Dangerous chemical risk dynamic management and control system and method

Abstract

A dangerous chemical risk dynamic management and control system comprises a dynamic detection subsystem, an information management subsystem and a risk prediction subsystem. The dynamic detection subsystem comprises a dangerous chemical detection mechanism and an environment detection mechanism, the dangerous chemical detection mechanism is used for detecting the use dynamic state of dangerous chemicals to obtain state information, and the environment detection mechanism is used for detecting the use environment of the dangerous chemicals to obtain environment information; the information management subsystem is used for collecting and managing process information generated in the use process of the dangerous chemicals; the risk prediction subsystem is used for generating a risk prediction warning based on the state information and the environment information, generating a management and control strategy based on the risk prediction warning, and generating a management and control record based on the risk prediction warning and the process information. According to the dangerous chemical risk dynamic management and control system and method provided by the invention, whether a risk accident possibly occurs can be predicted based on dynamic information and environmental information generated in a dangerous chemical use process, and personnel and property safety is ensured.

Description

Technical Field

[0001] This invention relates to the field of hazardous chemicals management technology, specifically a dynamic risk control system and method for hazardous chemicals. Background Technology

[0002] With the rapid development of petrochemical technology, chemicals are widely used in enterprise production. However, some chemicals are hazardous and are a key focus of safety management in petrochemical enterprises. Once a hazardous chemical accident occurs, it can easily lead to mass casualties, significant economic losses, and social impact. Therefore, the safety management of hazardous chemicals has always been highly valued. In units using hazardous chemicals, the management process mainly includes four aspects: procurement, storage, use, and disposal of hazardous chemicals. Common causes of risk accidents caused by hazardous chemicals include: (1) Environmental factors: inadequate supervision of hazardous storage, random stacking of various hazardous chemicals, leading to environmental pollution; (2) Human factors: lack of safety knowledge and awareness; (3) Material factors: lack of classified management of various hazardous materials, inability to control them in a timely manner, and failure to detect hazards in a timely manner; (4) Management factors: lack of effective training and management for operators.

[0003] Because there are many types of hazardous chemicals and their usage processes are very complex, automated management is difficult. Therefore, current hazardous chemical management technologies mainly involve moving offline data online to achieve automated acquisition, storage, and processing. However, risk prediction is still done manually, which results in low efficiency and poor prediction results. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a dynamic risk management system and method for hazardous chemicals, which can predict the possibility of potential risks and accidents based on dynamic and environmental information generated during the use of hazardous chemicals, thereby ensuring the safety of personnel and property.

[0005] To achieve the above objectives, the specific solution adopted by the present invention is as follows: a dynamic risk management and control system for hazardous chemicals, comprising a dynamic detection subsystem, an information management subsystem, and a risk prediction subsystem; The dynamic detection subsystem includes a hazardous chemical detection mechanism and an environmental detection mechanism. The hazardous chemical detection mechanism is used to detect the dynamic use of hazardous chemicals to obtain status information, while the environmental detection mechanism is used to detect the environment in which hazardous chemicals are used to obtain environmental information. The information management subsystem is used to collect and manage process information generated during the use of hazardous chemicals; The risk prediction subsystem is used to generate risk prediction warnings based on status information and environmental information, generate control strategies based on risk prediction warnings, and generate control records based on risk prediction warnings and process information.

[0006] As a further optimization of the aforementioned dynamic risk management system for hazardous chemicals: the hazardous chemical detection mechanism includes multiple hazardous chemical detection sensors, which are used to detect the status information of at least one hazardous chemical; the environmental detection mechanism includes multiple environmental detection sensors and several video acquisition devices, which are used to detect at least one environmental parameter, and the video acquisition devices are used to capture environmental images of the environment in use; the environmental parameters and environmental images constitute environmental information.

[0007] A method for dynamic risk management of hazardous chemicals includes the following steps: The system detects the usage dynamics of hazardous chemicals to obtain status information and detects the usage environment of hazardous chemicals to obtain environmental information. The status information and environmental information are then combined into detection information. Collect process information generated during the use of hazardous chemicals, including multiple business steps. The risk logic model and risk prediction model are used to calculate the first risk probability based on the detection information, and the image recognition model is used to calculate the second risk probability based on the detection information. The risk logic model includes a top event to characterize the risk event, basic events obtained by logical inversion based on the top event, and intermediate events to associate the top event and the basic events. The basic events have a first correlation with the detection information, and the intermediate events have a second correlation with the risk prediction model. Based on process information, the probability of the first risk, and the probability of the second risk, a risk prediction warning is issued, and a control strategy is generated. Once the process information indicates that all business steps have been completed, all risk predictions and warnings are recorded as control and management records.

[0008] As a further optimization of the above-mentioned dynamic risk management method for hazardous chemicals: the environmental information in the detection information includes environmental parameters and environmental images; the risk logic model and risk prediction model calculate the first risk probability based on the environmental parameters and state information; and the image recognition model calculates the second risk probability based on the environmental images.

[0009] As a further optimization of the above-mentioned dynamic risk management method for hazardous chemicals, the method for constructing a risk logic model includes: Identify the potential risks that hazardous chemicals may cause and designate these risks as top events. Based on empirical data, risk factors that can trigger risk events are identified, and these risk factors are treated as intermediate events. The first correlation is determined by logical inversion of risk factors using logical AND, logical OR, and logical NOT, and the detected information is taken as the basic event based on the first correlation. Combine top events, intermediate events, and basic events into a risk logic model.

[0010] As a further optimization of the above-mentioned dynamic risk management method for hazardous chemicals: in the construction method of the risk logic model, after determining the top event, intermediate events and basic events, the minimum cut set corresponding to the top event is determined. The minimum cut set includes at least one intermediate event and at least one basic event. When generating control strategies, several alternative strategies are determined based on the risk failure model, and one of the alternative strategies is determined as the optimal strategy based on the minimum cut set.

[0011] As a further optimization of the above-mentioned dynamic risk management method for hazardous chemicals: in the construction method of the risk logic model, after determining the top event, intermediate events and basic events, the importance of the basic events is assessed based on the results of logical inversion. When determining alternative strategies, alternative strategies are determined based on the importance of the basic events.

[0012] As a further optimization of the above-mentioned dynamic risk management method for hazardous chemicals: the risk prediction model is set as a Bayesian network model, and the basic events in the risk logic model are mapped to the root nodes of the Bayesian network model based on the second correlation relationship.

[0013] As a further optimization of the above-mentioned method for dynamic risk management of hazardous chemicals, the following steps are also included: The risk logic model and risk prediction model are updated and optimized based on the generated first risk probability. Methods for optimizing risk prediction models include: The generated posterior probability and the current likelihood estimate generated based on the probe information are used to calculate a new posterior probability, and the parameters of the Bayesian network model for generating the posterior probability are optimized.

[0014] As a further optimization of the above-mentioned dynamic risk management method for hazardous chemicals: the image recognition model is set as a convolutional neural network model, and the method for calculating the second risk probability using the image recognition model includes: Environmental images are input into the image recognition model; Extract spectral features from environmental images and input these features into the layer structure of the image recognition model; The second risk probability is obtained by using an image recognition model to extract features and classify environmental images.

[0015] Beneficial effects: By constructing a risk logic model, a risk prediction model, and an image recognition model, this invention can accurately predict whether a risk accident will occur based on the state and environmental information generated during the use of hazardous chemicals. When a risk accident is predicted, a risk prediction warning is issued to remind staff to quickly handle the situation or evacuate, thus fully ensuring the safety of personnel and the economy. Attached Figure Description

[0016] Figure 1 This is an overall structural block diagram of the dynamic control system of the present invention; Figure 2 This is a flowchart of the dynamic control method of the present invention; Figure 3 This is a schematic diagram illustrating the construction principle of the risk logic model. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] like Figure 1 As shown, a dynamic risk management system for hazardous chemicals includes a dynamic detection subsystem, an information management subsystem, and a risk prediction subsystem.

[0019] The dynamic detection subsystem includes a hazardous chemical detection agency and an environmental detection agency. The hazardous chemical detection agency is used to detect the dynamic use of hazardous chemicals to obtain status information, while the environmental detection agency is used to detect the environment in which hazardous chemicals are used to obtain environmental information.

[0020] The information management subsystem is used to collect and manage process information generated during the use of hazardous chemicals.

[0021] The risk prediction subsystem is used to generate risk prediction warnings based on status information and environmental information, generate control strategies based on risk prediction warnings, and generate control records based on risk prediction warnings and process information.

[0022] During the use of hazardous chemicals, a dynamic detection subsystem is first used to continuously monitor the dynamic information of the hazardous chemicals and environmental information. At the same time, an information management subsystem collects process information. Then, a risk prediction subsystem is used to predict the possibility of a risk accident based on the status information and environmental information. When a risk accident is likely to occur, the risk prediction subsystem generates a risk prediction warning and a corresponding control strategy, thereby warning personnel related to hazardous chemicals and providing control strategies for reference. After all the use processes of hazardous chemicals have ended, the risk prediction subsystem generates a control record based on the process information and all risk prediction warnings. The control record may also include the control strategies corresponding to the risk prediction warnings.

[0023] The specific structure of the hazardous chemical detection mechanism is as follows: The mechanism includes multiple hazardous chemical detection sensors, which are used to detect the state information of at least one hazardous chemical. The environmental detection mechanism includes multiple environmental detection sensors and several video acquisition devices. The environmental sensors are used to detect at least one environmental parameter, and the video acquisition devices are used to capture environmental images of the environment. The environmental parameters and environmental images constitute environmental information. In this invention, the hazardous chemical detection sensors include combustible gas sensors, hazardous gas sensors, wastewater hazardous substance sensors, and exhaust gas sensors. The combustible gas sensors are mainly used to detect the concentration of combustible gases in hazardous chemicals; the hazardous gas sensors are mainly used to detect the concentration of hazardous gases in hazardous chemicals; the wastewater hazardous substance sensors are mainly used to detect the content of hazardous substances in discharged wastewater; and the exhaust gas sensors are mainly used to detect the composition and content of exhaust gas. Because there are many types of hazardous chemicals, the types of hazardous chemicals that each sensor can detect need to be determined according to actual needs, and then the appropriate model should be selected. This will not be elaborated further here. Environmental detection sensors include oxygen concentration sensors, temperature sensors, pressure sensors, humidity sensors, and wind speed sensors. The oxygen concentration sensor detects the oxygen content in the environment, the temperature sensor detects the ambient temperature, the humidity sensor detects the ambient humidity, and the wind speed sensor detects the ambient wind speed. These sensors are primarily used to determine whether hazardous and harmful gases will be transferred and dispersed by wind. Considering that the application scenarios for hazardous chemicals vary, and the environmental information that needs to be detected differs in different application scenarios, the specific selection and deployment of environmental detection sensors must be determined based on actual needs, and will not be elaborated further here.

[0024] In hazardous chemical detection facilities, hazardous chemical detection sensors and environmental sensors can detect the state information of hazardous chemicals and the environmental information of the usage environment, reflecting to some extent whether a risk accident may occur. However, when a small-scale risk accident has already occurred, the values ​​of these sensors may change significantly, even to the point that it may be impossible to determine that a risk accident has occurred. For example, in an open environment, a small-scale combustion of combustible gas may occur. However, due to wind, the concentration of combustible gas, ambient temperature, and oxygen content in the environment may remain within normal ranges, making it difficult to determine that a risk accident has occurred. Therefore, hazardous chemical detection facilities also include video acquisition equipment. This equipment continuously captures environmental images of the usage environment, and based on these images, it is possible to determine whether such a small-scale risk accident has occurred. For example, when a small-scale combustion of combustible gas occurs in a localized area, flames will appear in the environmental images. Even if the data from the hazardous chemical detection sensors and environmental sensors do not reflect that a risk accident has occurred, it is still possible to identify and determine that a risk accident has occurred by analyzing the environmental images.

[0025] The information management subsystem comprises multiple information management computers. These computers can connect to hazardous chemical information collectors capable of acquiring process information. These collectors can collect process information by scanning barcodes or other markings on the packaging of hazardous chemicals, or through manual registration. Because hazardous chemicals undergo multiple business processes during use, the collected process information can include MSDS information, inbound and outbound information, usage records, disposal records, and return procedures—information corresponding to each business process. The information management subsystem can generate process information based on this collected process information and the actual usage of hazardous chemicals. Since different production entities use hazardous chemicals differently, the corresponding process information will also differ. Appropriate process information can be determined based on actual needs, which will not be elaborated further here.

[0026] like Figure 2 and 3 As shown, based on the above-mentioned dynamic risk management system for hazardous chemicals, the present invention further provides a dynamic risk management method for hazardous chemicals, including S1 to S6.

[0027] S1. Status information is obtained by detecting the dynamic use of hazardous chemicals, and environmental information is obtained by detecting the environment in which hazardous chemicals are used. The status information and environmental information are combined into detection information. The environmental information includes environmental parameters and environmental images. The risk logic model and risk prediction model calculate the first risk probability based on the environmental parameters and status information, and the image recognition model calculates the second risk probability based on the environmental images. The status information is obtained using hazardous chemical detection sensors from the hazardous chemical detection agency. The environmental parameters are obtained using environmental detection sensors from the environmental detection agency, and the environmental images are captured using image acquisition equipment from the environmental detection agency. The specific detection process has already been described in the hazardous chemical risk management system and will not be repeated here.

[0028] S2. Collect process information generated during the use of hazardous chemicals, including multiple business processes.

[0029] S3. Calculate the first risk probability based on the detection information using a risk logic model and a risk prediction model; calculate the second risk probability using an image recognition model based on the detection information. The risk logic model includes a top event characterizing a risk event, basic events obtained through logical inversion based on the top event, and intermediate events used to correlate the top event and basic events. The basic events have a first correlation with the detection information, and the intermediate events have a second correlation with the risk prediction model. The first risk probability characterizes the probability of a possible risk event based on state information and environmental parameters, while the second risk probability characterizes the probability of a possible risk event based on environmental images, as well as the probability of a small-scale risk event already occurring.

[0030] The risk logic model is used to determine the causal relationship between detection information and risk events, thereby enabling the prediction of the probability of risk events occurring based on the risk logic model, namely the first risk probability and the second risk probability. In the risk logic model, the top event corresponds to the risk event, the basic events correspond to the detection information, and the intermediate events are used to link the top event and the basic events. More specifically, the top event is the risk event itself, the intermediate events are the conditions for the occurrence of the risk event, and the basic events are the factors of the risk event.

[0031] The methods for constructing risk logic models include S31 to S34.

[0032] S31. Identify the risk events that hazardous chemicals can cause and designate them as top events. The identification of risk events can be based on actual conditions and empirical data. For example, risk events can include fire, explosion, poisoning, asphyxiation, and pollution. Furthermore, risk events can be classified, for example, into particularly serious accidents, major accidents, relatively serious accidents, and general accidents. The combination of risk events and levels is then used as the top event.

[0033] S32. Based on empirical data, identify the risk factors that can trigger risk incidents, and treat these risk factors as intermediate events. Different risk incidents can determine different risk migration rates. For example, for fire and explosion risk incidents, the associated risk factors may be increased ambient temperature, ventilation failure, or fire protection failure.

[0034] S33. Utilize logical AND, OR, and NOT to perform logical inversion on risk factors to determine the primary correlation, and use the detection information as the basic event based on this primary correlation. Because the causes of risk factors are complex, during the use of hazardous chemicals, risk factors may be related to multiple detection information. For example, a risk factor of rising ambient temperature may be related to a decrease in both ambient temperature and ambient wind speed. Therefore, it is necessary to determine the primary correlation based on the risk factors, and then identify all detection information related to the risk factors. Logical AND, OR, and NOT are conventional logical relationships and will not be elaborated upon here.

[0035] S34. Combine the top event, intermediate events and basic events into a risk logic model.

[0036] Furthermore, in the risk logic model construction method, after determining the top event, intermediate events, and basic events, the minimum cut set corresponding to the top event is determined. The minimum cut set includes at least one intermediate event and at least one basic event. In the failure logic model, there are many combinations of basic events that can cause the risk event corresponding to the top event. The minimum cut set is the smallest combination of basic events that leads to the risk event corresponding to the top event. Among different combinations, because the minimum cut set contains the fewest basic events, its probability of occurrence is also higher. Therefore, by determining the minimum cut set, risk events can be predicted more accurately. On the other hand, in the risk logic model construction method, after determining the top event, intermediate events, and basic events, the importance of basic events is assessed based on the results of logical inversion. When determining alternative strategies, the importance of basic events is used to determine alternative strategies. The importance assessment can be based on the logical relationships related to basic events. Specifically, when a basic event has more logical relationships with more intermediate events, it will also have more correlations with top events. That is, the triggering of this basic event is more likely to lead to the occurrence of a certain top event, and in this case, the importance of this basic event is higher.

[0037] On the other hand, in this embodiment, the risk prediction model is set as a Bayesian network model. The intermediate events in the risk logic model are mapped to nodes in the Bayesian network model based on the second correlation relationship. Since the intermediate events and the basic events have a logical relationship, the basic events in the risk logic model can be further mapped to the root nodes of the Bayesian network model based on the second correlation relationship.

[0038] S4. Based on process information, the probability of the first risk, and the probability of the second risk, issue a risk prediction warning and generate a control strategy. When the probability of the first risk and / or the probability of the second risk exceed a threshold, it indicates that a risk incident may occur. Issuing a risk prediction warning can remind staff to handle the situation quickly or take timely evasive action. On the other hand, when a risk incident occurs, it may be caused by some unexpected factors, or it may be caused by problems in some business processes. For example, when there is a flammable gas or harmful gas leak, it may be due to inadequate or missing inspections and maintenance. Such problems in business processes may lead to more frequent risk incidents. Therefore, the risk prediction warning can also include business process warnings to remind staff to optimize business processes.

[0039] In addition, risk prediction warnings also include dynamic risk warnings based on a first risk probability and a second risk probability. For example, when the first risk probability indicates that an accident may occur at a certain location due to a flammable gas leak, and the second risk probability indicates that an open flame has appeared at another location, if the flammable gas may spread from the current location to the location of the open flame, it may cause a larger-scale accident. In this case, the first risk probability and the second risk probability can be combined to determine a dynamic risk warning that can remind staff of the possibility of the accident escalating.

[0040] When generating control strategies, several alternative strategies are determined based on a risk failure model, and the optimal strategy is selected from these alternatives based on a minimum cut set. Control strategies primarily include methods for processing relevant equipment based on detection information related to basic events. For example, when a combustible gas leak causes a localized increase in combustible gas concentration, the leak needs to be interrupted, such as by shutting down the gas supply or accelerating ventilation to rapidly disperse the gas and reduce its concentration. Because the risk incident corresponding to the top event can be associated with a combination of multiple basic events, the occurrence of a top event may require simultaneous processing of relevant equipment based on detection information corresponding to multiple basic events. Higher-importance basic events may trigger more top events. Therefore, when determining control strategies, alternative strategies need to be determined based on the importance of basic events; higher-importance basic events require faster processing to prevent further risk incidents. On the other hand, because a minimum cut set is the smallest combination of basic events that can trigger a risk incident, the alternative strategy determined based on the minimum cut set may terminate the risk incident most quickly. Therefore, the optimal strategy is determined based on the minimum cut set.

[0041] S5. Once all business processes represented by the process information have been completed, a control record is generated based on all risk prediction warnings. The control record also includes the control strategies corresponding to the risk prediction warnings.

[0042] S6. Update and optimize the risk logic model and risk prediction model based on the generated first risk probability.

[0043] The specific methods for updating and optimizing the risk logic model mainly involve adjusting its structure, which may include adding or removing basic events, modifying logical relationships, and adjusting the importance of basic events. The updated and optimized risk logic model can be verified through simulation using an existing simulation system to determine whether it accurately reflects the current system state. The risk probability model can also be optimized based on environmental images corresponding to the second risk probability. If an open flame is detected when no risk prediction warning is issued based on the first risk probability, it indicates that the current risk logic model needs further optimization. Optimization methods may include adjusting the logical relationships within the risk logic model and the detection information associated with the top event, intermediate events, and basic events.

[0044] The method for optimizing the risk prediction model includes: calculating a new posterior probability from the generated posterior probability and the likelihood estimate generated based on the current probe information, and optimizing the parameters of the Bayesian network model that generate the posterior probability. More specifically, the first step is to acquire probe information; the second step is to calculate the probability of the risk event occurring based on the probe information, i.e., the likelihood estimate; the third step is to update the posterior probability using Bayes' theorem, combining the prior probability and the likelihood estimate. Bayes' theorem is expressed as P(θ|X)=\frac{P(X|θ)P(θ)}{P(X)}, where P(θ|X) is the posterior probability, P(X|θ) is the likelihood estimate, P(θ) is the prior probability, and P(X) is the evidence factor; the fourth step is to use the updated posterior probability as the prior probability for the next round of updates, and then apply Bayes' theorem again for updates. The number of updates can be determined based on the acquisition frequency of probe information, i.e., an update is performed every time new probe information is acquired. Accordingly, the detection frequency of all probe information can be pre-set.

[0045] In addition to the risk logic model and risk prediction model, this invention uses a convolutional neural network (CNN) model for image recognition. The method for calculating the second risk probability using the image recognition model includes: inputting environmental images into the image recognition model; extracting spectral features from the environmental images, which can be obtained using a wavelet-convolutional neural network method through a two-dimensional Haar transform, and inputting the spectral features into the layer structure of the image recognition model; and using the image recognition model to perform feature extraction and classification on the environmental images to obtain the second risk probability. CNNs are existing technology and will not be elaborated upon here.

[0046] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A dynamic risk management and control system for hazardous chemicals, characterized in that, It includes a dynamic detection subsystem, an information management subsystem, and a risk prediction subsystem; The dynamic detection subsystem includes a hazardous chemical detection mechanism and an environmental detection mechanism. The hazardous chemical detection mechanism is used to detect the dynamic use of hazardous chemicals to obtain status information, while the environmental detection mechanism is used to detect the environment in which hazardous chemicals are used to obtain environmental information. The information management subsystem is used to collect and manage process information generated during the use of hazardous chemicals; The risk prediction subsystem is used to generate risk prediction warnings based on status information and environmental information, generate control strategies based on risk prediction warnings, and generate control records based on risk prediction warnings and process information.

2. The dynamic risk management system for hazardous chemicals as described in claim 1, characterized in that, The hazardous chemical detection agency includes multiple hazardous chemical detection sensors, which are used to detect the status information of at least one hazardous chemical. The environmental detection agency includes multiple environmental detection sensors and several video acquisition devices. The environmental detection sensors are used to detect at least one environmental parameter, and the video acquisition devices are used to capture environmental images of the environment. The environmental parameters and environmental images constitute environmental information.

3. A method for dynamic risk management of hazardous chemicals, characterized in that, Includes the following steps: The system detects the usage dynamics of hazardous chemicals to obtain status information and detects the usage environment of hazardous chemicals to obtain environmental information. The status information and environmental information are then combined into detection information. Collect process information generated during the use of hazardous chemicals, including multiple business steps. The risk logic model and risk prediction model are used to calculate the first risk probability based on the detection information, and the image recognition model is used to calculate the second risk probability based on the detection information. The risk logic model includes a top event to characterize the risk event, basic events obtained by logical inversion based on the top event, and intermediate events to associate the top event and the basic events. The basic events have a first correlation with the detection information, and the intermediate events have a second correlation with the risk prediction model. Based on process information, the probability of the first risk, and the probability of the second risk, a risk prediction warning is issued, and a control strategy is generated. Once the process information indicates that all business steps have been completed, all risk predictions and warnings are recorded as control and management records.

4. The method for dynamic risk management of hazardous chemicals as described in claim 3, characterized in that, The detection information includes environmental parameters and environmental images. The risk logic model and risk prediction model calculate the first risk probability based on the environmental parameters and state information, and the image recognition model calculates the second risk probability based on the environmental images.

5. The method for dynamic risk management of hazardous chemicals as described in claim 4, characterized in that, Methods for constructing risk logic models include: Identify the potential risks that hazardous chemicals may cause and designate these risks as top events. Based on empirical data, risk factors that can trigger risk events are identified, and these risk factors are treated as intermediate events. Logical inversion is performed on the risk factors using logical AND, logical OR, and logical NOT to determine the first correlation relationship, and the detection information is treated as the basic event based on the first correlation relationship. Combine top events, intermediate events, and basic events into a risk logic model.

6. The method for dynamic risk management of hazardous chemicals as described in claim 5, characterized in that, In the construction method of the risk logic model, after determining the top event, intermediate events and basic events, the minimum cut set corresponding to the top event is determined. The minimum cut set includes at least one intermediate event and at least one basic event. When generating control strategies, several alternative strategies are determined based on the risk failure model, and one of the alternative strategies is determined as the optimal strategy based on the minimum cut set.

7. The method for dynamic risk management of hazardous chemicals as described in claim 6, characterized in that, In the construction method of the risk logic model, after determining the top event, intermediate events and basic events, the importance of the basic events is evaluated based on the results of logical inversion. When determining alternative strategies, alternative strategies are determined based on the importance of the basic events.

8. The method for dynamic risk management of hazardous chemicals as described in claim 3, characterized in that, The risk prediction model is set as a Bayesian network model, and the basic events in the risk logic model are mapped to the root nodes of the Bayesian network model based on the second correlation.

9. The method for dynamic risk management of hazardous chemicals as described in claim 7, characterized in that, It also includes the following steps: updating and optimizing the risk logic model and risk prediction model based on the generated first risk probability; Methods for optimizing risk prediction models include: The generated posterior probability and the current likelihood estimate generated based on the probe information are used to calculate a new posterior probability, and the parameters of the Bayesian network model for generating the posterior probability are optimized.

10. The method for dynamic risk management of hazardous chemicals as described in claim 3, characterized in that, The image recognition model is set as a convolutional neural network model. Methods for calculating the second risk probability using the image recognition model include: Environmental images are input into the image recognition model; Extract spectral features from environmental images and input these features into the layer structure of the image recognition model; The second risk probability is obtained by using an image recognition model to extract features and classify environmental images.

Images