Laboratory dangerous chemical whole-process risk management and control method and system based on AI large model

Abstract

The invention relates to the technical field of risk management and control, in particular to a laboratory dangerous chemical whole-process risk management and control method and system based on an AI large model. Under the condition that the gas concentration does not exceed the standard, the potential chemical reaction risk is predicted by depending on the gas component proportion, the reaction triggering condition and the environmental data, the potential safety hazard problem caused by the potential reaction is solved, the gas diffusion speed and the toxicity grade of the reaction product are combined through real-time monitoring and analysis of each control sub-region, and the safety of the reaction product is improved. According to the method, the reaction risk probability and the delay risk probability can be accurately calculated, an effective early warning signal and a management and control instruction are generated before the risk occurs, and the real-time performance and accuracy of laboratory safety management are remarkably improved, so that the pre-judgment and response capability of a laboratory to dangerous chemical leakage and reaction risk is effectively improved, and the safety of the laboratory is improved. Preventive measures can be taken when potential dangers do not appear, so that safety accidents are avoided.

Description

Technical Field

[0001] This invention relates to the field of risk management technology, and in particular to a method and system for full-process risk management of hazardous chemicals in laboratories based on an AI big data model. Background Technology

[0002] Hazardous chemicals used in laboratories may pose serious safety risks if used or managed improperly, affecting personnel health, environmental safety, and even the integrity of facilities. Therefore, it is necessary to conduct full-process risk management of hazardous chemicals. Full-process risk management of hazardous chemicals in laboratories refers to a comprehensive management method that identifies, assesses, controls, monitors, and optimizes the risks of hazardous chemicals throughout their entire life cycle, from procurement, storage, use, disposal to emergency response.

[0003] Currently, automated monitoring technology is used to manage the safety of hazardous chemicals in laboratories throughout the entire process. This mainly involves real-time monitoring of the concentration of chemical gases in the laboratory air. When the monitored concentration exceeds the safety threshold, the system will automatically trigger an alarm, issue a warning, and initiate corresponding safety measures. However, the above method fails to consider the safety hazards that may arise when the concentration of chemical gases in the air does not exceed the safety threshold, but the leaked chemical gases may react with each other. Summary of the Invention

[0004] The main objective of this invention is to provide a method for full-process risk management of hazardous chemicals in laboratories based on an AI-based large model, aiming to solve the technical problems in the prior art.

[0005] This invention proposes a method for full-process risk management of hazardous chemicals in laboratories based on an AI-powered large-scale model, comprising: Obtain full-process node information and basic attribute data of laboratory hazardous chemicals, and construct a chemical attribute database based on the basic attribute data; Based on the full-process node information, the laboratory is divided into multiple control sub-areas, and gas monitoring data and environmental status data of each control sub-area are acquired in real time. The gas monitoring data includes the proportion of gas components, real-time gas concentration, and gas diffusion rate. The proportion of each gas component is input into a pre-trained AI model to obtain the corresponding predicted potential reaction type, and the corresponding reaction triggering condition is obtained according to each predicted potential reaction type. The predicted potential reaction type includes one of oxidation reaction, reduction reaction, neutralization reaction and polymerization reaction. The reaction risk probability is obtained based on each reaction triggering condition, environmental data, and the real-time gas concentrations of any two gases, and it is determined whether the reaction risk probability exceeds a first preset threshold. If the probability of the reaction risk exceeds the first preset threshold, the toxicity level of the reaction product is obtained according to each of the predicted potential reaction types and chemical property databases, and the corresponding delay risk probability is obtained according to each of the reaction product toxicity levels and gas diffusion rates. Determine whether the probability of the delay risk exceeds a second preset threshold; If the probability of delay risk exceeds the second preset threshold, it is determined that there is a risk in the controlled sub-area, and a corresponding risk warning signal level is generated. For each of the aforementioned risk warning signal levels, a corresponding control instruction is generated for the corresponding control sub-area, and control is carried out according to the control instruction.

[0006] Preferably, the step of inputting the proportion of each gas component into a pre-trained AI model to obtain the corresponding predicted potential reaction type includes: The proportion of each gas component is input into a pre-trained AI model to identify and obtain all target gases within the corresponding control sub-region; Obtain the first characteristic parameter of each target gas, and extract the incompatible substance association table corresponding to the target gas from the chemical property database based on the first characteristic parameter; Based on the target gas, the incompatible gas is determined from the incompatible gas association table, and the corresponding reactivity correlation degree is obtained between the incompatible gas and the target gas. The percentage change rate and percentage fluctuation coefficient of each gas component are obtained within a preset time period, and the percentage change rate, percentage fluctuation coefficient and reactivity correlation of each component are input into a pre-trained AI large model to obtain the basic probability of each potential reaction. Extract the key reactive gas pairs and their correlation for each potential reaction from the chemical property database, and obtain the proportion of the two key reactive gases corresponding to each key reactive gas pair; The correction coefficient is obtained based on the correlation degree and the proportion of the two key reactant gases, and the correction probability of the corresponding potential reaction is obtained based on the correction coefficient and the corresponding basic probability. The potential response corresponding to the highest correction probability is selected as the predicted potential response type.

[0007] Preferably, the step of obtaining the reaction risk probability based on each reaction triggering condition, environmental data, and the real-time gas concentrations of any two gases includes: Based on the reaction triggering conditions, obtain the critical temperature, critical humidity, critical ventilation volume, and the critical concentration thresholds of the corresponding two gases; Real-time temperature, real-time humidity, and real-time ventilation volume are obtained based on the environmental data, and temperature influence factors are obtained based on the real-time temperature and critical temperature. The humidity influencing factor is obtained based on the real-time humidity and the critical humidity, and the ventilation influencing factor is obtained based on the real-time ventilation volume and the critical ventilation volume. The environmental impact coefficient is obtained based on the ventilation impact factor, humidity impact factor, and temperature impact factor. The concentration ratio factor is obtained based on the real-time concentrations of any two gases and their corresponding critical concentration thresholds, and the reaction risk probability is obtained based on the concentration ratio factor and the environmental impact coefficient.

[0008] Preferably, the step of obtaining the toxicity level of the reaction product based on each of the predicted potential reaction types and chemical property databases includes: Based on the predicted potential reaction types, reaction condition correlation features and reactant types are obtained, wherein the reaction condition correlation features include reaction temperature range, reaction pressure range, reaction medium type, and reaction catalyst; The average reaction temperature is obtained based on the reaction temperature range, and the average reaction pressure is obtained based on the reaction pressure range. Based on the types of reactants, reaction media, and reaction catalysts, all related product records are extracted from the chemical property database, and the corresponding record standard temperature and record standard pressure are obtained from the chemical property database for each related product record. The first matching degree of the corresponding associated product record is obtained based on the standard temperature, standard pressure, average reaction temperature and average reaction pressure of each record, and the associated product record corresponding to the highest first matching degree is selected as the target product record. Determine whether the first matching degree of the target product record is less than the first preset matching degree; If the first matching degree of the target product record is not less than the first preset matching degree, then the target product record is determined as the final target product record, and the main product and by-product are determined from the chemical property database based on the final target product record; If the first matching degree of the target product record is less than the first preset matching degree, then return to the step of selecting the associated product record corresponding to the highest first matching degree as the target product record, until the first matching degree of the target product record is not less than the first preset matching degree; Select the toxicity levels of the main product and by-product corresponding to the main product and by-product from the chemical property database, respectively, and obtain the toxicity level of the reaction product based on the toxicity levels of the main product and by-product.

[0009] Preferably, the step of obtaining the corresponding delayed risk probability based on the toxicity level of each reaction product and the gas diffusion rate includes: Obtain the spatial volume and second characteristic parameters of the ventilation equipment in each controlled sub-area, wherein the second characteristic parameters include the minimum air change rate, current equipment power, rated equipment power, and rated air volume; The theoretical minimum ventilation volume of the corresponding control sub-area is obtained based on each of the minimum air exchange rates and space volumes, and the actual ventilation volume of the corresponding control sub-area is obtained based on each of the current equipment power, rated equipment power and rated air volume. The ventilation efficiency of the corresponding controlled sub-area is obtained based on the actual ventilation volume and the theoretical minimum ventilation volume for each of the aforementioned sub-areas. The real-time temperature is obtained based on the environmental data, and the gas diffusion coefficient is obtained based on the real-time temperature, ventilation efficiency, and gas diffusion rate. The first initial reaction concentration of the main product and the second initial reaction concentration of the by-product are obtained respectively, and the total initial reaction concentration is obtained based on the first initial reaction concentration and the second initial reaction concentration. The actual reaction product concentration at a preset time point is obtained based on the gas diffusion coefficient, the total initial reaction concentration, and the spatial volume. The corresponding delay risk probability is obtained based on the actual reaction product concentration, the preset reaction product concentration, and the toxicity level of the reaction product.

[0010] Preferably, the step of generating corresponding control instructions for each of the risk warning signal levels for the corresponding control sub-areas, and performing control according to the control instructions, includes: According to each of the aforementioned risk warning signal levels, a corresponding control instruction is determined from a preset warning and control database, wherein the control instruction includes basic control instructions and special control instructions; The corresponding control sub-area is initially controlled according to the basic control instructions, and feedback information after the initial control is obtained in real time. Determine whether the feedback information meets the preset standard requirements; If the feedback information meets the preset standard requirements, it is determined that the controlled sub-area is out of risk; If the feedback information does not meet the preset standard requirements, then select the corresponding control measures from the special control instructions based on the feedback information to carry out special control until the feedback information meets the preset standard requirements.

[0011] This application also provides a laboratory hazardous chemicals full-process risk management system based on an AI large model, including: The module is used to acquire full-process node information and basic attribute data of laboratory hazardous chemicals, and to build a chemical attribute database based on the basic attribute data; The partitioning module is used to divide the laboratory into multiple control sub-areas based on the full-process node information, and to acquire gas monitoring data and environmental status data of each control sub-area in real time. The gas monitoring data includes the proportion of gas components, real-time gas concentration, and gas diffusion rate. The input module is used to input the proportion of each gas component into a pre-trained AI large model to obtain the corresponding predicted potential reaction type, and to obtain the corresponding reaction triggering condition according to each predicted potential reaction type, wherein the predicted potential reaction type includes one of oxidation reaction, reduction reaction, neutralization reaction and polymerization reaction; The acquisition module is used to acquire the reaction risk probability based on each reaction triggering condition, environmental data, and the real-time gas concentrations of any two gases, and to determine whether the reaction risk probability exceeds a first preset threshold. If the probability of the reaction risk exceeds the first preset threshold, the toxicity level of the reaction product is obtained according to each of the predicted potential reaction types and chemical property databases, and the corresponding delay risk probability is obtained according to each of the reaction product toxicity levels and gas diffusion rates. The judgment module is used to determine whether the probability of the delay risk exceeds a second preset threshold. If the probability of delay risk exceeds the second preset threshold, it is determined that there is a risk in the controlled sub-area, and a corresponding risk warning signal level is generated. The control module is used to generate corresponding control instructions for the corresponding control sub-area based on each of the risk warning signal levels, and to carry out control according to the control instructions.

[0012] Preferably, the acquisition module includes: The first acquisition unit is used to acquire the critical temperature, critical humidity, critical ventilation volume, and critical concentration thresholds of the corresponding two gases according to the reaction triggering conditions. The second acquisition unit is used to acquire real-time temperature, real-time humidity and real-time ventilation volume based on the environmental data, and to acquire temperature influence factor based on the real-time temperature and critical temperature. The third acquisition unit is used to acquire a humidity influence factor based on the real-time humidity and critical humidity, and to acquire a ventilation influence factor based on the real-time ventilation volume and critical ventilation volume. The fourth acquisition unit is used to acquire the environmental impact coefficient based on the ventilation impact factor, humidity impact factor and temperature impact factor; The fifth acquisition unit is used to acquire a concentration ratio factor based on the real-time concentrations of any two gases and their corresponding critical concentration thresholds, and to acquire a reaction risk probability based on the concentration ratio factor and the environmental impact coefficient.

[0013] The present invention also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the above-described method for full-process risk management of laboratory hazardous chemicals based on an AI big data model.

[0014] The present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the above-described method for full-process risk management of laboratory hazardous chemicals based on an AI large model.

[0015] The beneficial effects of this invention are as follows: When the gas concentration does not exceed the standard, this invention predicts potential chemical reaction risks based on the proportion of gas components, reaction triggering conditions, and environmental data, solving the safety hazards caused by potential reactions. Through real-time monitoring and analysis of each controlled sub-area, combined with the gas diffusion rate and the toxicity level of the reaction products, it can accurately calculate the probability of reaction risk and the probability of delayed risk, and generate effective early warning signals and control instructions before the risk occurs. This significantly improves the real-time performance and accuracy of laboratory safety management, thereby effectively enhancing the laboratory's ability to predict and respond to hazardous chemical leaks and reaction risks. It enables preventative measures to be taken before potential hazards manifest, thus avoiding safety accidents. By utilizing a chemical property database combined with an AI large model, this invention not only improves the intelligence of prediction but also adapts to the diversity of different laboratory environments, possessing strong applicability and scalability. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of a method flow according to an embodiment of the present invention.

[0017] Figure 2 This is a schematic diagram of the system structure according to an embodiment of the present invention.

[0018] Figure 3 This is a schematic diagram of the internal structure of a computer device according to an embodiment of this application.

[0019] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0020] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0021] like Figure 1 As shown, this application provides a method for full-process risk management of hazardous chemicals in laboratories based on an AI large-scale model, including: S1. Obtain full-process node information and basic attribute data of laboratory hazardous chemicals, and construct a chemical attribute database based on the basic attribute data; S2. Based on the full-process node information, the laboratory is divided into multiple control sub-areas, and gas monitoring data and environmental status data of each control sub-area are acquired in real time. The gas monitoring data includes the proportion of gas components, real-time gas concentration, and gas diffusion rate. S3. Input the proportion of each gas component into the pre-trained AI large model to obtain the corresponding predicted potential reaction type, and obtain the corresponding reaction triggering condition according to each predicted potential reaction type, wherein the predicted potential reaction type includes one of oxidation reaction, reduction reaction, neutralization reaction and polymerization reaction; S4. Obtain the reaction risk probability based on each reaction triggering condition, environmental data, and the real-time gas concentrations of any two gases, and determine whether the reaction risk probability exceeds a first preset threshold. If the probability of the reaction risk exceeds the first preset threshold, the toxicity level of the reaction product is obtained according to each of the predicted potential reaction types and chemical property databases, and the corresponding delay risk probability is obtained according to each of the reaction product toxicity levels and gas diffusion rates. S5. Determine whether the probability of the delay risk exceeds a second preset threshold; If the probability of delay risk exceeds the second preset threshold, it is determined that the controlled sub-area has a risk, and the corresponding risk warning signal level is determined from the preset risk warning table according to the probability of delay risk. The preset risk warning table includes the delay risk probability range corresponding to different risk warning signal levels. S6. Generate corresponding control instructions for the corresponding control sub-area according to each of the risk warning signal levels, and carry out control according to the control instructions.

[0022] As described in steps S1-S6 above, the full-process node information includes storage nodes, retrieval nodes, usage nodes, recycling nodes, and disposal nodes. The basic attribute data includes chemical name, chemical formula, flash point, explosion limits, reactivity correlation, related product records (including record standard temperature, record standard pressure, main / byproducts and their toxicity levels), incompatible substance correlation table, key reactive gas pairs and their correlation. This invention obtains the full-process node information and basic attribute data of laboratory hazardous chemicals, and constructs a chemical attribute database based on the basic attribute data. By obtaining the full-process node information and basic attribute data of laboratory hazardous chemicals, a comprehensive understanding of all aspects of chemicals from procurement, storage, use to disposal is ensured. This helps to more accurately assess the potential risks of chemicals and to refine the management and operation processes of the laboratory. By constructing a chemical attribute database, information such as the basic attributes, usage methods, and storage conditions of chemicals is centrally managed, enabling real-time monitoring, querying, and analysis of chemical risk information. This data-driven control method is more intelligent and accurate than traditional monitoring methods. Constructing a chemical attribute database helps to centrally manage the basic information of all hazardous chemicals, providing data support for subsequent analysis, prediction, and decision-making. Compared to the decentralized management methods in existing technologies, this method reduces information omissions and errors, improves system reliability, and provides a chemical property database as the foundation for potential risk prediction and decision-making based on big data analysis. By integrating multi-dimensional chemical data, it supports efficient data processing and accurate risk assessment. The laboratory is divided into multiple control sub-areas through end-to-end node information, and real-time environmental status data and gas monitoring data such as gas composition ratio, real-time gas concentration, and gas diffusion rate of each control sub-area are acquired. By dividing the laboratory into multiple control sub-areas, the system can independently manage and monitor the environment of each sub-area, avoiding blind spots in global monitoring, thereby improving monitoring accuracy and response speed. This regionalized control method can quickly identify risks in different areas and generate more precise control instructions based on the actual situation of each area. By acquiring data such as gas composition ratio, real-time concentration, and gas diffusion rate in real time, dynamic laboratory environmental monitoring can be achieved, and abnormal changes can be detected in a timely manner. Compared to traditional periodic monitoring methods, this method provides more accurate and timely risk warnings. Real-time monitoring of the environmental status of each sub-area enables the system to quickly identify potential reaction risks and improve the ability to prevent sudden chemical reactions.

[0023] By inputting the proportion of each gas component into a pre-trained AI model, corresponding predicted potential reaction types (including oxidation, reduction, neutralization, and polymerization reactions) are obtained. Based on each predicted potential reaction type, the corresponding reaction triggering conditions are acquired. By analyzing the gas component proportions through the AI ​​model, chemical reaction types, including oxidation, reduction, neutralization, and polymerization reactions, can be accurately predicted. This effectively avoids the inaccurate or missed reaction type identification issues of traditional monitoring methods, resulting in higher accuracy. The pre-trained AI model can continuously optimize prediction results as it acquires new data, enhancing the system's intelligence level and thus improving laboratory safety management. The system obtains the triggering conditions for each predicted reaction type and combines them with… Predicting reactions based on environmental data and gas concentrations allows for the early identification of potential reaction conditions. Compared to traditional methods based on static parameters, this approach is more adaptable to dynamic experimental environments. By calculating the probability of reaction risk and comparing it with a first preset threshold, an efficient early warning mechanism can be established. This avoids the problems of insufficient risk assessment or slow response in traditional safety measures, enhancing the system's sensitivity. The probability of reaction risk is obtained by considering each reaction trigger condition, environmental data, and the real-time concentrations of any two gases. By combining factors such as the toxicity level of reaction products and gas diffusion rates, the probability of delayed risk is obtained. This allows for a comprehensive assessment of potential safety hazards in the laboratory. Compared to simple gas concentration monitoring, this multi-dimensional assessment method offers a significant advantage. This system can consider more environmental variables and safety factors, improving the comprehensiveness of risk prediction. It assesses delayed risks after chemical reactions, preventing secondary hazards caused by gas diffusion, product toxicity, etc., ensuring long-term laboratory safety management. It determines whether the reaction risk probability exceeds a first preset threshold. If so, it obtains the toxicity level of the reaction product based on each predicted potential reaction type and chemical property database, and obtains the corresponding delayed risk probability based on the toxicity level of each reaction product and gas diffusion rate. It then determines whether the delayed risk probability exceeds a second preset threshold. If the delayed risk probability exceeds the second preset threshold, it determines that a risk has occurred in the controlled sub-area and generates a corresponding risk warning. The signal level, by acquiring the probability of delayed risk in real time and comparing it with a second preset threshold, enables timely assessment of the existence of serious risks after chemical leaks and reactions. Compared to traditional safety monitoring methods that fail to consider delayed reactions and secondary hazards, this invention provides a more efficient and accurate response strategy. The system automatically generates risk warning signals and issues control commands, which not only improves response speed but also reduces human intervention and improves the overall system efficiency. Corresponding control commands are generated for each risk warning signal level and implemented according to these commands. Based on the specific situation and risk level of each control sub-area, corresponding control commands are generated and automatically executed, avoiding the lag and misoperation inherent in traditional manual intervention.This invention ensures faster and more accurate implementation of safety measures. By automatically generating control instructions based on risk levels, it can automatically adjust control strategies according to real-time data, improving the flexibility and efficiency of laboratory management. Through multi-level and multi-dimensional information acquisition and analysis, combined with AI technology, this invention can predict potential reaction risks and issue timely warnings even when chemical gas concentrations do not exceed safety thresholds. Compared to traditional technologies, this method not only considers real-time gas concentrations but also dynamically assesses factors such as the triggering conditions of chemical reactions, the toxicity of reaction products, and gas diffusion, comprehensively improving the accuracy and reliability of laboratory chemical safety control.

[0024] In one embodiment, step S3, which inputs the proportion of each gas component into a pre-trained AI large model to obtain the corresponding predicted potential reaction type, includes: S31. Input the proportion of each gas component into the pre-trained AI large model to identify and obtain all target gases in the corresponding control sub-region; S32. Obtain the first characteristic parameter of each target gas, and extract the incompatible substance association table corresponding to the target gas from the chemical property database according to the first characteristic parameter; S33. Determine the incompatible gas from the incompatible gas association table based on the target gas, and obtain the corresponding reactivity correlation between the incompatible gas and the target gas; S34. Obtain the percentage change rate and percentage fluctuation coefficient within a preset time according to the percentage of each gas component, and input each percentage change rate, percentage fluctuation coefficient and reactivity correlation degree into the pre-trained AI large model to obtain the basic probability of each potential reaction. S35. Extract the key reactive gas pairs and their correlation degree for each potential reaction from the chemical property database, and obtain the proportion of the two key reactive gases corresponding to each key reactive gas pair. S36. The correction coefficient is obtained by weighted summation of the proportions of the two key reactant gases and their correlation, and the correction probability of the corresponding potential reaction is obtained by multiplying the correction coefficient and the corresponding basic probability. S37. Select the potential response corresponding to the largest correction probability as the predicted potential response type.

[0025] As described in steps S31-S37 above, the CAS Registry Number, also known as the CAS Accession Number or CAS Registration Number, is a unique numerical identification number for a substance (compound, polymer, biological sequence, mixture, or alloy). Specifically, the step of extracting the incompatible substance association table corresponding to the target gas from the chemical property database based on the first characteristic parameter involves inputting the first characteristic parameter of the target gas into the index engine of the chemical property database. First, using the CAS number as a unique identifier, a precise search is performed in the index engine of the chemical property database. If a completely matching CAS number exists, the corresponding attribute record is directly retrieved. If the CAS number match fails, a fuzzy match is performed using the chemical formula, combined with the bond energy characteristics of the molecular structure formula (such as the presence of hydroxyl groups, etc.). The structure similarity is calculated by weighted summation of functional groups (such as carboxyl groups) and bond length / bond angle deviation rates (functional group matching degree is the proportion of identical functional groups between the target gas and gases in the database, and bond length / bond angle deviation rate is obtained through molecular structure comparison tools). This determines the attribute record of the final target gas. Then, the incompatible substances association table corresponding to the target gas is extracted from the chemical attribute database based on the attribute record. This invention identifies and obtains all target gases within the corresponding control sub-region by inputting the proportion of each gas component into a pre-trained AI model. By inputting the proportion of each gas component into the pre-trained AI model, it can accurately predict potential hazards using existing big data and pattern recognition capabilities. This invention, leveraging the multidimensional data analysis capabilities of AI models, enables efficient real-time prediction of reactions in complex laboratory environments. Compared to traditional methods relying on single concentration data to determine safety, this invention considers not only gas concentration but also the proportion of gas components. This allows for more accurate identification of potential reactions between different gases, avoiding the safety risks associated with relying solely on single concentration data. The method involves obtaining the chemical formula, molecular structure, and CAS number of each target gas as its first characteristic parameter, and extracting the corresponding incompatible substance association table from a chemical property database based on these first characteristic parameters. Incompatible gases are then identified from the incompatible substance association table using the target gas. Detailed chemical parameters of the target gas are then extracted. Chemical formulas, molecular structures, and CAS numbers can precisely pinpoint the properties of gases. These parameters are crucial for predicting reaction probabilities. The CAS number, as a unique chemical identifier, allows for rapid retrieval of complete chemical information, avoiding the risk of incomplete or erroneous information. Compared to traditional monitoring methods that rely solely on gas concentration, in-depth chemical characterization can more accurately identify gas reactivity, reducing the probability of missed or false positives. Especially when gas concentrations are within limits, it can accurately determine the presence of potential hazards. By extracting incompatible substance association tables from chemical property databases, the errors and inefficiencies of manual searches can be avoided, thus helping to determine which gases pose a reaction risk under specific conditions.This provides data support for subsequent early warning and safety measures. Compared to traditional methods that rely on human experience or single threshold alarms, this invention combines scientific data and machine learning models to intelligently assess the reactivity between different chemicals, reducing human oversight and errors, and improving the accuracy and reliability of the early warning system.

[0026] The pre-trained AI model includes a gas identification module and a reaction prediction module. The gas identification module acquires the initial gas type identified by gas chromatography-mass spectrometry (GC-MS), then extracts the real-time characteristic peak surface and retention time of that gas. Next, it uses a pre-stored standard gas feature library within the pre-trained AI model to obtain the standard characteristic peak area and standard retention time. Finally, it calculates the second matching degree using the real-time characteristic peak surface, real-time retention time, standard characteristic peak area, and standard retention time. The calculation formula is as follows: ; Wherein, P(PD2) represents the second matching degree, S(TF) represents the real-time feature peak surface, B(TF) represents the standard feature peak area, S(BL) represents the real-time retention time, and B(BL) represents the standard retention time. If the second matching degree is not less than the second preset matching degree, the gas type is confirmed as the initial gas type. If the second matching degree is less than the second preset matching degree, the gas type is corrected by combining the spectral data of the infrared gas sensor to obtain the final gas type. The reaction probability is output through the feedforward neural network of the reaction prediction module. The feedforward neural network adopts a three-layer fully connected structure, including an input layer, a hidden layer, and an output layer. The parameters of the input layer include the percentage change rate, the percentage fluctuation coefficient, and the reaction activity correlation degree. The output layer contains four neurons, corresponding to the probability output of oxidation reaction, reduction reaction, neutralization reaction, and polymerization reaction, respectively.

[0027] By obtaining the reactivity correlation between the incompatible gas and the target gas, and calculating this correlation, the potential for hazardous reactions between the gases can be assessed, thus providing a quantitative risk assessment. This allows for more accurate identification of highly reactive gas combinations. Traditional monitoring systems typically cannot quantitatively assess gas reactivity, relying solely on concentration data to trigger alarms. This invention, however, quantifies gas reactivity through reactivity correlation, enabling early identification of potential hazards and avoiding the limitations of relying solely on concentration alarms. It obtains the percentage change rate and fluctuation coefficient of each gas component over a preset time period, and inputs these rates, fluctuation coefficients, and reactivity correlations into a pre-trained AI model. In this process, the basic probability of each potential reaction is obtained. By acquiring the rate of change and fluctuation coefficient of gas component proportion, the trend of gas concentration change can be analyzed in real time, rather than just static concentration values. The rate of change reflects the speed of gas concentration change, while the fluctuation coefficient indicates the amplitude and regularity of concentration fluctuation. The combination of these two can provide dynamic monitoring data for potential chemical reactions. Compared with traditional static concentration monitoring, this invention can reflect gas concentration fluctuations in real time, especially when the gas concentration has not yet reached the danger threshold. It can detect potential risks of reactions through parameters such as the fluctuation coefficient. The dynamic analysis method is more effective than static concentration monitoring in predicting and avoiding potential chemical reaction risks. By inputting the rate of change, fluctuation coefficient, and reactivity correlation, the potential risks of reactions can be identified. In large-scale AI models, multiple influencing factors can be comprehensively considered to predict the probability of potential reactions, thereby reducing the bias of single factors and improving the accuracy of predictions. Existing technologies typically only focus on the concentration of a single gas, ignoring the impact of the rate of change and fluctuation coefficient on the probability of a reaction. This invention, by comprehensively considering these factors, can more accurately assess the interactions and reaction risks between gases, avoiding the limitations of traditional methods that easily overlook potential reactions. It extracts the key reactive gas pairs and their correlations for each potential reaction from a chemical property database, and obtains the proportions of the two key reactive gases for each key reactive gas pair. A correction coefficient is obtained by weighted summation of the sum of the proportions of the two key reactive gases and the correlation. This is achieved by analyzing chemical properties... Extracting potential reactive gas pairs and their correlations from a database allows for accurate identification of the participating gas pairs and the intensity of the reaction. Unlike existing technologies that rely solely on concentration exceedances to determine whether a reaction has occurred, this invention utilizes precise data from a chemical property database to identify potential reaction risks earlier. This avoids the blind reliance on alarms triggered by excessive concentrations. By calculating correction coefficients and obtaining correction probabilities, the prediction results can be dynamically adjusted, making reaction predictions more accurate. The correction coefficients comprehensively consider the gas proportions and the correlations between reactive gas pairs, providing a more realistic reflection of the probability of a potential reaction. Unlike traditional linear threshold methods, this invention adjusts the prediction probability through correction coefficients, enabling more flexible and accurate prediction of reaction occurrence probabilities.Especially when gas concentrations are within limits, this invention can effectively warn of potential hazardous reactions. It obtains the corrected probability of a potential reaction by multiplying the correction coefficient by the corresponding base probability. By selecting the potential reaction with the highest corrected probability, the system ensures that the warning system selects the most likely reaction type, thus prioritizing appropriate safety measures. This invention intelligently identifies the most dangerous reaction by comprehensively considering multiple factors. Unlike existing technologies that trigger safety measures through simple concentration alarms, this invention calculates the most likely hazardous reaction type based on multiple parameters and takes targeted measures, avoiding the blindness and lag of a single-standard triggering mechanism. By introducing the analytical capabilities of large AI models and multi-dimensional data analysis (proportion change rate, proportion fluctuation coefficient, and reactivity correlation), this invention can intelligently predict potential reactions even when concentrations are within limits, promptly identifying and avoiding potential safety hazards. Compared to the limitations of existing technologies that rely on simple concentration threshold alarms, this invention provides more detailed, comprehensive, and dynamic risk prediction, greatly improving the safety and intelligence level of laboratory gas control systems.

[0028] In one embodiment, step S4, which involves obtaining the reaction risk probability based on each reaction triggering condition, environmental data, and the real-time gas concentrations of any two gases, includes: S41. Obtain the critical temperature, critical humidity, critical ventilation volume, and critical concentration thresholds of the two corresponding gases based on the reaction triggering conditions. S42. Obtain real-time temperature, real-time humidity, and real-time ventilation volume based on the environmental data, and obtain the temperature influence factor based on the ratio of the difference between the real-time temperature and the critical temperature to the critical temperature. S43. Obtain the humidity influence factor based on the ratio of the difference between the real-time humidity and the critical humidity to the critical humidity, and obtain the ventilation influence factor based on the ratio of the difference between the real-time ventilation volume and the critical ventilation volume to the critical ventilation volume. S44. The environmental impact coefficient is obtained by weighted summation of the ventilation impact factor, humidity impact factor and temperature impact factor. S45. Obtain the concentration ratio factor by the ratio of the product of the real-time concentrations of any two gases to the product of the corresponding two critical concentration thresholds, and calculate the reaction risk probability by weighted summation based on the concentration ratio factor and the environmental impact coefficient.

[0029] As described in steps S41-S45 above, this invention obtains critical temperature, critical humidity, critical ventilation volume, and corresponding critical concentration thresholds for two gases through reaction triggering conditions. Obtaining these critical parameters (critical temperature, humidity, ventilation volume, and concentration thresholds) is to establish an accurate environmental safety benchmark. The reaction characteristics of different gases are affected by environmental conditions. Understanding these critical parameters can provide early warning of potential hazardous reactions. Existing technologies mostly rely on monitoring the concentration of a single gas. When the concentration does not exceed the threshold, there is a lack of comprehensive consideration of environmental factors (temperature, humidity, ventilation, etc.). By pre-setting and obtaining critical values, it is possible to more accurately predict when the gas will react under different environmental conditions, thereby effectively preventing the risk of chemical reactions caused by changes in environmental conditions. Real-time temperature, real-time humidity, and real-time ventilation volume are obtained through environmental data, and the temperature influence factor is obtained based on the ratio of the difference between the real-time temperature and the critical temperature to the critical temperature. Temperature, humidity, and ventilation volume affect the occurrence and rate of chemical reactions. Real-time acquisition of these data can more accurately reflect the current environmental conditions, thereby assessing reaction risks. Traditional monitoring methods often only focus on gas concentration data, ignoring the potential impact of environmental factors on the reaction. By monitoring environmental parameters in real time, not only can a more complete risk assessment be provided, but also a dynamic response to changes in environmental conditions can be achieved, improving the accuracy of laboratory safety management. Temperature is a key factor affecting the rate of chemical reactions. By calculating the temperature influence factor, safety thresholds can be dynamically adjusted to ensure effective risk warnings even under temperature fluctuations. The influence of temperature on the reaction is quantified by the ratio of temperature to the critical value, providing a more sensitive and accurate reaction risk assessment. Dynamically calculating the temperature influence factor can reflect the impact of environmental changes on gas reactions in a timely manner, avoiding dangerous reactions caused by temperature changes, and improving the sensitivity and safety of monitoring.

[0030] Humidity influence factors are obtained by comparing the difference between real-time humidity and critical humidity with the ratio of the critical humidity. Humidity has a significant impact on the occurrence and rate of chemical reactions, especially the hygroscopic reactions of certain chemicals in gases. Introducing humidity influence factors provides an effective monitoring basis for real-time environmental adjustments. Dynamically adjusting reaction risk assessments using humidity influence factors compensates for the shortcomings of existing technologies. The introduction of humidity influence factors enables the monitoring system to maintain high accuracy and reaction early warning capabilities in high or low humidity environments, enhancing the system's adaptability. Ventilation influence factors are obtained by comparing the difference between real-time ventilation volume and critical ventilation volume with the ratio of the critical ventilation volume. These factors, along with temperature... The environmental impact coefficient is obtained by weighted summation of various factors. Ventilation volume directly affects the dilution effect of gases, thus influencing their concentration and reactivity. By calculating the ventilation impact factor, the reaction risk under insufficient or excessive ventilation can be accurately assessed. The calculation of the ventilation impact factor provides real-time feedback to the system under different ventilation conditions, effectively warning of chemical reaction hazards under insufficient ventilation, thereby improving the effectiveness of laboratory safety management. By comprehensively calculating temperature, humidity, and ventilation volume into an environmental impact coefficient, their combined impact on the reaction can be comprehensively assessed, making the risk assessment more comprehensive. This invention incorporates multiple environmental factors through weighted summation, making the assessment more comprehensive and dynamically adjustable. By comprehensively assessing environmental factors, the risk misjudgment caused by fluctuations in a single factor can be reduced, improving system stability and early warning accuracy. The concentration ratio factor is obtained by dividing the product of the real-time concentrations of any two gases by the product of their corresponding critical concentration thresholds. In some cases, the interaction between gases can lead to chemical reactions, even if their individual concentrations do not exceed safety thresholds. The concentration ratio factor calculation can identify this potential chemical reaction risk. By considering the interaction between gas concentrations, it overcomes the deficiency of existing technologies that ignore gas mixing reactions. The concentration ratio factor allows for real-time monitoring of the interaction between two gases, thereby preventing potential risks even when gas concentrations do not exceed thresholds. To mitigate reaction risks and further enhance safety, this invention employs a weighted summation calculation based on concentration ratio factors and environmental impact coefficients to obtain the probability of reaction occurrence. The final reaction risk probability comprehensively considers both the concentration ratio factor and the environmental impact coefficient to accurately assess the likelihood of a reaction. By comprehensively evaluating multiple factors, it provides more reliable risk predictions. By comprehensively considering concentration and environmental factors, it can significantly improve the early warning capability for reaction risks in the event of hazardous material leaks in laboratories, providing more precise guidance for accident prevention and control. This invention, by introducing environmental factors such as temperature, humidity, and ventilation volume, and combining them with the interactions between gases, can more accurately predict and prevent potential chemical reaction risks, thereby greatly improving laboratory safety.

[0031] In one embodiment, step S4, which involves obtaining the toxicity level of the reaction product based on each of the predicted potential reaction types and chemical property databases, includes: S46. Obtain reaction condition correlation features and reactant types based on the predicted potential reaction types, wherein the reaction condition correlation features include reaction temperature range, reaction pressure range, reaction medium type, and reaction catalyst; S47. Obtain the average reaction temperature based on the upper and lower temperature limits of the reaction temperature range, and obtain the average reaction pressure based on the upper and lower pressure limits of the reaction pressure range. S48. All related product records are screened and extracted from the chemical property database in the order of reactant type, reaction medium type and reaction catalyst, and the corresponding record standard temperature and record standard pressure are obtained from the chemical property database for each related product record. S49. Obtain the first matching degree of the corresponding associated product record based on each recorded standard temperature, recorded standard pressure, average reaction temperature, and average reaction pressure, wherein the calculation formula is: ; Wherein, P(PD1) represents the first matching degree, α represents the temperature weight, F(WP) represents the average reaction temperature, J(BW) represents the recording standard temperature, β represents the average reaction pressure, F(YP) represents the average reaction pressure, and J(BY) represents the recording standard pressure; and the associated product record corresponding to the highest first matching degree is selected as the target product record; S410. Determine whether the first matching degree of the target product record is less than the first preset matching degree; If the first matching degree of the target product record is not less than the first preset matching degree, then the target product record is determined as the final target product record, and the main product and by-product are determined from the chemical property database based on the final target product record; If the first matching degree of the target product record is less than the first preset matching degree, then return to the step of selecting the associated product record corresponding to the highest first matching degree as the target product record, and then select the associated product record corresponding to the second highest first matching degree as the target product record, until the first matching degree of the target product record is not less than the first preset matching degree. S411. Select the toxicity levels of the main product and the by-product corresponding to the main product and by-product from the chemical property database respectively, and obtain the toxicity level of the reaction product by weighted summation based on the toxicity levels of the main product and the by-product.

[0032] As described in steps S46-S411 above, the recording standard temperature and recording standard pressure of the associated product record are specific condition parameters of a known reaction, which are definite single values ​​(not ranges) used to characterize the known fact that the reactant combination will generate a specific product under these conditions. By finding the known record from the chemical property database that most closely matches the predicted conditions, the most likely main / byproduct is determined. This invention obtains the reaction temperature range, reaction pressure range, reaction medium type, and reaction catalyst by predicting potential reaction types and the correlation characteristics of reactant types and reaction conditions. In traditional technology, the reactivity of chemical gas leaks in the laboratory is not predicted and controlled in advance, and often relies on human observation and basic concentration. Traditional monitoring methods, such as temperature monitoring, cannot predict the risk of hazardous reactions in a timely manner. By predicting potential reaction types and obtaining correlation characteristics between reactant types and reaction conditions (such as temperature ranges and pressure ranges), it is possible to identify which chemicals may cause hazardous reactions in advance, thereby better preventing safety accidents. Traditional monitoring methods only rely on gas concentration monitoring and do not consider the potential hazardous reactions between reactants. By predicting potential reaction types, safety can be improved from the source, reducing laboratory safety hazards. The average reaction temperature is obtained by using the upper and lower temperature limits of the reaction temperature range, and the average reaction pressure is obtained by using the upper and lower pressure limits of the reaction pressure range. Reaction temperature and pressure are factors that affect chemical... Reaction rate and reactivity are crucial factors. In the laboratory, different reactants exhibit different chemical reaction characteristics under different temperatures and pressures. This invention, by accurately obtaining the average values ​​of reaction temperature and pressure, can more accurately simulate chemical reaction conditions, further improving the controllability and safety of the reaction process. By obtaining the average values ​​of reaction temperature and pressure, this invention can provide more accurate reaction condition parameters for subsequent prediction models, thereby improving the prediction accuracy of potentially hazardous reactions in the laboratory. By screening and extracting all related product records from a chemical property database according to the order of reactant type, reaction medium type, and reaction catalyst, it is found that different reactants, reaction media, and catalysts can significantly affect the reaction results. By filtering related product records for these key parameters from the chemical attribute database, the system can ensure more accurate selection and evaluation of reaction products. This avoids the problem of incomplete product evaluation caused by relying on only a single or local parameter in traditional methods. By simultaneously considering reactants, reaction media, and catalysts, the system can comprehensively evaluate the reaction process and product characteristics from multiple dimensions, rather than being limited to the influence of a single factor. This provides the system with higher prediction accuracy and safety. For each related product record, the corresponding record standard temperature and pressure are obtained from the chemical attribute database. The first matching degree of the corresponding related product record is obtained by using the record standard temperature, record standard pressure, average reaction temperature, and average reaction pressure.The first matching degree helps the system determine whether a certain reaction condition matches the conditions of the target product, further screening potential reaction products. By comparing with standard temperature and standard pressure, it ensures that the reaction process is within a safe range, reducing potential risks caused by excessively high or low reaction conditions. Existing technologies often cannot meticulously consider the impact of each reaction condition on the product matching degree. By introducing the first matching degree, the accuracy of the reaction product assessment can be improved, further enhancing the reliability and safety of the system's predicted reaction results. By judging whether the first matching degree of the target product record is less than the first preset matching degree, if the first matching degree of the target product record is not less than the first preset matching degree, the target product record is determined as the final target product record. The process involves identifying the main product and byproducts from a chemical property database based on the final target product record. If the first match degree of the target product record is less than the first preset match degree, the process returns to the step of selecting the associated product record corresponding to the highest first match degree as the target product record. Then, the associated product record corresponding to the second highest first match degree is selected as the target product record, and so on, until the first match degree of the target product record is not less than the first preset match degree. By screening the target product records based on match degree, the safety and feasibility of the selected product under specific reaction conditions can be ensured. Identifying the main product and byproducts allows for further evaluation of the side effects and potential hazards of the reaction. Especially in the event of a leak, the impact of byproduct formation on safety can be foreseen. Traditional techniques... Previously, without distinguishing between products, only preliminary monitoring of the reaction process was considered. Identifying main products and byproducts allows for a more detailed assessment of the complexity of the reaction process and potential secondary reaction risks, reducing laboratory safety hazards. By selecting the toxicity levels of the main product and byproducts from a chemical property database, and then weighting and summing these levels to obtain the toxicity grade of the reaction product, it becomes clear that different reaction products may have different toxicity levels. In particular, byproducts can pose a threat to the safety of laboratory personnel. Weighted summation of the toxicity grades of the main product and byproducts more accurately reflects the overall safety of the reaction process, ensuring the safety of gases in the laboratory environment. Leaks do not typically produce fatal toxic effects. Existing technologies often only focus on gas concentration, neglecting the safety impact of product toxicity. By considering toxicity levels and weighted summation, a more comprehensive assessment of the risk of leaked chemicals can be achieved, allowing for proactive safety measures to prevent hazards. This invention, by comprehensively considering potential reaction factors, ensures that even when gas concentrations do not exceed thresholds, safety hazards caused by gas reactions can still be prevented. By establishing a more comprehensive safety monitoring system that not only relies on concentration monitoring but also considers the likelihood of reactions, it can better address the safety risks posed by chemical reactions in the laboratory, significantly improving safety control capabilities. This invention introduces steps such as precise reaction prediction, matching degree assessment, and weighted summation of toxicity levels.This technology enables the identification of potential hazardous reactions and the early prevention and handling of safety hazards, even when gas concentrations are within acceptable limits. It not only improves the accuracy of reaction prediction but also allows for a more comprehensive assessment of the hazards of reaction products, significantly enhancing the safety management capabilities following laboratory chemical gas leaks.

[0033] In one embodiment, step S4, which involves obtaining the corresponding probability of delayed risk based on the toxicity level of each reaction product and the gas diffusion rate, includes: S412. Obtain the spatial volume of each controlled sub-area and the second characteristic parameters of the ventilation equipment, wherein the second characteristic parameters include the minimum number of air changes, the current equipment power, the rated equipment power, and the rated air volume; S413. Obtain the theoretical minimum ventilation volume of the corresponding control sub-area based on the product of each minimum air exchange rate and space volume, and obtain the actual ventilation volume of the corresponding control sub-area based on the product of the ratio of the current equipment power to the rated equipment power and the rated air volume of each device. S414. Obtain the ventilation efficiency of the corresponding controlled sub-area based on the ratio of each actual ventilation volume to the theoretical minimum ventilation volume; S415. Obtain the real-time temperature based on the environmental data, and calculate the gas diffusion coefficient based on the product of the real-time temperature, ventilation efficiency, and gas diffusion rate. S416. Obtain the first initial reaction concentration of the main product and the second initial reaction concentration of the by-product respectively, and obtain the total initial reaction concentration based on the first initial reaction concentration and the second initial reaction concentration; S417. Obtain the actual reaction product concentration at a preset time point based on the gas diffusion coefficient, total initial reaction concentration, and spatial volume, wherein the calculation formula is: ; Where S(CN) represents the actual reaction product concentration, Z(CN) represents the total initial reaction concentration, Q(KS) represents the gas diffusion coefficient, T represents the preset time, and K(TJ) represents the space volume; S418. Obtain the corresponding delayed risk probability by multiplying the ratio of the actual reaction product concentration to the preset reaction product concentration and the toxicity level of the reaction product.

[0034] As described in steps S412-S418 above, the calculation of the gas diffusion coefficient, the probability of delayed risk, and the concentration of actual reaction products requires prior normalization of their respective parameters to eliminate dimensional differences between different variables. The aim is to ensure that all variables are on the same order of magnitude, thereby making the calculation more stable and effective. This invention obtains the spatial volume of each controlled sub-region and the minimum air exchange rate, current equipment power, rated equipment power, and rated air volume of the second characteristic parameters of the ventilation equipment. By accurately obtaining the spatial volume of each controlled sub-region and the key parameters of the ventilation equipment (minimum air exchange rate, equipment power, air volume, etc.), it ensures precise control and dynamic adjustment of ventilation, unlike existing technologies that rely solely on a single parameter (such as equipment power). Unlike other ventilation control methods (such as power or air volume), this invention comprehensively considers both space volume and equipment performance parameters to provide personalized ventilation solutions for each area. This effectively avoids resource waste and improves system safety and efficiency. In existing technologies, ventilation volume is typically calculated using simple equipment power or air volume, neglecting the influence of space volume. This invention calculates the theoretical minimum ventilation volume by multiplying space volume by the number of air changes, making ventilation volume calculation more scientific and reasonable. The theoretical minimum ventilation volume for each controlled sub-area is obtained by multiplying the minimum number of air changes by the space volume. The actual ventilation volume for each controlled sub-area is obtained by multiplying the ratio of current equipment power to rated equipment power by the rated air volume. The actual ventilation volume for each controlled sub-area is then calculated by combining the theoretical minimum ventilation volume with the rated air volume. The ventilation efficiency of the corresponding controlled sub-area is obtained by calculating the ratio of minimum ventilation volume. The theoretical minimum ventilation volume is obtained by multiplying the space volume by the minimum air exchange rate, ensuring a minimum guaranteed ventilation volume. This invention considers the actual size of the laboratory space and the requirements of the ventilation equipment, effectively reducing the safety risks of insufficient ventilation and preventing the accumulation of chemical gas concentrations. Traditional methods often rely on set standards or uniform parameters, which can easily lead to uneven or insufficient ventilation. By independently calculating the minimum ventilation volume for each sub-area, the air quality and safety in the laboratory can be more accurately guaranteed. The actual ventilation volume is calculated through a ratio relationship, making ventilation volume adjustment more flexible and consistent with real-time operating conditions. Unlike existing technologies that operate equipment solely based on set values, this invention provides real-time monitoring... By controlling the changes in the relationship between equipment power and airflow, the ventilation effect can be dynamically adjusted to minimize overloading or underloading of equipment. The ratio of current equipment power to rated equipment power directly reflects the actual load of the equipment. Calculating the actual ventilation volume by considering this factor is more accurate than traditional methods that rely solely on rated airflow. Ventilation can be adjusted according to the actual operating status of the equipment, improving energy efficiency and optimizing system operation. By calculating ventilation efficiency, the working effect of the ventilation system can be quantified. The calculation of ventilation efficiency allows for real-time evaluation of the current operating effect of the ventilation system, identifying potential equipment failures or system malfunctions, and taking timely corrective measures. By monitoring ventilation efficiency, not only is adequate ventilation ensured, but unnecessary energy consumption and resource waste are also avoided.This invention improves the system's operational reliability by acquiring real-time temperature data from environmental sources and calculating the gas diffusion coefficient based on the product of real-time temperature, ventilation efficiency, and gas diffusion rate. Real-time acquisition of temperature and ventilation efficiency data allows for more accurate prediction of gas diffusion behavior within the laboratory. The gas diffusion coefficient is a crucial factor influencing gas concentration changes; an accurate coefficient helps the system accurately assess gas propagation speed, adjust ventilation volume and other safety measures promptly, and prevent accidents caused by excessive gas concentration. Existing technologies often neglect the impact of gas diffusion rate on gas concentration changes. This invention combines real-time ambient temperature and ventilation efficiency to dynamically calculate the gas diffusion coefficient, resulting in more accurate gas concentration predictions. This invention enhances safety and controllability by separately obtaining the first initial reaction concentration of the main product and the second initial reaction concentration of the byproduct, and then calculating the total initial reaction concentration based on these concentrations. This allows for a comprehensive assessment of the material transformation during the reaction process. Accurately understanding the initial reaction concentrations provides more accurate data for predicting subsequent reaction progress and identifying potential reaction risks in advance. In existing technologies, only the concentration of a single product is typically considered, neglecting changes in byproducts and other reactants. This invention, by simultaneously monitoring both the main and byproducts, ensures a comprehensive understanding of the reaction process, facilitating better prediction of the toxicity of the reaction products. This invention addresses the risk by obtaining the actual reaction product concentration at a preset time point using gas diffusion coefficients, total initial reaction concentration, and spatial volume. By considering gas diffusion coefficients, reaction concentrations, and spatial volume, it dynamically predicts reaction product concentrations at different time points, effectively monitoring gas concentration changes during the reaction process. This helps laboratory personnel take appropriate safety measures before gas concentrations reach dangerous levels, thus preventing safety accidents caused by chemical reactions. Compared to existing technologies that simply monitor gas concentration without considering the influence of diffusion coefficients and spatial volume, this invention provides more accurate reaction product concentration predictions through comprehensive calculation of multiple factors, avoiding blind warnings of reaction risks. The prediction is achieved by using the ratio of the actual reaction product concentration to the preset reaction product concentration and the reaction product concentration... The product of toxicity levels yields the corresponding probability of delayed risk. By comprehensively considering the product of reaction product concentration and toxicity level, the probability of delayed risk at different time points is accurately assessed. Even when gas concentrations do not exceed safety thresholds, potential reaction risks can be effectively assessed, avoiding possible safety hazards. Existing technologies often overlook the potential risks brought by chemical reactions when monitoring gas concentrations within limits. This invention, through the calculation of delayed risk probability, provides a more comprehensive risk assessment mechanism, providing early warnings of potential safety issues in the laboratory. Through precise parameter settings and multi-dimensional real-time data acquisition, this invention optimizes the shortcomings of traditional safety monitoring systems, comprehensively improving the ability to manage safety risks that may arise from chemical gas leaks and reactions in the laboratory.By calculating key factors such as ventilation efficiency, gas diffusion coefficient, and actual reaction concentration, this innovative approach addresses the potential hazards of neglecting chemical gas reactions in existing technologies, significantly improving laboratory safety and equipment operating efficiency.

[0035] In one embodiment, step S6, which generates corresponding control instructions for each control sub-area based on each risk warning signal level and performs control according to the control instructions, includes: S61. Determine the corresponding control instruction from the preset early warning and control database according to each of the risk warning signal levels, wherein the control instruction includes basic control instructions and special control instructions; S62. Perform preliminary control on the corresponding control sub-area according to the basic control instructions, and obtain feedback information after preliminary control in real time; S63. Determine whether the feedback information meets the preset standard requirements; If the feedback information meets the preset standard requirements, it is determined that the controlled sub-area is out of risk; If the feedback information does not meet the preset standard requirements, then select the corresponding control measures from the special control instructions based on the feedback information to carry out special control until the feedback information meets the preset standard requirements.

[0036] As described in steps S61-S63 above, this invention determines the basic control instructions and specific control instructions corresponding to each risk warning signal level from a preset warning and control database. By selecting the corresponding control instructions from the preset warning and control database, it ensures that the control measures corresponding to each risk warning signal are customized and precise. Unlike traditional methods that rely on general instructions, this invention refines the association between warning signals and control instructions, enabling different control measures to be taken for different risk levels, thereby achieving more personalized safety control. By establishing a database and pre-setting control instructions, automatic response to risk signals can be achieved, avoiding manual intervention and delays, improving response speed and efficiency, and thus reducing the probability of accidents. Preliminary control is implemented in corresponding sub-areas through basic control commands, and feedback information is obtained in real time after the preliminary control. Preliminary control intervenes in the controlled areas through basic control commands, forming a hierarchical control structure. Through phased control, potential risks can be identified and mitigated more precisely, and initial problems can be addressed promptly. Compared to existing technologies that directly initiate comprehensive control, hierarchical management helps improve the flexibility and accuracy of security assurance. Preliminary control avoids unnecessary over-intervention, taking further control measures only when actual risks occur. This avoids resource waste and operational complexity caused by over-control. The system judges whether the feedback information meets preset standards; if the feedback information meets the preset standards, the control area is deemed safe. Once the controlled area is out of risk, if the feedback information does not meet the preset standard requirements, the corresponding control measures are selected from the special control instructions based on the feedback information for special control until the feedback information meets the preset standard requirements. Real-time acquisition of feedback information after control effectively ensures that the implementation effect of control measures is evaluated in a timely manner, enabling timely detection of problems and adjustment of measures. Compared with traditional methods that rely on manual detection and delayed reactions, real-time feedback can greatly improve the response speed to hazards and prevent accidents from escalating. Real-time data feedback helps shift the control of chemicals in the laboratory from intuitive decision-making to data-driven intelligent decision-making, improving the scientific nature and accuracy of control. Automated judgment of whether feedback information meets standard requirements avoids errors that may arise from manual judgment. Errors or delays can be quickly assessed by determining whether feedback meets preset standards, rapidly confirming laboratory safety, reducing ineffective control time and operations, optimizing resource allocation, and improving management efficiency. Specific control instructions provide targeted interventions. Compared to traditional "one-size-fits-all" control, this invention can select appropriate control schemes based on actual conditions, more effectively solving specific problems and avoiding unnecessary over-control. As feedback information continuously changes, this invention can dynamically adjust control strategies, ensuring control measures are always in optimal condition. Through intelligent selection of specific control measures, it can not only respond to sudden risks but also provide precise intervention for specific chemicals or leaks. This invention demonstrates the advantages of closed-loop control throughout the entire control process.The system continuously monitors feedback information and dynamically adjusts control measures based on the feedback to ensure that problems are thoroughly resolved. Compared to the potential lack of continuity in control measures in traditional methods, the closed-loop control of this invention guarantees the effectiveness and consistency of control measures throughout the entire process. By continuously collecting feedback and adjusting control measures, the system of this invention has self-optimization capabilities. Control measures are adjusted based on actual feedback data, continuously optimizing the control effect and further reducing laboratory risks.

[0037] like Figure 2 As shown, this application also provides a laboratory hazardous chemicals full-process risk management system based on an AI large model, including: The module is used to acquire full-process node information and basic attribute data of laboratory hazardous chemicals, and to build a chemical attribute database based on the basic attribute data; The partitioning module is used to divide the laboratory into multiple control sub-areas based on the full-process node information, and to acquire gas monitoring data and environmental status data of each control sub-area in real time. The gas monitoring data includes the proportion of gas components, real-time gas concentration, and gas diffusion rate. The input module is used to input the proportion of each gas component into a pre-trained AI large model to obtain the corresponding predicted potential reaction type, and to obtain the corresponding reaction triggering condition according to each predicted potential reaction type, wherein the predicted potential reaction type includes one of oxidation reaction, reduction reaction, neutralization reaction and polymerization reaction; The acquisition module is used to acquire the reaction risk probability based on each reaction triggering condition, environmental data, and the real-time gas concentrations of any two gases, and to determine whether the reaction risk probability exceeds a first preset threshold. If the probability of the reaction risk exceeds the first preset threshold, the toxicity level of the reaction product is obtained according to each of the predicted potential reaction types and chemical property databases, and the corresponding delay risk probability is obtained according to each of the reaction product toxicity levels and gas diffusion rates. The judgment module is used to determine whether the probability of the delay risk exceeds a second preset threshold. If the probability of delay risk exceeds the second preset threshold, it is determined that there is a risk in the controlled sub-area, and a corresponding risk warning signal level is generated. The control module is used to generate corresponding control instructions for the corresponding control sub-area based on each of the risk warning signal levels, and to carry out control according to the control instructions.

[0038] In one embodiment, the acquisition module includes: The first acquisition unit is used to acquire the critical temperature, critical humidity, critical ventilation volume, and critical concentration thresholds of the corresponding two gases according to the reaction triggering conditions. The second acquisition unit is used to acquire real-time temperature, real-time humidity and real-time ventilation volume based on the environmental data, and to acquire temperature influence factor based on the real-time temperature and critical temperature. The third acquisition unit is used to acquire a humidity influence factor based on the real-time humidity and critical humidity, and to acquire a ventilation influence factor based on the real-time ventilation volume and critical ventilation volume. The fourth acquisition unit is used to acquire the environmental impact coefficient based on the ventilation impact factor, humidity impact factor and temperature impact factor; The fifth acquisition unit is used to acquire a concentration ratio factor based on the real-time concentrations of any two gases and their corresponding critical concentration thresholds, and to acquire a reaction risk probability based on the concentration ratio factor and the environmental impact coefficient.

[0039] It should be noted that each module and unit in the AI-based big data model-based laboratory hazardous chemicals full-process risk management system corresponds one-to-one with the steps in the AI-based big data model-based laboratory hazardous chemicals full-process risk management method.

[0040] like Figure 3 As shown, this application also provides a computer device, which can be a server, and its internal structure can be as follows: Figure 3 As shown, the computer device includes a processor, memory, network interface, and database connected via a system bus. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and database. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores all data required for the process of a laboratory hazardous chemicals risk management method based on an AI-based large-scale model. The network interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements the laboratory hazardous chemicals risk management method based on an AI-based large-scale model.

[0041] Those skilled in the art will understand that Figure 3 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer equipment on which the present application is applied.

[0042] An embodiment of this application also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements any of the above-described methods for full-process risk management of laboratory hazardous chemicals based on an AI large model.

[0043] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in this application and in the embodiments can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual-speed SDRAM (SSRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0044] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, apparatus, article, or method that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, apparatus, article, or method. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, apparatus, article, or method that includes that element.

[0045] The above description is merely a preferred embodiment of the present invention and does not limit the patent scope of the present invention. Any equivalent structural or procedural transformations made based on the content of the present invention's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of the present invention.

Claims

1. A method for full-process risk management of hazardous chemicals in laboratories based on an AI-powered large-scale model, characterized in that, include: Obtain full-process node information and basic attribute data of laboratory hazardous chemicals, and construct a chemical attribute database based on the basic attribute data; Based on the full-process node information, the laboratory is divided into multiple control sub-areas, and gas monitoring data and environmental status data of each control sub-area are acquired in real time. The gas monitoring data includes the proportion of gas components, real-time gas concentration, and gas diffusion rate. The proportion of each gas component is input into a pre-trained AI model to obtain the corresponding predicted potential reaction type, and the corresponding reaction triggering condition is obtained according to each predicted potential reaction type. The predicted potential reaction type includes one of oxidation reaction, reduction reaction, neutralization reaction and polymerization reaction. The reaction risk probability is obtained based on each reaction triggering condition, environmental data, and the real-time gas concentrations of any two gases, and it is determined whether the reaction risk probability exceeds a first preset threshold. If the probability of the reaction risk exceeds the first preset threshold, the toxicity level of the reaction product is obtained according to each of the predicted potential reaction types and chemical property databases, and the corresponding delay risk probability is obtained according to each of the reaction product toxicity levels and gas diffusion rates. Determine whether the probability of the delay risk exceeds a second preset threshold; If the probability of delay risk exceeds the second preset threshold, it is determined that there is a risk in the controlled sub-area, and a corresponding risk warning signal level is generated. For each of the aforementioned risk warning signal levels, a corresponding control instruction is generated for the corresponding control sub-area, and control is carried out according to the control instruction.

2. The method for full-process risk management of laboratory hazardous chemicals based on an AI large-scale model according to claim 1, characterized in that, The step of inputting the proportion of each gas component into a pre-trained AI model to obtain the corresponding predicted potential reaction type includes: The proportion of each gas component is input into a pre-trained AI model to identify and obtain all target gases within the corresponding control sub-region; Obtain the first characteristic parameter of each target gas, and extract the incompatible substance association table corresponding to the target gas from the chemical property database based on the first characteristic parameter; Based on the target gas, the incompatible gas is determined from the incompatible gas association table, and the corresponding reactivity correlation degree is obtained between the incompatible gas and the target gas. The percentage change rate and percentage fluctuation coefficient of each gas component are obtained within a preset time period, and the percentage change rate, percentage fluctuation coefficient and reactivity correlation of each component are input into a pre-trained AI large model to obtain the basic probability of each potential reaction. Extract the key reactive gas pairs and their correlation for each potential reaction from the chemical property database, and obtain the proportion of the two key reactive gases corresponding to each key reactive gas pair; The correction coefficient is obtained based on the correlation degree and the proportion of the two key reactant gases, and the correction probability of the corresponding potential reaction is obtained based on the correction coefficient and the corresponding basic probability. The potential response corresponding to the highest correction probability is selected as the predicted potential response type.

3. The method for full-process risk management of laboratory hazardous chemicals based on an AI large-scale model according to claim 1, characterized in that, The step of obtaining the reaction risk probability based on each reaction triggering condition, environmental data, and the real-time gas concentrations of any two gases includes: Based on the reaction triggering conditions, obtain the critical temperature, critical humidity, critical ventilation volume, and the critical concentration thresholds of the corresponding two gases; Real-time temperature, real-time humidity, and real-time ventilation volume are obtained based on the environmental data, and temperature influence factors are obtained based on the real-time temperature and critical temperature. The humidity influencing factor is obtained based on the real-time humidity and the critical humidity, and the ventilation influencing factor is obtained based on the real-time ventilation volume and the critical ventilation volume. The environmental impact coefficient is obtained based on the ventilation impact factor, humidity impact factor, and temperature impact factor. The concentration ratio factor is obtained based on the real-time concentrations of any two gases and their corresponding critical concentration thresholds, and the reaction risk probability is obtained based on the concentration ratio factor and the environmental impact coefficient.

4. The method for full-process risk management of laboratory hazardous chemicals based on an AI large-scale model according to claim 1, characterized in that, The step of obtaining the toxicity level of the reaction product based on each of the predicted potential reaction types and chemical property databases includes: Based on the predicted potential reaction types, reaction condition correlation features and reactant types are obtained, wherein the reaction condition correlation features include reaction temperature range, reaction pressure range, reaction medium type, and reaction catalyst; The average reaction temperature is obtained based on the reaction temperature range, and the average reaction pressure is obtained based on the reaction pressure range. Based on the types of reactants, reaction media, and reaction catalysts, all related product records are extracted from the chemical property database, and the corresponding record standard temperature and record standard pressure are obtained from the chemical property database for each related product record. The first matching degree of the corresponding associated product record is obtained based on the standard temperature, standard pressure, average reaction temperature and average reaction pressure of each record, and the associated product record corresponding to the highest first matching degree is selected as the target product record. Determine whether the first matching degree of the target product record is less than the first preset matching degree; If the first matching degree of the target product record is not less than the first preset matching degree, then the target product record is determined as the final target product record, and the main product and by-product are determined from the chemical property database based on the final target product record; If the first matching degree of the target product record is less than the first preset matching degree, then return to the step of selecting the associated product record corresponding to the highest first matching degree as the target product record, until the first matching degree of the target product record is not less than the first preset matching degree; Select the toxicity levels of the main product and by-product corresponding to the main product and by-product from the chemical property database, respectively, and obtain the toxicity level of the reaction product based on the toxicity levels of the main product and by-product.

5. The method for full-process risk management of laboratory hazardous chemicals based on an AI large-scale model according to claim 1, characterized in that, The step of obtaining the corresponding delayed risk probability based on the toxicity level of each reaction product and the gas diffusion rate includes: Obtain the spatial volume and second characteristic parameters of the ventilation equipment in each controlled sub-area, wherein the second characteristic parameters include the minimum air change rate, current equipment power, rated equipment power, and rated air volume; The theoretical minimum ventilation volume of the corresponding control sub-area is obtained based on each of the minimum air exchange rates and space volumes, and the actual ventilation volume of the corresponding control sub-area is obtained based on each of the current equipment power, rated equipment power and rated air volume. The ventilation efficiency of the corresponding controlled sub-area is obtained based on the actual ventilation volume and the theoretical minimum ventilation volume for each of the aforementioned sub-areas. The real-time temperature is obtained based on the environmental data, and the gas diffusion coefficient is obtained based on the real-time temperature, ventilation efficiency, and gas diffusion rate. The first initial reaction concentration of the main product and the second initial reaction concentration of the by-product are obtained respectively, and the total initial reaction concentration is obtained based on the first initial reaction concentration and the second initial reaction concentration. The actual reaction product concentration at a preset time point is obtained based on the gas diffusion coefficient, the total initial reaction concentration, and the spatial volume. The corresponding delay risk probability is obtained based on the actual reaction product concentration, the preset reaction product concentration, and the toxicity level of the reaction product.

6. The method for full-process risk management of laboratory hazardous chemicals based on AI large-scale model according to claim 1, characterized in that, The step of generating corresponding control instructions for each control sub-area based on each risk warning signal level, and performing control according to the control instructions, includes: According to each of the aforementioned risk warning signal levels, a corresponding control instruction is determined from a preset warning and control database, wherein the control instruction includes basic control instructions and special control instructions; The corresponding control sub-area is initially controlled according to the basic control instructions, and feedback information after the initial control is obtained in real time. Determine whether the feedback information meets the preset standard requirements; If the feedback information meets the preset standard requirements, it is determined that the controlled sub-area is out of risk; If the feedback information does not meet the preset standard requirements, then select the corresponding control measures from the special control instructions based on the feedback information to carry out special control until the feedback information meets the preset standard requirements.

7. A laboratory hazardous chemicals full-process risk management system based on an AI large-scale model, characterized in that, include: The module is used to acquire full-process node information and basic attribute data of laboratory hazardous chemicals, and to build a chemical attribute database based on the basic attribute data; The partitioning module is used to divide the laboratory into multiple control sub-areas based on the full-process node information, and to acquire gas monitoring data and environmental status data of each control sub-area in real time. The gas monitoring data includes the proportion of gas components, real-time gas concentration, and gas diffusion rate. The input module is used to input the proportion of each gas component into a pre-trained AI large model to obtain the corresponding predicted potential reaction type, and to obtain the corresponding reaction triggering condition according to each predicted potential reaction type, wherein the predicted potential reaction type includes one of oxidation reaction, reduction reaction, neutralization reaction and polymerization reaction; The acquisition module is used to acquire the reaction risk probability based on each reaction triggering condition, environmental data, and the real-time gas concentrations of any two gases, and to determine whether the reaction risk probability exceeds a first preset threshold. If the probability of the reaction risk exceeds the first preset threshold, the toxicity level of the reaction product is obtained according to each of the predicted potential reaction types and chemical property databases, and the corresponding delay risk probability is obtained according to each of the reaction product toxicity levels and gas diffusion rates. The judgment module is used to determine whether the probability of the delay risk exceeds a second preset threshold. If the probability of delay risk exceeds the second preset threshold, it is determined that there is a risk in the controlled sub-area, and a corresponding risk warning signal level is generated. The control module is used to generate corresponding control instructions for the corresponding control sub-area based on each of the risk warning signal levels, and to carry out control according to the control instructions.

8. The laboratory hazardous chemicals full-process risk management system based on AI large model according to claim 7, characterized in that, The acquisition module includes: The first acquisition unit is used to acquire the critical temperature, critical humidity, critical ventilation volume, and critical concentration thresholds of the corresponding two gases according to the reaction triggering conditions. The second acquisition unit is used to acquire real-time temperature, real-time humidity and real-time ventilation volume based on the environmental data, and to acquire temperature influence factor based on the real-time temperature and critical temperature. The third acquisition unit is used to acquire a humidity influence factor based on the real-time humidity and critical humidity, and to acquire a ventilation influence factor based on the real-time ventilation volume and critical ventilation volume. The fourth acquisition unit is used to acquire the environmental impact coefficient based on the ventilation impact factor, humidity impact factor and temperature impact factor; The fifth acquisition unit is used to acquire a concentration ratio factor based on the real-time concentrations of any two gases and their corresponding critical concentration thresholds, and to acquire a reaction risk probability based on the concentration ratio factor and the environmental impact coefficient.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images