Risk early warning method for pharmaceutical production, electronic equipment and medium

Abstract

The embodiment of the invention provides a risk early warning method for medicine production, electronic equipment and a medium, and belongs to the technical field of medicine production and artificial intelligence. The method comprises the following steps: acquiring at least one of personnel behavior data, equipment operation data, process data, material data and environment data in a drug production area, performing risk detection on each source data, and generating a reference event according to other source data in multi-source data to predict the risk of a risk event, and executing an early warning operation matched with the risk event and the event risk degree corresponding to the source data. According to the embodiment of the invention, real-time risk event detection and early warning operation are carried out on the drug production area, so that the final influence of risk events such as non-standard operation on the drug quality can be reduced fundamentally.

Description

Technical Field

[0001] This application relates to the fields of pharmaceutical manufacturing and artificial intelligence technology, and in particular to a risk warning method, electronic device and medium for pharmaceutical manufacturing. Background Technology

[0002] Currently, in the pharmaceutical manufacturing process, quality risk management is a core element in ensuring drug safety, efficacy, and compliance. Drug manufacturing is a highly sensitive process involving multiple critical stages such as raw material handling, synthesis reactions, and aseptic packaging. Any minor operational error or deviation in environmental parameters can lead to substandard drug quality.

[0003] Current technologies primarily rely on manual analysis of adverse drug reaction (ADR) reports for drug risk monitoring, which fails to provide real-time early warning of production risks. For example, existing risk management models typically only identify risks during routine manual inspections, resulting in a significant lag in risk identification during drug production. Risks are often only discovered after they have already impacted drug quality (e.g., causing batch contamination, parameter drift, or cross-contamination) through finished product inspection or deviation investigations, by which time serious consequences such as material waste, production delays, or even product recalls have occurred. Therefore, how to achieve risk early warning in drug production to reduce the impact of risk events such as improper operation on drug quality has become an urgent technical problem to be solved. Summary of the Invention

[0004] The main objective of this application is to propose a risk warning method, electronic device, and medium for pharmaceutical production, aiming to achieve risk warning in pharmaceutical production and reduce the impact of risk events such as non-standard operation on pharmaceutical quality.

[0005] To achieve the above objectives, a first aspect of this application proposes a risk warning method for pharmaceutical production, the method comprising: Acquire multi-source data of the drug production area in the target pharmaceutical factory, wherein the multi-source data includes at least one of the following sources: production personnel behavior data, production equipment operation data, production process data, production material data, and production environment data; Risk detection is performed on each of the source data in the multi-source data to obtain the risk event corresponding to the source data, and reference events are generated based on other source data in the multi-source data. Based on the reference event, risk prediction is performed on the risk event to obtain the event risk level; Perform an event risk warning operation that matches the risk event and the degree of risk of the event.

[0006] In some embodiments, the step of predicting the risk level of the risk event based on the reference event includes: Acquire historical risk record data, identify the frequency of occurrence of the risk events from the historical risk record data, and generate a risk probability index based on the frequency of occurrence; Based on the reference event, the consequences of the risk event are predicted to obtain a risk severity index; Predict the sensitivity of the reference event to the risk event to obtain a risk sensitivity index; The risk level of the event is obtained by fusing the risk probability index, the risk severity index, and the risk sensitivity index.

[0007] In some embodiments, the reference event includes at least one of the following events: a first reference event representing the product quality dimension, a second reference event representing the personnel safety dimension, a third reference event representing the equipment damage dimension, and a fourth reference event representing the production delay dimension; The process of predicting the consequences of the risk event based on the reference event to obtain a risk severity index includes: When the reference event is the first reference event representing the product quality dimension, the consequences of the risk event are scored based on the first reference event to obtain a product quality impact score. When the reference event is the second reference event representing the personnel safety dimension, the consequences of the risk event are scored based on the second reference event to obtain a personnel safety impact score. When the reference event is the third reference event representing the dimension of equipment damage, the consequences of the risk event are scored based on the third reference event to obtain the equipment damage impact score. When the reference event is the fourth reference event representing the dimension of production delay, the consequences of the risk event are scored based on the fourth reference event to obtain the production delay impact score. The risk severity index is obtained by fusing the product quality impact score, the personnel safety impact score, the equipment damage impact score, and the production delay impact score.

[0008] In some embodiments, the risk event includes a production process control risk event; the step of performing risk detection based on each source data in the multi-source data to obtain the risk event corresponding to the source data includes: When the source data is the production process data, feature extraction is performed on the production process data to obtain the production process control parameter features; Obtain process control baseline features; Based on the process control baseline characteristics and the production process control parameter characteristics, the production process control risk events corresponding to the production process data are generated.

[0009] In some embodiments, generating the production process control risk event corresponding to the production process data based on the process control baseline characteristics and the production process control parameter characteristics includes: Based on the process control baseline characteristics, parameter deviation analysis is performed on the production process control parameter characteristics to obtain the first risk value; The distribution differences of the characteristic mean and standard deviation of the production process control parameter characteristics are compared based on the characteristic mean and standard deviation of the process control baseline characteristics to obtain the second risk value; Based on the first risk value and the second risk value, the production process control risk event corresponding to the production process data is generated.

[0010] In some embodiments, after performing an event risk warning operation matching the risk event and the degree of event risk, the method further includes: If the risk level of the event meets the predetermined conditions, a risk assessment is performed on the drug production area based on the risk event and the multi-source data to obtain the regional risk level. Perform regional risk warning operations that match the risk level of the region.

[0011] In some embodiments, the step of conducting a risk assessment of the drug production area based on the risk event and the multi-source data to obtain the regional risk level includes: Feature extraction is performed on the multi-source data to obtain multi-source data features; Based on the multi-source data features and the risk events, causal effects are identified in the drug production area to obtain the risk causal effect strength; wherein, the risk causal effect strength is used to characterize the degree to which the multi-source data features and the risk events have a risk impact on the drug production area; Based on the strength of the risk causal effect, risk source data features that have a risk impact on the drug production area are selected from the multi-source data features; Obtain the causal effect weights of the risk source data features, and calculate the regional risk of the drug production area based on the risk source data features and the causal effect weights to obtain the regional risk value; The degree of risk in a region is determined based on the region's risk value.

[0012] In some embodiments, performing a regional risk warning operation matched to the risk level of the region includes: Obtain regional risk description information that matches the risk level of the region, and retrieve intervention knowledge from a preset drug production knowledge graph based on the regional risk description information to obtain risk intervention knowledge; Based on the regional risk description information and the risk intervention knowledge, risk intervention instruction information is generated; Based on the risk intervention indication information, the pre-trained risk intervention generation model is instructed to generate an intervention plan, thereby obtaining the risk intervention plan; According to the risk intervention plan, the regional risk warning operation is performed in accordance with the risk level of the region.

[0013] To achieve the above objectives, a second aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the method described in the first aspect.

[0014] A third aspect of this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the method of the first aspect described above.

[0015] The risk warning method, electronic equipment, and media for pharmaceutical production proposed in this application firstly acquire multi-source data from the pharmaceutical production area of ​​the target pharmaceutical plant, including at least one of the following: production personnel behavior data, production equipment operation data, production process data, production material data, and production environment data. This achieves real-time and comprehensive data collection of all elements in the pharmaceutical production process, providing a timely and comprehensive data foundation for subsequent risk warning. Secondly, it performs independent risk detection on each source data to obtain the corresponding risk events. By combining other source data, it generates reference events, enabling in-depth identification of complex risk patterns through multi-dimensional information correlation analysis, significantly improving the accuracy of risk assessment in pharmaceutical production. Furthermore, based on the reference events, it predicts the risk of the identified risk events and quantifies their risk level, significantly shifting the risk control node from post-event traceability to during or even before the event. This effectively solves the problem of lag in traditional methods that rely on manual analysis of ADR reports and executes event risk warning operations matched to the risk events and their risk levels. This allows for timely resolution of already occurred risk events, fundamentally reducing the ultimate impact of risk events such as non-standard operations on drug quality. Attached Figure Description

[0016] Figure 1 This is a flowchart of a risk warning method for drug production provided in the embodiments of this application; Figure 2 yes Figure 1 The flowchart of step S102 in the document; Figure 3 yes Figure 2 The flowchart of step S203 in the process; Figure 4 yes Figure 1 The flowchart of step S103 in the process; Figure 5 yes Figure 4 The flowchart of step S402 in the document; Figure 6 This is another flowchart of the risk warning method for drug production provided in the embodiments of this application; Figure 7 yes Figure 6 The flowchart of step S601 in the process; Figure 8 yes Figure 6 The flowchart of step S602 in the document; Figure 9 This is a schematic diagram of the hardware structure of the electronic device provided in the embodiments of this application. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0018] It should be noted that although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first," "second," etc., in the specification, claims, and the aforementioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0020] First, let's analyze some of the terms used in this application: Artificial intelligence (AI) is a new branch of computer science that studies, develops, and applies theories, methods, technologies, and systems to simulate, extend, and expand human intelligence. It aims to understand the essence of intelligence and produce intelligent machines that can react in a way similar to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing, and expert systems. AI can simulate the information processes of human consciousness and thought. Furthermore, AI utilizes digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceiving the environment, acquiring knowledge, and using that knowledge to achieve optimal results.

[0021] Based on this, embodiments of this application provide a risk warning method, electronic device, and medium for drug production, aiming to achieve risk warning in drug production and reduce the impact of risk events such as non-standard operation on drug quality.

[0022] The risk warning method, electronic device and medium for drug production provided in this application are specifically described through the following embodiments. First, the risk warning method for drug production in this application is described.

[0023] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0024] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0025] The risk warning method for drug production provided in this application relates to the fields of drug production and artificial intelligence technology. This risk warning method for drug production can be applied to a terminal, a server, or software running on either a terminal or a server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, etc.; the server can be configured as an independent physical server, a server cluster composed of multiple physical servers, or a distributed device; it can also be configured as a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms; the software can be an application implementing the risk warning method for drug production, but is not limited to the above forms.

[0026] This application can be used in a wide variety of general-purpose or special-purpose computer device environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor devices, microprocessor-based devices, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above devices, etc. This application can be described in the general context of computer-executable instructions that are executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0027] Please see Figure 1 , Figure 1 This is a flowchart of a risk warning method for drug production provided in an embodiment of this application. Figure 1 The method may include, but is not limited to, steps S101 to S104.

[0028] Step S101: Obtain multi-source data of the drug production area in the target pharmaceutical factory. The multi-source data includes at least one of the following source data: production personnel behavior data, production equipment operation data, production process data, production material data, and production environment data.

[0029] Step S102: Perform risk detection based on each source data in the multi-source data to obtain the risk event corresponding to the source data, and generate reference events based on other source data in the multi-source data.

[0030] Step S103: Based on the reference event, perform risk prediction on the risk event to obtain the risk level of the event.

[0031] Step S104: Execute an event risk warning operation that matches the risk event and the degree of event risk.

[0032] Steps S101 to S104 of this application embodiment firstly acquire multi-source data from the target pharmaceutical factory's drug production area, including at least one of the following: production personnel behavior data, production equipment operation data, production process data, production material data, and production environment data. This achieves real-time and comprehensive data collection of all elements in the drug production process, providing a timely and comprehensive data foundation for subsequent risk warnings. Secondly, independent risk detection is performed on each source data to obtain the corresponding risk events. Reference events are generated by combining other source data, enabling in-depth identification of complex risk patterns through multi-dimensional information correlation analysis, significantly improving the accuracy of risk assessment in drug production. Furthermore, risk prediction and quantification of the risk level of the identified risk events are performed based on the reference events. This significantly shifts the risk control node from post-event traceability to during or even before the event, effectively solving the lag problem of traditional methods relying on manual analysis of ADR reports. Event risk warning operations are executed according to the risk events and their risk levels, enabling timely resolution of already occurred risk events and fundamentally reducing the ultimate impact of risk events such as non-standard operations on drug quality.

[0033] In step S101 of some embodiments, specifically, the target pharmaceutical factory refers to a pharmaceutical manufacturing site with complete information infrastructure, which is internally divided into several functional units in accordance with the Good Manufacturing Practice (GMP) for pharmaceuticals.

[0034] Specifically, the pharmaceutical production area is the physical space with a regional code within the aforementioned functional units, such as the Class B filling room numbered AB-07 within the aseptic preparation workshop.

[0035] Specifically, multi-source data refers to a collection of multi-source data collected from multiple independent sources and through various heterogeneous sensors under the same regional encoding. This multi-source data includes at least one of the following source data: production personnel behavior data, production equipment operation data, production process data, production material data, and production environment data.

[0036] Furthermore, production personnel behavior images or video frame sequences can be collected using deployed high-definition network cameras. These images or video frame sequences can be input into a target detection model (such as YOLO or Faster R-CNN) to locate the production personnel region and extract the directional gradient histogram features of the production personnel region. By tracking the changes in the bounding boxes or skeleton key points of the production personnel in multiple consecutive frames, the temporal behavior center-of-gravity trajectory features of the production personnel can be extracted to describe the movement patterns and rhythms of the behavior. The directional gradient histogram features and the temporal behavior center-of-gravity trajectory features are concatenated to obtain the target personnel behavior features. The target personnel behavior features are then classified using a support vector machine (SVM) to obtain production personnel behavior data such as changing clothes, cleaning, calibration, feeding materials, leaning, disinfection, and wearing personal protective equipment. Industrial communication protocols (such as OPC UA, Open Platform Communications Unified Architecture) can be used to access programmable logic controllers (PLCs) or distributed control systems (DCSs) in reactors, freeze dryers, and filling lines. Data on production equipment operation, such as temperature, pressure, speed, and current, can be obtained from the Manufacturing Execution System (MES); production process data, such as drug formulation steps, drug process setpoints, and drug production process control limits, can be obtained from the electronic batch records in the Manufacturing Execution System (MES); production material data, such as drug raw material batch numbers, supplier codes, and inspection and release status, can be obtained from the Laboratory Information Management System (LIMS); and production environment data, such as temperature, humidity, differential pressure, and suspended particles, can be obtained from temperature and humidity sensors, differential pressure sensors, and suspended particle counters deployed in the clean areas of the drug production area.

[0037] In this embodiment, by acquiring multi-source data including at least one of the following in the drug production area of ​​the target pharmaceutical factory: production personnel behavior data, production equipment operation data, production process data, production material data, and production environment data, real-time and comprehensive data collection of all elements in the drug production process is realized. This provides a timely and comprehensive data foundation for subsequent risk warning and avoids risk misjudgment and omission due to incomplete data collection or time misalignment.

[0038] Please see Figure 2 In some embodiments, the risk event includes a production process control risk event, and step S102 may include, but is not limited to, steps S201 to S203: Step S201: When the source data is production process data, feature extraction is performed on the production process data to obtain the production process control parameter features.

[0039] Step S202: Obtain process control baseline features.

[0040] Step S203: Based on the process control baseline characteristics and the production process control parameter characteristics, generate the production process control risk events corresponding to the production process data.

[0041] In step S201 of some embodiments, specifically, the production process control parameter features refer to the vector representation of production process data.

[0042] For example, in the production of oral solid dosage forms, the production process data can include the stirring power of the granulation process, the total mixing time and mixing speed of the mixing process, the main tableting force and tableting speed of the tableting process, and the spraying rate of the film coating process. The characteristics of the production process control parameters are the vector representations of the production process data.

[0043] In step S202 of some embodiments, specifically, the process control baseline feature refers to a reference vector representation used to determine whether the production process meets predetermined standards or is in a statistically controlled state.

[0044] For example, in the production of oral solid dosage forms, the process control baseline characteristics for the inlet air temperature control stage of the film coating process can be "temperature setpoint 50°C, upper limit of allowable fluctuation 53°C, lower limit 47°C". For the tableting process of oral solid dosage forms, the baseline characteristics of the main tableting force can be determined based on multiple historical qualified batch data, that is, the mean can be 40kN, the upper limit of the standard deviation can be 42kN, and the lower limit can be 38kN.

[0045] Please see Figure 3 In some embodiments, step S203 may include, but is not limited to, steps S301 to S303: Step S301: Based on the process control baseline characteristics, perform parameter deviation analysis on the production process control parameter characteristics to obtain the first risk value.

[0046] Step S302: Based on the characteristic mean and standard deviation of the process control baseline characteristics, compare the distribution differences of the characteristic mean and standard deviation of the production process control parameter characteristics to obtain the second risk value.

[0047] Step S303: Generate production process control risk events corresponding to the production process data based on the first risk value and the second risk value.

[0048] In step S301 of some embodiments, specifically, the first risk value refers to the value obtained after deviation analysis between the characteristics of the production process control parameters and the characteristics of the process control baseline, which is used to quantitatively characterize the degree of instantaneous exceedance of the production process control parameters.

[0049] Specifically, the standard score of the production process control parameter characteristic can be calculated using the following formula, and the standard score represents the first risk value:

[0050] Where Z represents the standard score of the production process control parameter characteristic T, and T represents the production process control parameter characteristic. The characteristic mean representing the baseline characteristics of process control. The characteristic standard deviation represents the baseline characteristics of process control.

[0051] For example, in the film coating process of oral solid dosage forms, if the spray rate of the production process control parameter characteristic is 121.8 mL / min, which exceeds the characteristic mean of 121.0 mL / min corresponding to the process control baseline characteristic, and the characteristic standard deviation corresponding to the process control baseline characteristic is 0.2, then the first risk value corresponding to the spray rate of the production process control parameter characteristic is determined to be 4.

[0052] In step S302 of some embodiments, specifically, the second risk value refers to the distribution difference value between the mean and standard deviation of the process control baseline characteristics and the characteristic mean and standard deviation of the production process control parameters, which is used to characterize the stability of the drug production process.

[0053] Specifically, the standard deviation of the process control baseline characteristic can be divided by the mean of the characteristic to obtain the baseline coefficient of variation of the process control baseline characteristic; the standard deviation of the production process control parameter characteristic can be divided by the mean of the characteristic to obtain the actual coefficient of variation of the production process control parameter characteristic; and the price baseline coefficient of variation can be divided by the actual coefficient of variation to obtain the second risk value.

[0054] In step S303 of some embodiments, specifically, a manufacturing process control risk event refers to a process abnormality that occurs during the drug manufacturing process.

[0055] Specifically, the first risk value and the second risk value can be weighted and merged to obtain a comprehensive production process risk value. It is then determined whether the comprehensive production process risk value meets the preset production process risk conditions (such as being greater than the preset production process parameter threshold). If it does, a production process control risk event corresponding to the production process data is generated.

[0056] Through steps S301 to S303, not only can obvious abnormalities in the production process that have occurred be reported, but also the distribution trend of characteristics can be captured based on the differences in the distribution of production process characteristics, so as to issue an early warning of instability in the production process. This shifts the risk intervention point in drug production from post-event correction to pre-event prevention, providing a direct decision-making trigger basis for avoiding sudden failures in the drug production process.

[0057] Through steps S201 to S203, it is possible to combine process control baseline characteristics to issue event signals in the early stage when the characteristics of production process control parameters deviate but have not yet exceeded the action limit, thus avoiding the lag caused by the post-event recording and review of traditional methods and improving the detection timeliness and accuracy of production process deviations.

[0058] In step S102 of some embodiments, specifically, risk detection is performed based on each source data in the multi-source data to obtain the risk event corresponding to the source data, and a reference event is generated based on other source data in the multi-source data.

[0059] Specifically, a risk event refers to a specific, describable instance of anomaly corresponding to each source of data.

[0060] For example, in the process of tableting oral solid dosage forms, the risk events corresponding to production personnel behavior data could be failure to check tablet weight at the prescribed intervals, the risk events corresponding to production equipment operation data could be a continuous drop in pressure, the risk events corresponding to production material data could be low material compressibility, and the risk events corresponding to production environment data could be high ambient humidity.

[0061] Specifically, for risk event detection of production personnel behavior data, if support vector machines detect pre-defined non-compliant behaviors such as operators not wearing masks correctly in clean areas or failing to perform disinfection procedures when crossing areas of different cleanliness levels, a risk event is generated, including a description of the behavior, the time of occurrence, and the location of occurrence. For risk event detection of production equipment operation data, it is determined whether the production equipment operation data meets the equipment's standard process curve (such as the upper and lower limit envelopes of adaptive updates for variety, batch, and process during the tableting process of oral solid dosage forms). If not, a risk event is generated, specifying the equipment operating parameters, the magnitude of parameter curve deviation, and the duration. For risk event detection of production material data, material constraints are used to verify material batches, expiration dates, and inspection status. If issues such as "using unreleased raw materials" or "a broken material traceability chain" are found, a risk event corresponding to the production material data is generated. For risk event detection of production environment data, it is determined whether the production environment data meets GMP regulations. If not, a risk event corresponding to the production environment data is generated.

[0062] Specifically, a reference event refers to a set of auxiliary event information obtained from other data sources that are temporally and spatially related to a risk event after identifying a risk event in any source data.

[0063] For example, if the risk event of production personnel behavior is "suspected material spillage at the batching station in the Class B clean area", then other reference events that may be associated with it could be "abnormal peak value of the suspended particle counter reading in the batching station area at the same time" (reference event corresponding to production equipment operation data), "material traceability system shows that the material is a highly active raw material" (reference event corresponding to production material data), "strict control of tablet weight difference" (reference event corresponding to production process data), "instantaneous fluctuation of differential pressure in the Class B clean area shown by the environmental monitoring system" (reference event corresponding to production environment data), and so on, as well as multiple pieces of information from different sources.

[0064] Specifically, when a risk event corresponding to any metadata is identified, the production personnel behavior features corresponding to the production personnel behavior data, the equipment operation data features corresponding to the production equipment operation data, the production material data features corresponding to the production material data, and the production environment data features corresponding to the production environment data can be extracted using a large language model. The production personnel behavior features, equipment operation data features, production process control parameter features, production material data features, and production environment data features are then time-aligned to obtain time-aligned multi-source data features. Within a few minutes before and after the occurrence of a risk event (such as a risk event corresponding to production personnel behavior data), the equipment operation data features, production process control parameter features, production material data features, and production environment data features corresponding to other source data in the pharmaceutical production area are then spatially aligned to obtain spatially aligned multi-source data features. Based on the spatially aligned multi-source data features and the time-aligned multi-source data features, reference events corresponding to other source data are generated.

[0065] Please see Figure 4 In some embodiments, step S103 may include, but is not limited to, steps S401 to S404: Step S401: Obtain historical risk record data, identify the frequency of occurrence of risk events from the historical risk record data, and generate a risk probability index based on the frequency of occurrence. Step S402: Based on the reference event, predict the consequences of the risk event to obtain the risk severity index; Step S403: Predict the sensitivity of the reference event to the risk event and obtain the risk sensitivity index.

[0066] Step S404: The risk level of the event is obtained by fusing the risk probability index, risk severity index and risk sensitivity index.

[0067] In step S401 of some embodiments, specifically, historical risk record data refers to the occurrence data of the same risk event recorded within a historical period.

[0068] Specifically, the risk probability index is calculated from the frequency of the same risk event in historical risk record data. The higher the value, the higher the probability of the risk event recurring.

[0069] For example, regarding the risk event of "operators on the aseptic filling line touching filling components without performing hand disinfection", historical risk record data can be queried to find that this risk event has occurred 5 times in the past 1000 batches of production. Based on this, its occurrence frequency is calculated to be 0.5%, and this frequency value is converted into a risk probability index between 0 and 10 through normalization or piecewise mapping function.

[0070] Please see Figure 5 In some embodiments, the reference event includes at least one of the following events: a first reference event representing the product quality dimension, a second reference event representing the personnel safety dimension, a third reference event representing the equipment damage dimension, and a fourth reference event representing the production delay dimension. Step S402 may include, but is not limited to, steps S501 to S505: Step S501: When the reference event is the first reference event representing the product quality dimension, the consequences of the risk event are scored based on the first reference event to obtain the product quality impact score.

[0071] Step S502: When the reference event is a second reference event representing the personnel safety dimension, the consequences of the risk event are scored based on the second reference event to obtain the personnel safety impact score.

[0072] Step S503: When the reference event is the third reference event representing the equipment damage dimension, the consequences of the risk event are scored based on the third reference event to obtain the equipment damage impact score.

[0073] Step S504: When the reference event is the fourth reference event representing the production delay dimension, the consequences of the risk event are scored based on the fourth reference event to obtain the production delay impact score.

[0074] Step S505: The risk severity index is obtained by integrating the product quality impact score, personnel safety impact score, equipment damage impact score, and production delay impact score.

[0075] In step S501 of some embodiments, specifically, the first reference event refers to accompanying information that is directly related to product quality and is generated from production material data or production process data within the same risk event occurrence time window.

[0076] For example, when the risk event is "filling needle blockage alarm", the first reference event associated with it can be "the number of particles detected by the online visible foreign object detector surges during this time period".

[0077] Specifically, a large language model can be used to match the specific information of risk events and first reference events (such as the affected processes, the criticality of materials, whether sterile core areas are involved, test results, etc.) with corresponding product quality rules for logical judgment and quantitative scoring (such as 1 to 10 points).

[0078] For example, a product quality rule could be that if a risk event occurs in the core area of ​​aseptic filling and is associated with an increase in visible foreign matter, the product quality impact score would be directly assigned the highest score (e.g., 10 points).

[0079] In step S502 of some embodiments, specifically, the second reference event refers to accompanying information related to personnel injury risk generated from production personnel behavior data or production environment data within the same risk event occurrence time window.

[0080] For example, when the risk event is "the pressure safety valve of the reactor malfunctions and releases pressure", the associated second reference event can be "the release area is a potential passageway for personnel".

[0081] Specifically, a large language model can be used to match the specific information of risk events and second reference events with corresponding rules on personnel safety impact for logical judgment and quantitative scoring (e.g., 1 to 10 points).

[0082] In step S503 of some embodiments, specifically, the third reference event refers to accompanying information related to equipment hardware damage generated from production equipment operation data within the same risk event occurrence time window.

[0083] For example, when the risk event is "centrifuge vibration value exceeds the standard", the associated third reference event can be "the equipment has been running at high load continuously for more than the design cycle".

[0084] Specifically, large language models can be used to analyze the current health status of equipment, historical statistics of failure modes, and maintenance cost and cycle information in the third reference event. By comprehensively assessing the severity of potential damage (such as whether it is core irreplaceable equipment) and the complexity of maintenance, an equipment damage impact score (such as 1 to 10 points) can be generated. The higher the score, the more severe the potential equipment damage, the higher the maintenance cost, and the greater the impact on long-term production capacity.

[0085] In step S504 of some embodiments, specifically, the fourth reference event refers to accompanying information related to production cycle time generated from equipment status data within the same risk event occurrence time window.

[0086] For example, when the risk event is "network interruption of production line automation control system", the associated fourth reference event could be "the estimated average repair time required to restore the network is 4 hours".

[0087] Specifically, drug production planning and scheduling data can be input into a large language model, and the model can be used to analyze factors such as the expected downtime caused by risk events, the urgency of affected products, whether there is spare capacity, and whether there is a violation of delivery agreements with customers. Based on the impact of these factors on the overall production and operation plan, a production delay impact score (e.g., 1 to 10 points) can be determined.

[0088] In step S505 of some embodiments, specifically, the risk severity index is used to quantify the consequences of risk events in the dimensions of product quality, personnel safety, equipment damage, and production delay.

[0089] Specifically, the risk severity index can be obtained by weighting and integrating the product quality impact score, personnel safety impact score, equipment damage impact score, and production delay impact score.

[0090] Through steps S501 to S505, the potential consequences of risk events can be comprehensively assessed from the dimensions of product quality, personnel safety, equipment damage, and production delay. This avoids the one-sidedness of only considering quality losses while ignoring safety or equipment damage, and significantly improves the reliability of risk severity assessment.

[0091] In step S403 of some embodiments, specifically, risk sensitivity is a quantified environmental vulnerability index used to characterize the strength of resistance or buffering capacity against identified risk events under the current drug production state and environment, that is, whether the current state makes it easy for the risk to evolve into actual loss.

[0092] Specifically, the analysis references all information related to the stability of the current drug production environment (such as the aging level of equipment and recent maintenance records, fluctuations in environmental control parameters, the training and proficiency level of operators, and whether production is in the early stages of process validation). This information is then input into a large language model to quantify risk amplification or suppression coefficients (e.g., assigning a higher sensitivity weighting value to information about aging equipment and a lower sensitivity weighting value to information about "stable environmental parameters"). The information related to the stability of the current drug production environment is then weighted to obtain a risk sensitivity index (e.g., between 1 and 10).

[0093] In step S404 of some embodiments, specifically, the event risk level is used to measure the severity of a risk event, and the event risk level includes low event risk level, medium event risk level and high event risk level.

[0094] Specifically, the risk probability index, risk severity index, and risk sensitivity index can be weighted and summed to obtain a comprehensive event risk value. This comprehensive event risk value can then be mapped to the corresponding event risk level to determine the degree of event risk.

[0095] For example, if the comprehensive event risk value is 0 to 0.3, it is a low event risk level; if the comprehensive event risk value is 0.3 to 0.7, it is a medium event risk level; and if the comprehensive event risk value is 0.7 to 1.0, it is a high event risk level.

[0096] Through steps S401 to S404, a three-dimensional risk assessment can be conducted on the probability of a risk event, the magnitude of its consequences, and its environmental sensitivity. This avoids the one-sidedness of assessing risk events based solely on frequency or consequences, providing a clear basis for subsequent differentiated early warning and resource scheduling, and ultimately reducing production interruptions or over-response caused by misjudgment of risk levels.

[0097] In step S104 of some embodiments, specifically, an event risk warning operation matching the risk event and the degree of event risk is performed.

[0098] Specifically, if the risk level of a risk event is high, an orange or red alert will be issued.

[0099] For example, taking the tableting process of oral solid dosage forms as an example, if there is an orange or red warning, the tableting machine control command can be executed to issue a warning intervention message to reduce the running speed or force a stop and suggest inspection. Based on the warning intervention message and the sending personnel, a tableting process deviation handling form is generated and sent to the tableting operator, process engineer and workshop director to realize the warning intervention.

[0100] Please see Figure 6 In some embodiments, the risk warning method for drug production may also include, but is not limited to, steps S601 to S602: Step S601: If the risk level of the event meets the predetermined conditions, perform a risk assessment on the drug production area based on the risk event and the multi-source data to obtain the regional risk level.

[0101] Step S602: Perform a regional risk warning operation that matches the risk level of the region.

[0102] In step S601 of some embodiments, if the level of event risk meets predetermined conditions, a risk assessment is performed on the drug production area based on the risk event and multi-source data to obtain the level of regional risk.

[0103] Specifically, the predetermined condition can be that the event risk level meets the medium event risk level or above.

[0104] Please see Figure 7 In some embodiments, step S601 may also include, but is not limited to, steps S701 to S705: Step S701: Extract features from the multi-source data to obtain multi-source data features.

[0105] Step S702: Identify the causal effects of the drug production area based on the characteristics of multi-source data and risk events to obtain the intensity of the risk causal effect; wherein, the intensity of the risk causal effect is used to characterize the degree to which the characteristics of multi-source data and risk events have a risk impact on the drug production area.

[0106] Step S703: Based on the strength of the risk causal effect, select the risk source data features that have a risk impact on the drug production area from the multi-source data features.

[0107] Step S704: Obtain the causal effect weights of the risk source data characteristics, and calculate the regional risk of the drug production area based on the risk source data characteristics and causal effect weights to obtain the regional risk value.

[0108] Step S705: Determine the degree of regional risk based on the regional risk value.

[0109] In step S701 of some embodiments, specifically, the multi-source data features refer to the production personnel behavior features, equipment operation data features, production process control parameter features, production material data features, and production environment data features corresponding to the production personnel behavior data, production equipment operation data, production process data, production material data, and production environment data in the drug production area.

[0110] In step S702 of some embodiments, specifically, the risk causal effect strength refers to the probability that multi-source data features and risk events will cause risks in the drug production area. The risk causal effect strength is used to characterize the degree to which multi-source data features and risk events have a risk impact on the drug production area.

[0111] Specifically, causal discovery algorithms (such as PC algorithm and Granger causality test) can be used to analyze the leading and lagging relationships and conditional dependencies between each multi-source data feature and risk event, so as to output a quantitative effect strength value (such as regression coefficient and contribution score), thereby determining the fundamental factors driving the risk in the drug production area and the magnitude of their influence.

[0112] In step S703 of some embodiments, specifically, the risk source data features are a set of features that have a risk impact on the drug production area.

[0113] For example, after causal identification, the "temperature and humidity synergistic fluctuation index of the A-line filling area", "long-term deviation of cooling water temperature of B reactor" and "compliance rate of personnel changing clothes in C workshop" can be selected as the main risk source data features.

[0114] In step S704 of some embodiments, specifically, the regional risk value is used to quantify a comprehensive scalar value that characterizes the current overall risk level of the drug production area.

[0115] Specifically, the causal effect weight is a weight coefficient obtained by normalizing the intensity of the causal effect of each risk source data feature.

[0116] Specifically, the risk source data characteristics and causal effect weights can be weighted and summed to obtain the regional risk value.

[0117] In step S705 of some embodiments, specifically, the degree of regional risk is determined by mapping the calculated continuous "regional risk values" to the final determination result of regional risk level (such as low, medium, high).

[0118] For example, a regional risk value in [0, 0.3) can be considered low risk, [0.3, 0.7) can be considered medium risk, and [0.7, ∞) can be considered high risk. If a region's calculated risk value is 0.82, then its regional risk level is determined to be high risk.

[0119] Through steps S701 to S705, comprehensive management of risks from a single risk event to the overall production area is achieved. This not only resolves the risk events that have occurred in a timely manner, but also enables further assessment and reduction of the overall risk level of the area, thereby reducing the ultimate impact of risk events such as non-standard operations on drug quality at the root.

[0120] Please see Figure 8 In some embodiments, step S602 may include, but is not limited to, steps S801 to S804: Step S801: Obtain regional risk description information that matches the regional risk level, and retrieve intervention knowledge from the preset drug production knowledge graph based on the regional risk description information to obtain risk intervention knowledge.

[0121] Step S802: Based on regional risk description information and risk intervention knowledge, generate risk intervention instruction information.

[0122] Step S803: Based on the risk intervention indication information, the pre-trained risk intervention generation model is instructed to generate an intervention plan to obtain the risk intervention plan.

[0123] Step S804: Perform regional risk warning operations that match the regional risk level according to the risk intervention plan.

[0124] In step S801 of some embodiments, specifically, the regional risk description information refers to a structured text description of a drug production area that has been determined to have a risk level (such as high risk), which includes the root causes of the risk and the scope of its impact.

[0125] Specifically, a pharmaceutical manufacturing knowledge graph refers to a large-scale semantic knowledge base that stores entities (such as equipment, processes, and regulations) in the pharmaceutical manufacturing field and their relationships (such as causing and alleviating).

[0126] Specifically, risk intervention knowledge refers to the entity and relationship subgraphs retrieved from the knowledge graph that are associated with the current risk description, used to provide a basis and experience for handling potential risks in the region.

[0127] For example, if the risk level of a region is determined to be "the A-line filling area is high-risk, and the risk is mainly due to the persistently low pressure differential and the decline in the compliance rate of personnel changing clothes", then this description is converted into a query, that is, the entity "low pressure differential" is retrieved from the drug production knowledge graph, and its associated nodes such as "potential consequences" (such as microbial contamination), "relevant production regulations and clauses" (such as GMP appendices) and "historical corrective measures" are extracted to form risk intervention knowledge.

[0128] Specifically, risk description information can be parsed using natural language processing technology, key entities can be extracted as query conditions, and graph query language can be used to traverse the knowledge graph to retrieve semantically related knowledge subgraphs, thereby providing traceable domain knowledge support for generating intervention plans.

[0129] In step S802 of some embodiments, specifically, the risk intervention instruction information refers to the structured instructions generated based on regional risk description information and risk intervention knowledge, which are used to guide the formulation of detailed intervention plans. The risk intervention instruction information provides intervention action objectives, directions of key intervention measures, and compliance constraints.

[0130] For example, risk intervention instructions generated based on risk intervention knowledge can be: for the high risk of low differential pressure and declining personnel compliance rate in the A-line filling area, the intervention measures can be to immediately implement emergency adjustments to environmental control to stabilize the differential pressure and initiate immediate retraining for production personnel. Moreover, the intervention measures must comply with the GMP aseptic appendix and refer to the historical case numbered Y.

[0131] Specifically, by using preset instruction templates, key elements (such as location and risk sources) in the regional risk description information can be logically linked and filled in with the core intervention points and legal basis in the risk intervention knowledge obtained from knowledge retrieval, forming risk intervention instructions with clear objectives and sufficient basis, providing precise input guidance for the generation of subsequent intervention plans.

[0132] In step S803 of some embodiments, specifically, the pre-trained risk intervention generation model refers to a large language model trained on large-scale pharmaceutical production texts (such as pharmaceutical production deviation reports, regulations, etc.).

[0133] Specifically, a risk intervention plan refers to a detailed, comprehensive, and actionable risk intervention plan document output by a risk intervention generation model, driven by risk intervention instruction information.

[0134] For example, using risk intervention instruction information as input, a pre-trained risk intervention generation model may output a complete risk intervention plan, which may include: specific operating steps for differential pressure adjustment, responsible persons, and completion deadlines; temporary course arrangements and assessment methods for personnel retraining; and the number of record forms that need to be filled out.

[0135] Specifically, structured risk intervention instructions can be input as prompts into a finely tuned domain risk intervention generation model. The model will then generate detailed risk intervention operation guidelines with standardized formats and clear steps, thereby achieving automatic transformation from principle-based instructions to executable risk intervention plans.

[0136] In step S804 of some embodiments, specifically, performing a regional risk warning operation that matches the regional risk level means triggering and executing a digital or physical action that matches the regional risk level according to the generated risk intervention plan.

[0137] For example, for areas classified as "high-risk," the risk intervention plan corresponding to that level can be automatically executed, including displaying a red alert on the central monitoring screen, breaking down tasks in the plan into work orders and pushing them to the mobile terminals of relevant responsible persons, locking the next batch of production instructions for that area, and sending emergency notifications to the quality manager, among other regional risk warning operations.

[0138] Specifically, it can parse the structured information in the risk intervention plan (such as the action to be performed, the person in charge, the time, etc.) and, based on the preset strategy library bound to the regional risk level, call the relevant interface (such as the communication platform) to automatically issue tasks, trigger interlocks or send notifications, thereby ensuring that the early warning information is forcibly converted into traceable action to be performed.

[0139] By combining domain knowledge graphs, large language models, and workflow automation technology through steps S801 to S804, the entire process from risk assessment to disposal action is ensured to have knowledge basis, scientific solution support, and efficient execution guarantee, which significantly improves the response speed, decision quality, and execution reliability of pharmaceutical smart factories in response to regional system risks.

[0140] Through steps S601 to S602, comprehensive management of risks from a single risk event to the overall production area can be achieved. This not only resolves the risk events that have occurred in a timely manner, but also helps to further assess and reduce the overall risk level of the area. This reduces the ultimate impact of risk events such as non-standard operations on drug quality at the root, and realizes a fundamental shift from passive risk response to proactive risk prevention in the drug production process.

[0141] This application first acquires multi-source data from the target pharmaceutical factory's drug production area, including at least one of the following: production personnel behavior data, production equipment operation data, production process data, production material data, and production environment data. This achieves real-time and comprehensive data collection of all elements in the drug production process, providing a timely and comprehensive data foundation for subsequent risk warnings. Second, it performs independent risk detection on each source data to obtain the corresponding risk events. By combining other source data, it generates reference events, enabling in-depth identification of complex risk patterns through multi-dimensional information correlation analysis, significantly improving the accuracy of risk assessment in drug production. Furthermore, based on the reference events, it predicts the risks of the identified risk events and quantifies their risk levels. This significantly shifts the risk control node from post-event traceability to during or even before the event, effectively solving the lag problem of traditional methods relying on manual analysis of ADR reports. It also executes event risk warning operations matched to the risk events and their risk levels, enabling timely resolution of already occurred risk events and fundamentally reducing the ultimate impact of risk events such as non-standard operations on drug quality.

[0142] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the aforementioned risk warning method for drug production. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0143] Please see Figure 9 , Figure 9 The hardware structure of an electronic device according to another embodiment is illustrated. The electronic device includes: The processor 901 can be implemented using a general-purpose CPU (Central Processing Unit), microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of this application. The memory 902 can be implemented as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store operating devices and other application programs. When the technical solutions provided in the embodiments of this specification are implemented through software or firmware, the relevant program code is stored in the memory 902 and is called and executed by the processor 901 to execute the risk warning method for drug production according to the embodiments of this application. The input / output interface 903 is used to implement information input and output; The communication interface 904 is used to enable communication and interaction between this device and other devices. Communication can be achieved through wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.). Bus 905 transmits information between various components of the device (e.g., processor 901, memory 902, input / output interface 903, and communication interface 904); The processor 901, memory 902, input / output interface 903, and communication interface 904 are connected to each other within the device via bus 905.

[0144] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0145] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0146] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0147] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0148] Those skilled in the art will understand that all or some of the steps, apparatuses, or functional modules / units in the methods disclosed above can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0149] The terms “first,” “second,” “third,” etc. (if present) in the specification and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms “comprising” and “having,” and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, apparatus, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0150] It should be understood that in this application, "at least one (item)" means one or more, and "more than" means two or more. "And / or" is used to describe the relationship between related objects, indicating that three relationships can exist. For example, "A and / or B" can represent three cases: only A exists, only B exists, and both A and B exist simultaneously, where A and B can be singular or plural. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship. "At least one (item) of the following" or similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one (item) of a, b, or c can represent: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0151] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. The coupling or direct coupling or communication connection between the shown or discussed units may be through some interfaces, or the indirect coupling or communication connection between the apparatus or units may be electrical, mechanical, or other forms.

[0152] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0153] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0154] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes multiple instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this application. The aforementioned storage medium includes various media capable of storing programs, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0155] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A risk warning method for pharmaceutical production, characterized in that, The method includes: Acquire multi-source data of the drug production area in the target pharmaceutical factory, wherein the multi-source data includes at least one of the following sources: production personnel behavior data, production equipment operation data, production process data, production material data, and production environment data; Risk detection is performed on each of the source data in the multi-source data to obtain the risk event corresponding to the source data, and reference events are generated based on other source data in the multi-source data. Based on the reference event, risk prediction is performed on the risk event to obtain the event risk level; Perform an event risk warning operation that matches the risk event and the degree of risk of the event.

2. The method according to claim 1, characterized in that, The step of predicting the risk level of the risk event based on the reference event includes: Acquire historical risk record data, identify the frequency of occurrence of the risk events from the historical risk record data, and generate a risk probability index based on the frequency of occurrence; Based on the reference event, the consequences of the risk event are predicted to obtain a risk severity index; Predict the sensitivity of the reference event to the risk event to obtain a risk sensitivity index; The risk level of the event is obtained by fusing the risk probability index, the risk severity index, and the risk sensitivity index.

3. The method according to claim 2, characterized in that, The reference event includes at least one of the following events: a first reference event representing the product quality dimension, a second reference event representing the personnel safety dimension, a third reference event representing the equipment damage dimension, and a fourth reference event representing the production delay dimension; The process of predicting the consequences of the risk event based on the reference event to obtain a risk severity index includes: When the reference event is the first reference event representing the product quality dimension, the consequences of the risk event are scored based on the first reference event to obtain a product quality impact score. When the reference event is the second reference event representing the personnel safety dimension, the consequences of the risk event are scored based on the second reference event to obtain a personnel safety impact score. When the reference event is the third reference event representing the dimension of equipment damage, the consequences of the risk event are scored based on the third reference event to obtain the equipment damage impact score. When the reference event is the fourth reference event representing the dimension of production delay, the consequences of the risk event are scored based on the fourth reference event to obtain the production delay impact score. The risk severity index is obtained by fusing the product quality impact score, the personnel safety impact score, the equipment damage impact score, and the production delay impact score.

4. The method according to claim 1, characterized in that, The risk events include production process control risk events; The step of performing risk detection based on each source data in the multi-source data to obtain the risk event corresponding to the source data includes: When the source data is the production process data, feature extraction is performed on the production process data to obtain the production process control parameter features; Obtain process control baseline features; Based on the process control baseline characteristics and the production process control parameter characteristics, the production process control risk events corresponding to the production process data are generated.

5. The method according to claim 4, characterized in that, The step of generating the production process control risk event corresponding to the production process data based on the process control baseline characteristics and the production process control parameter characteristics includes: Based on the process control baseline characteristics, parameter deviation analysis is performed on the production process control parameter characteristics to obtain the first risk value; The distribution differences of the characteristic mean and standard deviation of the production process control parameter characteristics are compared based on the characteristic mean and standard deviation of the process control baseline characteristics to obtain the second risk value; Based on the first risk value and the second risk value, the production process control risk event corresponding to the production process data is generated.

6. The method according to claim 1, characterized in that, After performing an event risk warning operation matching the risk event and the degree of risk of the event, the method further includes: If the risk level of the event meets the predetermined conditions, a risk assessment is performed on the drug production area based on the risk event and the multi-source data to obtain the regional risk level. Perform regional risk warning operations that match the risk level of the region.

7. The method according to claim 6, characterized in that, The step of conducting a risk assessment of the drug production area based on the risk events and the multi-source data to obtain the regional risk level includes: Feature extraction is performed on the multi-source data to obtain multi-source data features; Based on the multi-source data features and the risk events, causal effects are identified in the drug production area to obtain the risk causal effect strength; wherein, the risk causal effect strength is used to characterize the degree to which the multi-source data features and the risk events have a risk impact on the drug production area; Based on the strength of the risk causal effect, risk source data features that have a risk impact on the drug production area are selected from the multi-source data features; Obtain the causal effect weights of the risk source data features, and calculate the regional risk of the drug production area based on the risk source data features and the causal effect weights to obtain the regional risk value; The degree of risk in a region is determined based on the region's risk value.

8. The method according to claim 6, characterized in that, The execution of regional risk early warning operations matched to the regional risk level includes: Obtain regional risk description information that matches the risk level of the region, and retrieve intervention knowledge from a preset drug production knowledge graph based on the regional risk description information to obtain risk intervention knowledge; Based on the regional risk description information and the risk intervention knowledge, risk intervention instruction information is generated; Based on the risk intervention indication information, the pre-trained risk intervention generation model is instructed to generate an intervention plan, thereby obtaining the risk intervention plan; According to the risk intervention plan, the regional risk warning operation is performed in accordance with the risk level of the region.

9. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the risk warning method for pharmaceutical production as described in any one of claims 1 to 8.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the risk warning method for pharmaceutical production as described in any one of claims 1 to 8.

Images