Medical quality data management platform based on micro-service architecture

Abstract

The invention relates to a medicine quality data management platform based on a micro-service architecture, and the platform comprises the steps: collecting medicine production data of a time sequence of a production process and a corresponding medicine qualification condition by taking a specific time as a step length; dividing the medicine production data into a plurality of medicine production categories by adopting a clustering analysis method according to the collected large-scale time sequence data; establishing a Bayesian network model to quantify a response relationship between medicine production categories and medicine qualification conditions by taking the medicine production categories as characteristics and the medicine qualification conditions as labels; determining a corresponding medicine production category according to the medicine production data monitored in real time; and inputting a medicine production category into the constructed Bayesian network model, and predicting a real-time medicine qualification condition. The problems that in the prior art, less attention is paid to the relation between medicine production data and the medicine qualification rate, unqualified medicine production data are difficult to trace and adjust, and the utilization efficiency of the medicine quality data is low are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing and relates to a pharmaceutical quality data management platform based on a microservice architecture. Background Art

[0002] Pharmaceutical quality data refers to various types of data generated during the drug production and quality control process, which are used to ensure the safety, effectiveness and quality consistency of drugs, and mainly include pharmaceutical production data (raw material data, production equipment data, environmental monitoring data, etc.) and pharmaceutical qualification rate. The monitoring and analysis of these data are of great significance to protecting public health, meeting regulatory requirements, maintaining GMP standards and quality management systems, and achieving market feedback and continuous improvement.

[0003] However, existing technologies pay little attention to the relationship between pharmaceutical production data and pharmaceutical qualification rate. In the process of managing pharmaceutical quality data, it is difficult to trace the source of unqualified medicines and adjust the corresponding pharmaceutical production data. Summary of the invention

[0004] The present invention provides a pharmaceutical quality data management platform based on a microservice architecture, aiming to split the pharmaceutical quality data management platform into independent data acquisition modules, data analysis modules and quality feedback modules through the microservice architecture, thereby simplifying system design and development; wherein the data analysis module constructs the relationship between pharmaceutical production data and pharmaceutical qualification rate, provides a basis for the quality feedback module to trace the source of unqualified medicines, and solves the problem of low data utilization efficiency in pharmaceutical quality data management.

[0005] The purpose of the present invention can be achieved through the following technical solutions: The present application provides a pharmaceutical quality data management platform based on a microservice architecture, including a data acquisition module, a data analysis module, a quality feedback module and a problem tracing module, wherein the data acquisition module, the data analysis module, the quality feedback module and the problem tracing module are communicatively connected, wherein: The data collection module is used to collect pharmaceutical production data and corresponding pharmaceutical qualification status of the production process time series with a specific time as a step length, wherein the pharmaceutical production data includes raw material data, production equipment data and environmental monitoring data, and the pharmaceutical qualification status includes qualified quality and unqualified quality; The data analysis module is used to construct a response relationship model between pharmaceutical production data and pharmaceutical qualification status; The quality feedback module is used to monitor the pharmaceutical production data in real time and input it into the response relationship model, and output the predicted pharmaceutical qualification status; The problem tracing module traces and adjusts the unqualified medicine production data according to the predicted medicine qualification status.

[0006] Furthermore, the raw material data includes the type, source and quality of the raw materials; the production equipment data includes mixing time, stirring speed and tableting pressure; and the environmental monitoring data includes temperature, humidity and dust amount.

[0007] Furthermore, the construction of the response relationship model between the pharmaceutical production data and the pharmaceutical qualification status includes the following steps: S1. Use cluster analysis method to divide pharmaceutical production data into several pharmaceutical production categories; S2. Construct a Bayesian network model with the pharmaceutical production category as the parent node and the pharmaceutical qualification status as the child node.

[0008] Furthermore, in step S1, the cluster analysis method is configured as ward clustering, comprising the following steps: S11. Calculate distance: calculate the Euclidean distance between samples of pharmaceutical production data; S12, initial clustering: each data sample is considered as an independent cluster, and the number of clusters is the same as the number of samples; S13, merging clusters: gradually merging the two closest clusters, specifically traversing each pair of clusters merged from two samples, calculating the increase in the variance of the internal Euclidean distance after merging, and selecting the pair with the smallest increase to merge; repeating this process until the predetermined number of clusters is reached; S14. Result determination: Use the silhouette coefficient to evaluate the rationality of the clustering results, and determine the pharmaceutical production category corresponding to the cluster based on the scope of pharmaceutical production data.

[0009] Furthermore, in step S2, the Bayesian network model includes the following construction steps: S21. Determine the model structure: determine a Bayesian network structure according to the causal relationship between the drug production category and the drug qualification status, wherein the Bayesian network structure points from a parent node to a child node; S22. Estimation of model parameters: Determine the parameters of the pharmaceutical production category and the pharmaceutical qualification status, wherein the parameters of the pharmaceutical production category are the results of classification using the cluster analysis method, and the parameters of the pharmaceutical qualification status include qualified quality and unqualified quality; S23, constructing a conditional probability table: calculating the conditional probability between each model parameter and constructing a conditional probability table; S24. Model evaluation: Use the test set to evaluate the accuracy of the model and adjust the model structure.

[0010] Furthermore, in step S23, the conditional probability between the model parameters is calculated using the following formula: P(Q|C) = P(Q, C) / P(C), Among them: P(Q|C) represents the probability that the pharmaceutical qualification is Q when the pharmaceutical production category is C; P (Q,C) represents the probability that the pharmaceutical qualification is Q and the pharmaceutical production category is C; P (C) represents the probability that the pharmaceutical production category is C.

[0011] Furthermore, the data acquisition module and the data analysis module are both equipped with storage units for storing pharmaceutical production data and response relationship models, and can transmit the response relationship models offline to other pharmaceutical production lines to predict pharmaceutical qualification status.

[0012] Furthermore, the quality feedback module includes a monitoring module, a judgment module and a reporting module, wherein: The monitoring module is used to monitor the pharmaceutical production data in real time and obtain the pharmaceutical production data in time series; The judgment module is used to input the time series medicine production data into the response relationship model to predict the probability of the medicine quality being qualified; The reporting module is used to send a report to the problem tracing module when the probability of the medicine quality being qualified is lower than a preset threshold.

[0013] Furthermore, the quality feedback module also transmits the real-time monitored pharmaceutical production data to the data acquisition module; after receiving the data, the data acquisition module stores and transmits it to the data analysis module for analysis, thereby realizing dynamic updating of the response relationship model.

[0014] Furthermore, tracing and adjusting the unqualified pharmaceutical production data according to the predicted pharmaceutical qualification status includes the following steps: When the pharmaceutical qualification is unqualified, check the numerical range of all pharmaceutical production data; The normal value range of pharmaceutical production data corresponding to the pharmaceutical production category when determining the quality is qualified; Compare the differences between the numerical range of pharmaceutical production data and the normal numerical range one by one to determine the pharmaceutical production data with abnormalities; Adjustments will be made to any abnormal pharmaceutical production data until the pharmaceuticals are of qualified quality.

[0015] Beneficial effects of the present invention: By taking a specific time as a step length, collecting the pharmaceutical production data of the production process time series and the corresponding pharmaceutical qualification status; using the cluster analysis method to divide the pharmaceutical production data into several pharmaceutical production categories based on the collected large-scale time series data; using the pharmaceutical production category as a feature and the pharmaceutical qualification status as a label, constructing a Bayesian network model to quantify the response relationship between the two; determining the corresponding pharmaceutical production category based on the real-time monitored pharmaceutical production data; inputting the pharmaceutical production category into the constructed Bayesian network model to predict the real-time pharmaceutical qualification status. The present invention solves the problem that the prior art pays less attention to the relationship between pharmaceutical production data and pharmaceutical qualification rate, it is difficult to trace and adjust pharmaceutical production data of unqualified quality, and the utilization efficiency of pharmaceutical quality data is low. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to facilitate understanding by those skilled in the art, the present invention is further described below with reference to the accompanying drawings.

[0017] Figure 1 This is a structural diagram of a pharmaceutical quality data management platform based on a microservice architecture in the present invention.

[0018] Figure 2 The figure is a flow chart of cluster analysis in one embodiment of the present invention.

[0019] Figure 3 The flowchart of constructing a Bayesian network model in one embodiment of the present invention. DETAILED DESCRIPTION

[0020] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the specific implementation mode, structure, characteristics and effects of the present invention are described in detail below in combination with the accompanying drawings and preferred embodiments.

[0021] See also Figure 1-Figure 3 The present application provides a pharmaceutical quality data management platform based on a microservice architecture, including a data acquisition module, a data analysis module, a quality feedback module and a problem tracing module, wherein the data acquisition module, the data analysis module, the quality feedback module and the problem tracing module are communicatively connected, wherein: The data collection module is used to collect pharmaceutical production data and corresponding pharmaceutical qualification status of the production process time series with a specific time as a step length, wherein the pharmaceutical production data includes raw material data, production equipment data, and environmental monitoring data, and the pharmaceutical qualification status includes qualified quality and unqualified quality; In this embodiment, the microservice architecture is a software design method that improves the flexibility, maintainability and scalability of the system by decomposing a single application into multiple small, independent services. Each microservice focuses on a specific business function, runs in an independent process, and interacts through lightweight communication mechanisms (such as HTTP API). Its core features include independent deployment, technical diversity, small and focused, distributed data management and lightweight communication. The advantages of the microservice architecture are elasticity and scalability, rapid development and deployment, fault isolation and flexibility, and its advantages are widely used in large systems that require rapid iteration and expansion. This embodiment uses the microservice architecture to split the pharmaceutical quality data management platform into independent data acquisition modules, data analysis modules and quality feedback modules, thereby improving data management efficiency.

[0022] The data collection module is used to collect pharmaceutical production data and pharmaceutical qualification data, especially raw material data, production equipment data, environmental monitoring data and quality qualification data: 1) Raw material data includes the type, source, quality inspection results, etc. of raw materials. The material management system (MMS) can automatically record the raw material in and out warehouse conditions and quality inspection results.

[0023] 2) Production equipment data mainly includes equipment operating parameters, including mixing time, stirring speed and tableting pressure, as well as real-time monitoring of equipment operating status through industrial control systems (ICS) and Internet of Things sensors (IoTSensors), and recording of various equipment indicators.

[0024] 3) Environmental monitoring data, including the temperature, humidity, air quality (such as dust volume) of the production workshop, etc., is collected in real time by installing environmental monitoring sensors (such as temperature and humidity sensors) and uploaded to the data acquisition module.

[0025] 4) Medical qualification data: Qualified quality means that the product has passed quality inspection and meets all specified standards; conversely, unqualified quality means that the product is found not to meet the specified standards during quality inspection.

[0026] Furthermore, the raw material data includes the type, source and quality of the raw materials; the production equipment data includes mixing time, stirring speed and tableting pressure; and the environmental monitoring data includes temperature, humidity and dust amount.

[0027] In the pharmaceutical production process, raw material data, production equipment data and environmental monitoring data are key factors in ensuring drug quality. The following are the specific impacts of these data on pharmaceutical quality: Raw material data: 1) Raw material types: Different types of raw materials have different chemical and physical properties, which have a direct impact on the efficacy, stability and safety of the drug. For example, the selection of active ingredients and excipients will affect the solubility, absorption rate and efficacy of the drug.

[0028] 2) Source of raw materials: The source of raw materials and supplier quality control are directly related to the consistency and purity of raw materials. Reliable suppliers and strict quality control can ensure the high quality of raw materials and avoid contamination and variation.

[0029] 3) Raw material quality: The purity, content and impurity level of raw materials are crucial to the final quality of the drug. High-quality raw materials can ensure the purity and efficacy of the drug and reduce side effects and adverse reactions.

[0030] Production equipment data: 1) Mixing time: Mixing time affects the uniform distribution of raw materials. Too short a mixing time may lead to uneven distribution of ingredients, affecting the efficacy and consistency of the drug; too long a mixing time may lead to over-processing of materials, affecting the stability of the drug.

[0031] 2) Stirring speed: Stirring speed affects the efficiency and uniformity of the mixing process. Appropriate stirring speed can ensure that the raw materials are fully mixed and improve the uniformity and quality of the drug.

[0032] 3) Tableting pressure: Tableting pressure directly affects the hardness, dissolution rate and stability of the drug. Appropriate pressure can ensure the density and strength of the tablet and avoid tablet breakage or disintegration.

[0033] Environmental monitoring data: 1) Temperature: The temperature of the production environment has a significant impact on the physical and chemical properties of drugs. Too high a temperature may cause the drug to decompose or deteriorate, while too low a temperature may affect the solubility and stability of the drug.

[0034] 2) Humidity: Humidity control is crucial to the stability and physical properties of drugs. High humidity may cause drugs to absorb moisture and deteriorate, while low humidity may cause drugs to lose water and become hard, especially tablets and capsules, which have more stringent requirements on humidity.

[0035] 3) Dust amount: Dust and particles in the production environment may cause contamination of drugs, affecting the purity and safety of drugs. Strictly controlling the amount of dust can reduce the risk of contamination and improve the quality of drugs.

[0036] The data analysis module is used to construct a response relationship model between pharmaceutical production data and pharmaceutical qualification status; The quality feedback module is used to monitor the pharmaceutical production data in real time and input it into the response relationship model, and output the predicted pharmaceutical qualification status; The problem tracing module traces and adjusts the unqualified medicine production data according to the predicted medicine qualification status.

[0037] Furthermore, the construction of the response relationship model between the pharmaceutical production data and the pharmaceutical qualification status includes the following steps: S1. Use cluster analysis method to divide pharmaceutical production data into several pharmaceutical production categories; S2. Construct a Bayesian network model with the pharmaceutical production category as the parent node and the pharmaceutical qualification status as the child node.

[0038] It should be noted that the cluster analysis method is used to divide the pharmaceutical production data into several pharmaceutical production categories. The number here can be as large as possible. The pharmaceutical production data is subdivided into multiple pharmaceutical production category characteristics. Such quantitative results are helpful to strengthen the control of pharmaceutical production quality according to different characteristics.

[0039] Furthermore, in step S1, the cluster analysis method is configured as ward clustering, comprising the following steps: S11. Calculate distance: First, calculate the Euclidean distance between samples of pharmaceutical production data. Euclidean distance is a commonly used distance measurement method that reflects the geometric distance between two data points in space. For pharmaceutical production data, each sample may contain multiple features, such as the composition content of raw materials, parameters of production equipment, indicators of environmental monitoring, etc. By calculating the Euclidean distance, the degree of difference between different samples in these features can be measured. This step provides a basic distance measurement basis for the subsequent clustering process.

[0040] S12, initial clustering: Each data sample is considered as an independent cluster. At this time, the number of clusters is the same as the number of samples. This is the initial state of the clustering process. Each sample is a cluster, which means that there is only one data point in each cluster. The purpose of this is to start from the most fine-grained division and gradually merge similar samples to find a more reasonable clustering structure. At this stage, the center of each cluster is the corresponding sample point itself, because the center of a cluster with only one data point is naturally the data point.

[0041] S13, merge clustering: Step by step, merge the two closest clusters. The specific method is to traverse each pair of clusters merged by two samples and calculate the increase in the variance of the internal Euclidean distance after the merger. This variance increase reflects the change in the dispersion of the internal data points of the new cluster after the merger of the two clusters relative to the dispersion of the internal data points of the two clusters before the merger. Selecting the pair with the smallest increase to merge means selecting the two clusters that have the least impact on the overall data structure after the merger to merge. Repeat this process and keep looking for the two clusters that are most suitable for merging until the predetermined number of clusters is reached. For example, if there are N samples initially, that is, N clusters, as the merging process proceeds, the number of clusters gradually decreases until the predetermined number of clusters is met.

[0042] S14. Result determination: Use the silhouette coefficient to evaluate the rationality of the clustering results. The silhouette coefficient is an indicator used to measure the quality of clustering, which combines the concepts of cohesion and separation. For each sample point, calculate its silhouette coefficient, and then average the silhouette coefficients of all sample points to obtain the silhouette coefficient of the entire clustering result. If the silhouette coefficient is close to 1, it means that the clustering effect is good; if it is close to -1, it means that the clustering effect is poor; if it is close to 0, it means that the sample is on the boundary of two clusters. Determine the pharmaceutical production category of the corresponding cluster according to the range of pharmaceutical production data, that is, according to the distribution of data characteristics after clustering and the actual situation of pharmaceutical production, assign a pharmaceutical production category label to each cluster for use in the subsequent construction of the Bayesian network model.

[0043] Furthermore, in step S2, the Bayesian network model includes the following construction steps: S21. Determine the model structure: First, the Bayesian network structure should be determined based on the causal relationship between the pharmaceutical production category and the pharmaceutical qualification status. In the pharmaceutical production process, it is generally believed that the pharmaceutical production category will have an impact on the pharmaceutical qualification status. For example, different raw material combinations, production equipment status, and pharmaceutical production categories represented by environmental monitoring data may lead to different pharmaceutical qualification probabilities. Therefore, the Bayesian network structure is set to point from the parent node (pharmaceutical production category) to the child node (pharmaceutical qualification status) to reflect this causal relationship. By clarifying this structure, a clear framework can be provided for subsequent parameter estimation and probability calculation.

[0044] S22. Estimation of model parameters: Determine the parameters of pharmaceutical production categories and pharmaceutical qualification. For the parameters of pharmaceutical production categories, the results of the classification by cluster analysis method are used. As mentioned above, the pharmaceutical production data is divided into several categories through cluster analysis, and these categories become different states of the parent node in the Bayesian network. The parameters of pharmaceutical qualification include two states: qualified quality and unqualified quality. In this way, the possible value range of each node in the model can be clarified, preparing for the subsequent probability calculation.

[0045] S23. Construct a conditional probability table: Calculate the conditional probabilities between the model parameters and construct a conditional probability table. For each node in the Bayesian network, it is necessary to determine the conditional probability distribution given the state of its parent node. For example, for the child node of drug qualification, the conditional probabilities of qualified and unqualified quality are calculated when the parent node states of different drug production categories are known. These conditional probabilities can be estimated by analyzing historical data, statistical methods, or expert knowledge. The constructed conditional probability table will become the basis for the Bayesian network to perform reasoning and prediction.

[0046] S24. Model evaluation: Use the test set to evaluate the accuracy of the model and adjust the model structure. Use a portion of the data that is not used to build the model as a test set and input it into the built Bayesian network model to observe the degree of consistency between the medical qualification predicted by the model and the actual situation. Some evaluation indicators, such as accuracy, recall rate, F1 value, etc., can be used to measure the performance of the model. If the accuracy of the model does not meet the requirements, consider adjusting the model structure, such as re-examining the causal relationship, adjusting the connection between nodes, or re-estimating parameters. Improve the accuracy and reliability of the model through continuous evaluation and adjustment.

[0047] Furthermore, in step S23, the conditional probability between the model parameters is calculated using the following formula: P(Q|C) = P(Q, C) / P(C), Among them: P(Q|C) represents the probability that the pharmaceutical qualification is Q when the pharmaceutical production category is C; P (Q,C) represents the probability that the pharmaceutical qualification is Q and the pharmaceutical production category is C; P (C) represents the probability that the pharmaceutical production category is C.

[0048] Furthermore, the data acquisition module and the data analysis module are both equipped with storage units for storing pharmaceutical production data and response relationship models, and can transmit the response relationship models offline to other pharmaceutical production lines to predict pharmaceutical qualification status.

[0049] In this embodiment, in this pharmaceutical quality data management platform based on microservice architecture, the storage unit equipped with the data acquisition module and the data analysis module plays a vital role. The pharmaceutical production data stored in the data acquisition module provides a solid foundation for subsequent analysis, and can be accurately traced when quality problems occur. The response relationship model stored in the data analysis module is of great significance. The model is constructed by in-depth analysis of the pharmaceutical production data and pharmaceutical qualification of a specific pharmaceutical production line, and fully considers the complex relationship between various factors such as various raw material data, production equipment data, and environmental monitoring data and pharmaceutical qualification. In addition, due to the existence of the storage unit, this carefully constructed and strictly verified response relationship model can be transmitted offline to other pharmaceutical production lines. This greatly expands the applicability of the model. Regardless of the size of different pharmaceutical production lines and the characteristics of the production process, this model can be used to accurately predict the pharmaceutical qualification. Even in an environment with poor network conditions or independent operation, other pharmaceutical production lines can make full use of the model to quickly and effectively control the quality of pharmaceuticals, providing a powerful tool and guarantee for quality control of the entire pharmaceutical production industry, and effectively promoting the high-quality development of the industry.

[0050] Furthermore, the quality feedback module includes a monitoring module, a judgment module and a reporting module, wherein: The monitoring module is used to monitor the pharmaceutical production data in real time and obtain the pharmaceutical production data in time series; The judgment module is used to input the time series medicine production data into the response relationship model to predict the probability of the medicine quality being qualified; The reporting module is used to send a report to the problem tracing module when the probability of the medicine quality being qualified is lower than a preset threshold.

[0051] In this embodiment, the monitoring module continuously monitors the pharmaceutical production process in real time, keeping an eye on the dynamic changes in pharmaceutical production. Through advanced monitoring technology, it can accurately obtain time series pharmaceutical production data, covering various aspects such as raw material data, production equipment data, and environmental monitoring data. These real-time data provide the freshest and most accurate information source for subsequent quality assessment and prediction.

[0052] The judgment module makes full use of the time series pharmaceutical production data obtained by the monitoring module and inputs it into a carefully constructed response relationship model. This response relationship model is established after in-depth analysis of a large amount of pharmaceutical production data and pharmaceutical qualification conditions. It can accurately predict the probability of pharmaceutical quality qualification based on the input pharmaceutical production data. In this way, the judgment module can timely understand the possibility of current pharmaceutical product quality qualification during the production process, providing an important basis for quality control decision-making.

[0053] The reporting module plays a key alarm role in the entire quality feedback mechanism. When the probability of qualified pharmaceutical quality predicted by the judgment module is lower than the preset threshold, the reporting module will respond quickly and send a report to the problem tracing module. This preset threshold is set according to the quality requirements and experience of pharmaceutical production. Once it is lower than this threshold, it means that there may be a risk of quality problems in the pharmaceutical product. The timely report of the reporting module can trigger the work of the problem tracing module, prompting it to trace and analyze the pharmaceutical production data that may cause quality problems, so as to take timely adjustment measures to ensure that the quality of pharmaceutical production always meets the standards.

[0054] Furthermore, the quality feedback module also transmits the real-time monitored pharmaceutical production data to the data acquisition module; after receiving the data, the data acquisition module stores and transmits it to the data analysis module for analysis, thereby realizing dynamic updating of the response relationship model.

[0055] In the pharmaceutical quality data management platform, in addition to the functions mentioned above, the quality feedback module also plays an important role in transmitting the real-time monitored pharmaceutical production data to the data acquisition module. While the monitoring module in the quality feedback module monitors the pharmaceutical production data in real time, it will transmit the latest data to the data acquisition module in a timely manner.

[0056] After receiving the real-time monitoring data from the quality feedback module, the data acquisition module will store the data and continuously enrich the historical records of pharmaceutical production data. On the other hand, the data acquisition module will transmit the newly received data to the data analysis module.

[0057] After obtaining new data, the data analysis module will use this data for analysis. By continuously incorporating new data, the data analysis module can dynamically update the existing response relationship model. As time goes by and new data is continuously added, the response relationship model can more accurately reflect the relationship between pharmaceutical production data and pharmaceutical qualification status.

[0058] This dynamic update mechanism enables the pharmaceutical quality data management platform to adapt to the ever-changing pharmaceutical production environment and conditions, and always maintain accurate prediction and effective management of pharmaceutical quality. The continuously updated response relationship model can better provide decision-making basis for quality control in the pharmaceutical production process, improve the quality stability and reliability of pharmaceutical production, and enhance the applicability and practicality of the entire platform.

[0059] Furthermore, tracing and adjusting the unqualified pharmaceutical production data according to the predicted pharmaceutical qualification status includes the following steps: When the pharmaceutical qualification is unqualified, check the numerical range of all pharmaceutical production data; The normal value range of pharmaceutical production data corresponding to the pharmaceutical production category when determining the quality is qualified; Compare the differences between the numerical range of pharmaceutical production data and the normal numerical range one by one to determine the pharmaceutical production data with abnormalities; Adjustments will be made to any abnormal pharmaceutical production data until the pharmaceuticals are of qualified quality.

[0060] In this embodiment, in the problem tracing module of the pharmaceutical quality data management platform, when the predicted pharmaceutical qualification is unqualified, a series of tracing and adjustment steps will be initiated: First, when it is found that the quality of pharmaceutical products is unqualified, the numerical range of all pharmaceutical production data is checked immediately. This means a comprehensive review of the specific values ​​of various pharmaceutical production data such as raw material data, production equipment data, and environmental monitoring data to determine the data status of each link in the current production process.

[0061] Next, determine the normal value range of pharmaceutical production data corresponding to the pharmaceutical production category when the quality is qualified. This step is based on the different pharmaceutical production categories and their corresponding standard data ranges under qualified conditions determined by cluster analysis and Bayesian network model construction. By comparing the data range under the current unqualified state with the normal value range under the qualified state, the gap and problem can be clearly identified.

[0062] Then, the difference between the pharmaceutical production data value range and the normal value range is compared one by one. Through this careful comparison, it is possible to accurately determine which pharmaceutical production data are abnormal. Data points with large differences are likely to be the key factors that lead to quality failure.

[0063] Finally, adjust the abnormal pharmaceutical production data. This may involve adjusting the use of raw materials, optimizing the parameter settings of production equipment, or improving environmental monitoring conditions. Continue to adjust and verify until the pharmaceutical qualification situation becomes qualified quality. This process ensures that when quality problems occur, the root cause of the problem can be quickly found and effective adjustment measures can be taken to restore the quality of pharmaceutical products and protect public health and the quality stability of pharmaceutical production.

[0064] The above description is only a preferred embodiment of the present invention and does not limit the present invention in any form. Although the present invention has been disclosed as a preferred embodiment as above, it is not used to limit the present invention. Any technical personnel in this field can make some changes or modify the technical contents disclosed above into equivalent embodiments without departing from the scope of the technical solution of the present invention. However, any brief modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the present invention without departing from the content of the technical solution of the present invention are still within the scope of the technical solution of the present invention.

Claims

1. A pharmaceutical quality data management platform based on microservice architecture, characterized by: It includes a data acquisition module, a data analysis module, a quality feedback module and a problem tracing module, which are communicatively connected to each other, wherein: The data collection module is used to collect pharmaceutical production data and corresponding pharmaceutical qualification status of the production process time series with a specific time as a step length, wherein the pharmaceutical production data includes raw material data, production equipment data and environmental monitoring data, and the pharmaceutical qualification status includes qualified quality and unqualified quality; The data analysis module is used to construct a response relationship model between pharmaceutical production data and pharmaceutical qualification status; The quality feedback module is used to monitor the pharmaceutical production data in real time and input it into the response relationship model, and output the predicted pharmaceutical qualification status; The problem tracing module traces and adjusts the unqualified medicine production data according to the predicted medicine qualification status.

2. A pharmaceutical quality data management platform based on microservice architecture according to claim 1, characterized in that: The raw material data include the type, source and quality of the raw materials; the production equipment data include mixing time, stirring speed and tableting pressure; the environmental monitoring data include temperature, humidity and dust amount.

3. A pharmaceutical quality data management platform based on microservice architecture according to claim 1, characterized in that: The step of constructing a response relationship model between pharmaceutical production data and pharmaceutical qualification conditions includes the following steps: S1. Use cluster analysis method to divide the pharmaceutical production data into several pharmaceutical production categories; S2. Construct a Bayesian network model with the pharmaceutical production category as the parent node and the pharmaceutical qualification status as the child node.

4. A pharmaceutical quality data management platform based on microservice architecture according to claim 3, characterized in that: In step S1, the cluster analysis method is configured as ward clustering, including the following steps: S11. Calculate distance: calculate the Euclidean distance between samples of pharmaceutical production data; S12, initial clustering: each data sample is considered as an independent cluster, and the number of clusters is the same as the number of samples; S13, merging clusters: gradually merging the two closest clusters, specifically traversing each pair of clusters merged from two samples, calculating the increase in the variance of the internal Euclidean distance after merging, and selecting the pair with the smallest increase to merge; repeating this process until the predetermined number of clusters is reached; S14. Result determination: Use the silhouette coefficient to evaluate the rationality of the clustering results, and determine the pharmaceutical production category corresponding to the cluster based on the scope of pharmaceutical production data.

5. The pharmaceutical quality data management platform based on microservice architecture according to claim 3, characterized in that: In step S2, the Bayesian network model includes the following construction steps: S21, determining the model structure: determining a Bayesian network structure according to the causal relationship between the drug production category and the drug qualification status, wherein the Bayesian network structure points from a parent node to a child node; S22. Estimation of model parameters: Determine the parameters of the pharmaceutical production category and the pharmaceutical qualification status, wherein the parameters of the pharmaceutical production category are the results of classification using the cluster analysis method, and the parameters of the pharmaceutical qualification status include qualified quality and unqualified quality; S23, constructing a conditional probability table: calculating the conditional probability between each model parameter and constructing a conditional probability table; S24. Model evaluation: Use the test set to evaluate the accuracy of the model and adjust the model structure.

6. A pharmaceutical quality data management platform based on microservice architecture according to claim 5, characterized in that: In step S23, the conditional probability between the model parameters is calculated using the following formula: P(Q|C) = P(Q, C) / P(C), Among them: P(Q|C) represents the probability that the pharmaceutical qualification is Q when the pharmaceutical production category is C; P (Q, C) represents the probability that the pharmaceutical qualification is Q and the pharmaceutical production category is C; P (C) represents the probability that the pharmaceutical production category is C.

7. The pharmaceutical quality data management platform based on microservice architecture according to claim 1, characterized in that: The data acquisition module and the data analysis module are both equipped with storage units for storing pharmaceutical production data and response relationship models, and can transmit the response relationship models offline to other pharmaceutical production lines to predict pharmaceutical qualification status.

8. The pharmaceutical quality data management platform based on microservice architecture according to claim 1, characterized in that: The quality feedback module includes a monitoring module, a judgment module and a reporting module, wherein: The monitoring module is used to monitor the pharmaceutical production data in real time and obtain the pharmaceutical production data in time series; The judgment module is used to input the time series medicine production data into the response relationship model to predict the probability of the medicine quality being qualified; The reporting module is used to send a report to the problem tracing module when the probability of the medicine quality being qualified is lower than a preset threshold.

9. The pharmaceutical quality data management platform based on microservice architecture according to claim 1, characterized in that: The quality feedback module also transmits the real-time monitored pharmaceutical production data to the data acquisition module; After the data acquisition module receives the data, it stores and transmits it to the data analysis module for analysis to achieve dynamic update of the response relationship model.

10. A pharmaceutical quality data management platform based on microservice architecture according to claim 3, characterized in that: The method of tracing and adjusting the unqualified pharmaceutical production data according to the predicted pharmaceutical qualification status includes the following steps: When the pharmaceutical qualification is unqualified, check the numerical range of all pharmaceutical production data; The normal value range of pharmaceutical production data corresponding to the pharmaceutical production category when determining the quality is qualified; Compare the differences between the numerical range of pharmaceutical production data and the normal numerical range one by one to determine the pharmaceutical production data with abnormalities; Adjustments will be made to any abnormal pharmaceutical production data until the pharmaceuticals are of qualified quality.

Images