A dialysis liquid preparation stirring parameter self-adaptive adjusting method based on MPSO

Abstract

This invention discloses an adaptive adjustment method for dialysis solution mixing parameters based on MPSO (Particle Swarm Optimization), relating to the field of dialysate preparation technology. The method includes: initializing a digital twin model and a causal inference network; constructing a virtual mapping of the physical structure of the solution mixing device and calibrating parameters to form a calibrated virtual-real mapping decision support environment; preprocessing multi-dimensional process data from historical solution mixing batches based on the virtual-real mapping decision support environment; updating the causal inference network using a causal discovery algorithm; learning the causal relationship between dialysis solution mixing parameters and solution quality; and establishing a dynamically updated causal relationship network. Based on the dynamically updated causal relationship network, iteratively searching for the optimal solution of the fitness function through particle swarm optimization generates a pre-optimized parameter set for the virtual environment. This invention, through an MPSO-based particle swarm optimization mechanism in a dynamic virtual environment, enables adaptive searching and recalibration of key parameters such as pump power, turbulence angle, and mixing time under multiple operating conditions.

Description

Technical Field

[0001] This invention relates to the field of dialysate preparation technology, and in particular to an adaptive adjustment method for stirring parameters of dialysate preparation based on MPSO. Background Technology

[0002] With the development of intelligent manufacturing and automation control technologies, digital twins and intelligent optimization algorithms are widely used in the fields of fluid mixing and process control. By establishing virtual simulation models, the flow state and mass transport characteristics of complex fluids can be reproduced in virtual space, providing data support for the prediction and optimization of process parameters. Especially in the preparation of dialysate, high-precision control of stirring parameters, fluid dynamics behavior, and the dissolution process has become a key factor in improving the uniformity and stability of the solution preparation.

[0003] However, existing dialysis fluid preparation control systems are mostly based on empirical rules or single physical models, making it difficult to dynamically reflect the coupled changes in the fluid preparation device under multiple operating conditions. This leads to parameter adjustments relying on human experience and significant model prediction bias. In particular, under changes in parameters such as stirring power, turbulence angle, and stirring time, there is a lack of system analysis and adaptive optimization mechanisms to address their causal relationships, making it impossible to achieve real-time response and efficient adjustment to complex flow fields. Therefore, how to establish an adaptive adjustment method that integrates causal reasoning, virtual simulation, and multi-objective optimization has become a core issue in current intelligent control of dialysis fluid preparation. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides an adaptive adjustment method for stirring parameters in dialysis solution preparation based on MPSO to solve the technical problem of unclear causal relationship between stirring parameters and solution quality during dialysis solution preparation.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] This invention provides an adaptive adjustment method for dialysis solution mixing parameters based on MPSO (Multi-Purpose Stirring and Control), comprising: initializing a digital twin model and a causal inference network; constructing a virtual mapping of the physical structure of the solution mixing device and calibrating the parameters to form a calibrated virtual-real mapping decision support environment; based on the virtual-real mapping decision support environment, preprocessing multi-dimensional process data of historical solution mixing batches, and using a causal discovery algorithm to update the causal inference network, learning the causal relationship graph between dialysis solution mixing parameters and solution quality, and establishing a dynamically updated causal relationship network; based on the dynamically updated causal relationship network, generating a virtual environment pre-optimized parameter set by iteratively searching for the optimal solution of the fitness function through particle swarm optimization; using the virtual environment pre-optimized parameter set for counterfactual inference and intervention evaluation to obtain a causally verified robust parameter combination, which is then applied to the adaptive adjustment process of the solution mixing device to drive an online learning mechanism to dynamically optimize the digital twin model parameters.

[0008] As a preferred embodiment of the MPSO-based adaptive adjustment method for dialysis solution mixing parameters described in this invention, the specific steps for constructing a virtual mapping of the physical structure of the solution mixing device and calibrating its parameters are as follows: Collect the geometric structural parameters and operating data of the solution mixing device, and establish a three-dimensional digital twin model including the tank structure, fluid channels, and mixing components; Establish a three-dimensional digital twin model using the collected geometric structural parameters and operating data of the solution mixing device, and generate node connection relationships using the operating variables output by the three-dimensional digital twin model; determine the causal direction using conditional independence tests to initialize the causal inference network; Based on the initialized digital twin model, map the three-dimensional structure of the tank, fluid channels, and mixing components to the virtual environment, and compare and analyze the simulated operating state with the sensor measured data to identify parameter deviations; Correct the fluid dynamics parameters, boundary conditions, and physical property parameters in the digital twin model according to the deviation results, and synchronously update the corrected digital twin model data to the causal inference network, thus completing the virtual mapping of the physical structure of the solution mixing device and parameter calibration.

[0009] As a preferred embodiment of the MPSO-based adaptive adjustment method for dialysis solution mixing parameters described in this invention, the specific steps for forming a calibrated virtual-real mapping decision support environment are as follows: The virtual mapping parameters are calibrated to form a virtual-real mapping environment; real-time feedback data is preprocessed and process parameters are updated based on the virtual-real mapping environment; and the solution mixing process is dynamically adjusted and optimized by analyzing the relationship between the dialysis solution mixing parameters and the solution quality.

[0010] As a preferred embodiment of the MPSO-based adaptive adjustment method for dialysis solution mixing parameters described in this invention, the multi-dimensional process data of the historical solution batches is obtained by exploratory analysis to identify factors affecting solution mixing and mining the dialysis solution mixing parameters.

[0011] As a preferred embodiment of the MPSO-based adaptive adjustment method for dialysis solution mixing parameters described in this invention, the causal inference network is updated by preprocessing multi-dimensional process data of historical solution batches. The specific steps are as follows: collect multi-dimensional process data of historical solution batches and perform data preprocessing; data preprocessing includes removing outliers, filling missing values, data standardization, and data classification and coding; divide the preprocessed multi-dimensional process data of historical solution batches into multiple feature groups, analyze the correlation between feature groups, and update the causal inference network.

[0012] As a preferred embodiment of the MPSO-based adaptive adjustment method for dialysis solution mixing parameters described in this invention, the dialysis solution mixing parameters include circulating pump power, baffle cap angle, number of baffles, and mixing time. The specific steps for establishing a dynamically updated causal relationship network are as follows: by collecting and analyzing multi-dimensional process data from historical solution batches, the relevant factors between the dialysis solution mixing parameters and solution quality are identified; based on the multi-dimensional process data from historical solution batches, the interrelationships between these relevant factors are analyzed to obtain the causal relationship between the dialysis solution mixing parameters and solution quality; by updating the continuously input multi-dimensional process data from real-time solution batches, the causal relationship is dynamically adjusted to establish a dynamically updated causal relationship network.

[0013] As a preferred embodiment of the MPSO-based adaptive adjustment method for dialysis solution mixing parameters described in this invention, the specific process for generating the virtual environment pre-optimized parameter set is as follows: Dialysis solution mixing parameters are extracted from multi-dimensional process data of historical solution batches to construct a dynamic virtual environment characterizing fluid and mass transport; Particle swarm optimization (PSO) is used in the dynamic virtual environment to search for pump power, turbulence angle, and mixing time to obtain preliminary optimized parameters; Multi-objective evaluation of the preliminary optimized parameters is performed to generate accuracy correction factors; The PSO optimization process is iteratively corrected using the accuracy correction factors to output an optimized parameter set; Through virtual environment disturbance testing and real-time feedback comparison, parameters that maintain consistent output under multiple operating conditions are selected to form the virtual environment pre-optimized parameter set.

[0014] As a preferred embodiment of the MPSO-based adaptive adjustment method for dialysis solution mixing parameters described in this invention, the specific process for obtaining a causally validated robust parameter set is as follows: Based on a pre-optimized parameter set in a virtual environment, a dynamic deduction process integrating causal graph analysis and strategy optimization iteration is constructed; dialysis solution mixing parameters are input and solution quality feedback is received to identify causal paths; based on the causal paths, the dialysis solution mixing parameters are dynamically adjusted to generate a preliminary robust parameter set that conforms to causal constraints; using the preliminary robust parameter set, multi-condition perturbation and performance correction are carried out in a virtual twin environment to extract the causally validated robust parameter set.

[0015] As a preferred embodiment of the MPSO-based adaptive adjustment method for dialysis solution mixing parameters described in this invention, the adaptive adjustment process of the solution mixing device refers to the dynamic adaptive adjustment of the dialysis solution mixing process through online monitoring, state estimation and adaptive control.

[0016] The beneficial effects of this invention are as follows: By using the MPSO-based particle swarm iterative optimization mechanism in a dynamic virtual environment, key parameters such as pump power, turbulence angle, and stirring time can be adaptively searched and recalibrated under multiple operating conditions; through iterative correction guided by the accuracy correction factor, the optimization process gradually conforms to the real changes in the fluid flow field and mass transport, thereby generating a virtual environment pre-optimized parameter set that can maintain consistent output under different disturbance conditions. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 Flowchart of the measurement method for a fluxgate current sensor with low-loss proportional amplification.

[0019] Figure 2 This is a flowchart for the virtual mapping and virtual-real mapping environment calibration of the physical structure of the liquid preparation device.

[0020] Figure 3 To update the dynamic causal network flowchart.

[0021] Figure 4 Generate a flowchart for robust parameter combinations. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figures 1-4 This is one embodiment of the present invention, which provides an adaptive adjustment method for stirring parameters of dialysis solution preparation based on MPSO, comprising the following steps:

[0026] S1: Initialize the digital twin model and causal reasoning network, construct a virtual mapping of the physical structure of the liquid preparation device and calibrate the parameters to form a calibrated virtual-real mapping decision support environment.

[0027] S1.1: Collect the geometric structural parameters and operating data of the liquid preparation device, and construct a three-dimensional digital twin model including the tank structure, fluid channels and stirring components.

[0028] Specifically, when collecting the geometric structural parameters and operational data of the liquid preparation device, a 3D laser scanner is used to obtain the spatial dimensions and position coordinates of the liquid preparation tank, inlet and outlet water pipes, and stirring components, forming a complete structural parameter dataset. Operational data during the liquid preparation process, including fluid velocity distribution, stirring torque, and energy changes, are collected in real time through flow sensors, pressure sensors, conductivity sensors, and temperature sensors. The structural parameter dataset and operational data are then imported into a 3D modeling platform to construct a 3D digital twin model that includes the tank structure, fluid channels, and stirring components.

[0029] S1.2: Establish a three-dimensional digital twin model by collecting the geometric structural parameters and operating data of the liquid preparation device, and generate node connection relationships using the operating variables output by the three-dimensional digital twin model. Use conditional independence test to determine the causal direction to initialize the causal inference network.

[0030] Specifically, by collecting the geometric structural parameters and operational data of the liquid preparation device, a complete digital twin model is established using 3D modeling tools to realistically reproduce the structure and flow characteristics of the liquid preparation device. After the model is built, the operational variables output by the digital twin model are used as the node inputs of the causal inference network. The joint distribution among the variables is calculated, and the causal connection direction between the variables is determined by the conditional independence test, thus forming the initialization of the causal inference network.

[0031] S1.3: Based on the initialized digital twin model, the three-dimensional structure of the tank, fluid channel and stirring component is mapped to the virtual environment, and the simulation operation status is compared and analyzed with the sensor measured data to identify parameter deviations.

[0032] Specifically, based on the initialized digital twin model, the three-dimensional structure of the tank, fluid channel, and stirring components is completely imported into the virtual simulation environment to establish a virtual physical model corresponding to the actual liquid preparation device. Then, operating parameters such as the power of the circulating pump, the turbulence angle, and the stirring time, which are synchronized with the actual operation, are input into the virtual environment to generate virtual simulation operating status data. The virtual simulation operating status data is compared item by item with the actual operating data collected by the sensors to analyze the flow field distribution, concentration changes, and energy consumption differences, and to identify the operating parameters with deviations.

[0033] S1.4: Based on the deviation results, correct the fluid dynamics parameters, boundary conditions, and physical property parameters in the digital twin model, and synchronously update the corrected digital twin model data to the causal inference network to complete the virtual mapping and parameter calibration of the physical structure of the liquid preparation device.

[0034] Specifically, based on the deviation between the virtual simulation running state and the actual running data, the fluid dynamics parameters, boundary conditions, and physical property parameters in the digital twin model are corrected item by item. After correction, the simulation is rerun in the virtual environment to verify the matching degree of the flow field distribution and concentration change. When the deviation between the simulation results and the measured data is lower than the set threshold, the calibrated model data is synchronously input into the causal inference network to update the node features and edge weights, thereby completing the virtual mapping and parameter calibration of the physical structure of the liquid preparation device.

[0035] In contrast to conventional methods that rely solely on empirical parameters or single experimental results to statically correct digital twin models, this step introduces causal directed graph recognition results to perform directional calibration of key variables such as pump power, turbulence angle, and stirring time. Based on dynamic parameter feedback and model optimization using causal weights, the digital twin model can accurately reflect the comprehensive impact of each control variable on fluid dynamic distribution and liquid preparation quality, thereby improving the fitting accuracy and dynamic response capability of the virtual mapping to the real liquid preparation process.

[0036] S1.5: A virtual-real mapping environment is formed by calibrating the parameters of the virtual mapping.

[0037] Specifically, by utilizing the physical structure of the liquid mixing device, the operating parameters of the actual liquid mixing device are mapped to the virtual mapping environment, so that the spatial positions of the liquid mixing tank, stirring paddle, circulation pipeline and sensor in the virtual mapping environment are consistent with the components in the actual liquid mixing device. By matching the tank size, stirring shaft center position, pipeline inlet and outlet direction and sensor installation coordinates, the geometric layout, movement direction and measurement points in the virtual and real mapping environment are made consistent with the actual device.

[0038] Based on real-time flow rate, stirring speed, liquid level change, and conductivity data collected by sensors through a data acquisition device, the fluid dynamics parameters in the virtual mapping environment are adjusted to keep the virtual environment synchronized with the actual liquid preparation device in terms of operating characteristics, thus forming a virtual-real mapping environment that can reflect the actual operating status in real time.

[0039] It should be noted that parameter identification and model calibration based on particle swarm optimization algorithm were used during the adjustment process.

[0040] S1.6: Based on the virtual-real mapping environment, preprocess real-time feedback data and update process parameters. By analyzing the relationship between dialysis solution mixing parameters and solution quality, dynamically adjust and optimize the solution mixing process.

[0041] Specifically, after the virtual-real mapping environment is formed, based on the real-time feedback data continuously received by the virtual-real mapping environment, preprocessing operations such as noise filtering, time series smoothing and outlier removal are performed on the real-time feedback data to ensure that the data can be used for dynamic parameter updates. The preprocessed real-time feedback data is used to correct process parameters such as pump power, turbulence angle and stirring time in the liquid preparation process, so that the prediction results in the virtual-real environment are consistent with the actual observation results.

[0042] S2: Based on the virtual-real mapping decision support environment, preprocess multi-dimensional process data of historical solution batches to update the causal reasoning network; use the causal discovery algorithm to learn the causal relationship graph between dialysis solution mixing parameters and solution quality, and establish a dynamically updated causal relationship network;

[0043] S2.1: The multi-dimensional process data of historical solution preparation batches were obtained by exploratory analysis to identify factors affecting solution preparation and by mining the stirring parameters of dialysis solution preparation.

[0044] Specifically, during the operation of multiple batches of solution preparation, information such as stirring speed, circulating pump power, turbulence angle, liquid temperature, conductivity, liquid level change, and solution preparation time of the solution preparation device is collected to reflect the operating characteristics under different solution preparation conditions. After the information collection is completed, the operating characteristics of different batches of solution preparation are compared, and factors that significantly affect the uniformity of solution concentration and mixing stability are identified based on the variable change patterns. Based on these factors, dialysis solution preparation stirring parameters are extracted so that they can reflect the correspondence between stirring state and solution quality changes, forming multi-dimensional process data of historical solution preparation batches.

[0045] It should be noted that the variable variation law refers to the trend relationship of operating parameters such as stirring speed, circulating pump power, turbulence angle, liquid temperature, conductivity, liquid level change and dispensing time with time or operating conditions in different batches of liquid preparation. It is used to reflect the synergistic variation characteristics between parameters and their influence on the uniformity of liquid concentration and mixing stability.

[0046] S2.2: Collect multi-dimensional process data of historical solution batches and perform data preprocessing; data preprocessing includes removing outliers, filling missing values, data standardization, and data classification and coding.

[0047] Specifically, during the collection of multi-dimensional process data for historical batches of liquid preparation, the stirring speed, circulating pump power, turbulence angle, liquid temperature, conductivity, liquid level change, and preparation time of the liquid preparation device are obtained to reflect the characteristics of different batches of liquid preparation. After the information collection is completed, the multi-dimensional process data of historical batches of liquid preparation are preprocessed by removing outliers, filling missing values, standardizing data, and classifying and coding data to ensure consistency of process data for different batches.

[0048] S2.3: Divide the multi-dimensional process data of historical liquid preparation batches after data preprocessing into multiple feature groups, analyze the correlation between feature groups, and update the causal inference network.

[0049] Specifically, the multi-dimensional process data of historical liquid preparation batches after data preprocessing are divided into multiple feature groups according to stirring parameters, fluid characteristic parameters, and liquid preparation quality parameters. A unified feature table is established for each feature group for subsequent comparison. Joint correlation analysis and conditional independence tests are performed on each feature group. The correlation coefficient matrix is ​​calculated and sensitivity index is evaluated to identify obvious direct correlations and potential indirect correlations.

[0050] Based on the correlation coefficient matrix and the conditional dependency test results, the PC algorithm is used to adjust the connection direction and edge weight of the causal inference network and estimate the parameters. The reliability of the updated causal inference network is verified by cross-validation or out-of-batch historical validation.

[0051] S2.4: Dialysis solution mixing parameters include circulating pump power, turbulence cap angle, number of turbulence vanes, and mixing time.

[0052] Specifically, based on the updated causal reasoning network, in determining the stirring parameters for dialysis solution preparation, the power of the circulating pump is determined according to the operating characteristics of the solution preparation device. The circulation intensity of the liquid inside the solution preparation tank is obtained by adjusting the operating state of the circulating pump. After the circulating pump power is determined, the angle of the baffle cap is adjusted to coordinate the liquid flow direction with the circulation intensity. The number of baffles is determined according to the flow state formed by the baffle cap angle to form a uniform flow field inside the tank. Finally, after the number of baffles is stabilized, the stirring time is determined so that the stirring duration matches the circulating pump power, the baffle cap angle, and the number of baffles, thus forming the stirring parameters for dialysis solution preparation.

[0053] S2.5: By collecting and analyzing multi-dimensional process data from historical solution preparation batches, identify the relevant factors between dialysis solution mixing parameters and solution quality.

[0054] Specifically, by collecting multi-dimensional process data from historical batches of solution preparation, information such as stirring speed, circulating pump power, turbulence angle, liquid temperature, conductivity, liquid level change, and solution preparation time of the solution preparation device under different batches is obtained. This information is used to determine the stirring parameters that significantly affect concentration uniformity and mixing stability under different solution preparation conditions, and to form the correlation factors between dialysis solution preparation stirring parameters and solution preparation quality.

[0055] S2.6: Based on multi-dimensional process data of historical solution preparation batches, analyze the interrelationships between various relevant factors and obtain the causal relationship between dialysis solution preparation stirring parameters and solution quality.

[0056] Specifically, based on the relevant factors already obtained, the variation characteristics of circulating pump power, turbulence angle, stirring time and conductivity, concentration uniformity and dissolution rate in the multi-dimensional process data of historical solution batches are used to determine the mutual influence relationship between the factors and obtain the causal relationship between the stirring parameters of dialysis solution preparation and the solution quality.

[0057] S2.7: By updating the multi-dimensional process data of the continuously input real-time batches of solution preparation, the causal relationship is dynamically adjusted to establish a dynamically updated causal relationship network.

[0058] Specifically, after the causal relationship is determined, the causal relationship is dynamically adjusted by continuously receiving multi-dimensional process data of real-time batches of solution. This allows the causal path and direction of influence to be continuously corrected as the real-time data changes, thereby establishing a dynamically updated causal relationship network that can reflect the actual operating status of the solution preparation process.

[0059] It should be noted that to dynamically adjust the causal relationship, it is necessary to compare the changes in the variable associations of multi-dimensional process data and the causal relationship network in real time, and to correct the direction and weight of the causal edges according to the preset threshold, so as to realize the dynamic adjustment of the causal path.

[0060] S3: Based on a dynamically updated causal relationship network, the optimal solution of the fitness function is searched through particle swarm optimization to generate a pre-optimized parameter set for the virtual environment.

[0061] S3.1: Extract dialysis solution mixing parameters from multi-dimensional process data of historical solution batches to construct a dynamic virtual environment characterizing fluid and mass transport;

[0062] Specifically, dialysis mixing parameters such as circulating pump power, turbulence angle, number of turbulence vanes, and stirring time are extracted from multi-dimensional process data of historical batches of solution preparation. These parameters are used to characterize the fluid flow and mass transport characteristics under different operating conditions, and to generate a dynamic virtual environment that can reflect the mixing behavior of the solution preparation.

[0063] In contrast to conventional static modeling methods that rely solely on a single operating condition to establish a fluid dynamics model, this step utilizes multi-dimensional process data from historical batches of liquid preparation to achieve dynamic mapping of multi-operating-condition characteristics. This enables the virtual environment to accurately reflect the impact of changes in stirring parameters on the flow field distribution and mass transport, thereby improving the realism of the simulation and the accuracy of its response to complex stirring conditions.

[0064] S3.2: In a dynamic virtual environment, the particle swarm optimization method is used to search for pump power, turbulence angle and stirring time to obtain preliminary optimization parameters.

[0065] Specifically, in a dynamic virtual environment, based on the evaluation criteria of mixing uniformity, concentration stability, and energy consumption level, the particle swarm optimization method is used to search for pump power, turbulence angle, and mixing time to obtain preliminary optimized parameters that balance mixing effect and energy consumption.

[0066] S3.3: Perform multi-objective evaluation on the preliminary optimized parameters and generate accuracy correction factors.

[0067] Specifically, when substituting the preliminary optimized parameters into the multi-objective evaluation process, parameters such as circulating pump power, turbulence angle, number of turbulence vanes, and stirring time obtained through the optimization algorithm are input into the virtual simulation model to perform multi-dimensional simulation of the fluid flow state under different operating conditions. The spatial distribution uniformity of solute concentration, the concentration fluctuation amplitude in the time series, and the energy consumption level per unit volume of liquid are calculated respectively during the solution preparation process to obtain quantitative results of mixing uniformity index, concentration stability index, and energy efficiency index. The three indicators are weighted and analyzed through a multi-objective comprehensive evaluation function, and the deviation between the virtual simulation results and the experimental measured data is compared. The degree of matching between the output of the digital twin simulation model and the actual performance is considered. The accuracy correction factor is calculated based on the deviation ratio.

[0068] S3.4: Combine the accuracy correction factor to perform a re-iteration correction on the particle swarm optimization process.

[0069] Specifically, when iterating the particle swarm optimization process using a precision correction factor, the precision correction factor is introduced as a dynamic weight parameter into the fitness evaluation function of the particle swarm optimization algorithm to correct the velocity and position update magnitude of particles in the search space in real time. Based on the virtual simulation error trend reflected by the correction factor, the search range of key variables such as circulating pump power, turbulence angle, number of turbulence vanes, and stirring time in the particle swarm is adaptively adjusted, so that the particle swarm automatically reduces the step size when it is close to the minimum error interval and accelerates the global search speed when it deviates from the high precision interval. In each iteration, the updated global optimal solution is calculated and re-introduced into the digital twin simulation model for flow field distribution and concentration change verification. Through continuous multiple rounds of correction iterations, the true laws of fluid and material transport changes in the virtual environment are gradually approximated.

[0070] S3.5: Output optimization parameter set. By comparing virtual environment disturbance test with real-time feedback, the optimization parameters are selected to maintain output consistency under multiple operating conditions, forming a virtual environment pre-optimized parameter set.

[0071] Specifically, after the iteration is completed, the optimized parameter set is output, and the results are compared with the real-time feedback results through perturbation test in the virtual environment. The parameter combination that maintains the consistency of flow field distribution, concentration uniformity and mixing stability under different operating conditions is selected to form the virtual environment pre-optimized parameter set.

[0072] S4: Utilize the pre-optimized parameter set in the virtual environment for counterfactual reasoning and intervention evaluation to obtain a robust parameter combination verified by causality, and apply it to the adaptive adjustment process of the liquid preparation device to drive the online learning mechanism to dynamically optimize the parameters of the digital twin model.

[0073] S4.1: Based on the pre-optimized parameter set of the virtual environment, construct a dynamic deduction process that integrates causal graph analysis and strategy optimization iteration, input dialysis solution mixing parameters and receive solution quality feedback, and identify causal paths.

[0074] Specifically, during the simulation, the stirring parameters of the dialysis solution preparation are imported into the virtual environment as input variables. By receiving real-time feedback data on the solution preparation quality, the corresponding changes between the circulating pump power, turbulence angle, stirring time, conductivity, and concentration uniformity are recorded. The feedback results are used to gradually determine the direct and indirect effects between variables in the causal relationship network structure, thereby identifying the causal path of the dialysis solution preparation stirring parameters affecting the solution preparation quality.

[0075] S4.2: Based on causal paths, dynamically adjust the stirring parameters of dialysis solution preparation to generate a preliminary robust parameter set that conforms to causal constraints;

[0076] Specifically, based on the identified causal path, the circulating pump power, turbulence angle, and stirring time are adjusted adaptively in conjunction with real-time monitoring of the solution quality feedback and causal weight constraints. This ensures that the circulating pump power, turbulence angle, and stirring time are in a coordinated and matched state under causal constraints, thereby generating a preliminary robust parameter set that can maintain stable solution concentration and uniform mixing.

[0077] S4.3: Utilize the preliminary robust parameter set to conduct multi-condition perturbation and performance correction in a virtual twin environment, and extract the robust parameter combination verified by causality.

[0078] Specifically, a preliminary robust parameter set was used in a virtual twin environment for multi-condition perturbation testing and performance calibration. Typical perturbation conditions such as liquid level, flow rate, temperature, and solution viscosity were set, and the robust parameter set was substituted into the simulation model. Then, the flow field distribution, concentration gradient, and mixing uniformity under each condition were monitored, and performance indicators such as concentration root mean square error and energy consumption deviation were calculated. Based on the performance differences under different conditions, the parameters were locally corrected and re-verified, and a robust parameter combination that maintained stable concentration and uniform mixing under multiple conditions and was causally verified was selected.

[0079] S4.4: The adaptive adjustment process of the solution preparation device refers to the dynamic adaptive adjustment of the dialysis solution preparation process through online monitoring, state estimation and adaptive control.

[0080] Specifically, the robust parameter combination verified by causality is used to monitor the real-time operating status of the solution preparation device during the dialysis solution preparation process, and to obtain monitoring data such as circulating pump power, stirring speed, liquid level change, conductivity and temperature, which are used to reflect the dynamic changes in the solution preparation process;

[0081] After acquiring the monitoring data, the operating status is estimated by combining the virtual and real mapping environment and the causal reasoning network to ensure that the stirring status is consistent with the changes in the solution quality. Based on the state estimation results, the stirring parameters of the dialysis solution are adaptively adjusted. By adjusting variables such as the power of the circulating pump, the turbulence angle and the stirring time in real time, the dialysis solution preparation process maintains uniform concentration, thorough mixing and stable quality under different operating conditions, thereby achieving dynamic adaptive adjustment.

[0082] In summary, this invention, through an MPSO-based particle swarm optimization mechanism in a dynamic virtual environment, enables adaptive searching and recalibration of key parameters such as pump power, turbulence angle, and stirring time under multiple operating conditions. Through iterative correction guided by accuracy correction factors, the optimization process gradually conforms to the real changes in fluid flow field and mass transport, thereby generating a virtual environment pre-optimized parameter set that maintains consistent output under different disturbance conditions.

[0083] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for adaptive adjustment of stirring parameters for dialysis solution preparation based on MPSO, characterized in that: include, Initialize the digital twin model and causal inference network, construct a virtual mapping of the physical structure of the liquid preparation device and calibrate the parameters, correct the fluid dynamics parameters, boundary conditions and physical property parameters in the digital twin model based on the deviation results, and synchronously update the corrected digital twin model data to the causal inference network to complete the virtual mapping and parameter calibration of the physical structure of the liquid preparation device, forming a calibrated virtual-real mapping decision support environment; Based on the virtual-real mapping decision support environment, multi-dimensional process data of historical solution batches are preprocessed, and causal discovery algorithms are used to update the causal reasoning network, learn the causal relationship graph between dialysis solution mixing parameters and solution quality, and establish a dynamically updated causal relationship network. Based on a dynamically updated causal relationship network, a pre-optimized parameter set for the virtual environment is generated by iteratively searching for the optimal solution of the fitness function through particle swarm optimization. The specific process is as follows. Dialysis solution mixing parameters were extracted from multi-dimensional process data of historical solution batches to construct a dynamic virtual environment characterizing fluid and mass transport. In a dynamic virtual environment, the particle swarm optimization method is used to search for pump power, turbulence angle and stirring time to obtain preliminary optimization parameters; Perform multi-objective evaluation on the preliminary optimized parameters to generate accuracy correction factors; The particle swarm optimization process is iteratively corrected by incorporating a precision correction factor, and an optimized parameter set is output. By comparing virtual environment disturbance tests with real-time feedback, the parameters that maintain output consistency under multiple operating conditions are selected for optimization, forming a virtual environment pre-optimization parameter set; By using a pre-optimized parameter set in a virtual environment for counterfactual reasoning and intervention evaluation, robust parameter combinations validated by causality are obtained. The specific process is as follows. Based on the pre-optimized parameter set of the virtual environment, a dynamic deduction process integrating causal graph analysis and strategy optimization iteration is constructed. The dialysis solution mixing parameters are input and the solution quality feedback is received to identify the causal path. Based on causal paths, the stirring parameters for dialysis solution preparation are dynamically adjusted to generate a preliminary robust parameter set that conforms to causal constraints; Using a preliminary robust parameter set, multi-condition perturbation and performance correction are carried out in a virtual twin environment. A robust parameter combination verified by causality is extracted and applied to the adaptive adjustment process of the liquid preparation device, driving the online learning mechanism to dynamically optimize the parameters of the digital twin model.

2. The method for adaptive adjustment of stirring parameters for dialysis solution preparation based on MPSO as described in claim 1, characterized in that: The specific steps for constructing a virtual mapping of the physical structure of the solution preparation device and calibrating its parameters are as follows: Collect the geometric structural parameters and operating data of the liquid preparation device, and establish a three-dimensional digital twin model including the tank structure, fluid channels and stirring components; A three-dimensional digital twin model is established by collecting the geometric structural parameters and operating data of the liquid preparation device. The operating variables output by the three-dimensional digital twin model are used to generate node connection relationships. The conditional independence test is used to determine the causal direction in order to initialize the causal inference network. Based on the initialized digital twin model, the three-dimensional structure of the tank, fluid channel and stirring component is mapped to the virtual environment, and the simulation operation status is compared and analyzed with the sensor measured data to identify parameter deviations; Based on the deviation results, the fluid dynamics parameters, boundary conditions, and physical property parameters in the digital twin model are corrected, and the corrected digital twin model data is synchronously updated to the causal inference network to complete the virtual mapping and parameter calibration of the physical structure of the liquid preparation device.

3. The adaptive adjustment method for stirring parameters of dialysis solution preparation based on MPSO as described in claim 2, characterized in that: The specific steps for forming the calibrated virtual-real mapping decision support environment are as follows. A virtual-real mapping environment is formed by calibrating the parameters of the virtual mapping. Based on real-time feedback data from virtual-real mapping environment preprocessing and updating process parameters, the solution preparation process is dynamically adjusted and optimized by analyzing the relationship between dialysis solution mixing parameters and solution quality.

4. The method for adaptive adjustment of stirring parameters for dialysis solution preparation based on MPSO as described in claim 3, characterized in that: The multi-dimensional process data of the historical solution preparation batches were obtained by exploratory analysis to identify factors affecting solution preparation and by mining the stirring parameters of dialysis solution preparation.

5. The method for adaptive adjustment of stirring parameters for dialysis solution preparation based on MPSO as described in claim 4, characterized in that: The multi-dimensional process data of the pretreatment historical solution batches are updated using a causal reasoning network. The specific steps are as follows: Collect multi-dimensional process data from historical batches of solution preparation and perform data preprocessing; Data preprocessing includes removing outliers, filling in missing values, data standardization, and data classification and coding; The multi-dimensional process data of historical liquid preparation batches after data preprocessing are divided into multiple feature groups. The correlation between feature groups is analyzed, and the causal inference network is updated.

6. The method for adaptive adjustment of stirring parameters for dialysis solution preparation based on MPSO as described in claim 5, characterized in that: The dialysis solution mixing parameters include the circulating pump power, the angle of the turbulence cap, the number of turbulence vanes, and the mixing time.

7. The method for adaptive adjustment of stirring parameters for dialysis solution preparation based on MPSO as described in claim 6, characterized in that: The specific steps for establishing a dynamically updated causal network are as follows: By collecting and analyzing multi-dimensional process data from historical dialysis solution batches, the correlation between dialysis solution mixing parameters and solution quality was identified. Based on multi-dimensional process data of historical solution preparation batches, the interrelationships between various relevant factors were analyzed to obtain the causal relationship between dialysis solution preparation stirring parameters and solution quality. By updating the multi-dimensional process data of the continuously input real-time batches of solution preparation, the causal relationships are dynamically adjusted to establish a dynamically updated causal relationship network.

8. The method for adaptive adjustment of stirring parameters for dialysis solution preparation based on MPSO as described in claim 7, characterized in that: The adaptive adjustment process of the solution preparation device refers to the dynamic adaptive adjustment of the dialysis solution preparation process through online monitoring, state estimation and adaptive control.

Images