Comprehensive quality management method and robot system for pharmaceutical enterprise
By obtaining and analyzing the process of process, quality detection and environmental parameters in the process of medicinal materials, and using quality model and large language model optimization algorithm to generate control strategies, the problem of Chinese medicine properties deviation in medicinal materials processing is solved, and the comprehensive improvement of drug quality and cost optimization is achieved. , reducing manual intervention.
Patent Information
- Application Number
- CN202411091893.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-09
- Publication Date
- 2024-11-22
AI Technical Summary
Due to incomplete analysis during the process of medicinal materials, existing pharmaceutical companies have deviations in the properties of traditional Chinese medicines, and they are unable to effectively deal with the overall situation of medicinal materials.
Total quality management method is adopted, by obtaining process parameters, quality detection data and environmental parameters for cleaning and storage, using quality models for prediction and outlier judgment, using large language models to analyze the causes of abnormalities, and generating the best quality control strategy through optimization algorithms and parameters, automatically output control instructions through RPA technology.
A comprehensive and accurate data support for drug quality is achieved, factors affecting drug quality are discovered, drug quality is improved and cost is reduced, manual intervention is reduced, and visual decision support tools are provided.
Smart Images

Figure CN119005799A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of information and communication technology, and in particular to a comprehensive quality management method and a robot system for a pharmaceutical enterprise. Background Art
[0002] As a product directly related to the life safety of patients, the quality of drugs is of vital importance. Through strict monitoring and continuous improvement of the entire process of drug production, total quality management ensures that the raw materials, production, packaging and other links of drugs meet high quality requirements, thereby improving the overall quality level of drugs and ensuring the safety of patients' medication.
[0003] In recent years, the concept of intelligent manufacturing has been gradually promoted in the pharmaceutical industry, aiming to improve the operational efficiency and product quality of pharmaceutical companies through digitalization, networking and intelligent means. As an emerging automation technology, robotic process automation has begun to be applied to intelligent management in the manufacturing field.
[0004] At present, some pharmaceutical companies are already carrying out automated production. By setting up automated equipment in different production links, the processing and processing of medicinal materials are monitored, so that problems can be discovered in time during the processing of medicinal materials. However, in the current automated production, the data of each link is monitored by independent automated equipment, and when problems are found, alarms are usually issued and handled by manual intervention. However, in the processing of medicinal materials, simple adjustments to process parameters will also have a significant impact on the medicinal properties of the traditional Chinese medicines made from the medicinal materials. If problems are dealt with in a single link, it is impossible to analyze the overall processing of the medicinal materials, which leads to deviations in the medicinal properties of the traditional Chinese medicines produced. Summary of the invention
[0005] In order to solve the problem that incomplete analysis during the processing of medicinal materials leads to deviations in the medicinal properties of manufactured traditional Chinese medicines, the present invention provides a comprehensive quality management method and a robot system for a pharmaceutical enterprise.
[0006] The present invention provides a comprehensive quality management method for pharmaceutical enterprises, which adopts the following technical solutions:
[0007] A comprehensive quality management method for a pharmaceutical enterprise comprises the following steps:
[0008] Acquiring process parameters, quality inspection data and environmental parameters, cleaning the process parameters, the quality inspection data and the environmental parameters, and then storing them in a database;
[0009] The quality of the drug produced by the process parameters and the environmental parameters is predicted by the quality model, and abnormal values are determined by a preset index interval. If an abnormal value exists, the cause of the abnormality is analyzed by a large language model;
[0010] Optimizing the process parameters and the environmental parameters by using an optimization algorithm to obtain an optimal quality control strategy and optimal quality control parameters;
[0011] The large language model obtains the optimal quality control strategy and the optimal quality control parameters, and breaks them down into control instructions to output to production equipment.
[0012] In a specific implementation scheme, the method for obtaining the process parameters, the quality detection data and the environmental parameters is:
[0013] Screen shots of the production equipment and the testing equipment are obtained, and parameters in the screen shots are extracted through an image processing algorithm.
[0014] In a specific implementation scheme, the method for constructing the quality model is:
[0015] Acquire historical data, and cluster the historical data using a clustering algorithm, wherein the historical data includes historical process parameters, historical quality inspection data, and historical environmental parameters.
[0016] A group of the historical process parameters, the historical quality inspection data and the historical environmental parameters corresponding to each other is a data point, K data points are randomly selected as initial cluster centers, and the data points are assigned to the nearest initial cluster center to form a cluster;
[0017] For each of the clusters, the cluster center is recalculated until the maximum number of iterations.
[0018] In a specific implementation scheme, the formula:
[0019]
[0020] Calculate the Euclidean distance dist between the data points and assign the data points to the nearest initial cluster center, where x is the data point and c is the nearest initial cluster center. i is the i-th initial cluster center, i=1, 2, ..., K, x j is the value of the data point x in the jth dimension, c ij is the value of the i-th initial cluster center in the j-th dimension.
[0021] In a specific implementation scheme, the method for determining the abnormal value is:
[0022] Constructing a quality model of the process parameters, the environmental parameters and the quality inspection data, and using historical data to train the model;
[0023] After the process parameters, the environmental parameters and the quality inspection data are input into the quality model, the predicted quality data is output, and the predicted quality data is compared with the index range. If the predicted quality data is within the index range, it indicates that the quality of the traditional Chinese medicine is normal. If the predicted quality data is not within the index range, it indicates that there is excessive fluctuation in the quality of the traditional Chinese medicine, and the process parameters and the environmental parameters need to be adjusted.
[0024] In a specific implementation scheme, the quality model is a multimodal neural network model.
[0025] In a specific possible implementation scheme, it includes a mobile terminal, which responds to user operations, obtains the optimal quality control strategy, the process parameters, the quality detection data and the environmental parameters, generates charts and outputs them.
[0026] In a specific implementation scheme, the method for cleaning the process parameters, the quality detection data and the environmental parameters is:
[0027] First, denoise by median filtering, then delete duplicate data, and then perform maximum and minimum normalization.
[0028] The present invention provides a pharmaceutical enterprise comprehensive quality management robot system, which adopts the following technical solutions:
[0029] A pharmaceutical enterprise total quality management robot system, comprising:
[0030] A data acquisition module, which acquires historical data, obtains screenshots of production equipment and testing equipment, extracts process parameters, quality inspection data and environmental parameters from the screenshots through image processing algorithms, and uploads them;
[0031] The data processing and analysis module includes a data cleaning unit, a data storage unit and a data analysis unit. The data cleaning unit obtains the process parameters, the quality inspection data and the environmental parameters uploaded by the data acquisition module and performs data cleaning. The data storage unit obtains the process parameters, the quality inspection data and the environmental parameters after data cleaning and stores them in a preset database. The data analysis unit calls the process parameters, the quality inspection data and the environmental parameters in the database and performs analysis.
[0032] The quality decision and optimization module includes a quality model construction unit, an anomaly detection unit and an optimization algorithm unit. The quality model construction unit constructs a quality model and performs model training through the historical data; the anomaly detection unit calls the quality model and the process parameters, the quality detection data and the environmental parameters, calculates the predicted quality data, and compares the predicted quality data with a preset index interval. If the predicted quality data is within the index interval, it indicates that the quality of the traditional Chinese medicine is normal; if the predicted quality data is not within the index interval, it indicates that the quality of the traditional Chinese medicine has excessive fluctuations, and the process parameters and the environmental parameters need to be adjusted; the optimization algorithm unit optimizes the process parameters and the environmental parameters through an optimization algorithm with the optimization goals of improving quality and reducing costs, and generates an optimal quality control strategy and optimal quality control parameters;
[0033] The control execution module includes a decision generation unit and an equipment control unit. The decision generation unit obtains the optimal quality control strategy and decomposes the optimal quality control strategy and the optimal quality control parameters into control instructions through a large language model; the equipment control unit obtains the control instructions and outputs them to the production equipment.
[0034] In a specific possible implementation scheme, it also includes a mobile control module and a display module. The mobile control module responds to user operations, calls the process parameters, the quality inspection data, the environmental parameters and the control instructions to output to the display module; after the display module obtains the process parameters, the quality inspection data, the environmental parameters and the control instructions, it performs visual output.
[0035] In summary, the present invention includes at least one of the following beneficial technical effects:
[0036] 1. Through the automated collection and multi-source integration of process parameters, quality inspection data and environmental parameters, comprehensive and accurate data support is provided for quality management, thus improving the quality of drugs.
[0037] 2. Through comprehensive analysis of process parameters, quality inspection data and environmental parameters, factors affecting drug quality and potential problems in the drug production process can be discovered, facilitating timely optimization strategies.
[0038] 3. Take improving quality and reducing costs as the optimization goals of the multivariate linear regression model to achieve dual optimization of quality and cost.
[0039] 4. The control execution module automatically generates control instructions. By combining RPA technology, it can realize automated execution and closed-loop control of quality management and reduce manual intervention.
[0040] 5. The display module intuitively presents real-time data of production equipment and testing equipment, providing a visual, predictable and diagnosable decision support tool for quality management. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] Figure 1 It is a flowchart of the total quality management approach for pharmaceutical companies.
[0042] Figure 2 It is a block diagram of a total quality management robotic system for pharmaceutical companies.
[0043] Explanation of the accompanying drawings: 1. Data acquisition module; 2. Data processing and analysis module; 21. Data cleaning unit; 22. Data storage unit; 23. Data analysis unit; 3. Quality decision and optimization module; 31. Quality model building unit; 32. Anomaly detection unit; 33. Optimization algorithm unit; 4. Control execution module; 41. Decision making unit; 42. Equipment control unit; 5. Mobile control module; 6. Display module. DETAILED DESCRIPTION
[0044] The following is combined with Figure 1-2 The present invention is described in further detail.
[0045] RPA technology is an automation technology that automatically executes preset and repetitive business processes by simulating human operations on computers, such as keyboard input, mouse clicks, data copying and pasting, etc. Software robots (or "virtual labor") are used to simulate human operations to complete tasks that originally required manual execution.
[0046] Reference Figure 1 , the pharmaceutical company's total quality management approach includes the following steps:
[0047] S100, data acquisition.
[0048] Get historical data. Get screenshots of production equipment and testing equipment, and extract process parameters, quality inspection data, and environmental parameters from the screenshots through image processing algorithms. Among them, process parameters are parameters achieved during the processing of Chinese medicinal materials, such as drying temperature, drying time, boiling time, boiling temperature, etc. Quality inspection data are extract content, active ingredient content, etc. Environmental parameters refer to the environmental parameters around the processing of Chinese medicinal materials, such as workshop temperature, workshop humidity, workshop air circulation, etc. Historical data includes historical process parameters, historical quality inspection data, and historical environmental parameters under different circumstances in previous production processes.
[0049] S200, data cleaning, and analysis of historical data through clustering algorithms.
[0050] The acquired process parameters, quality inspection data and environmental parameters are first denoised through median filtering, then duplicate data are deleted, and then maximum and minimum normalization is performed to complete data cleaning.
[0051] The process parameters, quality inspection data and environmental parameters of the completed cleaning are stored in a preset database. The database is specifically a distributed database, thereby improving the reliability and accessibility of the process parameters, quality inspection data and environmental parameters.
[0052] The historical data is analyzed by clustering algorithm, which is specifically K-means algorithm. A set of process parameters, quality inspection data and environmental parameters in the historical data is taken as a data point, and K data points are randomly selected as the initial clustering centers. By formula:
[0053]
[0054] Calculate the distance between each data point and the initial cluster center, assign the data point to the nearest initial cluster center, and form a cluster. i is the i-th initial cluster center, i = 1, 2, ..., K, dist is the Euclidean distance between data points, x j is the value of the data point x in the jth dimension, c ij is the value of the i-th initial cluster center in the j-th dimension. By formula:
[0055]
[0056] Recalculate the cluster center of each cluster. In the above formula, s i is the set of data points of the cluster to which the i-th initial cluster center belongs. Repeatedly convert the data in the database into data points and assign them to the nearest cluster center, then recalculate the cluster center of the re-divided cluster, and repeat the above steps until the maximum number of iterations.
[0057] For each cluster, the cluster center is recalculated until the maximum number of iterations is reached. The data analyzed by the clustering algorithm is stored again in the database according to the clustering results.
[0058] S300, builds a quality model, judges outliers, and obtains the best quality control strategy and best quality control parameters.
[0059] Construct a quality model of process parameters, environmental parameters and quality inspection data. The quality model is specifically a multimodal neural network. Input the current process parameters, quality inspection data and environmental parameters into the multimodal neural network, predict the medicinal properties of the possible Chinese medicine through the multimodal neural network, and output the predicted quality data. Compare the predicted quality data output by the multimodal neural network with the preset index range. If the predicted quality data is within the index range, it indicates that the quality of the Chinese medicine is normal and production can continue; if the predicted quality data is not within the index range, it indicates that there are abnormal values in the process parameters and environmental parameters, resulting in excessive fluctuations in the quality of the Chinese medicine, which may further lead to a decrease in the medicinal properties and quality of the Chinese medicine, and the process parameters and environmental parameters need to be adjusted.
[0060] The process parameters and environmental parameters with abnormal values are input into the large language model through the API interface, and the large language model infers the abnormal reasons that lead to the abnormal values. The large language model can be ChatGPT, etc.
[0061] The optimization algorithm obtains the predicted quality data, process parameters and environmental parameters with abnormal values, and the abnormal reasons output by the large language model, optimizes the process parameters and environmental parameters with the optimization goals of improving quality and reducing costs, and outputs the best quality control strategy and best quality control parameters. The optimization algorithm is specifically a reinforcement learning algorithm.
[0062] Before using the multimodal neural network to predict the medicinal properties of traditional Chinese medicine, the multimodal neural network model is first trained using historical data. Although historical data is preprocessed before being used for model training, it is difficult to completely avoid outliers in the historical data. The clustering in S200 can clearly show the distribution of historical data around the cluster center. The farther the data point is from the cluster center, the higher the possibility that the data point has an outlier, which makes it easier to find data points that may have outliers.
[0063] S400, generating control instructions and outputting them to the production equipment.
[0064] The large language model obtains the best quality control strategy and the best quality control parameters, breaks them down into specific control instructions, and outputs them to the production equipment through RPA technology, so that the production equipment can carry out subsequent production according to the adjusted process parameters and environmental parameters.
[0065] S500, data visualization output.
[0066] The process parameters, quality inspection data and environmental parameters in the database, the control instructions obtained by disassembling the large language model, and the prediction of the changes in process parameters, quality inspection data and environmental parameters after the implementation of the control instructions are output in the form of charts.
[0067] It should be noted that a mobile terminal can be set up to respond to user operations and display chart data and control instructions to users. The mobile terminal can also be connected to production equipment, etc., so that users can control the production equipment through the mobile terminal.
[0068] Reference Figure 2 The present invention also discloses a pharmaceutical enterprise comprehensive quality management robot system, comprising:
[0069] Data acquisition module 1 obtains historical data, obtains screenshots of production equipment and testing equipment, extracts process parameters, quality inspection data and environmental parameters from the screenshots through image processing algorithms, and uploads them.
[0070] The data processing and analysis module 2 includes a data cleaning unit 21, a data storage unit 22 and a data analysis unit 23. The data cleaning unit 21 obtains the process parameters, quality inspection data and environmental parameters uploaded by the data acquisition module 1, and performs data cleaning functions such as median filtering, deleting duplicate data, and maximum and minimum normalization. The data storage unit 22 obtains historical data, as well as the process parameters, quality inspection data and environmental parameters that have completed data cleaning, and stores them in a distributed database. The data analysis unit 23 calls historical data, performs clustering analysis through the K-means algorithm, and re-stores the clustered historical data in the database.
[0071] The quality decision and optimization module 3 includes a quality model building unit 31, an anomaly detection unit 32 and an optimization algorithm unit 33. The quality model building unit 31 builds a multimodal neural network as a quality model and obtains historical data in a database to perform model training on the quality model.
[0072] The abnormality detection unit 32 calls the quality model, obtains the process parameters, quality detection data and environmental parameters in the database, predicts the medicinal properties of the medicine made from the Chinese medicine currently being processed, and outputs the predicted quality data. The predicted quality data is compared with the preset index interval. If the predicted quality data is within the index interval, it indicates that the quality of the Chinese medicine is normal. If the predicted quality data is not within the index interval, it indicates that there is too much fluctuation in the quality of the Chinese medicine, and there are abnormal values in the process parameters and environmental parameters, and the process parameters and environmental parameters need to be adjusted. The process parameters and environmental parameters with abnormal values are input into the large language model through the API interface, and the abnormal reasons for the abnormal values are inferred by the large language model.
[0073] The optimization algorithm unit 33 obtains the cause of the abnormality, and predicts the quality data, process parameters and environmental parameters, takes improving quality and reducing costs as the optimization goals, solves the optimization plan, and outputs the best quality control strategy and the best quality control parameters.
[0074] The control execution module 4 includes a decision generation unit 41 and an equipment control unit 42. The decision generation unit 41 obtains the optimal quality control strategy and decomposes the optimal quality control strategy and optimal quality control parameters into specific control instructions through a large language model; the equipment control unit 42 obtains the control instructions and outputs them to the production equipment through RPA technology.
[0075] The equipment control unit 42 outputs the control instruction to the data acquisition module 1, so that when the data acquisition module 1 obtains the process parameters, environmental parameters and quality inspection data of the production equipment and the inspection equipment, it can compare them with the control instruction. If the process parameters, environmental parameters and quality inspection data meet the control instruction, production continues; if the process parameters, environmental parameters and quality inspection data do not meet the control instruction, steps S200-S400 of the comprehensive quality management method for pharmaceutical enterprises are cycled until the process parameters, environmental parameters and quality inspection data meet the control instruction, thereby realizing loop control.
[0076] Mobile control module 5, mobile control module 5 responds to user operations, calls process parameters, quality inspection data, environmental parameters and control instructions and outputs them; and responds to user operations, analyzes user operations, generates manual instructions, and outputs the manual instructions to the equipment control unit 42, which outputs them to the production equipment through RPA technology, so as to facilitate remote control of the production equipment.
[0077] The display module 6 obtains the process parameters, quality inspection data, environmental parameters and control instructions of the mobile control module 5 or the data storage unit 22, and outputs them visually in the form of charts.
[0078] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.
Claims
1. A comprehensive quality management method for a pharmaceutical enterprise, characterized by: The steps include: Acquiring process parameters, quality inspection data and environmental parameters, cleaning the process parameters, the quality inspection data and the environmental parameters, and then storing them in a database; The quality of the drug produced by the process parameters and the environmental parameters is predicted by the quality model, and abnormal values are determined by a preset index interval. If an abnormal value exists, the cause of the abnormality is analyzed by a large language model; Optimizing the process parameters and the environmental parameters by using an optimization algorithm to obtain an optimal quality control strategy and optimal quality control parameters; The large language model obtains the optimal quality control strategy and the optimal quality control parameters, and breaks them down into control instructions to output to production equipment.
2. The pharmaceutical enterprise comprehensive quality management method according to claim 1, characterized in that: The method for obtaining the process parameters, the quality inspection data and the environmental parameters is: Screen shots of the production equipment and the testing equipment are obtained, and parameters in the screen shots are extracted through an image processing algorithm.
3. The pharmaceutical enterprise comprehensive quality management method according to claim 1, characterized in that: The method for constructing the quality model is: Acquire historical data, and cluster the historical data using a clustering algorithm, wherein the historical data includes historical process parameters, historical quality inspection data, and historical environmental parameters. A group of the historical process parameters, the historical quality inspection data and the historical environmental parameters corresponding to each other is a data point, K data points are randomly selected as initial cluster centers, and the data points are assigned to the nearest initial cluster center to form a cluster; For each of the clusters, the cluster center is recalculated until the maximum number of iterations.
4. The pharmaceutical enterprise total quality management method according to claim 3, characterized in that: By formula: Calculate the Euclidean distance dist between the data points and assign the data points to the nearest initial cluster center, where x is the data point and c is the nearest initial cluster center. i is the i-th initial cluster center, i=1,2,…,K, x j is the value of the data point x in the jth dimension, c ij is the value of the i-th initial cluster center in the j-th dimension.
5. The pharmaceutical enterprise comprehensive quality management method according to claim 1, characterized in that: The method for determining the abnormal value is as follows: Constructing a quality model of the process parameters, the environmental parameters and the quality inspection data, and using historical data to train the model; After the process parameters, the environmental parameters and the quality inspection data are input into the quality model, the predicted quality data is output, and the predicted quality data is compared with the index range. If the predicted quality data is within the index range, it indicates that the quality of the traditional Chinese medicine is normal. If the predicted quality data is not within the index range, it indicates that there is excessive fluctuation in the quality of the traditional Chinese medicine, and the process parameters and the environmental parameters need to be adjusted.
6. The pharmaceutical enterprise total quality management method according to claim 5, characterized in that: The quality model is a multimodal neural network model.
7. The pharmaceutical enterprise total quality management method according to claim 1, characterized in that: It includes a mobile terminal, which responds to user operations, obtains the optimal quality control strategy, the process parameters, the quality detection data and the environmental parameters, generates charts and outputs them.
8. The pharmaceutical enterprise total quality management method according to claim 1, characterized in that: The method for cleaning the process parameters, the quality inspection data and the environmental parameters is: First, denoise by median filtering, then delete duplicate data, and then perform maximum and minimum normalization.
9. A pharmaceutical enterprise total quality management robot system, characterized by: include: A data acquisition module (1) acquires historical data, obtains screenshots of production equipment and testing equipment, extracts process parameters, quality test data and environmental parameters from the screenshots through image processing algorithms, and uploads them; The data processing and analysis module (2) comprises a data cleaning unit (21), a data storage unit (22) and a data analysis unit (23), wherein the data cleaning unit (21) obtains the process parameters, the quality inspection data and the environmental parameters uploaded by the data acquisition module (1) and performs data cleaning; the data storage unit (22) obtains the process parameters, the quality inspection data and the environmental parameters after data cleaning and stores them in a preset database; A data analysis unit (23) calls the process parameters, the quality inspection data and the environmental parameters in the database and performs analysis; The quality decision and optimization module (3) comprises a quality model construction unit (31), an anomaly detection unit (32) and an optimization algorithm unit (33), wherein the quality model construction unit (31) constructs a quality model and performs model training through the historical data; the anomaly detection unit (32) calls the quality model and the process parameters, the quality detection data and the environmental parameters, calculates predicted quality data, and compares the predicted quality data with a preset index interval; if the predicted quality data is within the index interval, it indicates that the quality of the Chinese medicine is normal; if the predicted quality data is not within the index interval, it indicates that the quality of the Chinese medicine has excessive fluctuations and the process parameters and the environmental parameters need to be adjusted; the optimization algorithm unit (33) optimizes the process parameters and the environmental parameters through an optimization algorithm with the optimization goals of improving quality and reducing costs, and generates an optimal quality control strategy and optimal quality control parameters; A control execution module (4) comprises a decision generation unit (41) and an equipment control unit (42), wherein the decision generation unit (41) obtains the optimal quality control strategy and decomposes the optimal quality control strategy and the optimal quality control parameters into control instructions through a large language model; and the equipment control unit (42) obtains the control instructions and outputs them to the production equipment.
10. The pharmaceutical enterprise total quality management robot system according to claim 9, characterized in that: The invention also comprises a mobile control module (5) and a display module (6), wherein the mobile control module (5) responds to user operations, calls the process parameters, the quality inspection data, the environmental parameters and the control instructions, and outputs them to the display module (6); and after the display module (6) obtains the process parameters, the quality inspection data, the environmental parameters and the control instructions, it performs visual output.
Citation Information
Cited By
Intelligent sensing and self-adaptive control system for curcuma wenyujin processing equipment
CN120276262A
An intelligent perception and adaptive control system for Curcuma aromatica processing equipment
CN120276262B
Process parameter adjusting method and device and electronic equipment
CN120742804A
Process parameter adjustment method, apparatus, and electronic device
CN120742804B
Process parameter optimization method and device based on industrial internet, equipment and medium
CN120891805A
