer, and mapping the engineering parameters into high-dimensional feature vectors using a preset feature mapping operator; matching a physical model from the model resource layer based on the high-dimensional feature vectors, and constructing a pulsating flow underlying control equation based on the matched physical model; inputting the high-dimensional feature vectors into a pre-trained surrogate model to obtain scalar performance indicators and modal coefficient vectors, and based on... The modal coefficient vector is used to reconstruct the physical field data matrix, and the physical consistency of the physical field data matrix is ​​checked based on the underlying control equation of the pulsating flow to obtain the prediction uncertainty result. The scalar performance index, the physical field data matrix, and the prediction uncertainty result are integrated to obtain the performance prediction result. A fitness function is constructed based on the scalar performance index, and a multi-objective optimization algorithm is used to perform iterative optimization by combining the fitness function and the high-dimensional feature vector to obtain the optimal waveform control parameters. The optimal waveform control parameters and the performance prediction result are integrated to generate a visualization report, and the pulsating flow preheating process of the ink supply pipeline is controlled based on the visualization report.

[0009] Optionally, in a first implementation of the first aspect of the present invention, the pre-built heat transfer optimization system includes: constructing a data support layer, the data support layer including a sample database, a surrogate model pre-trained based on the Kriging algorithm, and an intrinsic orthogonal decomposition mode library; constructing a model resource layer, the model resource layer including a parameterized flow channel geometry model library, a hierarchical rheological model library, and a pulsating flow field master control equation library; constructing a parameter calling layer, the parameter calling layer including a standardized engineering parameter input interface, a data validity verification module, and a feature mapping operator execution module connected in sequence; and constructing an optimization decision layer, the optimization decision layer including a fitness function construction module and a multi-objective optimization algorithm execution module connected in sequence.

[0010] Optionally, in a second implementation of the first aspect of the present invention, the step of obtaining engineering parameters through the parameter calling layer and mapping the engineering parameters into high-dimensional feature vectors using a preset feature mapping operator includes: obtaining engineering parameters through the parameter calling layer, the engineering parameters including flow channel structure parameters, fluid property parameters, and thermodynamic boundary condition parameters; performing data validity verification on the engineering parameters, determining whether each type of parameter is within a preset physically reasonable value range; if the engineering parameters pass the data validity verification, then calling the preset feature mapping operator to uniformly map each type of parameter into a high-dimensional feature vector, the high-dimensional feature vector including geometric feature sub-vectors, rheological feature sub-vectors, and operating condition feature sub-vectors.

[0011] Optionally, in a third implementation of the first aspect of the present invention, the step of matching a physical model from the model resource layer based on the high-dimensional feature vector and constructing the underlying control equations for the pulsating flow based on the matched physical model includes: matching a corresponding flow channel topology model from the parametric flow channel geometry model library of the model resource layer according to the geometric feature sub-vectors in the high-dimensional feature vector, and performing real-time parametric fine-tuning of the matched flow channel topology model according to the geometric feature sub-vectors to obtain a fine-tuned flow channel topology model; matching a corresponding fluid constitutive model from the hierarchical rheological model library of the model resource layer according to the rheological feature sub-vectors in the high-dimensional feature vector; and calling the corresponding continuity equation, unsteady momentum equation, and energy equation considering viscous dissipation from the pulsating flow field master control equation library of the model resource layer according to the fine-tuned flow channel topology model and the matched fluid constitutive model to construct the underlying control equations for the pulsating flow.

[0012] Optionally, in a fourth implementation of the first aspect of the present invention, the step of inputting the high-dimensional feature vector into a pre-trained surrogate model to obtain scalar performance indices and modal coefficient vectors, reconstructing a physical field data matrix based on the modal coefficient vectors, and performing physical consistency verification on the physical field data matrix based on the underlying control equations of the pulsating flow to obtain prediction uncertainty results includes: inputting the high-dimensional feature vector into a surrogate model pre-trained based on the Kriging algorithm to obtain scalar performance indices and POD modal coefficient vectors corresponding to the operating condition requirements, wherein the scalar performance indices include the mean Nusselt average (NNUS) modal coefficient vector. The data includes the number of parameters, average friction factor, and comprehensive performance evaluation coefficients; pre-stored reference modes of each order are retrieved from the intrinsic orthogonal decomposition mode library of the data support layer, and the POD mode coefficient vector is linearly weighted and combined with the reference modes of the corresponding order to reconstruct the physical field data matrix, which includes the quasi-steady-state velocity field, pressure field, and temperature field; based on the underlying control equation of the pulsating flow, mass conservation and energy conservation checks are performed on the physical field data matrix, and the root mean square error is calculated; the check results and root mean square error are integrated to obtain the prediction uncertainty result.

[0013] Optionally, in the fifth implementation of the first aspect of the present invention, the step of constructing a fitness function based on the scalar performance index, employing a multi-objective optimization algorithm, and performing iterative optimization by combining the fitness function and the high-dimensional feature vector to obtain the optimal waveform control parameters includes: constructing a fitness function based on the heat transfer enhancement index and the flow loss index according to the scalar performance index, and loading preset constraints, the constraints including physical limit constraints of the pulsation generator and maximum allowable shear stress constraints of the ink; using the NSGA-II multi-objective optimization algorithm to initialize a decision vector population based on the high-dimensional feature vector, the decision vector population including multiple individuals, each individual corresponding to a set of waveform control parameters to be optimized; performing multi-objective optimization iteration on the decision vector population based on the fitness function and the preset constraints; when the preset maximum number of iterations or convergence accuracy threshold is reached, outputting the optimal waveform control parameters based on the operating condition type in the high-dimensional feature vector, the optimal waveform control parameters being a globally optimal single-objective decision vector or a Pareto non-dominated solution set.

[0014] Optionally, in the sixth implementation of the first aspect of the present invention, the step of integrating the optimal waveform control parameters and performance prediction results to generate a visualization report, and controlling the pulsating flow preheating process of the ink supply pipeline based on the visualization report, includes: integrating the optimal waveform control parameters and performance prediction results to generate a visualization report, the visualization report including a core control parameter table, a quantitative performance index radar chart, a pulsating period characteristic phase temperature distribution cloud map, and a flow field vector trajectory map; extracting the optimal waveform control parameters from the visualization report and converting them into standardized control commands recognizable by the pulsating generator of the ink supply pipeline, the optimal waveform control parameters including the optimal pulsating frequency, pulsating amplitude, and reference operating pressure parameters; and issuing the standardized control commands to the pulsating execution unit of the ink supply pipeline to adjust the operating parameters of the pulsating execution unit, thereby realizing the control of the pulsating flow preheating process of the OLED inkjet printing ink.

[0015] A second aspect of the present invention provides an ink heat transfer device based on a surrogate model, comprising: a construction module for pre-constructing a heat transfer optimization system, the heat transfer optimization system including a data support layer, a model resource layer, a parameter calling layer, and an optimization decision layer interconnected via interfaces; a mapping module for obtaining engineering parameters through the parameter calling layer and mapping the engineering parameters into high-dimensional feature vectors using a preset feature mapping operator; a matching module for matching a physical model from the model resource layer based on the high-dimensional feature vectors and constructing a pulsating flow underlying control equation based on the matched physical model; and a verification module for inputting the high-dimensional feature vectors into a pre-trained surrogate model to obtain scalar performance indicators and modal coefficient vectors. The system reconstructs the physical field data matrix based on the modal coefficient vector and performs a physical consistency check on the physical field data matrix based on the underlying control equation of the pulsating flow to obtain the prediction uncertainty result; the integration module is used to integrate the scalar performance index, the physical field data matrix and the prediction uncertainty result to obtain the performance prediction result; the optimization module is used to construct a fitness function based on the scalar performance index, and use a multi-objective optimization algorithm to perform iterative optimization by combining the fitness function and the high-dimensional feature vector to obtain the optimal waveform control parameters; the control module is used to integrate the optimal waveform control parameters and the performance prediction result, generate a visualization report, and control the pulsating flow preheating process of the ink supply pipeline based on the visualization report.

[0016] A third aspect of the present invention provides an ink heat transfer device based on a proxy model, the ink heat transfer device based on a proxy model comprising: a memory and at least one processor, the memory storing instructions; the at least one processor calling the instructions in the memory to cause the ink heat transfer device based on a proxy model to perform the various steps of the ink heat transfer method based on a proxy model as described above.

[0017] A fourth aspect of the present invention provides a computer-readable storage medium storing instructions that, when executed by a processor, implement the steps of the ink heat transfer method based on the proxy model described in any of the preceding claims.

[0018] This invention discloses an ink heat transfer method based on a surrogate model. By using standardized feature mapping operators, engineering parameters are transformed into computable high-dimensional feature vectors, enabling non-professionals to optimize parameters without complex technical knowledge. The system's built-in surrogate model replaces traditional time-consuming real-time numerical solutions, significantly compressing the simulation process and improving process iteration efficiency. Simultaneously, by integrating scalar performance indicators, physical field data matrices, and prediction uncertainty results, it balances flow resistance and ink rheological safety while pursuing enhanced heat transfer, avoiding process risks caused by optimizing a single indicator. The final output of a visual report and standardized control commands bridges the gap between theoretical simulation and industrial application, providing a technical path for viscosity control in OLED inkjet printing and achieving effective optimization of ink pulsating flow preheating parameters. Attached Figure Description

[0019] Figure 1 A flowchart illustrating the ink heat transfer method based on a proxy model provided in an embodiment of the present invention; Figure 2 A schematic diagram of the structure of an ink heat transfer device based on a proxy model provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the structure of an ink heat transfer device based on a proxy model, provided in an embodiment of the present invention. Detailed Implementation

[0020] This invention provides an ink heat transfer method, apparatus, device, and storage medium based on a proxy model. In this invention, the terms "first," "second," "third," "fourth," etc. (if present)," in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having," and any variations thereof, are intended to cover a non-exclusive inclusion; for example, a process, method, system, product, or device that includes a series of steps or units is not necessarily limited to those explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0021] This application discloses an ink heat transfer method based on a surrogate model. For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1One embodiment of the ink heat transfer method based on the surrogate model in this invention includes: 101. A pre-constructed heat transfer optimization system, wherein the heat transfer optimization system includes a data support layer, a model resource layer, a parameter calling layer, and an optimization decision layer that are interconnected through interfaces; In this embodiment, the application programming interface (API) communication protocols for each layer are first defined at the software architecture level. The parameter calling layer sends geometric and physical property parameter configuration instructions to the model resource layer through the forward interface. The model resource layer requests sample data and loads the physical model from the data support layer through the callback interface. The data support layer provides proxy model inference services to the optimization decision layer through the prediction service interface. The optimization decision layer feeds back the optimization progress and constraint status to the parameter calling layer through the control bus. By constructing a heat transfer optimization system, the decoupling of data flow and control flow is achieved. At the same time, each functional layer can be upgraded and maintained independently. For example, the underlying rheological model algorithm can be replaced without changing the upper-layer optimization logic, which significantly reduces the complexity of system iteration and improves the adaptability and long-term scalability for different process requirements of OLED production lines.

[0022] 102. Obtain engineering parameters through the parameter calling layer, and map the engineering parameters into high-dimensional feature vectors using a preset feature mapping operator; In this embodiment, the parameter calling layer receives heterogeneous engineering parameters such as the inner diameter of the flow channel, the bending diameter, the zero-shear viscosity of the ink, the flow index, the inlet velocity, and the wall heat flux density input by the user through a graphical user interface or a standardized data interface. The feature mapping operator has a built-in maximum and minimum value normalization algorithm, which can linearly map physical quantities of different dimensions to the interval of 0-1. Through the mapping process, the negative impact of the huge difference in numerical scale of different physical quantities on the convergence of the optimization algorithm is eliminated. The unstructured engineering language is transformed into structured mathematical features, providing a standardized data base for the rapid retrieval and interpolation calculation of the subsequent proxy model, which greatly improves the efficiency and stability of the system in handling high-dimensional complex working conditions.

[0023] 103. Based on the high-dimensional feature vector, match the physical model from the model resource layer, and construct the underlying control equations of the pulsating flow based on the matched physical model; In this embodiment, the construction of the underlying control equation for pulsating flow solves the problems of easy errors in manually setting material properties and inability to adapt to non-Newtonian fluid characteristics in the prior art. In particular, it accurately models the viscosity dynamic changes of OLED ink during shear thinning, ensuring the authenticity and accuracy of the simulation physical model.

[0024] 104. Input the high-dimensional feature vector into the pre-trained surrogate model to obtain scalar performance index and modal coefficient vector. Reconstruct the physical field data matrix based on the modal coefficient vector, and perform physical consistency verification on the physical field data matrix based on the underlying control equation of the pulsating flow to obtain the prediction uncertainty result. In this embodiment, the pre-trained Kriging surrogate model receives high-dimensional feature vectors and outputs scalar indices such as the average Nusselt number, average friction factor, and comprehensive performance evaluation coefficients. It also outputs the coefficient vectors of each order of intrinsic orthogonal decomposition modes. Then, it calls the pre-stored baseline mode matrix and reconstructs the physical field data matrix through a linear weighted combination formula. Based on the underlying control equation of pulsating flow, it performs physical consistency verification, effectively eliminating non-physical solutions caused by extrapolation in the surrogate model. While maintaining prediction speed, it ensures that the prediction results follow the physical conservation laws, balancing computational efficiency and result accuracy.

[0025] 105. Integrate the scalar performance index, physical field data matrix, and prediction uncertainty results to obtain the performance prediction results; In this embodiment, the system encapsulates the discrete scalar data, three-dimensional spatial field distribution data, and error quantization data generated in step 104 into a complete data packet according to a predefined JSON data structure. The data packet includes a copy of the input parameters, output response indicators, full-field cloud map data, and confidence assessment values. Through data integration, the single numerical output is transformed into a multi-dimensional holographic data output, which not only provides conclusions on the performance, but also provides the physical field basis for the performance and the reliability of the results. This provides rich data dimensions for the subsequent optimization decision layer, supports the algorithm to make trade-offs among different physical mechanisms, and enhances the scientific and comprehensive nature of the decision.

[0026] 106. Construct a fitness function based on the scalar performance index, and use a multi-objective optimization algorithm to perform iterative optimization by combining the fitness function and high-dimensional feature vectors to obtain the optimal waveform control parameters; In this embodiment, multi-objective optimization solves the problem of difficulty in manually finding the optimal solution for multi-parameter strongly coupled nonlinear systems, automatically discovering the optimal balance between heat transfer gain and pump power consumption, and significantly improving the energy efficiency ratio of the OLED ink preheating process.

[0027] 107. Integrate the optimal waveform control parameters and performance prediction results to generate a visualization report, and control the pulsating flow preheating process of the ink supply pipeline based on the visualization report; In this embodiment, by generating a visual report to guide the pulsed flow preheating process, the data link between the virtual simulation environment and the physical actuator is established, realizing a closed loop from algorithm decision-making to physical control. This eliminates the errors that may be caused by manual transcription of parameters, and through precise pulsed flow control, online real-time optimization and adjustment of ink viscosity are achieved, improving the yield and process consistency of OLED inkjet printing.

[0028] The ink heat transfer method based on a surrogate model disclosed in this application transforms engineering parameters into computable high-dimensional feature vectors through standardized feature mapping operators, enabling non-professionals to perform parameter optimization without needing complex professional knowledge. The system's built-in surrogate model replaces the traditional time-consuming real-time numerical solution, significantly compressing the simulation process and improving the efficiency of process iteration. At the same time, by integrating scalar performance indicators, physical field data matrices, and prediction uncertainty results, it balances flow resistance and ink rheological safety while pursuing enhanced heat transfer, avoiding process risks caused by optimizing a single indicator. The final output of the visualization report and standardized control commands realizes the connection between theoretical simulation and industrial field, providing a technical path for viscosity control in OLED inkjet printing and achieving effective optimization of ink pulsating flow preheating parameters.

[0029] Furthermore, in this embodiment of the invention, the pre-built heat transfer optimization system includes: 201. Construct a data support layer, which includes a sample database, a surrogate model pre-trained based on the Kriging algorithm, and an intrinsic orthogonal decomposition mode library; In this embodiment, a sample database is first constructed in the storage medium. A uniformly distributed sample point is generated in the multidimensional parameter space using the Latin hypercube sampling method. The sample point refers to a discrete parameter combination generated in the multidimensional parameter space through Latin hypercube sampling. For each sample point, a full-physics transient simulation is performed using computational fluid dynamics software, recording the input parameters, output scalar performance (such as the Nusselt number), and a full-field snapshot (velocity, temperature, and pressure matrices) corresponding to the sample point. Subsequently, a Kriging Gaussian process regression model is trained based on the sample data to establish a mapping relationship between input parameters and output response. Intrinsic orthogonal decomposition is performed on the full-field snapshot data to extract the top few dominant modes and store them in a mode library. The implementation process of intrinsic orthogonal decomposition is as follows: N full-field snapshots in the sample database are concatenated column-wise to form a data matrix, the covariance matrix is ​​calculated, the calculated covariance matrix is ​​decomposed into eigenvalues, and the eigenvectors corresponding to the top k largest eigenvalues ​​are extracted as intrinsic orthogonal decomposition modes. By constructing a data support layer, the combination of offline high-fidelity simulation and online fast prediction is realized, which solves the problem of huge computational load in the simulation of non-Newtonian fluid pulsating heat transfer, making real-time optimization possible on conventional industrial computers.

[0030] 202. Construct a model resource layer, which includes a parametric flow channel geometry model library, a hierarchical rheological model library, and a library of master equations for pulsating flow fields; In this embodiment, a geometric model library including straight tubes, spiral tubes, serpentine tubes, and corrugated tubes is constructed using parametric modeling scripts. The scripts support real-time updates of the geometry by modifying feature parameters. A hierarchical rheological model library encapsulates Newtonian fluid models, power-law models, Carreau models, and viscoelastic models with Arrhenius temperature correction. The pulsating flow field master equation library pre-sets the discretization format and solver configuration of unsteady continuity equations, momentum equations, and energy equations. By constructing a model resource layer, modular encapsulation of physical resources is achieved, enabling the system to automatically match the most suitable constitutive equations according to different ink characteristics. This solves the defect of existing technologies that cannot handle complex non-Newtonian fluid rheological characteristics, ensuring that preheating simulation can accurately reflect the actual physical behavior of OLED inks.

[0031] 203. Construct a parameter calling layer, which includes a standardized engineering parameter input interface, a data validity verification module, and a feature mapping operator execution module connected in sequence. In this embodiment, the parameter calling layer deploys a standardized data input interface to receive external input. A data validity verification module is connected in series behind the data input interface. This module has a built-in physical rule engine that sets verification rules based on physical principles, such as requiring the characteristic pipe diameter to be greater than zero, the flow index to be between 0.1 and 1, and the temperature to be non-negative. If an illegal parameter is detected, the process is immediately interrupted and an alarm is triggered. If any engineering parameter exceeds its limit, the verification fails and the user is prompted to correct it. After successful verification, the data flows into the feature mapping operator execution module for normalization processing, effectively intercepting the risk of computational crashes caused by absurd parameters and ensuring the robustness of the system operation.

[0032] 204. Construct an optimization decision layer, which includes a fitness function construction module and a multi-objective optimization algorithm execution module connected in sequence; In this embodiment, the fitness function construction module presets the allocation ratio of heat transfer enhancement weight and flow resistance weight, and sets constraint boundaries for shear stress and temperature rise; the multi-objective optimization algorithm execution module incorporates the selection, crossover, and mutation operators of the NSGA-II algorithm, and configures the maximum number of iterations and convergence accuracy threshold; the preset parameters are set based on engineering experience and algorithm convergence tests, typically with the population size set to 10 times the variable dimension, and the number of iterations set to ensure that the fitness function value changes by less than 1 / 2 for 20 consecutive generations. The threshold is set; by constructing an optimization decision layer, the system is given decision-making capabilities. Through modular fitness definition, process engineers can flexibly adjust the optimization objectives according to actual needs, realizing dynamic adjustment and customized output of optimization strategies.

[0033] Furthermore, in this embodiment of the invention, the step of obtaining engineering parameters through the parameter calling layer and mapping the engineering parameters into high-dimensional feature vectors using a preset feature mapping operator includes: 301. Obtain engineering parameters through the parameter calling layer, the engineering parameters including flow channel structure parameters, fluid property parameters and thermodynamic boundary condition parameters; In this embodiment, the system specifically collects geometric parameters such as characteristic pipe diameter, characteristic bend diameter, and characteristic spacing of the flow channel through an interactive interface or data input interface; collects fluid property parameters such as zero shear viscosity, flow index, relaxation time, and flow activation energy of the ink; and collects thermodynamic boundary condition parameters such as inlet pressure, reference flow rate, initial temperature, wall heat flux density, or wall temperature. By collecting engineering parameters, the system clarifies the complete data scope of engineering requirements, covers key variables affecting pulsating flow heat transfer, and ensures the comprehensiveness of the input information.

[0034] 302. Perform data validity verification on the engineering parameters, and determine whether each type of parameter is within the preset physical reasonable value range; In this embodiment, the data validity verification module sets verification rules based on physical common sense. It compares parameters one by one, such as determining whether the characteristic pipe diameter is greater than 0 and less than the flow channel length, whether the flow index is between 0.1 and 1 to meet the shear thinning characteristics, and whether the initial temperature is a positive number. If any parameter exceeds the limit, the system immediately throws an error code and blocks the process. Through physical logic filtering, illegal data is prevented from entering the calculation engine, avoiding solution divergence caused by input errors, and significantly improving the fault tolerance of the system.

[0035] 303. If the engineering parameters pass the data validity check, then the preset feature mapping operator is called to uniformly map all types of parameters into high-dimensional feature vectors. The high-dimensional feature vectors include geometric feature sub-vectors, rheological feature sub-vectors, and working condition feature sub-vectors. In this embodiment, for parameters that pass the verification, the parameter invocation layer has a built-in feature mapping operator. The feature mapping operator This is used to transform discrete or interval-type physical quantities (including geometric topology, fluid properties, and boundary conditions) into high-dimensional feature vectors that drive subsequent surrogate model retrieval and computation. Its expression is as follows:

[0036] Among them, superscript This represents the transpose operation of a vector. This mapping logic realizes the transformation from the physical parameter space to the mathematical feature space. In this mapping logic, the output feature vector is... It consists of decoupling sub-vectors in three dimensions: first, The set of input flow channel structure parameters and geometric feature vectors The system will input the flow channel topology type and the characteristic pipe diameter describing the cross-sectional dimensions, corresponding to the mathematical representation of the flow channel structure. Characteristic diameter describing the curvature of a spatial path and the characteristic spacing describing the periodic distribution of the structure. Equal flow channel structure parameters, through the characteristic mapping operator It is converted into a numerical vector of a unified dimension; through the standardized encapsulation of common geometric features, the system can achieve parameterized description and real-time mapping of various heat exchange structures such as spiral tubes and straight tubes without changing the underlying algorithm, thereby improving the system's adaptability to different ink supply pipeline structures. Secondly For the set of input fluid property parameters, rheological feature vectors The system takes the corresponding mathematical representation of the physical properties of ink as input, which is the flow index used to describe shear-thinning characteristics. Consistency coefficient and relaxation time And the flow activation energy used to describe the evolution of viscosity with temperature. The isophysical parameters are mapped to feature vectors that describe the shear thinning, viscoelasticity and temperature change characteristics of OLED inks. These feature vectors serve as the core basis for subsequent fluid constitutive model matching, ensuring that the simulation process can reflect the actual rheological behavior of inks with different components. at last, The input set of thermodynamic boundary condition parameters, and the characteristic vector of the working condition. The system will input the initial pressure based on the mathematical representation of the corresponding operational boundary conditions. Initial temperature thermal boundary Boundary conditions, such as pulsating feature parameters, are converted into standardized numerical vectors. According to process requirements, each parameter can be set to a fixed physical value or specified as the boundary of the optimization space that allows fluctuations. In addition, the system also supports the default parameter execution for some optimization variables, and the kernel automatically derives the optimal optimization range. Through a compatibility mode that depends on the specific working conditions, the connection from discrete boundary definition to standardized feature extraction is realized. Through the feature mapping operator The standardized encapsulation enables the homogenization of heterogeneous data, eliminates the dimensional differences between different parameters, and allows subsequent optimization algorithms to perform optimization within a unified mathematical space. This solves the problem of optimization algorithm oscillation or premature convergence caused by the mixing of multiple physical dimensions, and improves the convergence speed and stability of the optimization process.

[0037] Furthermore, in this embodiment of the invention, the step of matching a physical model from the model resource layer based on the high-dimensional feature vector, and constructing the underlying control equations of the pulsating flow based on the matched physical model, includes: 401. Based on the geometric feature sub-vectors in the high-dimensional feature vector, match the corresponding flow channel topology model from the parameterized flow channel geometry model library of the model resource layer, and perform real-time parameterized fine-tuning of the matched flow channel topology model based on the geometric feature sub-vectors to obtain the fine-tuned flow channel topology model. In this embodiment, the topology type identifier refers to the first element in the geometric feature sub-vector, which is an integer encoding. For example, 1 represents a straight pipe, 2 represents a spiral pipe, and 3 represents a serpentine pipe. Based on the topology type identifier, the system calls the corresponding parametric script from the parametric flow channel geometry model library and assigns the subsequent values ​​in the geometric feature sub-vector to the corresponding variables in the script. This drives the script to regenerate the three-dimensional geometric model and update the mesh node coordinates, realizing the automated generation and fine-tuning of the geometric model. This eliminates the need for tedious manual modeling operations, supports rapid variant design for specific process requirements, and improves the efficiency and flexibility of simulation model construction.

[0038] 402. Based on the rheological feature sub-vectors in the high-dimensional feature vector, match the corresponding fluid constitutive model from the hierarchical rheological model library of the model resource layer; In this embodiment, the system reads the flow index from the rheological feature vector. If the flow index is equal to 1, a Newtonian fluid constitutive model is automatically loaded. If the flow index is less than 1, a Carreau non-Newtonian constitutive model is loaded. The parameters in the rheological feature vector are then assigned to the model material property card. This ensures that the most accurate rheological description equation can be automatically selected for different types of OLED inks, solving the problem of not being able to automatically distinguish fluid properties in the prior art and ensuring the accuracy of the material property settings.

[0039] 403. Based on the fine-tuned flow channel topology model and the matched fluid constitutive model, call the corresponding continuity equation, unsteady momentum equation and energy equation considering viscous dissipation from the pulsating flow field master control equation library of the model resource layer to construct the underlying control equation of the pulsating flow. In this embodiment, based on the configured geometry and physical properties, a continuity equation is generated to ensure mass conservation, an unsteady momentum equation is generated to ensure momentum conservation, and an energy equation is generated and the viscous dissipation term is explicitly activated to calculate the internal frictional heat generation under high shear rates. The aforementioned three equations are all existing techniques in the field of computational fluid dynamics (common knowledge). Their combination constitutes a closed set of partial differential equations, namely the underlying governing equations of pulsating flow, which completes the fully automatic configuration of the physics solver. By activating the viscous dissipation term, the additional heat generated by severe shear in the pulsating flow is captured, significantly improving the consistency between the simulation results and the actual physical process.

[0040] Further, in this embodiment of the invention, the step of inputting the high-dimensional feature vector into a pre-trained surrogate model to obtain scalar performance indicators and modal coefficient vectors, reconstructing a physical field data matrix based on the modal coefficient vectors, and performing a physical consistency check on the physical field data matrix based on the underlying control equations of the pulsating flow to obtain prediction uncertainty results includes: 501. Input the high-dimensional feature vector into the surrogate model pre-trained based on the Kriging algorithm to obtain scalar performance indicators and POD modal coefficient vectors corresponding to the working condition requirements. The scalar performance indicators include the average Nusselt number, the average friction factor, and the comprehensive performance evaluation coefficient. In this embodiment, a high-dimensional feature vector is input into a pre-trained Kriging model. The model outputs the average Nusselt number under the current operating conditions to characterize the heat transfer intensity, the average friction factor to characterize the flow resistance, and the comprehensive performance evaluation coefficient to characterize the heat transfer efficiency per unit pump work, based on a Gaussian process regression algorithm. Simultaneously, it outputs the linear combination coefficients of each order of intrinsic orthogonal decomposition modes. The training process of the surrogate model aims to establish a mapping function from the input (high-dimensional feature vector) to the output (performance indicators and modal coefficients). The Kriging surrogate model assumes that the response value follows a Gaussian process and estimates the hyperparameters by maximizing the likelihood function, thereby achieving prediction for any input point. By pre-training the surrogate model, replacing time-consuming fluid solutions, millisecond-level performance prediction is achieved, making real-time scanning of massive parameter combinations possible. This changes the inefficient mode of traditional trial-and-error methods and significantly shortens the process development cycle.

[0041] 502. Recall the pre-stored reference modes of each order from the intrinsic orthogonal decomposition mode library of the data support layer, and perform linear weighted combination of the POD mode coefficient vector and the reference mode of the corresponding order to reconstruct the physical field data matrix. The physical field data matrix includes the quasi-steady-state velocity field, pressure field and temperature field. In this embodiment, the system reads the stored POD baseline mode matrices of each order, where the order represents the ranking of energy proportions, with the first order corresponding to the mode with the largest energy and the kth order corresponding to the kth largest energy mode. The modal coefficients predicted by the surrogate model are used as weights, and the three-dimensional velocity vector, pressure scalar, and temperature scalar distribution matrices of the entire flow field are reconstructed through a linear superposition formula. This achieves rapid inversion from low-dimensional coefficients to high-dimensional physical fields, and provides intuitive flow field and temperature field distributions without rerunning the fluid simulation. This provides engineers with visual physical insights and helps to judge the rationality of optimization schemes. The linear superposition formula is:

[0042] in, For physical field data matrix, For mean field, Let be the i-th modal coefficient. This is the i-th reference mode.

[0043] 503. Based on the underlying control equation of the pulsating flow, perform mass conservation and energy conservation checks on the physical field data matrix respectively, and calculate the root mean square error. In this embodiment, the reconstructed velocity field is substituted into the continuity equation to calculate the divergence, which should ideally be 0. The difference between the actual value and 0 is calculated as the mass residual, used to verify mass conservation. The temperature field is substituted into the energy equation to calculate the difference between the left and right sides as the energy residual, and the root mean square error is calculated based on the distance between the prediction point and the sample database. By introducing physical constraints as a filter, the authenticity of the reconstructed physical field data matrix is ​​verified using the underlying physical laws, effectively eliminating non-physical false solutions that may be generated by the surrogate model, and ensuring the physical credibility of the rapid prediction results.

[0044] 504. Integrate the verification results and root mean square error to obtain the prediction uncertainty result; In this embodiment, the quality residual, energy residual, and root mean square error are integrated into an uncertainty quantification index. If the uncertainty quantification index exceeds a preset threshold, the prediction point is marked as a low-confidence prediction. By rating the reliability of the uncertainty results, the system is prevented from blindly trusting unreliable extrapolation data, providing a risk assessment basis for optimization decisions and enhancing the robustness of the system in actual industrial applications.

[0045] Furthermore, in this embodiment of the invention, the step of constructing a fitness function based on the scalar performance index, employing a multi-objective optimization algorithm, and combining the fitness function and high-dimensional feature vectors to perform iterative optimization to obtain the optimal waveform control parameters includes: 601. Based on the scalar performance index, construct a fitness function with heat transfer enhancement index and flow loss index as the core, and load preset constraints, including physical limit constraints of the pulsation generator and maximum allowable shear stress constraints of the ink. In this embodiment, the fitness function construction module reads scalar performance indicators, including the average Nusselt number, the average friction factor, and the comprehensive performance evaluation coefficient; the fitness function aims to maximize the Nusselt number and minimize the friction factor, specifically expressed as:

[0046] In the above formula, The fitness evaluation index to be optimized has the following input items: This is the decision vector for the current search iteration step. The system dynamically encapsulates combinations of physical components representing the characteristics of the target or variable operating conditions; during the optimization process, the system utilizes a surrogate model prediction engine to obtain the decision vector for each iteration step in real time. Corresponding heat transfer performance With resistance performance ; in the formula and This serves as a benchmark value for quantitative comparison. and These are the weighting coefficients preset according to process requirements; by constructing a fitness function, a clear convergence objective is provided for the subsequent global optimization process, namely, finding the one that makes the fitness function... The optimal solution that approaches the extreme value state; Furthermore, during the optimization process, the upper and lower physical limits of the frequency and amplitude of the pulsation generator are applied as boundary constraints, while the critical shear stress derived from the ink rheological properties is applied as a safety constraint. Once the predicted stress exceeds this value, a penalty term is applied. By applying the preset constraints, the rheological safety of the ink is incorporated into the optimization objective, preventing the excessive pursuit of heat exchange efficiency from causing damage to the ink molecular structure or nozzle clogging, thus ensuring the safety and feasibility of the optimization scheme in the industrial field.

[0047] 602. The NSGA-II multi-objective optimization algorithm is adopted to initialize a decision vector population based on the high-dimensional feature vector. The decision vector population includes multiple individuals, and each individual corresponds to a set of waveform control parameters to be optimized. In this embodiment, the NSGA-II algorithm is used to generate an initial population of N individuals. Each individual is a vector containing control variables such as pulsation frequency and pulsation amplitude. The population distribution range is determined by the upper and lower limits defined in the high-dimensional feature vector. The parameter space is explored through population evolution, avoiding the trap of single-point search falling into local optima, and the optimal combination of pulsation waveform parameters can be found from a global perspective.

[0048] 603. Based on the fitness function and combined with preset constraints, perform multi-objective optimization iteration on the decision vector population; In this embodiment, the multi-objective iterative optimization process is as follows: S1, for each individual in the decision vector population, the fitness value is calculated by calling the pre-trained surrogate model; S2, the population is divided into different frontiers according to the non-dominated sorting, and the crowding distance is calculated to maintain the diversity of solutions; S3, the offspring population is generated by tournament selection, simulated binary crossover, and polynomial mutation; S4, the parent and offspring generations are merged, a fast non-dominated sorting is performed, and the top N individuals are selected to enter the next generation; S5, S1 to S4 are repeated until the maximum number of iterations is reached or the fitness value converges; through the global search capability of the multi-objective optimization algorithm, the contradictory relationship between heat transfer and drag is effectively handled, and complex multi-objective trade-offs can be completed without manual intervention, realizing the full automation of the optimization process.

[0049] 604. When the preset maximum number of iterations or convergence accuracy threshold is reached, the optimal waveform control parameters are output based on the working condition type in the high-dimensional feature vector. The optimal waveform control parameters are the globally optimal single-objective decision vector or the Pareto non-dominated solution set. In this embodiment, if the operating condition is a single fixed-value requirement (such as fixed flow rate or fixed temperature), the system outputs a unique optimal solution, which is the globally optimal single-objective decision vector with the highest fitness, including the specific optimal frequency and amplitude values; if the operating condition is a range requirement (such as flow rate fluctuating within a certain range), the system outputs a Pareto non-dominated solution set, including multiple combinations of frequencies and amplitudes, corresponding to different schemes from extreme heat exchange to minimum energy consumption. The flexible decision-making output mode based on operating conditions adapts to the needs of different stages from R&D experiments to mass production, providing users with a wide range of process options.

[0050] Furthermore, in this embodiment of the invention, the step of integrating the optimal waveform control parameters and performance prediction results to generate a visualization report, and controlling the pulsating flow preheating process of the ink supply pipeline based on the visualization report, includes: 701. Integrate the optimal waveform control parameters and performance prediction results to generate a visualization report, which includes a core control parameter table, a quantitative performance index radar chart, a pulsating period characteristic phase temperature distribution cloud map, and a flow field vector trajectory map; In this embodiment, the data in the core control parameter table comes from the optimal waveform control parameters output by the optimization decision layer; the data in the quantified performance index radar chart comes from scalar performance indices such as the average Nusselt number, average friction factor, and comprehensive performance evaluation coefficient predicted by the surrogate model; the data in the pulsating period characteristic phase temperature distribution cloud map and flow field vector trajectory map come from the reconstructed quasi-steady-state temperature field matrix and velocity field matrix, and are generated by extracting four characteristic phases of 0°, 90°, 180°, and 270° in the pulsating period; by transforming the data into intuitive visualization maps, process engineers can directly observe how the pulsating flow destroys the thermal boundary layer, verify the physical mechanism of enhanced heat transfer, lower the understanding threshold, and assist in engineering decision-making.

[0051] 702. Extract the optimal waveform control parameters from the visualization report and convert them into standardized control commands that can be recognized by the ink supply pipeline pulsation generator. The optimal waveform control parameters include the optimal pulsation frequency, pulsation amplitude, and reference operating pressure parameters. In this embodiment, the system reads the optimal waveform control parameters and converts them into an instruction format supported by the pulse generator through a digital-to-analog conversion protocol; for example, the frequency is converted into the duty cycle of the pulse width modulation signal, and the amplitude is converted into an analog voltage signal, thus realizing the translation of algorithm parameters into device control signals and eliminating errors that may be generated by manual transcription.

[0052] 703. The standardized control command is sent to the pulse execution unit of the ink supply pipeline to adjust the working parameters of the pulse execution unit and realize the control of the pulse flow preheating process of OLED inkjet printing ink. In this embodiment, the pulsation execution unit includes a piezoelectric ceramic pump or a servo piston pump. The system sends standardized control commands to the pulsation execution unit via industrial Ethernet to drive a proportional valve or metering pump to generate pulsating flow that conforms to waveform parameters. For example, the driving voltage frequency of the piezoelectric ceramic is controlled to generate pulsations of a specific frequency, and the stroke of the servo motor is controlled to generate flow fluctuations of a specific amplitude, thereby efficiently preheating the ink as it flows through the pipeline. This achieves closed-loop control from virtual optimization to physical execution, and actively regulates ink temperature and viscosity through a precise pulsating flow field, thereby improving printing quality and production yield.

[0053] The above describes the ink heat transfer method based on the surrogate model in the embodiments of the present invention. The following describes the ink heat transfer device based on the surrogate model in the embodiments of the present invention. Please refer to [link / reference]. Figure 2 One embodiment of the ink heat transfer device based on the surrogate model in this invention includes: Module 801 is used to pre-build a heat transfer optimization system, which includes a data support layer, a model resource layer, a parameter calling layer and an optimization decision layer that are interconnected through interfaces. The mapping module 802 is used to obtain engineering parameters through the parameter calling layer and map the engineering parameters into high-dimensional feature vectors using a preset feature mapping operator. The matching module 803 is used to match the physical model from the model resource layer based on the high-dimensional feature vector, and to construct the underlying control equation of the pulsating flow based on the matched physical model. The verification module 804 is used to input the high-dimensional feature vector into the pre-trained surrogate model to obtain scalar performance indicators and modal coefficient vectors, reconstruct the physical field data matrix based on the modal coefficient vectors, and perform physical consistency verification on the physical field data matrix based on the underlying control equation of the pulsating flow to obtain the prediction uncertainty result. The integration module 805 is used to integrate the scalar performance index, the physical field data matrix and the prediction uncertainty results to obtain the performance prediction results. Optimization module 806 is used to construct a fitness function based on the scalar performance index, and to perform iterative optimization by using a multi-objective optimization algorithm in combination with the fitness function and high-dimensional feature vector to obtain the optimal waveform control parameters. The control module 807 is used to integrate the optimal waveform control parameters and performance prediction results, generate a visualization report, and control the pulsating flow preheating process of the ink supply pipeline based on the visualization report.

[0054] Based on the same ideas as the methods in the above embodiments, the apparatus provided in this application can implement the methods in the above embodiments.

[0055] above Figure 2 The ink heat transfer device based on the proxy model in the embodiments of the present invention will be described in detail from the perspective of modular functional entities. The ink heat transfer device based on the proxy model in the embodiments of the present invention will be described in detail from the perspective of hardware processing.

[0056] Figure 3This is a schematic diagram of the structure of an ink heat transfer device 900 based on a proxy model provided in an embodiment of the present invention. The proxy model-based ink heat transfer device 900 can vary significantly due to different configurations or performance characteristics. It may include one or more central processing units (CPUs) 910 and a memory 920, and one or more storage media 930 (e.g., one or more mass storage devices) storing application programs 933 or data 932. The memory 920 and storage media 930 can be temporary or persistent storage. The program stored in the storage media 930 may include one or more modules (not shown in the diagram), each module including a series of instruction operations performed on the proxy model-based ink heat transfer device 900. Furthermore, the processor 910 may be configured to communicate with the storage media 930 and execute the series of instruction operations in the storage media 930 on the proxy model-based ink heat transfer device 900 to implement the steps of the proxy model-based ink heat transfer method provided in the above-described method embodiments.

[0057] The agent-based ink heat transfer device 900 may also include one or more power supplies 940, one or more wired or wireless network interfaces 950, one or more input / output interfaces 960, and / or one or more operating systems 931, such as Windows Server, Mac OS X, Unix, Linux, FreeBSD, etc. Those skilled in the art will understand that... Figure 3 The illustrated surrogate model-based ink heat transfer device structure does not constitute a limitation on surrogate model-based ink heat transfer devices, and may include more or fewer components than illustrated, or combine certain components, or have different component arrangements.

[0058] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when executed on a computer, cause the computer to perform the steps of the ink heat transfer method based on the proxy model.

[0059] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the system, device or unit described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0060] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0061] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An ink heat transfer method based on a surrogate model, characterized in that, include: A pre-built heat transfer optimization system is provided, which includes a data support layer, a model resource layer, a parameter calling layer, and an optimization decision layer that are interconnected through interfaces. Engineering parameters are obtained through the parameter calling layer, and the engineering parameters are mapped into high-dimensional feature vectors using a preset feature mapping operator. The high-dimensional feature vectors include geometric feature sub-vectors, rheological feature sub-vectors, and working condition feature sub-vectors. Based on the high-dimensional feature vector, a physical model is matched from the model resource layer, and the underlying control equations for the pulsating flow are constructed based on the matched physical model. Specifically, according to the geometric feature sub-vectors in the high-dimensional feature vector, a corresponding flow channel topology model is matched from the parametric flow channel geometry model library of the model resource layer, and the matched flow channel topology model is fine-tuned in real time according to the geometric feature sub-vectors to obtain a fine-tuned flow channel topology model; according to the rheological feature sub-vectors in the high-dimensional feature vector, a corresponding fluid constitutive model is matched from the hierarchical rheological model library of the model resource layer; based on the fine-tuned flow channel topology model and the matched fluid constitutive model, the corresponding continuity equation, unsteady momentum equation, and energy equation considering viscous dissipation are called from the pulsating flow field master control equation library of the model resource layer to construct the underlying control equations for the pulsating flow. The high-dimensional feature vector is input into a pre-trained surrogate model to obtain scalar performance indices and modal coefficient vectors. A physical field data matrix is ​​reconstructed based on the modal coefficient vectors, and the physical consistency of the physical field data matrix is ​​checked based on the underlying control equations of the pulsating flow to obtain the prediction uncertainty result. Specifically, the high-dimensional feature vector is input into a surrogate model pre-trained based on the Kriging algorithm to obtain scalar performance indices and POD modal coefficient vectors corresponding to the operating conditions. The scalar performance indices include the average Nusselt number, average friction factor, and comprehensive performance evaluation coefficient. Pre-stored reference modes of various orders are retrieved from the intrinsic orthogonal decomposition mode library of the data support layer. The POD modal coefficient vectors are linearly weighted and combined with the reference modes of the corresponding orders to reconstruct the physical field data matrix. The physical field data matrix includes a quasi-steady-state velocity field, pressure field, and temperature field. Based on the underlying control equations of the pulsating flow, mass conservation checks and energy conservation checks are performed on the physical field data matrix, and the root mean square error is calculated. The check results and the root mean square error are integrated to obtain the prediction uncertainty result. By integrating the scalar performance index, the physical field data matrix, and the prediction uncertainty results, the performance prediction results are obtained. A fitness function is constructed based on the scalar performance index. A multi-objective optimization algorithm is used, combining the fitness function and high-dimensional feature vectors to perform iterative optimization to obtain the optimal waveform control parameters. Specifically, based on the scalar performance index, a fitness function with heat transfer enhancement index and flow loss index as the core is constructed, and preset constraints are loaded. The constraints include physical limit constraints of the pulsation generator and maximum allowable shear stress constraints of the ink. The NSGA-II multi-objective optimization algorithm is used to initialize a decision vector population based on the high-dimensional feature vectors. The decision vector population includes multiple individuals, each corresponding to a set of waveform control parameter combinations to be optimized. Based on the fitness function and the preset constraints, multi-objective optimization iterations are performed on the decision vector population. When the preset maximum number of iterations or convergence accuracy threshold is reached, the optimal waveform control parameters are output based on the working condition type in the high-dimensional feature vector. The optimal waveform control parameters are the global optimal single-objective decision vector or the Pareto non-dominated solution set. The optimal waveform control parameters and performance prediction results are integrated to generate a visualization report, and the pulse flow preheating process of the ink supply pipeline is controlled based on the visualization report.

2. The ink heat transfer method according to claim 1, characterized in that, The pre-built heat transfer optimization system includes: A data support layer is constructed, which includes a sample database, a surrogate model pre-trained based on the Kriging algorithm, and an intrinsic orthogonal decomposition mode library; A model resource layer is constructed, which includes a parametric flow channel geometry model library, a hierarchical rheological model library, and a library of master equations for pulsating flow fields. A parameter calling layer is constructed, which includes a standardized engineering parameter input interface, a data validity verification module, and a feature mapping operator execution module connected in sequence. An optimization decision layer is constructed, which includes a fitness function construction module and a multi-objective optimization algorithm execution module connected in sequence.

3. The ink heat transfer method according to claim 2, characterized in that, The step of obtaining engineering parameters through the parameter calling layer and mapping the engineering parameters into high-dimensional feature vectors using a preset feature mapping operator includes: Engineering parameters are obtained through the parameter calling layer, including flow channel structure parameters, fluid property parameters, and thermodynamic boundary condition parameters. The engineering parameters are validated for data validity, and each parameter is determined to be within the preset range of physically reasonable values. If the engineering parameters pass the data validity check, a preset feature mapping operator is invoked to uniformly map all types of parameters into high-dimensional feature vectors.

4. The ink heat transfer method according to claim 1, characterized in that, The process of integrating the optimal waveform control parameters and performance prediction results to generate a visualization report, and controlling the pulsating flow preheating process of the ink supply pipeline based on the visualization report, includes: The optimal waveform control parameters and performance prediction results are integrated to generate a visualization report, which includes a core control parameter table, a quantitative performance index radar chart, a pulsating period characteristic phase temperature distribution cloud map, and a flow field vector trajectory map. The optimal waveform control parameters are extracted from the visualization report and converted into standardized control commands that can be recognized by the ink supply pipeline pulsation generator. The optimal waveform control parameters include the optimal pulsation frequency, pulsation amplitude, and reference operating pressure parameters. The standardized control command is sent to the pulsation execution unit of the ink supply pipeline to adjust the working parameters of the pulsation execution unit and realize the control of the pulsation flow preheating process of OLED inkjet printing ink.

5. An ink heat transfer device based on a surrogate model, characterized in that, include: A construction module is used to pre-build a heat transfer optimization system, which includes a data support layer, a model resource layer, a parameter calling layer, and an optimization decision layer that are interconnected through interfaces. The mapping module is used to obtain engineering parameters through the parameter calling layer and use a preset feature mapping operator to map the engineering parameters into high-dimensional feature vectors. The high-dimensional feature vectors include geometric feature sub-vectors, rheological feature sub-vectors, and working condition feature sub-vectors. The matching module is used to match physical models from the model resource layer based on the high-dimensional feature vectors, and to construct the underlying control equations for the pulsating flow based on the matched physical models. Specifically, based on the geometric feature sub-vectors in the high-dimensional feature vectors, it matches the corresponding flow channel topology model from the parametric flow channel geometry model library of the model resource layer, and performs real-time parametric fine-tuning of the matched flow channel topology model based on the geometric feature sub-vectors to obtain a fine-tuned flow channel topology model; based on the rheological feature sub-vectors in the high-dimensional feature vectors, it matches the corresponding fluid constitutive model from the hierarchical rheological model library of the model resource layer; based on the fine-tuned flow channel topology model and the matched fluid constitutive model, it calls the corresponding continuity equation, unsteady momentum equation, and energy equation considering viscous dissipation from the pulsating flow field master control equation library of the model resource layer to construct the underlying control equations for the pulsating flow. The verification module is used to input the high-dimensional feature vector into a pre-trained surrogate model to obtain scalar performance indicators and modal coefficient vectors. Based on the modal coefficient vectors, a physical field data matrix is ​​reconstructed, and the physical consistency of the physical field data matrix is ​​verified based on the underlying control equations of the pulsating flow to obtain the prediction uncertainty result. Specifically, the high-dimensional feature vector is input into a surrogate model pre-trained based on the Kriging algorithm to obtain scalar performance indicators and POD modal coefficient vectors corresponding to the operating conditions. The scalar performance indicators include the average Nusselt number and the average friction factor. The data includes comprehensive performance evaluation coefficients; pre-stored reference modes of each order are called from the intrinsic orthogonal decomposition mode library of the data support layer; the POD mode coefficient vector is linearly weighted and combined with the reference modes of the corresponding order to reconstruct the physical field data matrix, which includes quasi-steady-state velocity field, pressure field and temperature field; based on the underlying control equation of the pulsating flow, mass conservation check and energy conservation check are performed on the physical field data matrix respectively, and the root mean square error is calculated; the check results and root mean square error are integrated to obtain the prediction uncertainty result; The integration module is used to integrate the scalar performance index, the physical field data matrix, and the prediction uncertainty results to obtain the performance prediction results. The optimization module is used to construct a fitness function based on the scalar performance index, and to perform iterative optimization using a multi-objective optimization algorithm, combining the fitness function and high-dimensional feature vectors to obtain the optimal waveform control parameters. Specifically, based on the scalar performance index, a fitness function with heat transfer enhancement index and flow loss index as its core is constructed, and preset constraints are loaded. The constraints include physical limit constraints for the pulsation generator and the maximum allowable shear stress constraint for the ink. The NSGA-II multi-objective optimization algorithm is used to initialize a decision vector population based on the high-dimensional feature vectors. The decision vector population includes multiple individuals, each corresponding to a set of waveform control parameter combinations to be optimized. Based on the fitness function and the preset constraints, multi-objective optimization iterations are performed on the decision vector population. When the preset maximum number of iterations or convergence accuracy threshold is reached, the optimal waveform control parameters are output based on the working condition type in the high-dimensional feature vector. The optimal waveform control parameters are the global optimal single-objective decision vector or the Pareto non-dominated solution set. The control module is used to integrate the optimal waveform control parameters and performance prediction results, generate a visualization report, and control the pulsating flow preheating process of the ink supply pipeline based on the visualization report.

6. An ink heat transfer device based on a surrogate model, characterized in that, The ink heat transfer device based on the proxy model includes: a memory and at least one processor, wherein the memory stores instructions; At least one of the processors invokes the instructions in the memory to cause the surrogate-based ink heat transfer device to perform the steps of the surrogate-based ink heat transfer method as described in any one of claims 1-4.

7. A computer-readable storage medium storing instructions thereon, characterized in that, When the instructions are executed by the processor, they implement the various steps of the ink heat transfer method based on the proxy model as described in any one of claims 1-4.

Images