Liquid milk UHT cleaning model construction method based on multiple monitoring parameters

Abstract

The invention discloses a liquid milk UHT cleaning model construction method based on multiple monitoring parameters, and the method comprises the steps: setting cleaning parameters, carrying out the multi-parameter monitoring through a plurality of monitoring probes in a CIP cleaning process, and carrying out the sampling at preset time intervals, so as to measure the protein concentration; by adopting a response surface analysis method, analyzing and fitting monitoring data, sampling measurement values and cleaning parameters of each probe to obtain an equation model of the cleaning time to the cleaning temperature, the cleaning flow speed and the cleaning liquid concentration; a symbolic regression method is adopted, and an equation model of the cleaning speed to the cleaning time, the cleaning temperature, the cleaning flow speed and the cleaning liquid concentration is obtained by analyzing and fitting the monitoring data, the sampling measurement values and the cleaning parameters of all the probes. The cleaning time equation and the cleaning rate equation adopt multiple parameters to carry out modeling research on the cleaning process, and reference is provided for optimization of the pipeline cleaning process in the dairy industry.

Description

Technical Field

[0001] This invention relates to the field of dairy equipment cleaning technology, specifically to a method for constructing a liquid milk UHT cleaning model based on multiple monitoring parameters. Background Technology

[0002] Ultra-high temperature instantaneous sterilization (UHT) is a core technology in the dairy processing industry, and it has achieved over 90% large-scale application in the global dairy industry. However, the high temperature and pressure of the UHT process, coupled with the complex components of dairy products (milk proteins, fats, lactose, minerals, etc.), make it extremely easy for scale to form on the inner walls of pipelines. Scale formation has become a major bottleneck restricting the efficient operation of UHT dairy production lines. To solve the scale problem, dairy companies generally use Cleaning In Place (CIP), a multi-step process involving the sequential delivery of hot water, alkali solution, clean water, acid solution, and clean water to remove scale. However, existing CIP cleaning solutions are mostly based on "empirical parameters," which have significant blind spots and limitations. For example, the cleaning parameters are fixed and are not dynamically adjusted according to the actual amount of scale, scale composition and pipeline operating conditions, resulting in "under-cleaning" or "over-cleaning." The evaluation of cleaning effect is lagging behind, mainly relying on visual inspection or sampling of the inner wall of the pipeline after cleaning, and cannot provide real-time feedback on the scale removal rate during the cleaning process, making it difficult to achieve closed-loop optimization of the cleaning process.

[0003] The dairy industry has high requirements for microbial content, and disassembling pipelines can easily lead to microbial infection. Therefore, there is an urgent need for online in-situ technologies and methods for predicting and evaluating scaling and CIP cleaning in UHT production pipelines. In existing technologies, scaling prediction and CIP cleaning optimization are mostly independent modules, failing to form an integrated closed-loop system of "parameter acquisition - scaling prediction - cleaning scheme generation - cleaning process monitoring - model correction." This results in a disconnect between predicted results and actual cleaning needs, hindering intelligent control of the production line. As the dairy industry transforms towards "high efficiency, energy saving, safety, and intelligence," there is an urgent need to develop integrated devices and technologies that integrate multi-source parameter monitoring, accurate scaling status prediction, dynamic optimization of cleaning schemes, and real-time monitoring of the cleaning process. This would address the core technological bottlenecks in existing UHT process pipeline scaling and CIP cleaning, improving the operating efficiency and economy of the production line. Summary of the Invention

[0004] This application provides a method for constructing a liquid milk UHT cleaning model based on multiple monitoring parameters, in order to solve the technical problems of blindness and limitations in existing UHT process pipeline scaling and CIP cleaning.

[0005] One embodiment of the present invention provides a method for constructing a liquid milk UHT cleaning model based on multiple monitoring parameters, the method comprising: Set cleaning parameters and use the CIP cleaning method to clean the liquid milk UHT production equipment. The cleaning parameters include cleaning temperature, cleaning flow rate, and cleaning solution concentration. During the CIP cleaning process, multiple monitoring probes are used to monitor multiple parameters, and samples are taken at preset time intervals to measure protein concentration. Response surface methodology was employed to analyze and fit the monitoring data from each probe, sampled measurements, and cleaning parameters to obtain an equation model for the relationship between cleaning time and cleaning temperature, cleaning flow rate, and cleaning solution concentration. Specifically, this involved fitting an equation with cleaning time as the dependent variable and cleaning temperature, cleaning flow rate, and cleaning solution concentration as independent variables. The Box-Behnken experimental design method was used to design the experimental scheme, and multiple regression analysis was employed to establish the mapping function between the independent and response variables. Symbolic regression was used to analyze and fit the monitoring data of each probe, the sampled measurement values, and the cleaning parameters to obtain an equation model of the cleaning rate on the cleaning time, cleaning temperature, cleaning flow rate, and cleaning solution concentration. Specifically, the time difference between the probe monitoring data and the sampling measurement time difference are the same. The difference between the probe monitoring data or the sampling measurement value in adjacent time intervals is used as the characterization value of the cleaning rate and as the dependent variable. The cleaning time, cleaning temperature, cleaning flow rate, and cleaning solution concentration are used as independent variables to obtain the equation of the cleaning rate on the cleaning time, cleaning temperature, cleaning flow rate, and cleaning solution concentration.

[0006] Furthermore, during the CIP cleaning process, multiple monitoring probes are used for multi-parameter monitoring, specifically including: A flow meter is used to monitor milk flow rate; a pressure gauge is used to monitor pipeline pressure; a thermometer is used to monitor fluid temperature; and turbidity probes, pH probes, and conductivity probes integrated into the water quality analysis device are used to monitor turbidity, pH, and conductivity values, respectively.

[0007] Furthermore, the CIP cleaning specifically includes: After the dairy products are sterilized, a CIP cleaning process is performed, in the following order: a) water wash; b) alkaline wash; c) water wash; d) acid wash; e) water wash. During the alkaline and acid washing processes, a preset volume of liquid flowing back into the cleaning tank is taken at preset time intervals to measure the protein concentration in the solution.

[0008] Furthermore, the CIP cleaning specifically includes: The cleaning parameters for each cleaning process are as follows: water washing uses room temperature and a flow rate of 50-300 L / h; alkaline washing uses a temperature of 50-150℃, a cleaning flow rate of 50-300 L / h, and a cleaning solution concentration of 0.5-5% by mass; acid washing uses a temperature of 50-120℃, a cleaning flow rate of 50-300 L / h, and a cleaning solution concentration of 0.5-5% by mass.

[0009] Furthermore, using cleaning time as the dependent variable and cleaning temperature, cleaning flow rate, and cleaning solution concentration as independent variables, equations were fitted to obtain the relationship between cleaning time and cleaning temperature, cleaning flow rate, and cleaning solution concentration, specifically including: Based on the turbidity, pH, and conductivity values ​​monitored by the probe, as well as the protein concentration values ​​measured by the sample, a curve showing the change of data over time is plotted. When the data tends to stabilize, it indicates that the cleaning is complete, and the cleaning time is determined accordingly.

[0010] Furthermore, multiple regression analysis is used to establish a mapping function between independent variables and response variables, specifically including: The mapping function is as follows: ; In the formula, Y: the response variable, i.e., the dependent variable; β0: a constant; β i β ii β ij : These represent the linear, quadratic, and second-order terms of the model, respectively, X i X j : Independent variable; When the system reaches its optimal operating condition, the mapping function converges to a steady state, at which point the interaction between the independent variables reaches a dynamic equilibrium.

[0011] Furthermore, response surface methodology was employed to analyze and fit the monitoring data from each probe, the sampled measurements, and the cleaning parameters, resulting in an equation model for the relationship between cleaning time and cleaning temperature, cleaning flow rate, and cleaning solution concentration. This model specifically includes: The polynomial equation model obtained through modeling analysis is used with the coefficient of determination R. 2 The evaluation was conducted using the degree of misfit and the p-values ​​of significance for each factor.

[0012] Furthermore, using symbolic regression, by analyzing and fitting the monitoring data from each probe, sampled measurements, and cleaning parameters, an equation model was obtained for the cleaning rate on the basis of cleaning time, cleaning temperature, cleaning flow rate, and cleaning solution concentration. Specifically, this includes: Symbolic regression fitting was performed using Eureqa software, and the evaluation metrics included mean absolute error (MAE), mean squared error (MSE), and root mean squared error (RMSE).

[0013] Furthermore, the functional relationship between the cleaning rate and the cleaning time, cleaning temperature, cleaning flow rate, and cleaning solution concentration is expressed as follows: ; In the formula, y is the cleaning rate characterized by the difference between pH, conductivity, turbidity or protein concentration in adjacent monitoring time intervals, t is the cleaning time, T is the cleaning temperature, v is the cleaning flow rate, and c is the cleaning solution concentration.

[0014] This application provides a method for constructing a liquid milk UHT cleaning model based on multiple monitoring parameters. The method specifically includes: setting cleaning parameters; during the CIP cleaning process, using multiple monitoring probes to monitor multiple parameters and taking samples at preset time intervals to measure protein concentration; the cleaning parameters include cleaning temperature, cleaning flow rate, and cleaning solution concentration; using response surface methodology, by analyzing and fitting the monitoring data from each probe, the sampled measurements, and the cleaning parameters, obtaining an equation model of cleaning time versus cleaning temperature, cleaning flow rate, and cleaning solution concentration; and using symbolic regression, by analyzing and fitting the monitoring data from each probe, the sampled measurements, and the cleaning parameters, obtaining an equation model of cleaning rate versus cleaning time, cleaning temperature, cleaning flow rate, and cleaning solution concentration. In this invention, the cleaning time equation and cleaning rate equation use multiple parameters to model the cleaning process, providing a reference for optimizing pipeline cleaning processes in the dairy industry. Attached Figure Description

[0015] Figure 1 A flowchart illustrating a method for constructing a liquid milk UHT cleaning model based on multiple monitoring parameters, provided in one embodiment of the present invention; Figure 2 A schematic diagram of the flow meter installation position in a liquid milk UHT cleaning model construction method based on multiple monitoring parameters provided in an embodiment of the present invention; Figure 3 A schematic diagram of the pressure gauge installation position in a liquid milk UHT cleaning model construction method based on multiple monitoring parameters provided in an embodiment of the present invention; Figure 4 A schematic diagram of the thermometer installation position in a liquid milk UHT cleaning model construction method based on multiple monitoring parameters provided in an embodiment of the present invention; Figure 5 A schematic diagram of the installation location of the water quality analysis device in a method for constructing a liquid milk UHT cleaning model based on multiple monitoring parameters, provided in an embodiment of the present invention; Figure 6 A diagram of alkaline washing turbidity data in a liquid milk UHT cleaning model construction method based on multiple monitoring parameters provided in an embodiment of the present invention; Figure 7The diagram shows the conductivity data of alkaline washing in a liquid milk UHT cleaning model construction method based on multiple monitoring parameters, provided in one embodiment of the present invention. Detailed Implementation

[0016] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings. Similar elements in different embodiments are referred to by associated similar element reference numerals. In the following embodiments, many details are described to facilitate a better understanding of this application. However, those skilled in the art will readily recognize that some features may be omitted in different situations, or may be replaced by other elements, materials, or methods. In some cases, certain operations related to this application are not shown or described in the specification. This is to avoid obscuring the core parts of this application with excessive description. For those skilled in the art, detailed description of these related operations is not necessary; they can fully understand the related operations based on the description in the specification and general technical knowledge in the art.

[0017] Furthermore, the features, operations, or characteristics described in the specification can be combined in any suitable manner to form various embodiments. At the same time, the steps or actions in the method description can be rearranged or adjusted in a manner obvious to those skilled in the art. Therefore, the various orders in the specification and drawings are only for the clear description of a particular embodiment and do not imply a necessary order, unless otherwise stated that a particular order must be followed.

[0018] The cleaning process involves multiple factors. The classic Sinner ring identifies four key influencing factors for CIP cleaning: time, temperature, chemical action, and mechanical force—namely, cleaning time, cleaning temperature, cleaning agent concentration, and cleaning agent flow rate. Monitoring the cleaning process and examining the impact of each factor on the cleaning effect yields a cleaning model, which can then be used to evaluate and predict the cleaning process. Modeling can be achieved using statistical optimization methods and machine learning algorithms. Response Surface Methodology (RSM) constructs a full factorial model by designing a finite number of experiments. It quantitatively describes the nonlinear relationship between independent variables and response values, thereby analyzing the dynamic impact of multivariate interactions on the system. Symbolic regression has achieved significant breakthroughs in engineering modeling and the derivation of physical laws. As a supervised machine learning paradigm, symbolic regression searches the symbolic space to obtain explicit expressions that can resolve the underlying mathematical relationship between feature variables and target variables. It is not only suitable for modeling the quantitative relationship between features and labels but can also reveal dimensionless correlations in complex physical phenomena by constructing nonlinear mappings between dimensionless number groups.

[0019] The first embodiment of this invention provides a method for constructing a liquid milk UHT cleaning model based on multiple monitoring parameters. The following is in conjunction with... Figure 1 Please provide a detailed explanation.

[0020] like Figure 1 As shown, in step S100, cleaning parameters are set. During the CIP cleaning process, multiple monitoring probes are used to monitor multiple parameters, and samples are taken at preset time intervals to measure protein concentration.

[0021] This embodiment is based on a UHT sterilization equipment for dairy products, the basic structure of which is detailed in patent (ZL202422712251.4). Multiple probes are used to monitor relevant data. Temperature, pressure, and flow probes are used to determine and adjust pipeline fluid conditions and operating status. Turbidity, pH, and conductivity probes measure fluid data for monitoring and analyzing the cleaning process, and for subsequent model building.

[0022] The following describes the probe placement: A flow meter is used to determine the milk flow rate and is installed between the pump and the sterilization pipeline. Valves 1 and 2 are used to discharge liquid from the tank, and valve 3 controls the liquid flow between the storage tank and the cleaning tank. Figure 2 As shown.

[0023] The pressure gauge is used to determine pipeline pressure. Scaling in the pipeline reduces pressure, and the pressure gauge indicates whether the pipeline is operating normally. Valve 4 is a pressure regulating valve, adjusting the pipeline pressure and pump flow. Valve 5 is a three-way valve, controlling whether the outlet liquid enters the cleaning tank or water quality analysis device and the flow rate of the corresponding pipeline. Valve 6 controls whether the liquid enters the storage tank. Figure 3 As shown.

[0024] Thermometers determine the temperature of each fluid component to verify the proper functioning of the heat exchange and sterilization processes. All bends in the pipes are flanged connections; detachable bends serve as experimental verification piping. Figure 4 As shown.

[0025] Turbidity, pH, and conductivity probes are integrated into the water quality analysis device to measure relevant fluid data for subsequent analysis of the cleaning process and the establishment of a cleaning model. Valve 7 is a drain valve to discharge residual liquid from the water quality analysis device. Valves 8 and 9 regulate the flow rate into and out of the liquid storage tank of the water quality analysis device. Data is recorded once at preset time intervals, and the corresponding data can be saved on the data panel. Figure 5 As shown.

[0026] like Figure 1As shown, in step S200, response surface methodology is used to analyze and fit the monitoring data of each probe, the sampled measurement values ​​and the cleaning parameters to obtain the equation model of cleaning time on cleaning temperature, cleaning flow rate and cleaning fluid concentration.

[0027] Specifically, the cleaning time equation is as follows: Response surface methodology is used to analyze and fit the monitoring data from each probe, the sampled measurement values, and various cleaning parameters. The probe data (turbidity value, pH value, conductivity value) and the sampled measurement values ​​(protein concentration) are used to determine the cleaning time (the time when the probe data or sampled data no longer increases or decreases and tends to stabilize) as dependent variables. The cleaning parameters (cleaning temperature, cleaning flow rate, cleaning solution concentration, with the cleaning solution concentration expressed as hydroxide and hydrogen ion concentrations) are used as independent variables, resulting in an equation for cleaning time versus cleaning temperature, cleaning flow rate, and cleaning solution concentration.

[0028] like Figure 1 As shown, in step S300, the symbolic regression method is used to analyze and fit the monitoring data of each probe, the sampled measurement values ​​and the cleaning parameters to obtain the equation model of the cleaning rate on the cleaning time, cleaning temperature, cleaning flow rate and cleaning solution concentration.

[0029] Specifically, the cleaning rate equation is calculated using symbolic regression. Based on the monitoring data from each probe, the sampled measurements, and various cleaning parameters, an analysis and fitting process is performed. The time difference between the probe monitoring data and the sampling measurement time difference are the same. The difference between adjacent time intervals of the probe monitoring data or sampled measurements is used as the dependent variable, with time and various cleaning parameters as independent variables. This yields an equation for the cleaning rate against cleaning time, cleaning temperature, cleaning flow rate, and cleaning solution concentration. For example, the difference between adjacent time intervals of the probe monitoring parameter values ​​or sampling parameter values ​​is used to calculate the cleaning rate at 5 minutes as the conductivity value at 5 minutes minus the conductivity value at 4.5 minutes. Here, 0.5 minutes represents the adjacent time interval.

[0030] Specific application examples are as follows: I. Cleaning Time Equation 1. Experimental steps: A small-scale UHT device was used to simulate the liquid milk sterilization process, and model milk residue was prepared under fixed sterilization conditions.

[0031] Simulate the pipeline cleaning process. After dairy product sterilization, a CIP cleaning process is performed, in the following order: water wash (to remove residual dairy products from the pipeline), alkaline wash (to remove organic dirt from the pipeline), water wash (to remove residual alkaline solution from the pipeline), acid wash (to remove inorganic dirt from the pipeline), and water wash (to remove residual acid solution from the pipeline).

[0032] The detailed steps are as follows: Water washing: towards Figure 2Add a certain amount of water to the storage tank, close valve 1, adjust valve 3 to allow the water in the storage tank to flow through the pump; close... Figure 3 Adjust valve 6 and valve 5 to direct the liquid flow to the cleaning tank, then open valve 6. Figure 2 Valve 2; set temperature, flow rate, and time, and wash with water until finished; after finishing, open valve 1 to drain the liquid from the storage tank.

[0033] Alkaline washing: Rinse Figure 1 After the cleaning tank is clean, close valve 2, adjust valve 3 so that the cleaning tank flows through the pump, and add alkaline solution to the cleaning tank; adjust... Figure 3 Valve 6 directs the liquid flow to the water quality analysis device; closing... Figure 5 Open valve 7 and valve 8; set temperature, flow rate, and time; alkaline washing begins; adjust... Figure 5 Valve 9 controls the liquid level in the water quality analysis device; alkaline washing continues until completion; after completion, it is opened. Figure 2 Use valve 2 to rinse and clean the tank until it is clean.

[0034] Water wash: Close Figure 2 Use valve 1 to add a certain amount of water to the storage tank, and adjust valve 3 to allow the water in the storage tank to flow through the pump; set the temperature, flow rate, and time, and wash until the end; after the end, open valve 1 to drain the liquid from the storage tank.

[0035] Pickling: The process is the same as alkaline washing, except that the alkaline solution is replaced with an acid solution, and the corresponding parameters are changed. Washing: Same as the previous washing procedure.

[0036] Cleaning complete. The relevant cleaning parameters are shown in Table 1.

[0037] During the alkaline and acid washing processes, a predetermined volume of liquid flowing back into the washing tank is taken at preset time intervals to measure the protein concentration in the solution for subsequent analysis.

[0038] Table 1 CIP Cleaning Process Table

[0039] 2. The fitting process for the cleaning time equation is as follows: 1) Data collection: pH, conductivity, and turbidity curves were obtained from a water quality analysis device; protein concentration curves were obtained from sampling measurements.

[0040] 2) Determination of the cleaning endpoint: Plot the pH, conductivity, turbidity, and protein concentration curves. Once the data points stabilize, the cleaning process can be considered complete, and the corresponding time is recorded as the cleaning time.

[0041] 3) Response surface analysis: Key data were obtained through a Box-Behnken Design (BBD) experimental scheme. A mapping function between the variable space and the response value was established using multiple regression analysis. A visualization method combining three-dimensional response surface and two-dimensional contour line analysis was used to achieve quantitative analysis of the synergistic effect of multiple parameters. When the system reaches the optimal operating condition, the response function can converge to the steady state represented by equation (1), at which point the interaction between the variables reaches dynamic equilibrium.

[0042] The design is shown in Table 2 and Formula (1):

[0043] In the formula, Y: the response variable, i.e., the dependent variable; β0: a constant; β i β ii β ij The linear, quadratic, and second-order terms of the model, X i X j Independent variable. A multinomial model is obtained through modeling analysis and prediction, and the coefficient of determination (R²) is used. 2 The model was evaluated using factors such as the degree of misfit and the significance of each factor. The smaller the p-value in the experimental results, the more accurate the fitted model was.

[0044] Table 2. Response surface equation design table (specific parameters differ for acid and base).

[0045] Based on the data, the corresponding times for each experiment were obtained for pH, conductivity, turbidity, and protein concentration, and equation (2) was obtained by fitting the data.

[0046] 3. Implementation Case (Alkali Washing Process, Time Equation Using Turbidity Meter): See data chart Figure 6 The timetable is shown in Table 3 below: Table 3. Statistical Table of Turbidity Parameters and Cleaning Time in Alkaline Washing Process

[0047] Input the data into the response surface analysis software Design-Expert to calculate the response equations, with the following conditions set: Temperature: Minimum 110 ℃, maximum 130 ℃; Flow rate: minimum 110 L / h, maximum 130 L / h; Concentration: minimum 1.5%, maximum 2.5%.

[0048] The resulting equation is as follows: Y=-123.0000+2.1813×A+1.2938×B-0.5000×C+0.0063×A×B+0.0500×A×C-1.3198×10 -17 ×B×C-0.0144×A²-0.0094×B²-3.2500×C² In the formula, Y is the cleaning time, A is the temperature (°C), B is the flow rate (L / h), C is the concentration (% w / w), the adjusted determination coefficient (Adjusted R²) is 0.9921, and R² is 0.9965.

[0049] II. Cleaning Rate Equation 1. The detailed experimental steps are the same as those for the cleaning time equation.

[0050] 2. Fitting the cleaning time equation: Data collection and determination of the cleaning endpoint are the same as those described above for the cleaning time equation.

[0051] Symbolic Regression Analysis: This embodiment uses Eureqa software for symbolic regression fitting. Eureqa data analysis software employs automated modeling technology, using intelligent algorithms to find the optimal mathematical expression from the experimental data. This software can automatically filter thousands of possible equation forms, and by comparing the fitting accuracy and structural complexity of different equations, it gradually optimizes to obtain the best mathematical model that accurately reflects the data patterns while possessing a concise form. This modeling method effectively avoids the subjective limitations of traditional manual modeling. The evaluation metrics are MAE (mean absolute error), MSE (mean squared error), and RMSE (root mean squared error). The fitted data is shown in Table 4.

[0052] Table 4. Symbolic Regression Fitting Rate Equation Table

[0053] The fitting yields equation (2):

[0054] Where y is the difference between pH, conductivity, turbidity, and protein concentration in adjacent monitoring time intervals, and t, T, v, and c represent time, temperature, flow rate, and concentration, respectively. The corresponding values ​​are determined by the above-mentioned parameters, and the rate equation is finally obtained by fitting.

[0055] 3. Implementation Case (Alkali Washing Process, Rate Equation in Conductivity).

[0056] See data chart Figure 7 The timetable is shown in Table 5 below: Table 5. Statistics of Cleaning Parameters and Time in Alkaline Washing Process

[0057] The difference between adjacent time intervals of the corresponding conductivity monitoring data was used as the characterization value for the cleaning rate for fitting. The data was input into Eureqa software for symbolic regression prediction, and the fitted equation for the cleaning rate was obtained as follows: (Cleaning rate) = 0.0155501079700582 × (concentration) + -1.57582564901858 × (concentration) / (temperature) + -4.9252544512714 / (-4.37228134071692 - (time)^2) In the formula, T is temperature (°C), t is time (min), c is concentration (% w / w), v is flow rate (L / h), and R is... 2 The correlation coefficient is 0.96674226, the maximum error is 0.14769293, the mean squared error (MSE) is 0.0002138842, and the mean squared error (MAE) is 0.0095770774. The fitted data may not include all parameters and should be determined based on actual process conditions.

[0058] The above examples illustrate the present invention only to aid in understanding it and are not intended to limit the scope of the invention. Those skilled in the art can make various simple deductions, modifications, or substitutions based on the principles of this invention.

Claims

1. A method for constructing a liquid milk UHT cleaning model based on multiple monitoring parameters, characterized in that, The method includes: Set cleaning parameters and use the CIP cleaning method to clean the liquid milk UHT production equipment. The cleaning parameters include cleaning temperature, cleaning flow rate, and cleaning solution concentration. During the CIP cleaning process, multiple monitoring probes are used to monitor multiple parameters, and samples are taken at preset time intervals to measure protein concentration. Response surface methodology was employed to analyze and fit the monitoring data from each probe, sampled measurements, and cleaning parameters to obtain an equation model for the relationship between cleaning time and cleaning temperature, cleaning flow rate, and cleaning solution concentration. Specifically, this involved fitting an equation with cleaning time as the dependent variable and cleaning temperature, cleaning flow rate, and cleaning solution concentration as independent variables. The Box-Behnken experimental design method was used to design the experimental scheme, and multiple regression analysis was employed to establish the mapping function between the independent and response variables. Symbolic regression was used to analyze and fit the monitoring data, sampling measurements, and cleaning parameters of each probe to obtain an equation model of the cleaning rate on cleaning time, cleaning temperature, cleaning flow rate, and cleaning solution concentration. Specifically, the difference between probe monitoring data or sampling measurements in adjacent time intervals was used as the characterization value of the cleaning rate and as the dependent variable, while cleaning time, cleaning temperature, cleaning flow rate, and cleaning solution concentration were used as independent variables to obtain the equation of the cleaning rate on cleaning time, cleaning temperature, cleaning flow rate, and cleaning solution concentration.

2. The method for constructing a liquid milk UHT cleaning model based on multiple monitoring parameters as described in claim 1, characterized in that, During the CIP cleaning process, multiple monitoring probes are used to monitor various parameters, including: A flow meter is used to monitor milk flow rate; a pressure gauge is used to monitor pipeline pressure; a thermometer is used to monitor fluid temperature; and turbidity probes, pH probes, and conductivity probes integrated into the water quality analysis device are used to monitor turbidity, pH, and conductivity values, respectively.

3. The method for constructing a liquid milk UHT cleaning model based on multiple monitoring parameters as described in claim 1, characterized in that, The CIP cleaning specifically includes: After the dairy products are sterilized, a CIP cleaning process is performed, in the following order: a) water wash; b) alkaline wash; c) water wash; d) acid wash; e) water wash. During the alkaline and acid washing processes, a preset volume of liquid flowing back into the cleaning tank is taken at preset time intervals to measure the protein concentration in the solution.

4. The method for constructing a liquid milk UHT cleaning model based on multiple monitoring parameters as described in claim 3, characterized in that, The CIP cleaning specifically includes: The cleaning parameters for each cleaning process are as follows: water washing uses room temperature and a flow rate of 50-300 L / h; alkaline washing uses a temperature of 50-150℃, a cleaning flow rate of 50-300 L / h, and a cleaning solution concentration of 0.5-5% by mass; acid washing uses a temperature of 50-120℃, a cleaning flow rate of 50-300 L / h, and a cleaning solution concentration of 0.5-5% by mass.

5. The method for constructing a liquid milk UHT cleaning model based on multiple monitoring parameters as described in claim 2, characterized in that, Using cleaning time as the dependent variable and cleaning temperature, cleaning flow rate, and cleaning solution concentration as independent variables, equations were fitted to obtain the relationship between cleaning time and cleaning temperature, cleaning flow rate, and cleaning solution concentration. Specifically, these equations include: Based on the turbidity, pH, and conductivity values ​​monitored by the probe, as well as the protein concentration values ​​measured by the sample, a curve showing the change of data over time is plotted. When the data tends to stabilize, it indicates that the cleaning is complete, and the cleaning time is determined accordingly.

6. The method for constructing a liquid milk UHT cleaning model based on multiple monitoring parameters as described in claim 1, characterized in that, Multiple regression analysis is used to establish a mapping function between independent variables and response variables, specifically including: The mapping function is as follows: ； In the formula, Y: the response variable, i.e., the dependent variable; β0: a constant; β i β ii β ij : These represent the linear, quadratic, and second-order terms of the model, respectively, X i X j : Independent variable; When the system reaches its optimal operating condition, the mapping function converges to a steady state, at which point the interaction between the independent variables reaches a dynamic equilibrium.

7. The method for constructing a liquid milk UHT cleaning model based on multiple monitoring parameters as described in claim 6, characterized in that, Response surface methodology was employed to analyze and fit the monitoring data from each probe, sampled measurements, and cleaning parameters. This yielded an equation model relating cleaning time to cleaning temperature, cleaning flow rate, and cleaning solution concentration. Specifically, the model includes: The polynomial equation model obtained through modeling analysis is used with the coefficient of determination R. 2 The evaluation was conducted using the degree of misfit and the p-values ​​of significance for each factor.

8. The method for constructing a liquid milk UHT cleaning model based on multiple monitoring parameters as described in claim 1, characterized in that, Using symbolic regression, the monitoring data from each probe, sampled measurements, and cleaning parameters were analyzed and fitted to obtain an equation model for the cleaning rate on the basis of cleaning time, cleaning temperature, cleaning flow rate, and cleaning solution concentration. Specifically, this model includes: Symbolic regression fitting was performed using Eureqa software, and the evaluation metrics included mean absolute error (MAE), mean squared error (MSE), and root mean squared error (RMSE).

9. The method for constructing a liquid milk UHT cleaning model based on multiple monitoring parameters as described in claim 8, characterized in that, The functional relationship between cleaning rate and cleaning time, cleaning temperature, cleaning flow rate, and cleaning solution concentration is expressed as follows: ； In the formula, y is the cleaning rate characterized by the difference between pH, conductivity, turbidity or protein concentration in adjacent monitoring time intervals, t is the cleaning time, T is the cleaning temperature, v is the cleaning flow rate, and c is the cleaning solution concentration.

Images