Drug production quality identification method and platform based on big data

Abstract

The application discloses a drug production quality identification method and platform based on big data, relates to the technical field of drug identification, and comprises the following steps: for each batch of drugs, obtaining equipment fluctuation information in the production process to calculate an equipment fluctuation value; obtaining drug raw material change information in the production process to calculate a raw material change value; obtaining production scale information in the production process to calculate a production scale value; calculating a quality identification change value according to the equipment fluctuation value, the raw material change value and the production scale value, and judging whether the production quality identification frequency of the corresponding batch of drugs needs to be changed according to the quality identification change value. In this way, the identification frequency can be adjusted according to the actual fluctuation, process complexity or potential risk of different batches of products, and the dynamic identification strategy can ensure that high-risk batches are identified sufficiently and quality hidden dangers are not omitted, so that the big data quality identification method can be further improved in production flexibility and quality guarantee capability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of drug identification, in particular to a drug production quality identification method and platform based on big data. BACKGROUND

[0002] The drug production quality identification method based on big data refers to collecting, fusing and analyzing the multi-dimensional quality data formed by the raw drug powder particles after production, using machine learning, statistical modeling and other technical means to comprehensively evaluate the key indicators such as particle size distribution, morphological characteristics and physical and chemical properties, so as to judge whether the product meets the preset quality standard; this method can realize efficient, data-driven quality control and quality traceability without relying on traditional manual detection, and is especially suitable for batch production quality evaluation of powder and particle raw materials.

[0003] At present, most of the drug production quality identification methods based on big data generally adopt fixed identification times in actual application, that is, one or more quality identifications are carried out according to the preset process, and the identification frequency is not adjusted according to the actual volatility, process complexity or potential risk of different batches of products. This static identification strategy may lead to insufficient identification of some high-risk batches and omission of quality risks, which limits the further improvement of the production flexibility and quality assurance capability of the big data quality identification method. SUMMARY

[0004] The purpose of the present application is to solve the above-mentioned problems and provide a drug production quality identification method and platform based on big data.

[0005] In the first aspect of the present application, a drug production quality identification method based on big data is first proposed, which comprises:

[0006] The drug after production is divided into several batches according to the corresponding production batch, for each batch of drug, the equipment fluctuation information in the production process is obtained, and the equipment fluctuation value is calculated according to the equipment fluctuation information; for evaluating the fluctuation degree of the equipment in the production process of the drug;

[0007] The drug raw material change information in the production process is obtained, and the raw material change value is calculated according to the drug raw material change information; for evaluating the change degree of the raw material in the production process of the drug;

[0008] The production scale information in the production process is obtained, and the production scale value is calculated according to the production scale information; for evaluating the size of the production scale in the production process of the drug;

[0009] The quality identification change value is calculated according to the equipment fluctuation value, the raw material change value and the production scale value, and whether the production quality identification frequency of the corresponding batch of drug is changed is judged according to the quality identification change value.

[0010] Optionally, the step of obtaining the equipment fluctuation information in the production process and calculating the equipment fluctuation value according to the equipment fluctuation information is:

[0011] Obtaining the time-sequential production vibration signal of each equipment in the production process of the drug, and pre-processing the production vibration signal, and converting the production vibration signal from the time domain to the frequency domain through the fast Fourier algorithm to obtain the frequency spectrum of the vibration signal;

[0012] Extracting the maximum standard vibration signal corresponding to each equipment during normal production of the drug from the historical data, and recording the frequency of the maximum standard vibration signal in the frequency domain as the vibration frequency threshold;

[0013] Taking the vibration frequency threshold as the limit of the frequency spectrum of the current production vibration signal, recording the region corresponding to the frequency greater than the vibration frequency threshold in the frequency spectrum as the high-frequency region; and calculating the total energy of the high-frequency region and the total energy of the frequency spectrum, and dividing the energy of the high-frequency region by the total energy of the frequency spectrum to obtain the vibration abnormality degree of the corresponding equipment;

[0014] Comparing the vibration abnormality degree of each equipment with the corresponding preset vibration abnormality degree minimum threshold, if the vibration abnormality degree is not less than the corresponding preset vibration abnormality degree minimum threshold, the corresponding equipment is recorded as a fluctuation equipment, and the total number of fluctuation equipment is divided by the total number of equipment in the production process of the drug to obtain the equipment fluctuation value.

[0015] Optionally, the step of calculating the raw material change value according to the raw material transformation information is:

[0016] Obtaining the multi-dimensional chemical attribute data of each batch of raw materials before production in the production process of the drug, and normalizing the multi-dimensional chemical attribute data so that the chemical attributes of each raw material are standardized to a unified scale;

[0017] Based on the normalized data, the multi-dimensional distribution of the current batch of raw materials is established by kernel density estimation method;

[0018] Obtaining a plurality of historical raw material multi-dimensional distributions during normal production of the drug in the historical data, and calculating the KL divergence between each historical raw material multi-dimensional distribution and the multi-dimensional distribution of the current batch of raw materials, and taking the inverse of the KL divergence as the raw material difference value between the corresponding historical raw material multi-dimensional distribution and the multi-dimensional distribution of the current batch of raw materials;

[0019] Taking the minimum raw material difference value as the raw material change value.

[0020] Optionally, the step of calculating the production scale value according to the production scale information is:

[0021] For each batch of medicine, the total weight of the actual production of the corresponding batch of medicine is obtained, and the total weight of the actual production is compared with the preset standard total weight of production to obtain a production weight ratio;

[0022] For each batch of medicine, the time from the start of production to the end of production of the corresponding batch of medicine is obtained to obtain the total duration of production, and the total duration of production is divided by the preset standard production time to obtain a production time ratio;

[0023] For each batch of medicine, the temperature and humidity during the production of the corresponding batch of medicine are obtained, and the time ratio of the temperature and humidity deviating from the preset normal temperature and humidity range is calculated;

[0024] The production weight ratio, the production time ratio, and the time ratio of the temperature and humidity deviating from the preset normal temperature and humidity range are added to obtain a production scale value.

[0025] Optionally, the step of calculating the quality identification change value according to the equipment fluctuation value, the raw material change value, and the production scale value is:

[0026] The equipment fluctuation value, the raw material change value, and the production scale value are converted into a comprehensive feature vector, the comprehensive feature vector is taken as an input of a machine learning model, the machine learning model takes each set of comprehensive feature vector as a prediction target to predict each quality identification change value label, minimizes the sum of prediction errors of all quality identification change value labels as a training target, trains the machine learning model until the sum of prediction errors converges to stop model training, and determines the quality identification change value according to the model output result, wherein the machine learning model is a polynomial regression model.

[0027] Optionally, the step of judging whether the number of times of changing the production quality identification of the corresponding batch of medicine needs to be changed according to the quality identification change value is:

[0028] The quality identification change value is compared with a preset quality identification change value threshold, if the quality identification change value is less than the preset quality identification change value threshold, the number of times of changing the production quality identification of the corresponding batch of medicine does not need to be changed, and the medicine production quality can still be identified according to the original set fixed identification number;

[0029] If the quality identification change value is not less than the preset quality identification change value threshold, the number of times of changing the production quality identification of the corresponding batch of medicine needs to be changed, and the number of times of medicine quality identification is increased.

[0030] In the second aspect of the implementation of the present application, a medicine production quality identification platform based on big data is proposed, which comprises:

[0031] The device fluctuation module: divides the produced drugs into several batches of drugs according to corresponding production batches, obtains device fluctuation information in the production process of each batch of drugs, and calculates a device fluctuation value according to the device fluctuation information; and is used for evaluating the fluctuation degree of the device in the production process of the drugs.

[0032] The raw material change module: obtains drug raw material change information in the production process, and calculates a raw material change value according to the drug raw material change information; and is used for evaluating the change degree of the raw material in the production process of the drugs.

[0033] The production scale module: obtains production scale information in the production process, and calculates a production scale value according to the production scale information; and is used for evaluating the size of the production scale in the production process of the drugs.

[0034] The identification module: calculates a quality identification change value according to the device fluctuation value, the raw material change value and the production scale value, and judges whether the number of production quality identification changes for the corresponding batch of drugs needs to be changed according to the quality identification change value.

[0035] The beneficial effects of the present application are as follows:

[0036] The present application provides a drug production quality identification method and platform based on big data. The produced drugs are divided into several batches of drugs according to corresponding production batches. The device fluctuation value is calculated according to the device fluctuation information in the production process of each batch of drugs. The raw material change value is calculated according to the drug raw material change information in the production process. The production scale value is calculated according to the production scale information in the production process. The quality identification change value is calculated according to the device fluctuation value, the raw material change value and the production scale value. Whether the number of production quality identification changes for the corresponding batch of drugs needs to be changed is judged according to the quality identification change value. In this way, the identification frequency can be adjusted according to the actual fluctuation, process complexity or potential risk of different batches of products. The dynamic identification strategy can ensure that the high-risk batches are identified sufficiently and quality hidden dangers are not omitted, so that the big data quality identification method can be further improved in production flexibility and quality guarantee capability. BRIEF DESCRIPTION OF DRAWINGS

[0037] The present application will be further described below with reference to the accompanying drawings.

[0038] Figure 1 The flowchart of the drug production quality identification method based on big data;

[0039] Figure 2 The framework diagram of the drug production quality identification platform based on big data. DETAILED DESCRIPTION

[0040] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0041] Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0042] This invention provides a method for identifying drug manufacturing quality based on big data. See also... Figure 1 , Figure 1 A flowchart illustrating a drug production quality identification method based on big data, provided in an embodiment of the present invention. The method includes the following steps:

[0043] After production, the drugs are divided into several batches according to their corresponding production batches. For each batch, equipment fluctuation information during the production process is obtained, and equipment fluctuation value is calculated based on the equipment fluctuation information; this is used to assess the degree of equipment fluctuation during drug production.

[0044] It acquires information on changes in drug raw materials during the production process and calculates the changes in raw material values ​​based on this information; this is used to assess the degree of change in raw materials during drug production.

[0045] To obtain production scale information during the production process and calculate the production scale value based on the production scale information; used to assess the extent of production scale in the drug production process;

[0046] The quality identification change value is calculated based on the equipment fluctuation value, raw material change value, and production scale value. Based on the quality identification change value, it is determined whether the production quality identification frequency should be changed for the corresponding batch of drugs.

[0047] Based on the big data-based drug production quality identification method provided in this invention, the identification frequency can be adjusted according to the actual volatility, process complexity, or potential risks of different batches of products. This dynamic identification strategy can ensure that high-risk batches are adequately identified and that no quality hazards are missed, thereby further improving the big data quality identification method in terms of production flexibility and quality assurance capabilities.

[0048] In one embodiment, the produced drugs are divided into several batches according to the corresponding production batches. For each batch of drugs, equipment fluctuation information during the production process is obtained, and equipment fluctuation value is calculated based on the equipment fluctuation information; this is used to assess the degree of equipment fluctuation during the drug production process.

[0049] Specifically, the steps for obtaining equipment fluctuation information during the production process and calculating equipment fluctuation values ​​based on this information are as follows:

[0050] The production vibration signals of each piece of equipment in the drug production process are obtained in time sequence, and the production vibration signals are preprocessed and converted from the time domain to the frequency domain by the fast Fourier algorithm to obtain the frequency spectrum of the vibration signals.

[0051] Extract the maximum standard vibration signal corresponding to each piece of equipment during normal drug production from historical data, and record the frequency of the maximum standard vibration signal in the frequency domain as the vibration frequency threshold.

[0052] The frequency spectrum of the current production vibration signal is defined with the vibration frequency threshold as the boundary. The region in the frequency spectrum with frequencies greater than the vibration frequency threshold is defined as the high-frequency region. The total energy of the high-frequency region and the total energy of the frequency spectrum are calculated. The energy of the high-frequency region is divided by the total energy of the frequency spectrum to obtain the vibration anomaly degree of the corresponding equipment.

[0053] The vibration anomaly degree of each device is compared with the corresponding preset minimum vibration anomaly degree threshold. If the vibration anomaly degree is not less than the corresponding preset minimum vibration anomaly degree threshold, the corresponding device is recorded as a fluctuating device. The total number of fluctuating devices is divided by the total number of devices in the drug production process to obtain the device fluctuation value.

[0054] It should be noted that the preset minimum threshold for vibration anomaly is set by professionals based on actual conditions, and specific details are not limited or elaborated upon. During the calculation of equipment fluctuation values, data acquisition methods include real-time monitoring and historical data storage. First, vibration signals from each piece of equipment are collected in real time using vibration sensors (such as accelerometers and vibration sensors) installed on the production equipment, and recorded synchronously with the equipment's operating time. Second, historical data can be obtained through the equipment's management system or production record database. This data includes vibration signals, operating frequency, and load conditions of the equipment under normal production conditions. All data should be transmitted to a central database or cloud platform for storage and analysis via standardized communication protocols (such as Industrial Internet of Things or SCADA systems) to ensure data accuracy and real-time performance.

[0055] It's important to note that equipment fluctuation value refers to the proportion of equipment whose vibration abnormality exceeds a preset threshold during drug production. In short, it measures the frequency of abnormal equipment fluctuations during production. A higher equipment fluctuation value indicates a higher proportion of equipment experiencing abnormal vibrations during production, reflecting overall instability or potential malfunctions in the production equipment. For example, if multiple pieces of equipment exceed the set threshold during a particular batch, it suggests potential operational problems, such as equipment aging, improper maintenance, or changes in the production environment, potentially leading to instability in the production process or fluctuations in product quality. Therefore, a higher equipment fluctuation value means an increased risk of equipment failure or abnormalities during production, directly impacting drug quality. To ensure final drug quality, increasing the frequency of quality checks in cases of high equipment fluctuation helps identify potential quality issues earlier and reduces the production of substandard products. For instance, a high equipment fluctuation value may necessitate more frequent quality inspections, such as sample testing or full-batch testing, to ensure that every step of production meets quality standards. This approach allows for timely identification and correction of potential production quality problems, preventing the production of defective products and quality incidents.

[0056] In one embodiment, information on changes in drug raw materials during the production process is obtained, and the raw material change value is calculated based on the drug raw material change information; this is used to assess the degree of change in raw materials during drug production.

[0057] Specifically, the steps for calculating the change value of raw materials based on the information on the change of drug raw materials are as follows:

[0058] The process involves acquiring multidimensional chemical property data of raw materials for each batch of drugs before production, and then normalizing the multidimensional chemical property data to standardize the chemical properties of each raw material to a uniform scale.

[0059] Based on the normalized data, a multidimensional distribution of the raw materials for the current batch is established using the kernel density estimation method.

[0060] Obtain several historical raw material multidimensional distributions during normal drug production from historical data, calculate the KL divergence between each historical raw material multidimensional distribution and the current batch's raw material multidimensional distribution, and use the reciprocal of the KL divergence as the raw material difference between the corresponding historical raw material multidimensional distribution and the current batch's raw material multidimensional distribution;

[0061] The minimum raw material difference value is taken as the raw material change value.

[0062] It is important to note that the data acquisition method in calculating raw material variation values ​​primarily relies on real-time monitoring during the production process and the accumulation of historical records. Firstly, real-time data acquisition involves collecting chemical property data of raw materials through sensors installed in the raw material storage and processing stages (such as temperature sensors, humidity sensors, particle size analyzers, and spectrometers). This data includes multi-dimensional chemical properties such as raw material composition, particle size, humidity, moisture content, temperature, and pH value, which directly affect the production quality of the drug. Raw material data is typically uploaded to the production monitoring system or database in real time to ensure its timeliness and accuracy. Secondly, historical data is obtained from past production batch records. This data is usually stored in the company's quality management system or enterprise resource planning (ERP) system and contains the chemical property distribution and quality information of all batches of raw materials under normal production conditions. Historical data provides a stable benchmark for raw material properties, allowing for comparison and analysis of current batch data. By combining real-time monitoring data with historical data, the differences between the current batch and the standard batch can be accurately analyzed and assessed, thereby calculating the raw material variation value and ensuring the stability of the production process and product quality control.

[0063] It's important to note that the raw material variation value refers to the degree of difference between the current batch of raw materials and historical normal production batches during drug manufacturing. It measures the degree of change in the chemical property distribution of the current batch of raw materials compared to historical standard raw materials. By calculating the Kullback-Leibler (KL) gradient between the multidimensional distribution of raw materials in each historical batch and the current batch, a raw material variation value is obtained, representing the change in multiple chemical property dimensions of the raw material. If the current batch of raw materials differs significantly from historical batches, the KL divergence is large, and the raw material variation value is also large. A larger raw material variation value indicates a more significant difference in chemical properties between the current batch of raw materials and the standard production batch, which usually foreshadows potential instability or quality problems in the production process. For example, if the particle size, moisture content, or impurity content of the current batch of raw materials differs significantly from historical batches, this may lead to fluctuations in key properties such as drug solubility, stability, or bioavailability. Therefore, to ensure the stability of the drug manufacturing process and product quality, it is necessary to increase the frequency of quality identification and conduct more frequent and in-depth quality testing. For example, if the raw material variation is high, it may be necessary to conduct more rigorous sampling inspections on each production batch or perform more laboratory analyses to prevent potential quality issues from affecting the final product's pass rate.

[0064] In one implementation method, the primary purpose of calculating raw material variation values ​​is to accurately assess the differences between the current batch of raw materials and historical standard production batches, thereby providing a scientific basis for quality control in the drug manufacturing process. Using normalized multidimensional chemical property data and combining it with kernel density estimation to construct a multidimensional distribution model of the raw materials can effectively capture the full picture of raw material properties, avoiding the bias that may result from single-attribute data. By calculating the KL divergence between the current batch of raw materials and historical batches, the differences between the two can be quantified, reflecting the changes in raw material properties across multiple dimensions.

[0065] The advantage of this method lies in its ability to deeply analyze the differences in raw materials across multiple dimensions, rather than focusing solely on a single chemical property. This multi-dimensional analysis comprehensively assesses the potential impact of raw material variations on the production process and product quality. Furthermore, using the advanced statistical method of KL divergence, it can accurately measure differences between probability distributions. Compared to traditional simple comparison methods, it can more accurately capture subtle changes in raw materials, effectively reducing the risk of quality problems caused by raw material fluctuations. Ultimately, by using the minimized raw material variance as a benchmark, it ensures that raw materials in the production process better meet expectations, reduces the occurrence of defective products, and improves the stability and controllability of the production process.

[0066] In one embodiment, production scale information during the production process is obtained, and a production scale value is calculated based on the production scale information; this is used to assess the extent of production scale during drug production.

[0067] Specifically, the steps for calculating the production scale value based on production scale information are as follows:

[0068] For each batch of drugs, obtain the total actual production weight of the corresponding batch of drugs, and compare the total actual production weight with the preset standard total production weight to obtain the production weight ratio;

[0069] For each batch of drugs, obtain the time from the start to the end of production for the corresponding batch of drugs to get the total production duration, and divide the total production duration by the preset standard production time to get the production time ratio;

[0070] For each batch of drugs, obtain the temperature and humidity during the production process of the corresponding batch of drugs, and calculate the time ratio of temperature and humidity deviating from the preset normal temperature and humidity range.

[0071] The production scale value is obtained by adding the production weight ratio, production time ratio, and the time ratio of temperature and humidity deviating from the preset normal temperature and humidity range.

[0072] It's important to note that the data acquisition methods used in calculating production scale primarily rely on real-time monitoring systems and historical production data. First, the total weight of actual production is obtained through automated weighing equipment, typically installed at different stages of the production line. This equipment accurately records the production weight of each batch of drugs and uploads it to the production management system in real time. Second, the total production duration is obtained through the production line's time tracking system. This system automatically records the time from the start to the end of production and compares it with the preset standard production time to ensure that the production process is completed within the specified timeframe. Furthermore, temperature and humidity data are monitored in real time by temperature and humidity sensors installed in the production environment. These sensors record the temperature and humidity at every moment during production, and the data is synchronized to the production monitoring system. Based on this real-time data, the system can calculate the percentage of time that temperature and humidity deviate from the preset normal range, further assessing the stability of the production environment. By integrating all this data, the system can comprehensively evaluate the production scale of each batch of drugs, ensuring that the production process meets expected standards and adjusting production parameters promptly when anomalies are detected, thus guaranteeing the consistency and stability of drug quality.

[0073] It's important to note that production scale refers to a comprehensive indicator measuring the difference between the batch production scale and standard production conditions during drug manufacturing. It reflects the degree of deviation in different aspects of the production process through a comprehensive assessment of the production weight ratio, production time ratio, and temperature and humidity deviation ratio. For example, if the total production weight of a batch of drugs significantly exceeds the standard production weight, or the production time is significantly prolonged, or the temperature and humidity of the production environment deviate from the standard range for an extended period, these factors can all lead to instability in the quality of the drug during production. A larger production scale means that the production process of that batch of drugs deviates more from standard conditions, and potentially involves more production problems. For example, a larger production weight ratio may indicate improper handling of equipment or raw materials during production, a longer production time ratio may indicate low production efficiency, and temperature and humidity deviations may affect the quality and stability of the drug. These deviations can lead to fluctuations in drug quality, thereby increasing quality risks. Therefore, the larger the production scale, the more frequent the identification of drug production quality should be, in order to conduct more frequent and detailed monitoring and inspection of potential quality problems, thereby ensuring the pass rate and safety of the drug. For example, for batches with large temperature and humidity deviations, more laboratory testing may be required after production, or additional stability monitoring may be needed to ensure that the product meets quality standards.

[0074] In one embodiment, the step of calculating the quality identification change value based on equipment fluctuation value, raw material change value, and production scale value is as follows:

[0075] The equipment fluctuation value, raw material change value, and production scale value are converted into a comprehensive feature vector. The comprehensive feature vector is used as the input of the machine learning model. The machine learning model uses the prediction of each quality identification change value label of each set of comprehensive feature vectors as the prediction objective and minimizes the sum of prediction errors for all quality identification change value labels as the training objective. The machine learning model is trained until the sum of prediction errors converges and the model training stops. The quality identification change value is determined based on the model output. The machine learning model is a multinomial regression model.

[0076] It's important to note the following steps for data preparation and feature extraction: First, the equipment fluctuation value, raw material change value, and production scale value for each batch of drugs are used as features to form a comprehensive feature vector. These features represent fluctuations and changes in different aspects of the production process, fully reflecting the potential risks to drug quality. Machine learning model construction: Using these comprehensive feature vectors as input, a multinomial regression model is built. Multinomial regression models can handle non-linear relationships and are suitable for complex data patterns. The model's goal is to predict the quality identification change value for each batch of drugs using these features, indicating whether more quality checks or adjustments are needed for that batch. Training and optimization: During training, the machine learning model predicts the corresponding quality identification change value label based on each set of feature vectors and optimizes the model parameters by minimizing the prediction error. The goal of minimizing the prediction error is to adjust the model parameters by comparing the difference between the model's predicted value and the actual label until the error converges. This process typically uses gradient descent or other optimization algorithms for training. Output of quality identification change values: When the model training is complete and convergence is achieved, the trained model is used to predict the quality identification change value. Based on the model's output, it is determined which batches of drugs require increased quality checks. If the quality identification change value is high, it indicates that there is a significant possibility of quality fluctuation in this batch of drugs, requiring more quality control and inspection.

[0077] In one implementation, the advantage of this method is that it can automatically and dynamically adjust the quality inspection strategy according to the fluctuation characteristics of different dimensions in the production process, thereby improving the quality control efficiency of the production process, avoiding rough human judgment, and improving the reliability and compliance of drug production.

[0078] In one embodiment, the step of determining whether to change the number of production quality identifications for the corresponding batch of drugs based on the quality identification change value is as follows:

[0079] Compare the quality identification change value with the preset quality identification change value threshold. If the quality identification change value is less than the preset quality identification change value threshold, then there is no need to change the number of times the production quality identification is changed for the corresponding batch of drugs. The drug production quality can still be identified according to the original fixed number of identifications.

[0080] If the quality identification change value is not less than the preset quality identification change value threshold, then the number of times the production quality identification is changed for the corresponding batch of drugs needs to be changed, and the number of drug quality identifications needs to be increased.

[0081] It should be noted that the preset quality identification change value threshold is set by professionals based on the actual situation, and no specific limitations or details are provided.

[0082] In one implementation, the process of determining whether to adjust the frequency of quality identification based on the quality identification change value is actually done by comparing the quality identification change value with a pre-set quality identification change value threshold. When the quality identification change value is less than the threshold, it indicates that the production quality variation of this batch of drugs is relatively small and the production process is relatively stable. Therefore, there is no need to increase the number of quality identifications; the existing fixed number of quality tests can continue. For example, if equipment fluctuations, raw material changes, and production scale deviations in the production process of a batch of drugs are all within acceptable limits, then the quality risk of this batch of drugs is low, and no additional testing is required. However, when the quality identification change value is greater than or equal to the preset threshold, it means that there are significant fluctuations or anomalies in the production process of this batch of drugs, which may lead to quality instability. Therefore, it is necessary to increase the number of quality identifications and conduct more frequent and detailed quality tests. For example, if there are large equipment fluctuations or large fluctuations in raw material composition during the production process, it may affect the final quality of the drug. In this case, it is necessary to increase the number of tests based on the quality identification change value of this batch of drugs to ensure that the drug quality meets the standards and prevent unqualified products from entering the market. In this way, quality control can be more refined and flexible, thereby improving the safety and reliability of drug production.

[0083] Based on the same inventive concept, embodiments of the present invention also provide a drug production quality identification platform based on big data. See also Figure 2 , Figure 2 This is a framework diagram of a big data-based drug production quality identification platform provided in an embodiment of the present invention. The platform includes:

[0084] Equipment fluctuation module: After production, the drugs are divided into several batches according to the corresponding production batches. For each batch, the equipment fluctuation information during the production process is obtained, and the equipment fluctuation value is calculated based on the equipment fluctuation information; this is used to assess the degree of equipment fluctuation during drug production.

[0085] Raw material change module: Acquires information on changes in drug raw materials during the production process and calculates the raw material change value based on the information; used to assess the degree of change in raw materials during drug production.

[0086] Production Scale Module: Acquires production scale information during the production process and calculates the production scale value based on the production scale information; used to assess the extent of production scale during drug production.

[0087] Identification module: Calculates the quality identification change value based on equipment fluctuation value, raw material change value, and production scale value, and determines whether to change the number of times the production quality identification should be changed for the corresponding batch of drugs based on the quality identification change value.

[0088] Based on the big data-based drug production quality identification platform provided in this invention, the identification frequency can be adjusted according to the actual volatility, process complexity, or potential risks of different batches of products through the above-mentioned method. This dynamic identification strategy may ensure that high-risk batches are identified sufficiently and that no quality hazards are missed, thereby further improving the big data quality identification method in terms of production flexibility and quality assurance capabilities.

[0089] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the present invention should still fall within the patent coverage of the present invention.

Claims

1. A drug manufacturing quality identification method based on big data, characterized in that, Includes the following steps: After production, the drugs are divided into several batches according to their corresponding production batches. For each batch, equipment fluctuation information during the production process is obtained, and equipment fluctuation value is calculated based on the equipment fluctuation information; this is used to assess the degree of equipment fluctuation during drug production. It acquires information on changes in drug raw materials during the production process and calculates the changes in raw material values ​​based on this information; this is used to assess the degree of change in raw materials during drug production. To obtain production scale information during the production process and calculate the production scale value based on the production scale information; used to assess the extent of production scale in the drug production process; The quality identification change value is calculated based on the equipment fluctuation value, raw material change value, and production scale value, and the quality identification change value is used to determine whether the production quality identification number of the corresponding batch of drugs needs to be changed. The steps for obtaining equipment fluctuation information during the production process and calculating equipment fluctuation values ​​based on this information are as follows: The production vibration signals of each piece of equipment in the drug production process are obtained in time sequence, and the production vibration signals are preprocessed and converted from the time domain to the frequency domain by the fast Fourier algorithm to obtain the frequency spectrum of the vibration signals. Extract the maximum standard vibration signal corresponding to each piece of equipment during normal drug production from historical data, and record the frequency of the maximum standard vibration signal in the frequency domain as the vibration frequency threshold. The frequency spectrum of the current production vibration signal is defined with the vibration frequency threshold as the boundary. The region in the frequency spectrum with frequencies greater than the vibration frequency threshold is defined as the high-frequency region. The total energy of the high-frequency region and the total energy of the frequency spectrum are calculated. The energy of the high-frequency region is divided by the total energy of the frequency spectrum to obtain the vibration anomaly degree of the corresponding equipment. The vibration anomaly of each device is compared with the corresponding preset minimum threshold for vibration anomaly. If the vibration anomaly is not less than the corresponding preset minimum threshold for vibration anomaly, the corresponding device is recorded as a fluctuating device. The total number of fluctuating devices is divided by the total number of devices in the drug production process to obtain the device fluctuation value. The steps for calculating the raw material transformation value based on the drug raw material transformation information are as follows: The process involves acquiring multidimensional chemical property data of raw materials for each batch of drugs before production, and then normalizing the multidimensional chemical property data to standardize the chemical properties of each raw material to a uniform scale. Based on the normalized data, a multidimensional distribution of the raw materials for the current batch is established using the kernel density estimation method. Obtain several historical raw material multidimensional distributions during normal drug production from historical data, calculate the KL divergence between each historical raw material multidimensional distribution and the current batch's raw material multidimensional distribution, and use the reciprocal of the KL divergence as the raw material difference between the corresponding historical raw material multidimensional distribution and the current batch's raw material multidimensional distribution; The minimum raw material difference value is taken as the raw material change value; The steps for calculating the production scale value based on production scale information are as follows: For each batch of drugs, obtain the total actual production weight of the corresponding batch of drugs, and compare the total actual production weight with the preset standard total production weight to obtain the production weight ratio; For each batch of drugs, obtain the time from the start to the end of production for the corresponding batch of drugs to get the total production duration, and divide the total production duration by the preset standard production time to get the production time ratio; For each batch of drugs, obtain the temperature and humidity during the production process of the corresponding batch of drugs, and calculate the time ratio of temperature and humidity deviating from the preset normal temperature and humidity range. The production scale value is obtained by adding the production weight ratio, production time ratio, and the time ratio of temperature and humidity deviating from the preset normal temperature and humidity range. The steps for calculating the quality identification change value based on equipment fluctuation value, raw material change value, and production scale value are as follows: The equipment fluctuation value, raw material change value, and production scale value are converted into a comprehensive feature vector. The comprehensive feature vector is used as the input of the machine learning model. The machine learning model uses the prediction of each quality identification change value label of each set of comprehensive feature vectors as the prediction objective and minimizes the sum of prediction errors for all quality identification change value labels as the training objective. The machine learning model is trained until the sum of prediction errors converges and the model training stops. The quality identification change value is determined based on the model output. The machine learning model is a multinomial regression model.

2. The drug production quality identification method based on big data according to claim 1, characterized in that, The steps for determining whether to change the number of production quality identifications for the corresponding batch of drugs based on the quality identification change value are as follows: Compare the quality identification change value with the preset quality identification change value threshold. If the quality identification change value is less than the preset quality identification change value threshold, then there is no need to change the number of times the production quality identification is changed for the corresponding batch of drugs. The drug production quality can still be identified according to the original fixed number of identifications. If the quality identification change value is not less than the preset quality identification change value threshold, then the number of times the production quality identification is changed for the corresponding batch of drugs needs to be changed, and the number of drug quality identifications needs to be increased.

3. A big data-based drug production quality identification cloud platform, used to implement the big data-based drug production quality identification method according to any one of claims 1-2, characterized in that: The gimbal includes: Equipment fluctuation module: After production, the drugs are divided into several batches according to the corresponding production batches. For each batch, the equipment fluctuation information during the production process is obtained, and the equipment fluctuation value is calculated based on the equipment fluctuation information; this is used to assess the degree of equipment fluctuation during the drug production process. Raw material change module: Acquires information on changes in drug raw materials during the production process and calculates the raw material change value based on the information; used to assess the degree of change in raw materials during drug production. Production Scale Module: Acquires production scale information during the production process and calculates the production scale value based on the production scale information; used to assess the extent of production scale during drug production. Identification module: Calculates the quality identification change value based on equipment fluctuation value, raw material change value, and production scale value, and determines whether to change the number of times the production quality identification should be changed for the corresponding batch of drugs based on the quality identification change value.

Images