Multi-objective optimization method and system for cell culture medium purification parameter tuning

Abstract

This invention discloses a multi-objective optimization method and system for tuning cell culture medium purification parameters, primarily relating to the field of cell culture medium purification technology. The method includes: acquiring equipment environmental information during the intermediate purification stage of the target cell culture medium; determining an optimization target set based on the equipment environmental information as constraints; obtaining real-time purification monitoring data and historical purification monitoring data sequences; obtaining latent purification impact prediction features; obtaining explicit purification anomaly features; obtaining target purification anomaly features; and performing multi-objective optimization of purification parameters based on the optimization target set and the target purification anomaly features to obtain the optimized purification parameters. The beneficial effect of this invention is that it solves the technical problem in the prior art of lacking analysis of the actual purification situation during the intermediate purification stage of cell culture medium, leading to the inability to obtain the required cell solution, thus achieving the technical effect of improving the quality of purification parameter optimization.

Description

Technical Field

[0001] This invention relates to the field of cell culture medium purification technology, specifically to a method and system for multi-objective optimization of cell culture medium purification parameters. Background Technology

[0002] Cell culture medium purification is a crucial step in the biopharmaceutical industry, aiming to extract pure target products from complex culture media. Different purification techniques are employed, including affinity chromatography, ion exchange chromatography, and molecular sieve chromatography. Affinity chromatography removes impurities by utilizing the affinity between the target product and specific ligands, while ion exchange chromatography separates based on the charge properties of the target product. These methods are often used in combination in multi-stage purification processes, especially in intermediate purification stages, optimized for removing small molecule impurities. However, in practice, the purification process involves multiple interdependent variables, such as temperature, pH, flow rate, and salt concentration. Current analytical depth for these parameters is insufficient, leading to unmet requirements for target product purity, yield, and activity.

[0003] The existing technology lacks analysis of the actual purification process during the intermediate purification stage of cell culture medium, resulting in the inability to obtain the required cell fluid. Summary of the Invention

[0004] This application provides a multi-objective optimization method and system for adjusting cell culture medium purification parameters, which is used to address the technical problem in the prior art that the lack of analysis of the actual purification situation in the intermediate purification stage of cell culture medium leads to the inability to obtain the required cell fluid.

[0005] In view of the above problems, this application provides a method and system for multi-objective optimization of cell culture medium purification parameters.

[0006] The first aspect of this application provides a method for optimizing cell culture medium purification parameters through multi-objective optimization, the method comprising:

[0007] The process involves: acquiring equipment environment information for the intermediate purification stage of the target cell culture medium; using this information as a constraint and the application scenario information of the target cell culture medium as a basis; matching and optimizing objectives to determine an optimization objective set, where each objective has an optimization weight; using a sensor monitoring array to monitor the intermediate purification stage of the target cell culture medium and storing the purification monitoring data in a processing database to obtain real-time purification monitoring data and historical purification monitoring data sequences; using a low-frequency slow network to predict implicit purification impacts on the historical purification monitoring data sequences to obtain implicit purification impact prediction features; using a high-frequency fast network to identify explicit purification anomaly features on the real-time purification monitoring data sequences to obtain explicit purification anomaly features; correcting the explicit purification anomaly features based on the implicit purification impact prediction features to obtain target purification anomaly features; and using the optimization objective set as a constraint and the target purification anomaly features as a basis to perform multi-objective optimization of purification parameters to obtain the purification parameter optimization results.

[0008] In one possible implementation, the sensor monitoring array includes at least a temperature sensor, a pH sensor, a flow rate sensor, a salt concentration sensor, and a charge change sensor.

[0009] In one possible implementation, a low-frequency slow network is invoked to predict the implicit purification impact of the historical purification monitoring data sequence to obtain implicit purification impact prediction features. This includes: invoking the low-frequency slow network to extract purification features for each historical purification monitoring data in the historical purification monitoring data sequence to obtain a historical purification feature sequence; performing iterative implicit purification analysis on the historical purification feature sequence in chronological order to determine iterative implicit purification features; and predicting the implicit purification impact of the iterative implicit purification features to obtain the implicit purification impact prediction features.

[0010] In one possible implementation, iterative implicit purification analysis is performed on the historical purification feature sequence in chronological order to determine iterative implicit purification features, including: extracting a first historical purification feature and a second historical purification feature from the historical purification feature sequence; performing iterative implicit association analysis on the first historical purification feature and the second historical purification feature to determine a first associated iterative implicit purification feature; and using the first associated iterative implicit purification feature to sequentially perform iterative implicit association analysis on the remaining historical purification features in the historical purification feature sequence to determine the iterative implicit purification feature.

[0011] In one possible implementation, performing iterative implicit correlation analysis on the first historical purification feature and the second historical purification feature to determine the first associated iterative implicit purification feature includes: performing iterative implicit correlation analysis on features of the same type in the first historical purification feature and the second historical purification feature respectively to obtain a first set of iterative implicit correlation coefficients; constructing a first iterative implicit correlation vector based on the first set of iterative implicit correlation coefficients; and performing convolution analysis on the second historical purification feature based on the first iterative implicit correlation vector to obtain the first associated iterative implicit purification feature.

[0012] In one possible implementation, the purification parameters are optimized using a set of optimization targets as constraints, combined with the target purification anomaly features, to obtain the purification parameter optimization results. This includes: pre-constructing a purification parameter identifier; using the purification parameter identifier to identify the target purification anomaly features to obtain multiple purification parameters; simulating each of the multiple purification parameters and scoring the simulation results against the set of optimization targets to obtain multiple score results; adjusting the multiple purification parameters according to a first adjustment method, using the purification parameter corresponding to the maximum value among the multiple score results as the direction, to obtain multiple adjusted purification parameters, wherein the first adjustment method involves randomly adjusting the purification parameters according to a preset adjustment quantity and preset adjustment range; scoring the multiple adjusted purification parameters again to determine whether the maximum value among the multiple adjusted score results is greater than the maximum value among the multiple score results; if so, the adjusted purification parameter corresponding to the maximum value among the multiple adjusted score results is taken as the purification parameter optimization result.

[0013] In one possible implementation, if the maximum value among the multiple adjustment achievement scores is less than or equal to the maximum value among the multiple achievement scores, a readjustment instruction is obtained, and the multiple purification parameters are randomly adjusted and optimized again according to the first adjustment method.

[0014] A second aspect of this application provides a multi-objective optimized cell culture medium purification parameter tuning system, the system comprising:

[0015] The optimization target set determination module is used to acquire equipment environment information during the intermediate purification stage of the target cell culture medium, and, based on the equipment environment information and the application scenario information of the target cell culture medium, match optimization targets to determine the optimization target set, wherein each optimization target has an optimization weight; the purification monitoring module is used to call a sensor monitoring array to perform purification monitoring during the intermediate purification stage of the target cell culture medium, and store the purification monitoring data in a processing database to obtain real-time purification monitoring data and historical purification monitoring data sequences; the implicit purification impact prediction module is used to call a low-frequency slow network to predict the implicit purification impact of the historical purification monitoring data sequences to obtain implicit purification impact prediction features; the explicit purification anomaly feature acquisition module is used to call a high-frequency fast network to identify explicit purification anomaly features of the real-time purification monitoring data sequences to obtain explicit purification anomaly features; the target purification anomaly feature acquisition module is used to correct the explicit purification anomaly features based on the implicit purification impact prediction features to obtain target purification anomaly features; the purification parameter optimization result acquisition module is used to perform multi-objective optimization of purification parameters, constrained by the optimization target set and combined with the target purification anomaly features, to obtain purification parameter optimization results.

[0016] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0017] This application acquires equipment environmental information during the intermediate purification stage of the target cell culture medium. Using this information as a constraint and the application scenario information of the target cell culture medium as a basis, it matches and optimizes objectives, determining a set of optimization objectives, where each objective has an optimization weight. Then, it uses a sensor monitoring array to monitor the intermediate purification stage of the target cell culture medium and stores the purification monitoring data in a processing database, obtaining real-time purification monitoring data and historical purification monitoring data sequences. Next, it uses a low-frequency slow network to predict implicit purification impacts on the historical purification monitoring data sequences, obtaining implicit purification impact prediction features. Then, it uses a high-frequency fast network to identify explicit purification anomaly features on the real-time purification monitoring data sequences, obtaining explicit purification anomaly features. Based on the implicit purification impact prediction features, it corrects the explicit purification anomaly features to obtain target purification anomaly features. Finally, using the set of optimization objectives as constraints and combining the target purification anomaly features, it performs multi-objective optimization of purification parameters to obtain the optimized purification parameter results. This achieves the technical effect of improving the quality of purification parameter optimization and enhancing the stability of purification effects. Attached Figure Description

[0018] Appendix Figure 1 This is a schematic diagram of the multi-objective optimization method for cell culture medium purification parameter tuning provided in an embodiment of the present invention.

[0019] Appendix Figure 2This is a schematic diagram of the structure of the multi-objective optimized cell culture medium purification parameter tuning system provided in the embodiment of the present invention.

[0020] The labels shown in the attached diagram:

[0021] The module includes: 11 for determining the target set, 12 for purification monitoring, 13 for predicting the impact of implicit purification, 14 for obtaining abnormal features of explicit purification, 15 for obtaining abnormal features of target purification, and 16 for obtaining the results of purification parameter optimization. Detailed Implementation

[0022] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined by the appended claims. It should be noted that the terms "comprising" and "having" are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules not explicitly listed or inherent to these processes, methods, products, or devices.

[0023] Example 1, as shown in the appendix Figure 1 As shown, this application provides a multi-objective optimization method for tuning cell culture medium purification parameters, wherein the method includes:

[0024] Step S100: Obtain the equipment environment information of the intermediate purification stage of the target cell culture medium. Based on the equipment environment information and the application scenario information of the target cell culture medium, match optimization targets and determine the set of optimization targets, wherein each optimization target has an optimization weight.

[0025] In one possible embodiment, equipment environment information refers to all environmental data related to the equipment used in the cell culture medium purification process, including temperature, humidity, pressure, flow rate, column length, and column diameter. Application scenario information refers to the actual application background and specific requirements of the cell culture medium, such as the properties of the target product, its application field (e.g., drug production, food processing), and specific requirements for purity, yield, etc. The optimization target set is a collection of multiple optimization targets matched based on the equipment environment information and application scenario information. For example, optimization targets may include the purity, yield, impurity content, and activity of the target product, each with a different optimization weight.

[0026] Preferably, the equipment environment is monitored in real time using multiple sensors (such as temperature sensors, flow rate sensors, pH sensors, and pressure sensors) deployed on the purification equipment. For example, the temperature sensor obtains the current operating temperature of the equipment, the flow rate sensor obtains the liquid flow rate, and the pH sensor monitors the liquid pH value. By utilizing a multi-objective optimization algorithm, taking the equipment environment information and application scenario information as input, the set of optimization objectives is obtained. Furthermore, the weights of the data in the set of optimization objectives are determined using the analytic hierarchy process (AHP), obtaining the optimization weight for each optimization objective. By determining the set of optimization objectives, the technical effect of providing constraints for efficient and precise parameter adjustment during cell culture medium purification is achieved.

[0027] Step S200: Call the sensor monitoring array to monitor the intermediate purification stage of the target cell culture medium, and store the purification monitoring data in the processing database to obtain real-time purification monitoring data and historical purification monitoring data sequences;

[0028] Furthermore, the sensor monitoring array includes at least a temperature sensor, a pH sensor, a flow rate sensor, a salt concentration sensor, and a charge change sensor.

[0029] In one embodiment, a sensor monitoring array continuously monitors the intermediate purification stage of the target cell culture medium, collecting data on environmental and process parameters during the purification process. This data includes temperature, pH, flow rate, salt concentration, and charge changes. Different purification monitoring data have different impacts on the intermediate purification stage. For example, changes in temperature and pH may affect the stability and separation efficiency of the target protein, while changes in flow rate, salt concentration, and charge directly affect the selectivity and purity of the separation medium.

[0030] The real-time purification monitoring data reflects the current level of purification of the target cell culture medium, while the historical purification monitoring data sequence reflects the level of purification already achieved in the target cell culture medium. Obtaining these two sets of data provides data support for subsequent analysis of both latent and overt purification processes, based on both real-time and historical perspectives.

[0031] Step S300: Call the low-frequency slow network to predict the implicit purification impact on the historical purification monitoring data sequence and obtain the implicit purification impact prediction features;

[0032] Furthermore, by invoking a low-frequency slow network to predict the implicit purification impact on the historical purification monitoring data sequence, and obtaining the implicit purification impact prediction features, step S300 of this embodiment further includes:

[0033] The low-frequency slow network is invoked to extract purification features from each historical purification monitoring data in the historical purification monitoring data sequence to obtain a historical purification feature sequence.

[0034] Iterative implicit purification analysis was performed on the historical purification feature sequence in chronological order to determine the iterative implicit purification features;

[0035] The implicit purification effect prediction feature is obtained by performing implicit purification effect prediction on the iterative implicit purification feature.

[0036] In one possible embodiment, the low-frequency slow network is a pre-built neural network model operating at a lower frequency. The input data is historical purification monitoring data, and the output data is historical purification features. The historical purification feature sequence is obtained by performing feature analysis on the historical purification monitoring data sequence using the low-frequency slow network. Then, iterative implicit correlation analysis is performed on the historical purification feature sequence to uncover implicit purification situations existing in the historical data. Based on the obtained iterative implicit purification features, implicit purification impact prediction is performed to obtain implicit purification impact prediction features based on historical situation analysis.

[0037] Preferably, multiple sample iterative implicit purification features and corresponding implicit purification effect prediction features are obtained as training data, and the framework constructed based on the feedforward neural network is trained under supervision until convergence is achieved, thereby obtaining the trained effect predictor. The effect predictor is then used to predict the implicit purification effect of the iterative implicit purification features to obtain the implicit purification effect prediction features.

[0038] Furthermore, in accordance with the time sequence, the historical purification feature sequences are subjected to iterative implicit purification analysis to determine the iterative implicit purification features. In this embodiment, step S300 further includes:

[0039] Extract the first historical purification feature and the second historical purification feature from the historical purification feature sequence;

[0040] An iterative implicit correlation analysis was performed on the first historical purification feature and the second historical purification feature to determine the first associated iterative implicit purification feature;

[0041] Using the first associated iterative implicit purification feature, iterative implicit association analysis is performed sequentially on the remaining historical purification features in the historical purification feature sequence to determine the iterative implicit purification feature.

[0042] Furthermore, an iterative implicit correlation analysis is performed on the first historical purification feature and the second historical purification feature to determine the first associated iterative implicit purification feature. In this embodiment, step S300 further includes:

[0043] Iterative implicit correlation analysis is performed on the same type of features in the first historical purification features and the second historical purification features respectively to obtain the first iterative implicit correlation coefficient set;

[0044] A first iterative implicit correlation vector is constructed based on the first iterative implicit correlation coefficient set, and convolution analysis is performed on the second historical purification feature based on the first iterative implicit correlation vector to obtain the first correlation iterative implicit purification feature.

[0045] In one embodiment, the implicit similarity between features of the same type in the first and second historical purified features is calculated using a cosine similarity function to obtain a first iterative implicit correlation coefficient set. The larger the first iterative implicit correlation coefficient, the higher the correlation between the corresponding feature types. Then, the first iterative implicit correlation coefficient set is normalized and added to an initially empty vector to obtain the first iterative implicit correlation vector.

[0046] A graph convolutional network is invoked to perform convolutional analysis on the first iterative implicit association vector and the second historical purification feature. Specifically, based on the association between the first and second historical purification features, the unrepresented features implicitly contained in the second historical purification feature are enhanced to obtain the first associated iterative implicit purification feature. Based on the principle of obtaining the first associated iterative implicit purification feature, iterative implicit association analysis is performed on the remaining historical purification features in the historical purification feature sequence in chronological order, using the first associated iterative implicit purification feature to determine the iterative implicit purification feature.

[0047] Step S400: Call the high-frequency fast network to identify the dominant purification abnormality features of the real-time purification monitoring data sequence, and obtain the dominant purification abnormality features;

[0048] Step S500: Based on the latent purification impact prediction features, the explicit purification anomaly features are corrected to obtain the target purification anomaly features;

[0049] In one possible embodiment, a high-frequency fast network is used to rapidly analyze real-time acquired purification monitoring data to identify obvious purification anomalies. These obvious anomalies are readily apparent features that can be directly extracted from the purification monitoring data. Examples include rapid temperature changes, abnormal flow rate fluctuations, and rapid deviations in pH from the target value. Utilizing the high-frequency processing capabilities of the fast network, real-time data is quickly analyzed to identify these obvious purification anomalies.

[0050] By calculating the mean of the latent purification impact prediction feature against the explicit purification anomaly feature, the purification status reflected by real-time data and historical data is fused to obtain the target purification anomaly feature that can reflect the current purification status.

[0051] Step S600: Using the set of optimization targets as constraints, and combining the abnormal characteristics of the target purification, perform multi-objective optimization of the purification parameters to obtain the purification parameter optimization results.

[0052] Furthermore, using the optimized target set as a constraint, and combining the target purification anomaly characteristics, multi-objective optimization of the purification parameters is performed to obtain the purification parameter optimization results. In this embodiment, step S600 further includes:

[0053] A pre-constructed purification parameter identifier is used to identify the target purification anomaly features and obtain multiple purification parameters.

[0054] The multiple purification parameters are simulated respectively, and the set of simulation results is compared with the set of optimization targets to obtain multiple achievement scores;

[0055] Using the purification parameter corresponding to the maximum value among the multiple standard achievement scores as the direction, the multiple purification parameters are adjusted according to the first adjustment method to obtain multiple adjusted purification parameters, wherein the first adjustment method is to randomly adjust the purification parameters according to a preset adjustment quantity and a preset adjustment range;

[0056] The multiple adjusted purification parameters are scored again to determine whether the maximum value among the multiple adjusted scores is greater than the maximum value among the multiple adjusted scores. If so, the adjusted purification parameter corresponding to the maximum value among the multiple adjusted scores is taken as the purification parameter optimization result.

[0057] Furthermore, if the maximum value among the multiple adjustment achievement scores is less than or equal to the maximum value among the multiple achievement scores, a readjustment instruction is obtained, and the multiple purification parameters are randomly adjusted and optimized again according to the first adjustment method.

[0058] In one embodiment, a purification parameter identifier is constructed to identify purification parameters that need optimization based on abnormal characteristics of the target purification process. The purification parameter identifier can determine adjustable purification parameters, such as temperature, pH, flow rate, and salt concentration, based on historical data, real-time data, and optimization targets. Simulations are then performed on multiple identified purification parameters. Preferably, multiple rounds of simulation are conducted using a computer simulation tool, based on a set range of purification parameters. Different combinations of purification parameters are input in each round of simulation, and the simulation results reflect the purification effect under that combination. Each combination of purification parameters generates a set of results (such as target purity, yield, and impurity removal). The similarity between the optimized target set and the simulation results is used as the achievement score. For example, different temperatures (e.g., 37°C, 40°C, 42°C), flow rates (e.g., 1 mL / min, 2 mL / min, 3 mL / min), and salt concentrations (e.g., 0.1 M, 0.2 M, 0.3 M) are input, and the simulation program is run to obtain data on purity, yield, and impurity removal under different combinations.

[0059] Using the purification parameter corresponding to the maximum value among the multiple target achievement scores as a direction, the parameters among the multiple purification parameters with a preset adjustment number are randomly adjusted according to a preset adjustment range, thereby obtaining multiple adjusted purification parameters. The multiple adjusted purification parameters are then simulated again, and target achievement scores are obtained based on the simulation results, yielding multiple adjusted target achievement scores. It is then determined whether the maximum value among the multiple adjusted target achievement scores is greater than the maximum value among the multiple target achievement scores. If so, it indicates that the corresponding adjusted purification parameter is more suitable for the current purification situation, and the adjusted purification parameter corresponding to the maximum value among the multiple adjusted target achievement scores is taken as the optimized purification parameter result.

[0060] If the maximum value among multiple adjustment achievement scores is less than or equal to the maximum value among multiple achievement scores, it indicates that the optimization has failed. In this case, a readjustment instruction is obtained, and the multiple purification parameters are randomly adjusted and optimized again according to the first adjustment method. Preferably, the number of readjustments is counted. When the number of readjustments is greater than or equal to a preset readjustment threshold, the purification parameter corresponding to the maximum value among the multiple achievement scores is taken as the purification parameter optimization result. By continuously simulating, evaluating, and adjusting the purification parameters, a balance is found among multiple optimization objectives, and finally, a purification parameter optimization result that meets the requirements of each objective is determined.

[0061] In summary, the embodiments of this application have at least the following technical effects:

[0062] This application dynamically determines the set of optimization targets, precisely sets the priorities among the targets, and combines real-time sensor data and historical data sequences to monitor the purification process in real time. It uses a low-frequency, slow network to predict latent purification effects and a high-frequency, fast network to identify explicit purification anomalies, thereby obtaining corrected target purification anomalies. Simultaneously, a multi-objective optimization algorithm is used to optimize purification parameters, balancing multiple optimization objectives (such as purity, yield, and activity) to achieve optimal purification results. This achieves the technical effects of improving the efficiency of the purification process and the quality of the target product, ensuring precise control and dynamic adjustment of the cell culture medium purification process.

[0063] Example 2, based on the same inventive concept as the multi-objective optimization method for cell culture medium purification parameter tuning in the foregoing examples, as shown in the appendix. Figure 2 As shown, this application provides a multi-objective optimized cell culture medium purification parameter tuning system. The system and method embodiments in this application are based on the same inventive concept. The system includes:

[0064] The optimization target set determination module 11 is used to obtain the equipment environment information of the intermediate purification stage of the target cell culture medium, and based on the equipment environment information and the application scenario information of the target cell culture medium, match optimization targets and determine the optimization target set, wherein each optimization target has an optimization weight.

[0065] The purification monitoring module 12 is used to call the sensor monitoring array to monitor the intermediate purification stage of the target cell culture medium, and store the purification monitoring data in the processing database to obtain real-time purification monitoring data and historical purification monitoring data sequences.

[0066] The implicit purification impact prediction module 13 is used to call a low-frequency slow network to predict the implicit purification impact of the historical purification monitoring data sequence and obtain the implicit purification impact prediction features.

[0067] The dominant purification anomaly feature acquisition module 14 is used to call a high-frequency fast network to identify dominant purification anomaly features in the real-time purification monitoring data sequence and obtain dominant purification anomaly features.

[0068] The target purification anomaly feature acquisition module 15 is used to correct the explicit purification anomaly feature based on the implicit purification impact prediction feature to obtain the target purification anomaly feature;

[0069] The purification parameter optimization result acquisition module 16 is used to perform multi-objective optimization of purification parameters with the set of optimization targets as constraints and in combination with the target purification anomaly characteristics, so as to obtain the purification parameter optimization result.

[0070] Furthermore, the sensor monitoring array includes at least a temperature sensor, a pH sensor, a flow rate sensor, a salt concentration sensor, and a charge change sensor.

[0071] Furthermore, the latent purification effect prediction module 13 is used to perform the following steps:

[0072] The low-frequency slow network is invoked to extract purification features from each historical purification monitoring data in the historical purification monitoring data sequence to obtain a historical purification feature sequence.

[0073] Iterative implicit purification analysis was performed on the historical purification feature sequence in chronological order to determine the iterative implicit purification features;

[0074] The implicit purification effect prediction feature is obtained by performing implicit purification effect prediction on the iterative implicit purification feature.

[0075] Furthermore, the latent purification effect prediction module 13 is used to perform the following steps:

[0076] Extract the first historical purification feature and the second historical purification feature from the historical purification feature sequence;

[0077] An iterative implicit correlation analysis was performed on the first historical purification feature and the second historical purification feature to determine the first associated iterative implicit purification feature;

[0078] Using the first associated iterative implicit purification feature, iterative implicit association analysis is performed sequentially on the remaining historical purification features in the historical purification feature sequence to determine the iterative implicit purification feature.

[0079] Furthermore, the latent purification effect prediction module 13 is used to perform the following steps:

[0080] Iterative implicit correlation analysis is performed on the same type of features in the first historical purification features and the second historical purification features respectively to obtain the first iterative implicit correlation coefficient set;

[0081] A first iterative implicit correlation vector is constructed based on the first iterative implicit correlation coefficient set, and convolution analysis is performed on the second historical purification feature based on the first iterative implicit correlation vector to obtain the first correlation iterative implicit purification feature.

[0082] Furthermore, the purification parameter optimization result acquisition module 16 is used to perform the following steps:

[0083] A pre-constructed purification parameter identifier is used to identify the target purification anomaly features and obtain multiple purification parameters.

[0084] The multiple purification parameters are simulated respectively, and the set of simulation results is compared with the set of optimization targets to obtain multiple achievement scores;

[0085] Using the purification parameter corresponding to the maximum value among the multiple standard achievement scores as the direction, the multiple purification parameters are adjusted according to the first adjustment method to obtain multiple adjusted purification parameters, wherein the first adjustment method is to randomly adjust the purification parameters according to a preset adjustment quantity and a preset adjustment range;

[0086] The multiple adjusted purification parameters are scored again to determine whether the maximum value among the multiple adjusted scores is greater than the maximum value among the multiple adjusted scores. If so, the adjusted purification parameter corresponding to the maximum value among the multiple adjusted scores is taken as the purification parameter optimization result.

[0087] Furthermore, if the maximum value among the multiple adjustment achievement scores is less than or equal to the maximum value among the multiple achievement scores, a readjustment instruction is obtained, and the multiple purification parameters are randomly adjusted and optimized again according to the first adjustment method.

[0088] It should be noted that the order of the embodiments described above is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. Furthermore, the above description focuses on specific embodiments of this specification. Additionally, the processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired results. In some implementations, multitasking and parallel processing are possible or may be advantageous.

[0089] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

[0090] This specification and accompanying drawings are merely illustrative examples of this application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Clearly, those skilled in the art can make various alterations and modifications to this application without departing from its scope. Therefore, if such modifications and modifications fall within the scope of this application and its equivalents, this application intends to include such modifications and modifications.

Claims

1. A multi-objective optimization method for tuning cell culture medium purification parameters, characterized in that, The method includes: Obtain the equipment environment information of the intermediate purification stage of the target cell culture medium. Using the equipment environment information as a constraint and the application scenario information of the target cell culture medium as a basis, match optimization targets and determine the set of optimization targets, wherein each optimization target has an optimization weight. The sensor monitoring array is invoked to monitor the intermediate purification stage of the target cell culture medium, and the purification monitoring data is stored in the processing database to obtain real-time purification monitoring data and historical purification monitoring data sequences. A low-frequency slow network is invoked to predict the implicit purification impact on the historical purification monitoring data sequence, and the implicit purification impact prediction features are obtained. A high-frequency fast network is invoked to identify explicit purification anomaly features in the real-time purification monitoring data sequence, thereby obtaining explicit purification anomaly features; Based on the implicit purification impact prediction features, the explicit purification anomaly features are modified to obtain the target purification anomaly features; By using the set of optimization objectives as constraints and combining the abnormal characteristics of the target purification, the purification parameters are optimized in multiple objectives to obtain the purification parameter optimization results.

2. The method for optimizing cell culture medium purification parameters according to claim 1, characterized in that, The sensor monitoring array includes at least a temperature sensor, a pH sensor, a flow rate sensor, a salt concentration sensor, and a charge change sensor.

3. The method for optimizing cell culture medium purification parameters according to claim 1, characterized in that, A low-frequency slow network is invoked to predict the implicit purification impact on the historical purification monitoring data sequence, obtaining the implicit purification impact prediction features, including: The low-frequency slow network is invoked to extract purification features from each historical purification monitoring data in the historical purification monitoring data sequence to obtain a historical purification feature sequence. Iterative implicit purification analysis was performed on the historical purification feature sequence in chronological order to determine the iterative implicit purification features; The implicit purification effect prediction feature is obtained by performing implicit purification effect prediction on the iterative implicit purification feature.

4. The method for optimizing cell culture medium purification parameters according to claim 3, characterized in that, Iterative implicit purification analysis was performed on the historical purification feature sequences in chronological order to determine the iterative implicit purification features, including: Extract the first historical purification feature and the second historical purification feature from the historical purification feature sequence; An iterative implicit correlation analysis was performed on the first historical purification feature and the second historical purification feature to determine the first associated iterative implicit purification feature; Using the first associated iterative implicit purification feature, iterative implicit association analysis is performed sequentially on the remaining historical purification features in the historical purification feature sequence to determine the iterative implicit purification feature.

5. The method for optimizing cell culture medium purification parameters according to claim 4, characterized in that, An iterative implicit association analysis was performed on the first historical purification feature and the second historical purification feature to determine the first associated iterative implicit purification feature, including: Iterative implicit correlation analysis is performed on the same type of features in the first historical purification features and the second historical purification features respectively to obtain the first iterative implicit correlation coefficient set; A first iterative implicit correlation vector is constructed based on the first iterative implicit correlation coefficient set, and convolution analysis is performed on the second historical purification feature based on the first iterative implicit correlation vector to obtain the first correlation iterative implicit purification feature.

6. The method for optimizing cell culture medium purification parameters according to claim 1, characterized in that, Using the set of optimization objectives as constraints, and combining the abnormal characteristics of the target purification, multi-objective optimization of purification parameters is performed to obtain the optimized purification parameters, including: A pre-constructed purification parameter identifier is used to identify the target purification anomaly features and obtain multiple purification parameters. The multiple purification parameters are simulated respectively, and the set of simulation results is compared with the set of optimization targets to obtain multiple achievement scores; Using the purification parameter corresponding to the maximum value among the multiple standard achievement scores as the direction, the multiple purification parameters are adjusted according to the first adjustment method to obtain multiple adjusted purification parameters, wherein the first adjustment method is to randomly adjust the purification parameters according to a preset adjustment quantity and a preset adjustment range; The multiple adjusted purification parameters are scored again to determine whether the maximum value among the multiple adjusted scores is greater than the maximum value among the multiple adjusted scores. If so, the adjusted purification parameter corresponding to the maximum value among the multiple adjusted scores is taken as the purification parameter optimization result.

7. The method for optimizing cell culture medium purification parameters according to claim 6, characterized in that, If the maximum value among multiple adjustment achievement scores is less than or equal to the maximum value among the multiple achievement scores, then a readjustment instruction is obtained, and the multiple purification parameters are randomly adjusted and optimized again according to the first adjustment method.

8. A multi-objective optimized cell culture medium purification parameter tuning system, characterized in that, The system is used to implement the multi-objective optimized cell culture medium purification parameter tuning method according to any one of claims 1-7, the system comprising: The optimization target set determination module is used to obtain the equipment environment information of the intermediate purification stage of the target cell culture medium, and based on the equipment environment information and the application scenario information of the target cell culture medium, match optimization targets and determine the optimization target set, wherein each optimization target has an optimization weight; The purification monitoring module is used to call the sensor monitoring array to monitor the intermediate purification stage of the target cell culture medium, and store the purification monitoring data in the processing database to obtain real-time purification monitoring data and historical purification monitoring data sequences. The implicit purification impact prediction module is used to call a low-frequency slow network to predict the implicit purification impact on the historical purification monitoring data sequence and obtain the implicit purification impact prediction features. The dominant purification anomaly feature acquisition module is used to call a high-frequency fast network to identify dominant purification anomaly features in the real-time purification monitoring data sequence and obtain dominant purification anomaly features. The target purification anomaly feature acquisition module is used to correct the explicit purification anomaly features based on the implicit purification impact prediction features to obtain the target purification anomaly features; The purification parameter optimization result acquisition module is used to perform multi-objective optimization of purification parameters by taking the set of optimization targets as constraints and combining the target purification anomaly characteristics to obtain the purification parameter optimization result.

Images