Method and device for predicting viscosity of fermented milk, electronic equipment and storage medium

Abstract

The application provides a fermented milk processing viscosity prediction method and device, electronic equipment and storage medium, relates to the dairy product processing technical field, and includes the following steps: coupling a fermented milk constitutive model and a computational fluid dynamics model to obtain a whole-process processing simulation model, and obtaining whole-process processing process data of target fermented milk under different process parameter combinations based on the whole-process processing simulation model; and training a preset machine learning model based on the whole-process processing process data to obtain a processing viscosity prediction model. The method and device provided by the application generate whole-process processing process data of target fermented milk under different process parameter combinations through the whole-process processing simulation model; the processing viscosity prediction model is trained through the whole-process processing process data, and the viscosity of the target fermented milk after processing is quickly predicted; the viscosity of the fermented milk product in the processing process is accurately and quickly predicted, and the research and development efficiency and product stability of the fermented milk product are improved.

Description

Technical Field

[0001] This invention relates to the field of dairy processing technology, and in particular to a method, apparatus, electronic device, and storage medium for predicting the viscosity of fermented milk. Background Technology

[0002] For fermented milk products (such as yogurt), viscosity, as a key rheological parameter, directly determines the product's sensory quality (such as smoothness and adhesion to the bottle) and consumer acceptance. The processing of fermented milk products is affected by factors such as the precision of equipment control and variations in process operation, resulting in differences in the heat intensity and shear strength experienced by the materials during processing. This leads to discrepancies between the viscosity of the processed fermented milk and the expected viscosity.

[0003] In the processing of fermented milk products, related technologies typically employ a trial-and-error R&D model. This model has significant limitations: a single fermentation experiment can take 6-12 hours, and the lack of quantitative research on the relationship between formula components, process equipment and parameters, and yogurt viscosity leads to large fluctuations in product viscosity. Furthermore, the absence of a quantitative model during the pilot-to-scale production process makes it impossible to reproduce the viscosity of the product during R&D. Simultaneously, the viscosity of the same product produced on different production lines varies, and even between different batches of the same product from the same production line. This results in a typical new fermented milk product development cycle of 6-8 months, requiring 200-300 sets of repeatable tests, significantly increasing R&D costs and time (room temperature fermented milk, due to its complex formula and process, has an even longer development cycle and higher costs).

[0004] Therefore, how to accurately and quickly predict the viscosity of fermented milk products during processing, and improve the R&D efficiency and product stability of fermented milk products, has become a technical problem that the industry urgently needs to solve. Summary of the Invention

[0005] This invention provides a method, apparatus, electronic device, and storage medium for predicting the viscosity of fermented milk during processing, which addresses the technical problem of how to accurately and quickly predict the viscosity of fermented milk products during processing, thereby improving the R&D efficiency and product stability of fermented milk products.

[0006] This invention provides a method for predicting the viscosity of fermented milk during processing, comprising:

[0007] A constitutive model of the target fermented milk is constructed; the constitutive model is used to describe the effects of shear rate and temperature on the viscosity of the target fermented milk.

[0008] Computational fluid dynamics models of each process device are constructed; the computational fluid dynamics models are used to describe the effects of the combination of process parameters of the process device on the shear rate and / or temperature of the target fermented milk;

[0009] By coupling the constitutive model of the fermented milk and the computational fluid dynamics model, a full-process processing simulation model of the target fermented milk is obtained, and based on the full-process processing simulation model, full-process processing data of the target fermented milk under different combinations of process parameters are obtained;

[0010] The preset machine learning model is trained based on the data from the entire processing process to obtain a processing viscosity prediction model for the target fermented milk, and the viscosity of the target fermented milk after processing is predicted based on the processing viscosity prediction model.

[0011] In some embodiments, before obtaining the full-process processing data of the target fermented milk under different combinations of process parameters based on the full-process processing simulation model, the method further includes:

[0012] Determine at least one target combination of process parameters;

[0013] Obtain the actual viscosity value of the target fermented milk under the at least one combination of target process parameters;

[0014] Determine the viscosity simulation value of the full-process simulation model under the at least one combination of target process parameters;

[0015] The simulation model of the entire process is verified based on the comparison between the simulated viscosity value and the actual viscosity value under the at least one combination of target process parameters.

[0016] If the comparison between the simulated viscosity value and the actual viscosity value is greater than a preset threshold, it is determined that the full-process processing simulation model has failed verification, and the constitutive model of fermented milk and / or the computational fluid dynamics model are corrected.

[0017] In some embodiments, before training a preset machine learning model based on the full-process processing data to obtain a processing viscosity prediction model for the target fermented milk, the method further includes:

[0018] Principal component analysis was performed on the entire process data to obtain key feature data.

[0019] In some embodiments, training a preset machine learning model based on the entire processing data to obtain a processing viscosity prediction model for the target fermented milk includes:

[0020] Using the key feature data as samples and the simulated viscosity of fermented milk corresponding to the key feature data as sample labels, the preset machine learning model is trained to obtain the processing viscosity prediction model.

[0021] The preset machine learning model includes at least one of artificial neural networks, process regression models, and support vector machines.

[0022] In some embodiments, the constitutive model of fermented milk includes a structural dynamics model and a temperature-stress relationship model.

[0023] In some embodiments, constructing the constitutive model of the target fermented milk includes:

[0024] Based on the stress constitutive equation and the rate equation, a structural dynamics model is established; the stress constitutive equation is used to describe the relationship between shear rate, structural parameters, and shear stress; the rate equation is used to describe the relationship between structural parameters and shear rate.

[0025] Based on the relationship between chemical reaction rate and temperature, a temperature-stress relationship model is established; this model is used to describe the effect of temperature change on stress.

[0026] Based on the structural dynamics model and the temperature-stress relationship model, the constitutive model of the fermented milk is determined.

[0027] In some embodiments, the process equipment includes at least one of a fermenter, a transfer pump, pipelines, and a heat exchanger; the process parameter combination includes control parameters and / or structural parameters of the process equipment.

[0028] This invention provides a fermented milk processing viscosity prediction device, comprising:

[0029] The first construction module is used to construct a constitutive model of the target fermented milk; the constitutive model of the fermented milk is used to describe the effects of shear rate and temperature on the viscosity of the target fermented milk;

[0030] The second construction module is used to construct computational fluid dynamics models for each process device; the computational fluid dynamics models are used to describe the effects of the process parameter combinations of the process devices on the shear rate and / or temperature of the target fermented milk;

[0031] The model coupling module is used to couple the constitutive model of the fermented milk and the computational fluid dynamics model to obtain the full-process simulation model of the target fermented milk, and to obtain the full-process processing data of the target fermented milk under different combinations of process parameters based on the full-process simulation model.

[0032] The processing prediction module is used to train a preset machine learning model based on the full-process processing data to obtain a processing viscosity prediction model for the target fermented milk, and to predict the viscosity of the target fermented milk after processing based on the processing viscosity prediction model.

[0033] The present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the fermented milk processing viscosity prediction method.

[0034] The present invention provides a non-transitory computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the fermented milk processing viscosity prediction method.

[0035] The present invention provides a method, apparatus, electronic device, and storage medium for predicting the viscosity of fermented milk processing. By constructing a constitutive model of the target fermented milk and computational fluid dynamics models of various process equipment, a full-process processing simulation model of the target fermented milk is obtained through coupling. This achieves a full-process digital mapping of the rheological properties of raw materials during processing, accurately predicting the impact of different equipment structural parameters and control parameters on the microstructure and macroscopic quality of the product. It can generate full-process processing data of the target fermented milk under different combinations of process parameters. The processing viscosity prediction model of the target fermented milk is obtained by training a preset machine learning model using the full-process processing data. This model is then used to rapidly predict the viscosity of the target fermented milk after processing, combining time-consuming physical simulation with efficient machine learning. This not only ensures the physical accuracy of the prediction but also enables rapid iteration and optimization of the design scheme. Overall, it achieves accurate and rapid prediction of the viscosity of fermented milk products during processing, improving the R&D efficiency and product stability of fermented milk products, reducing R&D costs, and increasing the success rate of quantitative scale-up from pilot-scale to industrial production. It has significant technical advantages and economic value. Attached Figure Description

[0036] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0037] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0038] Figure 1 This is a schematic diagram of the process for the research and development and production of fermented milk products provided by this invention.

[0039] Figure 2 This is one of the flowcharts illustrating the method for predicting the viscosity of fermented milk processing provided by this invention.

[0040] Figure 3This is the steady-state fitting curve of yogurt at 25°C provided by the present invention.

[0041] Figure 4 This is the stress-shear rate relationship curve provided by the present invention at different temperatures.

[0042] Figure 5 This is a fitting curve of the stress-temperature relationship of different types of yogurt at different temperatures, provided by the present invention.

[0043] Figure 6 This is the second schematic diagram of the process flow diagram for predicting the viscosity of fermented milk provided by the present invention.

[0044] Figure 7 This is a schematic diagram illustrating the process of constructing a full-process processing simulation model provided by the present invention.

[0045] Figure 8 This is a schematic diagram of the process of training the processing viscosity prediction model provided by the present invention.

[0046] Figure 9 This is a schematic diagram of the fermented milk processing viscosity prediction device provided by the present invention.

[0047] Figure 10 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0048] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0049] It should be noted that the terms "first," "second," etc., used in this invention 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 of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that comprises a series of steps, units, or modules is not necessarily limited to those explicitly listed, but may include other steps, units, or modules not explicitly listed or inherent to such processes, methods, products, or devices.

[0050] ‌ Figure 1This is a schematic diagram of the process for the research and development and production of fermented milk products provided by this invention, such as... Figure 1 As shown, the research and development and production process of fermented milk products in related technologies typically includes stages such as determining the fermented milk formula design, small-scale experiments, sample viscosity testing, pilot-scale amplification, reproducibility viscosity testing, commercial production line trial production, and process and equipment optimization. Among these, small-scale experiments refer to laboratory-scale production, while pilot-scale experiments refer to simulated industrial production.

[0051] The processing of fermented milk products is affected by factors such as the precision of equipment control and variations in process operation, resulting in differences in the heat and shear strength experienced by the materials during processing. This leads to discrepancies between the viscosity of the processed fermented milk and the expected viscosity. Related technologies lack quantitative research on the relationship between "formula components, process equipment and parameters, and yogurt viscosity," resulting in large viscosity fluctuations. The lack of quantitative models during small-scale, pilot-scale, and production scale-up processes makes it impossible to reproduce the viscosity of the product during R&D. The viscosity of the same product produced on different production lines varies, and even the viscosity of different batches of the same product from the same production line can differ.

[0052] To address the shortcomings of related technologies, this invention provides a method, apparatus, electronic device, and storage medium for predicting the viscosity of fermented milk processing. By constructing a coupled digital design platform that combines a constitutive model of fermented milk (characterizing its rheological properties) with a computational fluid dynamics (CFD) model of the process equipment (simulating the velocity field, temperature field, and shear field within the equipment), a full-process digital mapping from raw material rheological properties to processing dynamics can be achieved. This platform accurately predicts the impact of different equipment structural parameters and control parameters (such as stirring rate and temperature gradient) on the microstructure and macroscopic quality of the product. This assists R&D personnel in conducting scientific analysis more quickly and effectively, significantly reducing the number of tests and improving the R&D efficiency and product stability of fermented milk products.

[0053] Figure 2 This is one of the flowcharts illustrating the method for predicting the viscosity of fermented milk processing provided by this invention, such as... Figure 2 As shown, the method includes steps 210, 220, 230 and 240.

[0054] Step 210: Construct a constitutive model of the target fermented milk; the constitutive model of the fermented milk is used to describe the effects of shear rate and temperature on the viscosity of the target fermented milk.

[0055] Specifically, the fermented milk processing viscosity prediction method provided in this embodiment of the invention is executed by a fermented milk processing viscosity prediction device. This device can be implemented in software, such as a fermented milk processing viscosity prediction program running on a computer; or it can be implemented in hardware, such as a computer or server that executes the fermented milk processing viscosity prediction method.

[0056] Fermented milk refers to dairy products (such as yogurt) made from animal milk or plant-based raw materials, which have a lower pH after sterilization and fermentation. Target fermented milk refers to fermented milk for which viscosity prediction is required during processing; for example, it may be a fermented milk product with a specific formula (e.g., specific protein content, fat content, types and amounts of stabilizers, fermentation strains, etc.).

[0057] Fermented milk viscosity refers to the ability of fermented milk (fluid) to resist shear deformation, and can be used to quantify flow resistance. Viscosity determines the taste (thickness), filling efficiency, and shelf-life stability of fermented milk.

[0058] Shear stress refers to the stress component parallel to the cross section in a fluid or solid, representing the internal force per unit area resisting relative slippage between layers. Shear rate is the ratio of the relative velocity between adjacent layers in a fluid to the interlayer distance, quantifying the rate of fluid deformation and directly affecting the viscosity of fermented milk.

[0059] Temperature is a physical quantity that indicates the degree of hotness or coldness of an object, and it directly affects the rate of molecular motion and chemical reaction. Temperature affects the activity of lactic acid bacteria, protein denaturation, and gel formation rate, thereby altering the viscosity of fermented milk.

[0060] Constitutive models are mathematical equations that describe the stress-strain-temperature relationship of materials and are used to predict rheological (flow and deformation) properties. Constitutive models of fermented milk can establish quantitative relationships between shear rate, temperature, and viscosity, guiding process optimization.

[0061] A constitutive model of the target fermented milk can be constructed by starting from the mechanism of the influence of shear rate and temperature on the viscosity of fermented milk.

[0062] Step 220: Construct computational fluid dynamics models for each process device; the computational fluid dynamics models are used to describe the effects of the combination of process parameters of the process device on the shear rate and / or temperature of the target fermented milk.

[0063] Specifically, process equipment refers to all the unit equipment that may affect the shear rate and temperature of fermented milk from the end of fermentation to bottling. For example, process equipment may include fermentation tanks, transfer pumps (such as rotary pumps, screw pumps, etc.), pipelines connecting various units (such as straight pipes, elbows, valves, etc.), heat exchangers (such as plate heat exchangers, shell and tube heat exchangers, etc.), etc.

[0064] Computational fluid dynamics (CFD) models are digital models that use computers for numerical calculations and graphical displays to analyze systems containing physical phenomena such as fluid flow and heat conduction.

[0065] Based on Computer-Aided Design (CAD) drawings or optical scanning data, a precise 3D geometric model of the process equipment can be created using 3D modeling software. This geometric model is then discretized into a computational mesh composed of numerous tiny units (such as tetrahedrons and hexahedrons). Process parameters are then set for the inlet, outlet, and walls of the process equipment's geometric model. A combination of process parameters refers to a set of multiple process parameters, which may include the control parameters and / or structural parameters of the process equipment.

[0066] Control parameters, also known as operating parameters, refer to variables that operators or automated control systems can actively adjust and change during the production process without requiring physical modifications to the equipment. Examples include the rotational speed of the agitator in a fermenter, the duration of agitation, the flow rate of fermented milk delivered through pumps and pipelines, and the inlet temperature of the cooling medium in a plate heat exchanger. Adjusting these control parameters can influence the shear rate and temperature of the fermented milk.

[0067] Structural parameters refer to variables that describe the physical form, size, and inherent characteristics of process equipment. These structural parameters also affect the shear rate and temperature of fermented milk. For example, the inner diameter and total length of the delivery pipeline. At the same flow rate, the smaller the pipe diameter, the faster the flow velocity and the higher the shear rate. Other examples include the number of heat exchange plates, the area of ​​each plate, the spacing between plates, and the corrugation pattern on the plates. These all determine the temperature change of the fermented milk within the heat exchanger.

[0068] Taking a fermenter as an example, the control parameters of a fermenter include rotational speed, torque, power, and stirring time. Structural parameters include the tank's structural dimensions and the dimensions of the stirring blades.

[0069] Rotational speed affects the shear rate, which can be expressed by the formula: .in, For shear rate, It is a constant and can be obtained by calculation from the CFD model (depending on the type of impeller); The value is the rotational speed.

[0070] Power and tank structural dimensions affect the shear rate, which can be expressed by the formula: .in, , It is a constant and can be obtained by calculation from the CFD model (related to the design of the stirring blades). Power; This refers to the tank volume (tank structural dimensions).

[0071] In a fermenter, the effects of its control parameters and structural parameters on viscosity can include:

[0072] 1. The higher the rotation speed, the lower the viscosity of the fermented milk;

[0073] 2. The greater the torque, the lower the viscosity of the fermented milk;

[0074] 3. The higher the power, the lower the viscosity of the fermented milk;

[0075] 4. The longer the stirring time, the more significant the loss (reduction) in the viscosity of the fermented milk, until the texture network of the fermented milk is completely destroyed;

[0076] 5. The larger the tank volume, the slower the viscosity decreases;

[0077] 6. The effect of impeller structure on viscosity can be calculated using CFD to obtain constants. , , .

[0078] These process parameters constitute the combination of process parameters for the process equipment. For example, for a fermenter model, the process parameter could be the rotational speed of the agitator; for a pipeline model, it could be the inlet flow rate; and for a plate heat exchanger model, it could be the flow rate of the fermented milk and the temperature and flow rate of the cooling medium. Computational fluid dynamics (CFD) models can transform macroscopic, directly controllable process parameters (such as rotational speed and flow rate) into microscopic physical fields that directly affect the viscosity of the fermented milk. In other words, given a set of process parameters and a solution is performed, a CFD model can output the shear rate and temperature at any location inside the process equipment.

[0079] Step 230: Couple the constitutive model of fermented milk and the computational fluid dynamics model to obtain the full-process simulation model of the target fermented milk, and obtain the full-process processing data of the target fermented milk under different combinations of process parameters based on the full-process simulation model.

[0080] Specifically, coupling the constitutive model of fermented milk and the computational fluid dynamics model means embedding or integrating the constitutive model of fermented milk into the solver of the computational fluid dynamics model in the form of a program or function.

[0081] In one specific embodiment, this can be achieved through a user-defined function interface provided by the CFD software. In each computational iteration step, the CFD solver calculates the shear rate and temperature within each process unit, then calls the fermented milk constitutive model embedded through the user-defined function to calculate the real-time viscosity of the fermented milk within that unit using these shear rates and temperatures. The solver then uses this updated viscosity value to continue solving the fluid dynamics governing equations. This two-way coupling ensures high fidelity in the simulation.

[0082] The full-process simulation model refers to a comprehensive simulation model that couples and connects the computational fluid dynamics models of multiple process equipment that fermented milk passes through with the constitutive model of fermented milk, forming a complete simulation model that can simulate the entire physical processing process from the end of fermentation to before bottling.

[0083] By setting different combinations of process parameters and running a full-process simulation model, the full-process processing data of the target fermented milk under different combinations of process parameters can be obtained.

[0084] In a specific embodiment, an experiment can be designed containing hundreds of simulation cases, each corresponding to a different combination of process parameters. For each simulation case, the obtained full-process data is extremely rich, including not only the final simulated viscosity value, but more importantly, detailed records of the shear rate changes and temperature changes experienced by the fluid particles throughout the processing path.

[0085] Step 240: Train the preset machine learning model based on the data of the entire process to obtain the processing viscosity prediction model of the target fermented milk, and predict the viscosity of the target fermented milk after processing based on the processing viscosity prediction model.

[0086] Specifically, the preset machine learning model can be any of a variety of machine learning models that are well known to those skilled in the art and capable of handling regression problems. Examples include Artificial Neural Network (ANN), Support Vector Machine (SVM), and Gaussian Process Regression (GPR).

[0087] Some or all of the process parameters in the entire process data can be used as samples, and the final viscosity of the fermented milk after processing (which can be the viscosity obtained from online simulation or the viscosity verified by actual production) in the entire process data can be used as a label to train the preset machine learning model, thereby improving the preset machine learning model's ability to predict the viscosity of the target fermented milk after processing, and thus obtaining a processing viscosity prediction model.

[0088] Processing viscosity prediction models, as alternative or surrogate models, can simulate the inherent logic of complex physical simulation processes through data-driven methods. Their greatest advantage lies in their extremely fast prediction speed. When researchers need to evaluate a new combination of process parameters, they no longer need to spend hours or even days running CFD simulations; they can simply provide the new process parameters as input to the processing viscosity prediction model, which can then provide a prediction of the post-processing viscosity in a short time.

[0089] The fermented milk processing viscosity prediction method provided in this invention constructs a constitutive model of the target fermented milk and a computational fluid dynamics model of each process device, coupling them to obtain a full-process processing simulation model of the target fermented milk. This achieves a full-process digital mapping of the rheological properties of raw materials during processing, accurately predicting the impact of different equipment structural parameters and control parameters on the microstructure and macroscopic quality of the product. It can generate full-process processing data of the target fermented milk under different combinations of process parameters. By training a preset machine learning model with the full-process processing data, a processing viscosity prediction model of the target fermented milk is obtained. The processing viscosity prediction model is then used to quickly predict the viscosity of the target fermented milk after processing. This combines time-consuming physical simulation with efficient machine learning, ensuring not only the physical accuracy of the prediction but also enabling rapid iteration and optimization of the design scheme. Overall, it achieves accurate and rapid prediction of the viscosity of fermented milk products during processing, improving the R&D efficiency and product stability of fermented milk products, reducing R&D costs, and increasing the success rate of quantitative scale-up from pilot-scale to industrial production. It has significant technical advantages and economic value.

[0090] It should be noted that each embodiment of the present invention can be freely combined, rearranged, or executed individually, and does not need to rely on or depend on a fixed execution order.

[0091] In some embodiments, before obtaining the full-process processing data of the target fermented milk under different combinations of process parameters based on the full-process processing simulation model, the method further includes:

[0092] Determine at least one target combination of process parameters;

[0093] Obtain the actual viscosity value of the target fermented milk under at least one combination of target process parameters;

[0094] Determine the viscosity simulation value of the full-process simulation model under at least one combination of target process parameters;

[0095] The simulation model of the entire process is verified based on the comparison between the simulated viscosity value and the actual viscosity value under at least one combination of target process parameters.

[0096] If the comparison between the simulated viscosity value and the actual viscosity value is greater than the preset threshold, it is determined that the full-process processing simulation model has failed the verification, and the constitutive model and / or computational fluid dynamics model of fermented milk are corrected.

[0097] Specifically, the accuracy and reliability of the full-process simulation model can be verified, thereby providing a high-quality and reliable data source for the training of subsequent pre-set machine learning models. For example, the actual viscosity values ​​of fermented milk at the inlet and outlet of some key process equipment can be collected and compared with the viscosity simulation values ​​output by the full-process simulation model. Based on the comparison results, the constitutive model and / or computational fluid dynamics model of fermented milk can be corrected.

[0098] The target process parameter combination refers to the selected combination of process parameters used to verify the full-process processing simulation model. For example, it can be a combination of process parameters that will be run on actual physical equipment or that has already been run.

[0099] In a specific embodiment, one or more combinations of process parameters that are typical in actual production can be selected as the target process parameter combination.

[0100] In another specific embodiment, multiple combinations of target process parameters can be selected to represent different shearing conditions (e.g., low shearing, medium shearing, and high shearing) to more comprehensively verify the accuracy of the model under different conditions.

[0101] The actual viscosity value can be obtained through physical experiments. The specific procedure is as follows: In a laboratory, pilot plant, or production line, operate the corresponding process equipment strictly according to the determined combination of target process parameters to process the target fermented milk. After the processing is completed, take a sample from the final product, or take samples from the inlet and outlet of key process equipment during the processing, and immediately use a standard viscosity measuring instrument to measure its viscosity under preset measurement conditions. This measurement result is the actual viscosity value.

[0102] The same combination of target process parameters used to obtain the actual viscosity value is loaded into the full-process simulation model as input. Running the simulation model calculates the predicted viscosity of the fermented milk at the endpoint after flowing through the entire processing flow, or calculates the predicted viscosity of the fermented milk at the inlet and outlet of key process equipment. This result is the simulated viscosity value.

[0103] The simulated viscosity values ​​under the target process parameter combination are compared with the actual viscosity values. For example, the deviation between the simulated and actual viscosity values ​​can be calculated and compared with a preset threshold. This preset threshold is an acceptable upper limit of deviation, pre-set according to the different accuracy requirements of specific application scenarios. For example, for the final parameter determination of large-scale industrial production, a model error of less than 5% may be required; while for formula screening in the early R&D stage, an error of 10% may also be acceptable. Therefore, this preset threshold can be set according to actual needs, such as 5% or 10%. If the comparison result is less than or equal to the preset threshold, the full-process simulation model is considered to have passed the verification, and its accuracy meets the requirements; otherwise, the model is considered to have failed the verification.

[0104] If model validation fails, the full-process simulation model needs to be modified. This is because the full-process simulation model is obtained by coupling the fermented milk constitutive model and the computational fluid dynamics model. Deviations originate from the fermented milk constitutive model and / or the computational fluid dynamics model. Therefore, modifying the full-process simulation model essentially involves modifying both the fermented milk constitutive model and the computational fluid dynamics model.

[0105] To modify the constitutive model of fermented milk, the following measures can be taken: supplementing the data with rheological experimental data over a wider range (such as higher or lower shear rates, wider temperature ranges); changing or optimizing the mathematical form of the constitutive model; and refitting the parameters of the constitutive model using new data.

[0106] To correct computational fluid dynamics models, the following measures can be taken: check whether the three-dimensional geometric model is oversimplified or whether key structures that have a significant impact on the flow field are omitted; check whether the computational mesh is sufficiently fine in key regions and refine the mesh if necessary; check whether the process parameters are set realistically, such as whether the wall slip parameters and the heat transfer coefficient of the equipment are accurate.

[0107] After the correction is completed, the steps of determining the viscosity simulation value and verifying it need to be performed again to form an iterative closed loop of "simulation-experiment-comparison-correction" until the comparison result between the viscosity simulation value and the actual viscosity value meets the requirements of the preset threshold and the model is verified.

[0108] The fermented milk processing viscosity prediction method provided in this invention verifies the full-process processing simulation model by comparing the simulated viscosity value with the actual viscosity value under the target process parameter combination. This ensures the accuracy and reliability of the full-process processing simulation model, which serves as the basis for data generation, thereby providing a high-quality and reliable data source for the subsequent training of the preset machine learning model.

[0109] In some embodiments, before training a preset machine learning model based on full-process processing data to obtain a processing viscosity prediction model for the target fermented milk, the method further includes:

[0110] Principal component analysis was performed on the data from the entire processing flow to obtain key feature data.

[0111] Specifically, the data generated by the full-process simulation model is extremely large and complex. For each simulation case with a combination of process parameters, the output data may contain tens of thousands of variables, such as the shear rate, temperature, pressure, and residence time of each grid cell in the simulation model at multiple time steps. Directly using this raw, unprocessed high-dimensional data as input to a pre-defined machine learning model will require a massive amount of training samples for the model to learn effectively. Using high-dimensional data for training will significantly increase the consumption of computing resources and the time required for training.

[0112] Principal component analysis (PCA) is used to reconstruct a new set of lower-dimensional, linearly independent variables from original high-dimensional correlated variables through linear transformation. These new variables are called principal components. PCA can be used to analyze and identify the process parameters that have the most significant impact on the viscosity and overall rheological properties of the entire processing data, and these parameters can be used as key feature data.

[0113] In the key feature data, each feature (i.e., each principal component) is a linear combination of the original high-dimensional data features. It is no longer a single physical quantity, but a comprehensive representation of a certain change pattern throughout the entire processing.

[0114] The fermented milk processing viscosity prediction method provided in this invention uses principal component analysis to reduce the dimensionality of the entire processing data. This effectively reduces the dimensionality of the high-dimensional and redundant data, extracting the key feature data that has the most critical impact on the final viscosity. This not only greatly reduces the computational complexity and time cost of subsequent machine learning model training, but also filters out noise and redundant information in the data, reducing the risk of model overfitting.

[0115] In some embodiments, a preset machine learning model is trained based on data from the entire processing flow to obtain a processing viscosity prediction model for the target fermented milk, including:

[0116] Using key feature data as samples and the simulated viscosity of fermented milk corresponding to the key feature data as sample labels, a preset machine learning model is trained to obtain a processing viscosity prediction model.

[0117] The preset machine learning model includes at least one of artificial neural networks, process regression models, and support vector machines.

[0118] Specifically, key feature data is used as samples, and the simulated viscosity values ​​of fermented milk corresponding to the key feature data are used as sample labels to train a preset machine learning model, ultimately obtaining a processing viscosity prediction model.

[0119] The initial model for the preset machine learning model can be an artificial neural network, a process regression model, or a support vector machine.

[0120] Artificial neural networks possess powerful nonlinear fitting capabilities, enabling them to accurately learn the complex physical laws hidden behind data. Process regression models not only provide a point prediction value for viscosity (i.e., the predicted mean) but also a measure of uncertainty about that prediction value (i.e., the prediction variance or confidence interval). This is extremely valuable in practical applications. Support vector machines excel at handling small sample sizes and high-dimensional data, exhibiting excellent generalization ability and effectively avoiding model overfitting. In the scenario of this invention embodiment, even with a limited number of simulation cases, the support vector machine can learn a robust prediction model.

[0121] In practice, any of the above models can be used for training individually, or multiple models can be combined (e.g., through model ensemble) to achieve better prediction performance.

[0122] The fermented milk processing viscosity prediction method provided in this invention uses key feature data as samples to train a preset machine learning model. Through data dimensionality reduction, the training efficiency of the model is greatly improved and its generalization ability is enhanced. By selecting a high-performance machine learning model, it is ensured that the final processing viscosity prediction model can reproduce the results of complex physical simulations with extremely high accuracy. This makes the entire prediction method combine the accuracy of a physical model with the speed of a data-driven model.

[0123] In some embodiments, the constitutive model of fermented milk includes a structural dynamics model and a temperature-stress relationship model.

[0124] Specifically, the constitutive model of fermented milk includes a structural dynamics model and a temperature-stress relationship model. The structural dynamics model describes the effect of shear rate on the viscosity of fermented milk. The temperature-stress relationship model describes the effect of temperature on the viscosity of fermented milk.

[0125] In some embodiments, constructing a constitutive model of the target fermented milk includes:

[0126] A structural dynamics model is established based on the stress constitutive equation and the rate equation. The stress constitutive equation is used to describe the relationship between shear rate, structural parameters and shear stress. The rate equation is used to describe the relationship between structural parameters and shear rate.

[0127] Based on the relationship between chemical reaction rate and temperature, a temperature-stress relationship model is established; this model is used to describe the effect of temperature change on stress.

[0128] Based on the structural dynamics model and the temperature-stress relationship model, the constitutive model of fermented milk was determined.

[0129] Specifically, from the perspective of structural dynamics:

[0130] Viscosity loss is significantly affected by shearing. The underlying process is that shearing causes the microstructure of fermented milk to be destroyed. The destroyed microstructure is macroscopically manifested as changes in rheological properties such as viscosity.

[0131] Fermented milk is a thixotropic fluid. To characterize the transient structural changes of thixotropic fluids, scalar parameters are introduced. ( This is used to characterize the evolution of the internal structure of a fluid, that is: when the structure is intact, When the structure is completely destroyed, .

[0132] Fluids described based on the above scalar parameters can be called structural fluids. The constitutive equations of a structural fluid model consist of two parts: the first part is the stress constitutive equation, and the second part is the rate equation (i.e., the structural dynamics equation).

[0133] The stress constitutive equation applies instantaneous shear stress With structural parameters instantaneous value and shear rate Connecting them, the form is:

[0134] .

[0135] in, express Time-dependent shear stress; for The time-dependent scalar parameter of shear stress; express The shear rate over time; Indicates when the structure is completely destroyed ( viscosity, This indicates the contribution of structure to viscosity. The yield stress also depends on the degree of structuring. According to the Herschel-Bulkley model (a three-parameter non-Newtonian fluid constitutive equation), the corresponding structural fluid stress constitutive equation is expressed as:

[0136] .

[0137] in, The yield stress; The first viscosity coefficient represents the effect of structure on viscosity; The second viscosity coefficient represents the effect of complete structural failure on viscosity; This is the first power law exponent.

[0138] The rate equation introduces thixotropy through the derivatives of structural parameters, establishing a connection between shear rate, external stress, and discrete particle properties. Based on network correlation theory, it is assumed that the failure rate of the structure depends on the shear rate and the size of the flocs, while the recovery rate depends only on the size of the flocs. Based on this assumption, the rate equation can be expressed as:

[0139] 。

[0140] Among them, the first item The second term represents the structural rupture caused by shear rate. The third term represents orthokinetic structure generation caused by shear rate. This represents the peristaltic structure generation caused by Brownian motion. Indicates the structural damage impact coefficient. This represents the structural generation influence coefficient. Let represent the Brownian motion influence coefficient, which can be the reciprocal of the characteristic time of the Brownian motion. Introducing an additional time scale and a power-law exponent, the rate equation can be written as:

[0141] .

[0142] in, and There are two characteristic time scales. The time is the characteristic time of shear failure; For shear recovery feature time; The second power law exponent; It is the third power law exponent.

[0143] In summary, under transient conditions, assuming the structure is in a complete state for any initial moment, by simultaneously solving the stress constitutive equations and the rate equations, we can obtain the complete set of structural dynamic equations:

[0144] .

[0145] This set of structural dynamics equations constitutes a structural dynamics model. The stress constitutive equation describes the relationship between shear rate, structural parameters, and shear stress; the rate equation describes the relationship between structural parameters and shear rate.

[0146] From the perspective of temperature changes:

[0147] Considering the temperature dependence of fermented milk viscosity, an additional temperature change model is introduced. The temperature relationship is derived from the Arrhenius relation. The Arrhenius relation is a mathematical equation describing the relationship between chemical reaction rate and temperature. It is assumed that the Arrhenius relation acts on the stress equation, and is separate from the rate equation for structural parameters. For stress... The corresponding Arrhenius equation can be written as:

[0148] .

[0149] in, Activation energy Compared with ideal gas parameters The ratio can be considered a constant. . Initial temperature The corresponding stress. The equation is transformed into a linear relationship using the natural logarithm equation:

[0150] 。

[0151] The effect of temperature is only reflected in the stress constitutive equation, and it is proposed that... As a temperature characterization term:

[0152] .

[0153] in, It is a constant at a constant temperature.

[0154] The above formula can be used as a temperature-stress relationship model; the temperature-stress relationship model is used to describe the effect of temperature changes on stress.

[0155] Finally, the structural dynamics model and the temperature-stress relationship model are combined as the constitutive model of fermented milk.

[0156] In some embodiments, the constitutive model of fermented milk is parameter-fitted based on the rheological property test data of fermented milk samples with different formulations to obtain the structural dynamic parameters and temperature-affected parameters of fermented milk with different formulations, including:

[0157] Rheological tests were performed on fermented milk samples with different formulations to obtain rheological property test data;

[0158] Based on rheological property test data, the constitutive model of fermented milk was fitted with parameters to obtain the structural dynamic parameters and temperature-affected parameters of fermented milk with different formulations.

[0159] Specifically, samples of fermented milk with different formulations can be collected, and rheological tests can be performed on these samples to obtain rheological property data. Based on the rheological property data, parameter fitting can be performed on the constitutive model of the fermented milk to obtain the structural kinetic parameters and temperature-dependent parameters of the fermented milk with different formulations.

[0160] Taking beverage-grade yogurt as the target fermented milk, rheological tests were conducted on yogurt samples based on their rheological properties and processing characteristics. A rheometer was used on a parallel plate (PP25, 25mm in diameter). The test temperature range was 15–75℃, with each 10℃ increment serving as a gradient.

[0161] To ensure data reproducibility and minimize the impact of shear history, all tests employed a standardized experimental protocol that included a pre-shearing step and a shear rate of 100%. (Countdown seconds), lasting for 60 seconds; after pre-shearing, a 600-second dynamic time scan was performed at a frequency of 1 Hz and a strain of 1% to ensure the recovery of the sample structure.

[0162] Subsequently, fermented milk samples (yogurt) with different formulations were tested. The rheological tests included at least one of the following: steady-state shear test, structure formation test, shear failure test, and thixotropic ring test.

[0163] Steady-state shear test: at 1000 To 0.001 Steady-state shearing experiments were conducted within the range of shear rates: the changes in shear stress and viscosity were measured, and the test time for each measuring point (corresponding to a shear rate) was kept as long as possible.

[0164] Structural generation test: First, apply a large shear to destroy the original structure (shear rate 200). (120s), then the shear rate was controlled at 0.05 0.1 and 0.5 If the shear rate remains constant, the changes in shear stress and shear viscosity over time can be measured, i.e., a curve is obtained for each shear rate.

[0165] Shear failure test: controlling the shear rate (20 40 60 and 80 If the shear rate remains constant, the changes in shear stress and shear viscosity over time can be measured, i.e., a curve is obtained for each shear rate.

[0166] Thixotropic ring test: To evaluate the thixotropic properties of the sample, the thixotropic ring test method is first used. This test involves continuously increasing the shear rate from a rest state to a maximum value ( =200 The sample was then allowed to rest for 5 minutes (5 min) within the set scan time, while the response to shear stress was recorded. The area enclosed by the two curves in the thixotropic ring test represents the thixotropic level exhibited by the sample under these conditions. To further investigate the effect of the number of shear cycles on the thixotropic properties of the material, four consecutive thixotropic ring tests were performed without changing the sample.

[0167] According to the constitutive model of fermented milk in the above embodiments, it can be seen that:

[0168] The structural dynamics model has nine structural dynamic parameters, namely: , , , , , , , and The temperature-stress relationship model has two temperature-affected parameters, namely... and .

[0169] For the structural dynamics model, performing steady-state shearing, we have: According to the rate equation, we can obtain: By fitting steady-state shear data, the parameters in the stress constitutive equation are determined. , , , .

[0170] The rheological test data can be analyzed and fitted using one or more professional data analysis software such as MATLAB, Python, and R. The process is divided into two parts: fitting a complete structural dynamics model and fitting a temperature-stress relationship model. The adjustable parameter values ​​in the model parameters are determined by fitting the experimental data using the nonlinear least squares method.

[0171] For structural dynamics models:

[0172] Based on the characteristic that drinkable yogurt does not deform significantly during transient shearing, its structural equation can be simplified to the Herschel-Bulkley shear-thinning model: .

[0173] Through the above steps, the structural dynamic parameters of yogurt can be obtained by fitting, as shown in Table 1.

[0174] Table 1. Structural dynamic parameters of yogurt

[0175]

[0176] Figure 3 This is the steady-state fitting curve of yogurt at 25°C provided by the present invention, as shown below. Figure 3 As shown, black scatter dots represent the original data; brown solid lines represent the fitted data. Shear Stress and Shear Rate are shown in the figure. The fitting coefficient of determination and the fitting curves indicate that the model parameters fit well and converge. Due to differences in yogurt formulas, the constitutive model parameters differ for different yogurt formulas; the parameter range in the table above covers commonly available yogurts on the market.

[0177] For the temperature-stress relationship model:

[0178] Using the above method, the parameters of the temperature-stress relationship model are fitted to obtain... , and the coefficient of determination of fit As shown in Table 2.

[0179] Table 2 Temperature Influence Parameters

[0180]

[0181] The relationship between temperature and stress can be established by using a temperature-stress relationship model. Figure 4 The figure shows the relationship curves between stress and shear rate at different temperatures provided by this invention. In the figure, Shear Stress is the shear stress and Shear Rate is the shear rate. Figure 5 This invention provides fitting curves of the stress-temperature relationship for different types of yogurt at different temperatures. The subplots in the figure show the non-stress term parameters in the stress constitutive equation (for structural fluids: , and , for The graph shows the relationship between yield stress and temperature, where yield stress is the yield stress and temperature (K) is the temperature. Figure 4 and Figure 5 As shown, the non-stress parameters do not change significantly with temperature and can be considered independent of temperature. The effect of temperature on yogurt is only reflected in the change of stress, which is consistent with the assumptions made when building the model.

[0182] The fermented milk processing viscosity prediction method provided in this invention uses rheological property detection data of fermented milk samples to fit parameters of a fermented milk constitutive model, thereby obtaining the structural dynamic parameters and temperature influence parameters of the fermented milk. This results in a fermented milk constitutive model that can be specifically used for the target fermented milk, enabling quantitative characterization of the intrinsic quality differences of fermented milk and accurate prediction of the viscosity evolution law of the target fermented milk product.

[0183] Figure 6 This is the second schematic diagram of the process flow for predicting the viscosity of fermented milk processing provided by the present invention, as shown below. Figure 6 As shown, the method can include three parts: construction of a constitutive model of fermented milk, construction of a CFD model of the process equipment, process simulation, and viscosity prediction.

[0184] The constitutive model of fermented milk was constructed based on the structural dynamics model, the temperature-stress relationship model, and rheological test data.

[0185] The CFD modeling of process equipment can accurately reconstruct the three-dimensional structure of the equipment (including fermenters, impellers, valves, pipelines, pumps, plate heat exchangers, etc.) based on CAD drawings or 3D optical scanning data. Topology optimization is performed on complex components (such as static mixers and scraped heat exchangers) to retain key features (such as impeller tilt angles and clearance dimensions). Regular areas of the process equipment (cylindrical sections of the tank) are divided using structured hexahedral meshes, while complex areas (impellers, baffles) are divided using unstructured tetrahedral / polyhedral meshes, with boundary layer refinement.

[0186] Figure 7 This is a schematic diagram illustrating the process of constructing a full-process processing simulation model provided by the present invention, such as... Figure 7 As shown, this method utilizes the OpenFOAM (an open-source computational fluid dynamics software suite) platform to achieve deep function-level customization, enabling the construction of a full-process simulation model to simulate the rheological properties of complex materials. Specific steps include:

[0187] Step 1: Create a custom constitutive model in OpenFOAM. You can define a new class that inherits from an existing rheological model base class. Implement the structural dynamics equations by degenerating the parameters into a Herschel-Bulkley shear-thinning model and a temperature-stress relationship model.

[0188] Step 2: Modify the custom solver (such as simpleFoam or icoFoam, or other computational fluid dynamics solvers). Integrate the constitutive model into the solver loop, where fluid properties (such as viscosity) are used for the momentum transport equations. Ensure the solver passes shear rate and real-time temperature as inputs to the constitutive model.

[0189] Step 3: Set the process parameter range and determine the initial conditions.

[0190] The following example illustrates the setting of process parameter ranges, using the temperature range after demulsification and the initial viscosity of fermented milk as examples. The temperature range after demulsification can be set to 15℃-75℃. The initial viscosity range of fermented milk (refer to actual measured values) can be set to 10 mPas - 8000 mPas. mPas is millipasseconds.

[0191] The initial conditions were the actual temperature of the demulsification process and the initial viscosity of the fermented milk.

[0192] Step 4: Couple the constitutive model and the CFD model to construct a full-process processing simulation model. Utilize CFD to establish the relationship between shear rate, temperature, and time, achieving coupling between the shear field, temperature field, and constitutive model, thereby enabling simulation calculations of the fermented milk processing process.

[0193] Step 5: By detecting the viscosity data at the inlet and outlet of key equipment (such as demulsification, cooling, pumps, pipelines, pasteurization, filling, etc.) and comparing it with the model prediction results, the CFD model of the process equipment and the constitutive model of fermented milk can be corrected.

[0194] Taking a fermenter as an example, the structural parameters of the fermenter include: a tank volume of 3 cubic meters and a three-layer, two-blade impeller. The control parameters include: a rotation speed of 25 rpm, a torque of 573 N·m, a power of 1500 watts, and a stirring time of 1 minute.

[0195] Based on the simulation of the entire process, the simulated viscosity at the fermenter outlet is 2354 mPa·s, while the actual measured viscosity is 2335 mPa·s. A viscosity deviation ratio threshold can be set to measure whether the simulation result meets expectations. In this embodiment, the viscosity deviation ratio can be the ratio of the difference between the actual viscosity value and the simulated viscosity value to the actual viscosity value. A viscosity deviation ratio threshold of 5% is set; that is, if the viscosity deviation ratio is less than or equal to 5%, the entire process simulation model is considered to have met expectations; otherwise, it has not. In this embodiment, the deviation is 19 mPa·s, and the calculated viscosity deviation ratio is 0.8%, which is much less than the viscosity deviation ratio threshold. Therefore, the entire process simulation model can be considered to have met expectations, and the verification is successful. Otherwise, the constitutive model of the fermented milk or the CFD model of the fermenter can be corrected.

[0196] Figure 8 This is a schematic diagram of the process of training the processing viscosity prediction model provided by the present invention, as shown below. Figure 8 As shown, due to the long processing path of dairy products, using a full-process processing simulation model to simulate the fermented milk processing process and predict viscosity to obtain full-process processing data requires a huge amount of computation and high computing power.

[0197] This invention, through sensitivity analysis, determines that the process parameters (speed, torque, power, stirring time, tank volume, impeller type) for fermentation demulsification and constant-volume stirring before filling, the constitutive model parameters of fermented milk, the process parameters of the delivery pump (flow rate, rotor diameter, radial clearance between rotor and pump casing, pump displacement), the process parameters of pipeline bends (flow velocity of fermented milk in the pipeline, number of bends, pipeline inner diameter, radius of curvature of bends), the pasteurization intensity parameters of fermented milk after fermentation (flow rate, temperature, time), and the process parameters of the plate heat exchanger (pressure drop, cross-sectional area of ​​flow channel, number of flow channels, plate spacing) have the most significant impact on the viscosity and overall rheological properties of the processing. Furthermore, this embodiment of the invention uses Principal Component Analysis (PCA) to reduce the dimensionality of the parameter space by identifying the most important principal components, and determines the process parameters (power, stirring time, tank volume) for fermentation demulsification and final volume mixing in the tank, the constitutive model parameters of the fermented milk, the process parameters of the delivery pump (flow rate, rotor diameter), and the pasteurization intensity (flow rate, temperature, time) of the fermented milk after fermentation as key feature data. For thixotropic non-Newtonian fluid systems, key temporal features are extracted to capture the rheological behavior under different shear and temperature conditions, such as time constant, relaxation time, and yield stress.

[0198] This invention employs PCA technology to reduce the dimensionality of the data throughout the entire processing flow, which significantly reduces the number of variables that the simulation model needs to track in each iteration. Simultaneously, machine learning models such as Artificial Neural Networks (ANN), Gaussian Process Regression (GPR), or Support Vector Machines (SVM) are trained on key feature data to obtain an alternative model to the simulation model for predicting the viscosity of fermented milk during processing, without requiring a complete CFD calculation in each iteration. This reduces computation time while still providing accurate viscosity predictions.

[0199] By using the above methods, the complex multi-physics coupled full-process simulation model can be reduced to a reliable machine learning model, which is beneficial for rapid online viscosity prediction and optimized process design.

[0200] The apparatus provided in the embodiments of the present invention will be described below. The apparatus described below can be referred to in correspondence with the method described above.

[0201] Figure 9 This is a schematic diagram of the fermented milk processing viscosity prediction device provided by the present invention, as shown below. Figure 9 As shown, the device includes:

[0202] The first building module 910 is used to build a constitutive model of the target fermented milk; the constitutive model of the fermented milk is used to describe the effects of shear rate and temperature on the viscosity of the target fermented milk;

[0203] The second building module 920 is used to build computational fluid dynamics models for each process device; the computational fluid dynamics models are used to describe the effects of the combination of process parameters of the process device on the shear rate and / or temperature of the target fermented milk;

[0204] The model coupling module 930 is used to couple the constitutive model of fermented milk and the computational fluid dynamics model to obtain the full-process simulation model of the target fermented milk, and to obtain the full-process processing data of the target fermented milk under different combinations of process parameters based on the full-process simulation model.

[0205] The processing prediction module 940 is used to train a preset machine learning model based on the data of the entire processing process to obtain a processing viscosity prediction model for the target fermented milk, and to predict the viscosity of the target fermented milk after processing based on the processing viscosity prediction model.

[0206] The fermented milk processing viscosity prediction device provided in this invention constructs a constitutive model of the target fermented milk and a computational fluid dynamics model of each process device, coupling them to obtain a full-process processing simulation model of the target fermented milk. This achieves a full-process digital mapping of the rheological properties of raw materials during processing, accurately predicting the impact of different equipment structural parameters and control parameters on the microstructure and macroscopic quality of the product. It can generate full-process processing data of the target fermented milk under different combinations of process parameters. By training a preset machine learning model with the full-process processing data, a processing viscosity prediction model of the target fermented milk is obtained. The processing viscosity prediction model is then used to quickly predict the viscosity of the target fermented milk after processing. This combines time-consuming physical simulation with efficient machine learning, ensuring not only the physical accuracy of the prediction but also enabling rapid iteration and optimization of the design scheme. Overall, it achieves accurate and rapid prediction of the viscosity of fermented milk products during processing, improving the R&D efficiency and product stability of fermented milk products, reducing R&D costs, and increasing the success rate of quantitative scale-up from pilot-scale to industrial production. It has significant technical advantages and economic value.

[0207] Figure 10 This is a schematic diagram of the structure of the electronic device provided by the present invention, such as... Figure 10 As shown, the electronic device may include: a processor 1010, a communications interface 1020, a memory 1030, and a communications bus 1040, wherein the processor, communications interface, and memory communicate with each other via the communications bus. The processor can invoke logical commands stored in the memory to execute the methods described in the above embodiments, for example:

[0208] A constitutive model of the target fermented milk is constructed. This model describes the effects of shear rate and temperature on the viscosity of the target fermented milk. Computational fluid dynamics (CFD) models of each process device are constructed. These models describe the effects of combinations of process parameters on the shear rate and / or temperature of the target fermented milk. The constitutive model and the CFD models are coupled to obtain a full-process simulation model of the target fermented milk. Based on this model, full-process processing data of the target fermented milk under different combinations of process parameters are obtained. A pre-set machine learning model is trained using the full-process processing data to obtain a processing viscosity prediction model for the target fermented milk. The viscosity of the processed target fermented milk is then predicted based on this model.

[0209] Furthermore, when the logical commands in the aforementioned memory can be implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a 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 commands 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 this 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.

[0210] The processor in the electronic device provided in this embodiment of the invention can call logical instructions in the memory to implement the above method. Its specific implementation method is the same as the aforementioned method implementation method and can achieve the same beneficial effects, which will not be repeated here.

[0211] This invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to perform the methods provided in the above embodiments.

[0212] The specific implementation method is the same as the aforementioned method implementation method and can achieve the same beneficial effects, so it will not be repeated here.

[0213] This invention provides a computer program product, including a computer program that, when executed by a processor, implements the method described above.

[0214] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0215] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0216] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for predicting the viscosity of a fermented milk process, characterized in that, include: A constitutive model of the target fermented milk is constructed; the constitutive model is used to describe the effects of shear rate and temperature on the viscosity of the target fermented milk. Computational fluid dynamics models of each process device are constructed; the computational fluid dynamics models are used to describe the effects of the combination of process parameters of the process device on the shear rate and temperature of the target fermented milk; By coupling the constitutive model of the fermented milk and the computational fluid dynamics model, a full-process processing simulation model of the target fermented milk is obtained, and based on the full-process processing simulation model, full-process processing data of the target fermented milk under different combinations of process parameters are obtained; The preset machine learning model is trained based on the data from the entire processing process to obtain a processing viscosity prediction model for the target fermented milk, and the viscosity of the target fermented milk after processing is predicted based on the processing viscosity prediction model; the preset machine learning model includes at least one of artificial neural networks, process regression models, and support vector machines; The constitutive model of fermented milk includes a structural dynamics model and a temperature-stress relationship model; The construction of the constitutive model of the target fermented milk includes: Based on the stress constitutive equation and the rate equation, a structural dynamics model is established; the stress constitutive equation is used to describe the relationship between shear rate, structural parameters, and shear stress; the rate equation is used to describe the relationship between structural parameters and shear rate. Based on the relationship between chemical reaction rate and temperature, a temperature-stress relationship model is established; this model is used to describe the effect of temperature change on stress. Based on the structural dynamics model and the temperature-stress relationship model, the constitutive model of the fermented milk is determined; The stress constitutive equation is: ； The rate equation is: ； The temperature-stress relationship model is as follows: ； For stress; Scalar parameters characterizing the evolution of the internal structure of a fluid; Shear rate; The yield stress; The first viscosity coefficient represents the effect of structure on viscosity; The second viscosity coefficient represents the effect of complete structural failure on viscosity; The first power law exponent; For time; The coefficient representing the influence of Brownian motion; The time is the characteristic time of shear failure; The second power law exponent; For shear recovery feature time; The third power law exponent; It is a constant at a constant temperature.

2. The method of predicting viscosity of a fermented milk process according to claim 1, characterized in that, Before obtaining the full-process processing data of the target fermented milk under different combinations of process parameters based on the full-process processing simulation model, the method further includes: Determine at least one target combination of process parameters; Obtain the actual viscosity value of the target fermented milk under the at least one combination of target process parameters; Determine the viscosity simulation value of the full-process simulation model under the at least one combination of target process parameters; The simulation model of the entire process is verified based on the comparison between the simulated viscosity value and the actual viscosity value under the at least one combination of target process parameters. If the comparison between the simulated viscosity value and the actual viscosity value is greater than a preset threshold, it is determined that the full-process processing simulation model has failed verification, and the constitutive model of fermented milk and / or the computational fluid dynamics model are corrected.

3. The method of predicting viscosity of a fermented milk process according to claim 1, wherein Before training a preset machine learning model based on the entire processing data to obtain a processing viscosity prediction model for the target fermented milk, the method further includes: Principal component analysis was performed on the entire process data to obtain key feature data.

4. The method of predicting viscosity of a fermented milk process according to claim 3, characterized in that, The step of training a preset machine learning model based on the entire processing data to obtain a processing viscosity prediction model for the target fermented milk includes: Using the key feature data as samples and the simulated viscosity of fermented milk corresponding to the key feature data as sample labels, the preset machine learning model is trained to obtain the processing viscosity prediction model.

5. The method of predicting viscosity of a fermented milk process according to any one of claims 1 to 4, characterized in that, The process equipment includes at least one of a fermenter, a transfer pump, pipelines, and a heat exchanger; the process parameter combination includes the control parameters and / or structural parameters of the process equipment.

6. A fermented milk processing viscosity prediction device characterized by comprising: include: The first construction module is used to construct a constitutive model of the target fermented milk; the constitutive model of the fermented milk is used to describe the effects of shear rate and temperature on the viscosity of the target fermented milk; The second construction module is used to construct computational fluid dynamics models for each process device; the computational fluid dynamics models are used to describe the influence of the combination of process parameters of the process device on the shear rate and temperature of the target fermented milk; The model coupling module is used to couple the constitutive model of the fermented milk and the computational fluid dynamics model to obtain the full-process simulation model of the target fermented milk, and to obtain the full-process processing data of the target fermented milk under different combinations of process parameters based on the full-process simulation model. The processing prediction module is used to train a preset machine learning model based on the full-process processing data to obtain a processing viscosity prediction model for the target fermented milk, and to predict the viscosity of the target fermented milk after processing based on the processing viscosity prediction model; the preset machine learning model includes at least one of artificial neural networks, process regression models, and support vector machines. The constitutive model of fermented milk includes a structural dynamics model and a temperature-stress relationship model; The construction of the constitutive model of the target fermented milk includes: Based on the stress constitutive equation and the rate equation, a structural dynamics model is established; the stress constitutive equation is used to describe the relationship between shear rate, structural parameters, and shear stress; the rate equation is used to describe the relationship between structural parameters and shear rate. Based on the relationship between chemical reaction rate and temperature, a temperature-stress relationship model is established; this model is used to describe the effect of temperature change on stress. Based on the structural dynamics model and the temperature-stress relationship model, the constitutive model of the fermented milk is determined; The stress constitutive equation is: ； The rate equation is: ； The temperature-stress relationship model is as follows: ； For stress; Scalar parameters characterizing the evolution of the internal structure of a fluid; Shear rate; The yield stress; The first viscosity coefficient represents the effect of structure on viscosity; The second viscosity coefficient represents the effect of complete structural failure on viscosity; The first power law exponent; For time; The coefficient representing the influence of Brownian motion; The time is the characteristic time of shear failure; The second power law exponent; For shear recovery feature time; The third power law exponent; It is a constant at a constant temperature.

7. An electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the fermented milk processing viscosity prediction method according to any one of claims 1 to 5.

8. A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, When the computer program is executed by the processor, it implements the fermented milk processing viscosity prediction method according to any one of claims 1 to 5.

Images