A risk level-based dynamic classification management and control method and system for chemicals

Abstract

The application discloses a chemical dynamic classification management and control method and system based on risk levels. The method comprises the following steps: collecting inherent attribute data, real-time use scene data and environmental state data of a target chemical; calculating the current dynamic risk index, reaction evolution path and predicted risk index in a future preset period of the target chemical by using the preprocessed data; dividing the target chemical into a corresponding risk level according to the dynamic risk index and the predicted risk index; matching the management and control strategy based on the risk level and the reaction evolution path, and simultaneously monitoring the state changes of the real-time use scene data and the environmental state data in real time to trigger the dynamic adjustment of the risk level and the update of the dynamic classification label. The application realizes the spatio-temporal dynamic evaluation and forward-looking prediction of chemical risks, eliminates the response lag and resource mismatch problems caused by traditional static management and control, and significantly improves the safety and refinement level of chemical life cycle management and control.

Description

Technical Field

[0001] This application relates to the field of industrial safety and intelligent management and control technology of hazardous chemicals, and in particular to a method and system for dynamic classification and management of chemicals based on risk level. Background Technology

[0002] With the rapid development of modern chemical and manufacturing industries, the scale and variety of hazardous chemicals in production, storage, transportation, and use are increasing daily. The inherent flammability, explosiveness, toxicity, and corrosiveness of chemicals make them highly susceptible to serious safety accidents under uncontrolled conditions. Therefore, the scientific and precise risk classification and control of chemicals is of paramount importance in industrial safety management.

[0003] Traditional chemical risk classification primarily relies on Material Safety Data Sheets (MSDS / SDS) or the Globally Harmonized Classification and Labelling System (GHS) standards. This classification method is essentially a static classification based on the inherent properties of chemicals (such as flash point, explosion limits, and acute toxicity). However, in real-world industrial scenarios, the true risk of a chemical depends not only on its inherent physicochemical properties but also heavily on real-time usage scenarios (such as usage volume, operator qualifications, and process parameters) and environmental conditions (such as temperature, humidity, ventilation conditions, operating status of surrounding equipment, and personnel density). Traditional static classification methods fail to incorporate this real-time data for dynamic assessment, resulting in risk ratings that significantly lag behind actual hazard levels.

[0004] Furthermore, to address the limitations of static classification, existing technologies have introduced some monitoring systems based on IoT sensors. However, these systems have significant shortcomings: First, existing systems often treat various sensor data in isolation, failing to construct a unified spatiotemporal evolution state matrix from real-time usage scenario data and environmental state data, thus failing to reveal the coupling effects of risks in the spatiotemporal dimension. Second, in risk analysis, existing systems fail to combine the inherent property data of chemicals with dynamic spatiotemporal data for mechanism- and data-driven modeling, and cannot simultaneously calculate the current dynamic risk index, reaction evolution path, and predicted risk index for future periods. This results in the system lacking both an understanding of the intrinsic physicochemical evolution mechanism and the ability to predict future risk trends. Third, existing control strategies are mostly single threshold alarms, failing to combine risk levels and reaction evolution paths for differentiated matching, and unable to trigger dynamic adjustments to risk levels and control strategies based on real-time state changes.

[0005] Therefore, there is an urgent need for a management method that can integrate multi-source spatiotemporal data, calculate dynamic risks and predict risks based on a dual-driven model of mechanism and data, and dynamically match management strategies according to the reaction evolution path and risk level, and can be continuously updated dynamically. Summary of the Invention

[0006] This application provides a method and system for dynamic classification and control of chemicals based on risk levels, in order to solve the problems of static chemical risk classification, extensive control strategies, and delayed response in the prior art.

[0007] To solve the above-mentioned technical problems, one technical solution adopted in this application is: to provide a method for dynamic classification and control of chemicals based on risk levels. This method for dynamic classification and control of chemicals includes: S1: Collect data on the inherent properties of the target chemical, real-time usage scenario data, and environmental status data, and preprocess the collected data; S2: Construct a spatiotemporal evolution state matrix using the preprocessed real-time usage scenario data and environmental state data. Based on the preset mechanism and data dual-drive model, analyze the spatiotemporal evolution state matrix in combination with the inherent attribute data, and calculate the current dynamic risk index, reaction evolution path, and predicted risk index of the target chemical in the future preset period. S3: Based on the dynamic risk index and the predicted risk index, classify the target chemical into the corresponding risk level and generate a dynamic classification label containing a timestamp; S4: Based on the risk level and the reaction evolution path, match the control strategy and issue the control strategy and the dynamic classification label. At the same time, monitor the status changes of the real-time usage scenario data and the environmental status data in real time to trigger the dynamic adjustment of the risk level and the update of the dynamic classification label.

[0008] In some embodiments, the specific method for calculating the dynamic risk index, reaction evolution path, and predicted risk index based on the preset mechanism and data dual-driven model in step S2 is as follows: Based on the inherent attribute data, the theoretical hazard baseline value and the reaction evolution path are calculated using a physicochemical mechanism model; Based on the spatiotemporal evolution state matrix, the spatiotemporal disturbance correction coefficient is analyzed and calculated using a data-driven model; The theoretical risk benchmark value is fused with the spatiotemporal disturbance correction coefficient to obtain the current dynamic risk index; The spatiotemporal evolution state matrix is ​​input into the time series prediction network, and combined with the reaction evolution path, the evolution trend of environmental state data within a preset time period is deduced, and the predicted risk index is calculated.

[0009] In some embodiments, step S2 further includes a dynamic energy overlap verification step: When two or more target chemicals are detected in the same physical space, the energy field distribution of each chemical reaction exothermic or gas diffusion is calculated based on the reaction evolution path and usage of each chemical. If the energy fields overlap in time and space, a penalty coefficient is introduced to the calculated dynamic risk index and predicted risk index for a step-up adjustment.

[0010] In some embodiments, the risk level in step S3 includes four levels: low risk level, medium risk level, high risk level and extremely high risk level. The classification is based not only on the threshold range of the current dynamic risk index, but also on the probability of the predicted risk index exceeding the threshold of the next higher level. When the probability exceeds the preset confidence level, the target chemical is classified to the next higher level in advance.

[0011] In some embodiments, the control strategy includes a hierarchical requisition and approval strategy, which includes: The approval process is accelerated according to the risk level. When a target chemical is classified as high-risk or extremely high-risk, a dual online and offline review and approval mechanism is triggered. When the device is issued, the spatiotemporal identifier code in the dynamic classification label is read to verify whether the current environmental status data meets the issuance conditions. If it does not meet the conditions, the issuance operation is rejected.

[0012] In some embodiments, the control strategy includes a global resource game-theoretic scheduling strategy, which is used in scenarios where multiple chemicals coexist and emergency resources are limited. A game-theoretic scheduling model is constructed with the goal of minimizing the global dynamic risk index and the available emergency resource capacity in the real-time enterprise resource data as a constraint. This model allocates the optimal ventilation volume, spray cooling water volume, and emergency personnel intervention sequence to each target chemical.

[0013] In some embodiments, the control policy includes a storage isolation policy, which includes: Based on the risk level of the target chemical and the energy field distribution derived from the spatiotemporal evolution state matrix, the safe isolation distance of the chemical within the storage area is dynamically adjusted. When the spatial distance between a target chemical that has an incompatible relationship and another target chemical is less than the safe isolation distance, a spatial rearrangement warning is triggered.

[0014] In some embodiments, the control strategy includes a fire alarm linkage strategy, which includes: Based on the risk level and the response evolution path, match the pre-stored special response plans; When the dynamic risk index reaches the threshold of extremely high risk level, the emergency response plan, which includes the physicochemical properties of the target chemical and the reaction evolution path, will be automatically sent to the regional fire control system.

[0015] In some embodiments, the control strategy includes a ventilation interlock strategy, which includes: Based on the real-time temperature, humidity and gas concentration in the environmental status data, and combined with the predicted risk index, the operating frequency of ventilation equipment in the area is dynamically adjusted. When the predicted risk index approaches the threshold boundary, the ventilation equipment should be upgraded to the preset maximum power in advance.

[0016] To address the aforementioned technical problems, another technical solution adopted in this application is to provide a risk-level-based dynamic classification and control system for chemicals. This risk-level-based dynamic classification and control system is used to implement the method described above. The system includes: a data acquisition module for collecting inherent attribute data, real-time usage scenario data, and environmental status data of the target chemical; a risk assessment and prediction module for constructing a spatiotemporal evolution state matrix using preprocessed real-time usage scenario data and environmental status data, calculating the current dynamic risk index, reaction evolution path, and future predicted risk index based on a mechanism- and data-driven dual-drive model, combined with the inherent attribute data; a dynamic classification module for classifying chemicals into corresponding risk levels according to the dynamic risk index and predicted risk index, and generating dynamic classification labels; and a control execution module for matching control strategies based on the risk level and the reaction evolution path, issuing control strategies and dynamic classification labels, and monitoring the state changes of the real-time usage scenario data and the environmental status data to trigger dynamic adjustments and label updates.

[0017] The beneficial effect of this application is that, unlike the prior art, this application discloses a method and system for dynamic classification and control of chemicals based on risk level. This application's embodiments construct a spatiotemporal evolution state matrix, uniformly mapping isolated multi-source scene data and environmental data onto a spatiotemporal coordinate system, revealing the coupling effect of risk in the temporal and spatial dimensions. Through a pre-defined mechanism and data-driven dual-drive model, combined with inherent attribute data analysis, it not only calculates a dynamic risk index reflecting the current real danger state but also simultaneously deduces the reaction evolution path reflecting the logic of inherent physicochemical changes, and further calculates a predicted risk index for a pre-defined future period, achieving a deep integration of mechanism prior knowledge and data-driven prediction. Based on this, chemicals are classified into corresponding risk levels according to the dynamic risk index and predicted risk index, and dynamic classification labels are generated. Subsequently, control strategies are matched based on risk levels and reaction evolution paths, enabling control measures to be accurately matched to different reaction mechanisms and risk levels. Finally, by monitoring the state changes of scene data and environmental state data in real time, dynamic adjustments to risk levels and updates to dynamic classification labels are triggered, ensuring the real-time tracking of risk rating and control strategies, and realizing a closed loop of spatiotemporal dynamic assessment, forward-looking prediction, and refined control of chemical risks. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort, wherein: Figure 1 This is a flowchart illustrating an embodiment of the dynamic classification and control method for chemicals based on risk levels provided in this application. Figure 2 yes Figure 1 A flowchart illustrating an embodiment of step S2 in the dynamic classification and control method for chemicals shown below; Figure 3 This is a schematic diagram of an embodiment of the dynamic classification and control system for chemicals based on risk levels provided in this application. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0020] The terms "first," "second," and "third" used in the embodiments of this application are for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first," "second," or "third" may explicitly or implicitly include at least one of that feature. In the description of this application, "multiple" means at least two, such as two, three, etc., unless otherwise explicitly specified. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but may optionally include steps or units not listed, or may optionally include other steps or units inherent to these processes, methods, products, or devices.

[0021] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a mutually exclusive, independent, or alternative embodiment. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0022] To address the problems of static chemical risk classification, inability to reflect spatiotemporal coupling effects, and lack of forward-looking control in existing technologies, this application provides a dynamic chemical classification and control scheme based on risk levels. This scheme constructs a spatiotemporal evolution state matrix and introduces a dual-driven model of mechanism and data to achieve accurate calculation of dynamic and predicted risk indices. Furthermore, it achieves refined dynamic control throughout the entire process by dynamically matching control strategies based on reaction evolution paths and risk levels.

[0023] Specifically, see Figure 1 , Figure 1 This is a flowchart illustrating an embodiment of the dynamic classification and control method for chemicals based on risk levels provided in this application. The dynamic classification and control method for chemicals includes: S1: Collect data on the inherent properties of the target chemical, real-time usage scenario data, and environmental status data, and preprocess the collected data.

[0024] This step aims to comprehensively acquire various static and dynamic parameters affecting chemical risks through a multi-source data acquisition network, providing a data foundation for subsequent spatiotemporal evolution state matrix construction and risk calculation.

[0025] Specifically, the inherent property data of a target chemical refers to the physicochemical and toxicological parameters that are inherent to the chemical itself and do not change with the external environment.

[0026] In this embodiment, inherent property data includes, but is not limited to, the physicochemical properties of chemicals (such as flash point, ignition point, boiling point, melting point, density, vapor pressure), toxicological parameters (such as LD50, LC50, maximum allowable concentration (MAC), flammability and explosion characteristics (such as explosion limits, minimum ignition energy, heat of combustion), and reactivity parameters (such as adiabatic temperature rise, exothermic reaction rate, autoaccelerating decomposition temperature (SADT), and incompatible materials list). This data is typically stored in the chemical ledger database of the enterprise's EHS (Environmental, Health and Safety) management system or in the material master data of the enterprise's ERP system; it can be collected periodically from the ERP system via API interface or obtained by importing the parsed results of the Safety Data Sheet (SDS).

[0027] Real-time usage scenario data refers to the actual operational status data of chemicals in the current production, storage, or transfer process. This data reflects the dynamic disturbances to risk caused by human factors and process conditions. In this embodiment, real-time usage scenario data includes, but is not limited to, usage volume, usage frequency, operator qualification level, and process node parameters. Usage volume and frequency can be obtained through MES (Manufacturing Execution System) or WMS (Warehouse Management System), for example, the current batch input is 500 kg, and the average daily usage frequency is 3 times. Operator qualification level can be obtained by comparing the operator's employee number with a qualification database through an access control system or personnel positioning system, such as whether the operator possesses a special operation certificate for handling high-risk chemicals. Process node parameters reflect the process stage of the chemical, such as "reactor heating stage," "centrifugal separation stage," or "static storage stage," and these parameters can be provided by DCS (Distributed Control System).

[0028] Environmental status data refers to the real-time physical quantities and status parameters of the physical space where chemicals are located and their surrounding environment. Environmental status data directly reflects the spatiotemporal disturbances caused by the external environment to risks. In this embodiment, environmental status data includes, but is not limited to, real-time temperature and humidity data, ventilation status data, operating status data of surrounding equipment, and personnel density data of the chemical storage or usage area. Real-time temperature and humidity data can be collected by temperature and humidity sensors deployed within the area; ventilation status data includes the start / stop status, speed, or number of air changes of the ventilation fan, which can be obtained by reading the operating parameters of the HVAC system; operating status data of surrounding equipment includes whether nearby equipment is energized, whether there are open flames or high-temperature surfaces, etc., which can be obtained through a SCADA system; personnel density data within the area can be calculated by combining the area area with video recognition equipment (such as AI cameras), or by counting the number of UWB positioning tags within the area.

[0029] In this embodiment, preprocessing the collected data is a crucial step in ensuring the quality and usability of multi-source heterogeneous data. Because inherent attribute data, real-time usage scenario data, and environmental state data originate from different business systems and heterogeneous sensor networks, their data formats, sampling frequencies, communication protocols, and spatiotemporal references all differ significantly. For example, inherent attribute data consists of static relational data entries; usage scenario data is event-driven data with a low frequency, such as one entry per minute; while environmental state data, such as temperature and humidity, is time-series data with a high frequency, such as one entry per second.

[0030] Preprocessing mainly includes data cleaning, missing value imputation, and spatiotemporal alignment to transform multi-source heterogeneous data into standard format data with unified timestamps and spatial coordinates.

[0031] The specific data cleaning process includes removing outliers from sensor-collected data that are outside the physically reasonable range, such as temperature sensor readings of -200℃ or 999℃, and filtering out high-frequency noise, such as using moving average filtering or Kalman filtering to process random fluctuations in temperature and humidity data.

[0032] Missing value imputation compensates for data loss caused by network jitter or sensor failure. For short-term missing environmental status data, linear interpolation or the mean of historical data from the same period can be used to fill in the missing data. For long-term missing scene data, the previous valid value can be filled in, or a data acquisition anomaly warning can be triggered to prompt manual intervention.

[0033] Spatiotemporal alignment processing involves aligning low-frequency scene data and inherent attribute data forward based on the most recent timestamp, using the sampling period of high-frequency environmental data or a globally unified period preset by the system as the baseline time axis. Spatially, it uses the absolute coordinates of each sensor and device in the factory's three-dimensional coordinate system to assign spatial labels to the data. Through spatiotemporal alignment, each set of data possesses both temporal and spatial attributes, forming standardized data formats.

[0034] S2: Construct a spatiotemporal evolution state matrix using preprocessed real-time usage scenario data and environmental state data. Based on a pre-defined mechanism and data dual-drive model, analyze the spatiotemporal evolution state matrix in conjunction with inherent attribute data to calculate the current dynamic risk index, reaction evolution path, and predicted risk index for the target chemical within a pre-defined future time period.

[0035] By constructing a matrix and using a dual-drive model for analysis, static attributes, dynamic scenarios, and spatiotemporal environment are deeply integrated to output a series of risk quantification indicators that reflect the current state and future trends.

[0036] The method for constructing the spatiotemporal evolution state matrix includes: using the three-dimensional coordinates determined by the UWB location data of the target chemical as the spatial axis and the time series as the time axis, mapping the environmental state data and usage scenario data at different spatial locations to the spatiotemporal coordinate system to form a matrix characterizing the risk propagation path and spatiotemporal coupling effect.

[0037] Specifically, UWB (Ultra-Wideband) positioning base stations are deployed within the factory area, and UWB positioning tags are installed on the containers or transport carriers of the target chemicals. These tags can acquire the coordinates of the target chemicals in the three-dimensional space of the factory area in real time with centimeter-level accuracy. Using these three-dimensional coordinates as the origin or center, a spatial neighborhood is defined, such as a 10m × 10m × 5m cube, and the spatial positions of each sensor node and device within this range are used as the spatial dimensions of the matrix.

[0038] In terms of time dimension, the time series of the past T period is selected as the time axis, such as the past 30 minutes, and the time step can be set to 1 minute according to actual needs.

[0039] Environmental state data and usage scenario data at different spatial locations are mapped to this spatiotemporal coordinate system. For example, if a temperature sensor reading at a spatial location is 30°C at time t, this reading is used as an element value of the matrix at that spatiotemporal node. Through this mapping, a high-dimensional tensor containing spatial coordinates, timestamps, and multi-dimensional data features is constructed—that is, a spatiotemporal evolution state matrix. This matrix not only records the instantaneous state of the environment surrounding the chemicals at the current moment but also records the trend of environmental state evolution over time and the spatial correlation between data from different spatial locations, thus comprehensively characterizing the risk propagation path and spatiotemporal coupling effect.

[0040] For example, if a gas diffusion model shows that a certain harmful gas is diffusing from spatial point A to spatial point B where the chemical is located, then in the spatiotemporal evolution state matrix, the gas concentration at point A and the gas concentration at point B will show a time-delayed correlation change.

[0041] In this embodiment, the calculation of dynamic risk index, reaction evolution path and predicted risk index relies on a preset mechanism and data dual-drive model. The dual-drive model combines prior knowledge of physicochemical mechanisms with data-driven pattern recognition capabilities, making risk calculation both interpretable and scientific, as well as having the ability to adaptively fit complex environments.

[0042] See Figure 2 The specific methods for calculating the dynamic risk index, response evolution path, and predicted risk index based on a pre-defined mechanism and data-driven dual-drive model include: S21: Based on the inherent attribute data, calculate the theoretical hazard baseline value and reaction evolution path through a preset physicochemical mechanism model.

[0043] The physicochemical mechanism model is a set of mathematical equations based on the thermodynamics, kinetics and mass-energy conservation principles of chemicals. The theoretical hazard baseline value is the basic risk level calculated based on the inherent properties of chemicals under standard conditions.

[0044] For example, by combining flash point, explosion limits, toxicity parameters, etc., a dimensionless baseline value can be calculated using the DOW fire explosion index method or the ICI Mond method. This baseline value only reflects the absolute hazard level of the chemical itself.

[0045] The reaction evolution path is a trajectory derived from the chemical's reactivity parameters and a list of incompatible substances. For example, if the target chemical is hydrogen peroxide, its reaction evolution path might include nodes such as "slow decomposition at room temperature → violent catalytic decomposition upon heating → release of oxygen and heat → initiation of combustion of combustibles." The reaction evolution path, in the form of a directed graph or state machine, represents the logical trajectory and triggering conditions of a chemical's evolution from a stable state to an uncontrolled state.

[0046] S22: Based on the spatiotemporal evolution state matrix, analyze and calculate the spatiotemporal disturbance correction coefficient through a preset data-driven model.

[0047] Spatiotemporal perturbation correction coefficients are used to quantify the amplification or reduction effect of real-time usage scenarios and environmental state changes on theoretical hazard baseline values. Data-driven models can employ spatiotemporal graph convolutional networks or long short-term memory networks (LSTM). The spatiotemporal evolution state matrix is ​​input into the data-driven model, which extracts the deterioration trend of environmental parameters over time from the matrix, such as the rate of continuous temperature rise, and spatial coupling characteristics, such as the heat source superposition caused by the simultaneous operation of surrounding equipment, and outputs a correction coefficient greater than or equal to 1.

[0048] If the environment is safe and the scenario is normal, the correction coefficient is close to 1; if the environment is harsh or there is overlapping energy fields, the correction coefficient will increase significantly. For example, when the ambient temperature rises from 25℃ to 45℃, the model, based on the weight distribution trained on historical data, outputs a correction coefficient of 1.5, indicating that the risk has been amplified by 50%.

[0049] S23: The theoretical risk benchmark value is fused with the spatiotemporal disturbance correction coefficient to obtain the current dynamic risk index.

[0050] The fusion method can employ simple multiplication operations or a weighted fusion algorithm based on attention mechanisms to more reasonably balance the influence of inherent hazards and external disturbances. The resulting dynamic risk index is a quantitative indicator that comprehensively reflects the true degree of danger of the target chemical in a specific spatiotemporal environment at the current moment.

[0051] S24: Input the spatiotemporal evolution state matrix into the time series prediction network, combine it with the reaction evolution path, deduce the evolution trend of environmental state data within the future preset time period, and calculate the predicted risk index.

[0052] Temporal prediction networks can be based on Transformer architecture or Temporal Convolutional Network (TCN) models. The spatiotemporal evolution state matrix of past periods is input into the prediction network, which outputs a sequence of predicted environmental state data (such as temperature and gas concentration) for a preset future period. Based on this predicted environmental state data sequence, the aforementioned data-driven model is used again to calculate future spatiotemporal perturbation correction coefficients. Combined with theoretical hazard benchmarks, a future dynamic risk index sequence is calculated, and the maximum or final value in this sequence is taken as the predicted risk index. Simultaneously, by considering the reaction evolution path, it is determined whether future environmental conditions will trigger the next node of reaction evolution, such as when the temperature reaches the decomposition threshold. Therefore, the predicted risk index not only includes numerical magnitude but also implicitly predicts possible future chemical reaction evolution stages.

[0053] Furthermore, in this embodiment, in order to accurately assess the spatiotemporal coupling risk in a multi-chemical coexistence scenario, step S2 also includes a dynamic energy overlap verification step, the specific steps of which are as follows: When two or more target chemicals are detected in the same physical space, the energy field distribution of each chemical's reaction exothermic or gas diffusion is calculated based on the reaction evolution path and usage of each chemical. If the energy fields of various chemicals overlap in space and time, a penalty coefficient is introduced to adjust the calculated dynamic risk index and predicted risk index in a stepwise manner.

[0054] In chemical industrial parks or production workshops, multiple chemical storage tanks or reaction vessels are often clustered within the same physical area. In such cases, the dynamic risk index of a single chemical cannot fully reflect the overall risk of the area, as there may be superposition and coupling of energy fields or toxic gas diffusion fields between chemicals. The system uses UWB positioning data and spatial coordinate sets to determine the spatial proximity of multiple target chemicals in real time. When it is determined that two or more target chemicals are in a preset proximity space, such as a distance of less than 10 meters, a dynamic energy overlap check is triggered.

[0055] Based on the reaction evolution path of each chemical (such as exothermic reaction path, toxic gas release path) and the usage in the current real-time usage scenario data, the system uses fluid dynamics or thermodynamics models to calculate the energy field distribution formed by each chemical in space.

[0056] For example, for chemical A undergoing an exothermic reaction, the temperature gradient field it forms in space is calculated based on its exothermic reaction rate and the amount used; for volatile and toxic chemical B, the lethal concentration distribution field it forms in space is calculated based on its volatilization rate and gas diffusion model. Each energy field is a set of three-dimensional voxel data that decays outwards from its three-dimensional coordinates.

[0057] The system determines whether spatiotemporal overlap of energy fields occurs by calculating the intersection volume of each energy field in three-dimensional space. If overlap exists, it means that the heat or toxic gas concentration will have a cumulative effect within this spatial region, potentially leading to extreme events beyond the theoretical evolution path of a single chemical, such as cascading thermal runaway or mixed gas explosion. In this case, the system introduces a penalty coefficient to progressively increase the dynamic risk index and predicted risk index calculated for each target chemical involved.

[0058] The penalty coefficient can be set based on the overlap volume ratio and the incompatible relationship database. For example, if chemicals A and B are incompatible and the overlap volume of their energy fields exceeds 20%, the penalty coefficient is set to 1.5, which means the dynamic risk index is directly amplified by 50%. This mechanism effectively compensates for the deficiency in considering the spatiotemporal coupling and superposition risks when assessing a single chemical independently, and greatly improves the accuracy of risk quantification under complex operating conditions.

[0059] S3: Based on the dynamic risk index and the predicted risk index, classify the target chemicals into the corresponding risk levels and generate dynamic classification labels containing timestamps.

[0060] After completing the quantitative assessment of current and future risks, the continuous risk index is transformed into a discrete risk level that is easy to manage and execute, and a dynamic label that can be machine-readable and transferred is generated.

[0061] In this embodiment, the risk level includes four levels: low risk, medium risk, high risk, and extremely high risk. Traditional risk level classification often relies solely on static classification based on the current risk value. However, this application not only considers the threshold range of the current dynamic risk index but also the probability of the predicted risk index exceeding the threshold of the next higher level. When the probability exceeds a preset confidence level, the target chemical is preemptively classified into the next higher level.

[0062] Specifically, the system pre-defines dynamic risk index threshold ranges for each level. For example, 0-30 represents low risk, 30-60 represents medium risk, 60-85 represents high risk, and above 85 represents extremely high risk. When classifying risk, if the current dynamic risk index is 55, it should fall under medium risk. However, if the prediction network calculates that the probability of the risk index reaching 65 within the next hour is 95% due to the continued rise in ambient temperature, exceeding the pre-set confidence level of 90%, the system will not wait for the risk to occur but will immediately classify the target chemical as high-risk. This proactive transition mechanism provides valuable advance intervention time for subsequent control strategies.

[0063] After classifying the risk levels, the step of generating dynamic classification labels in step S3 includes: encoding the risk level, current dynamic risk index, predicted risk index and reaction evolution path into machine-readable spatiotemporal identifiers, and binding the identifiers to the physical container or digital twin model of the target chemical.

[0064] Dynamic classification labels contain rich, multi-dimensional information, encoded using lightweight data exchange formats such as JSON or XML to form machine-readable spatiotemporal identifiers. These identifiers not only include the risk level (e.g., "HIGH_RISK"), but also embed the current dynamic risk index (e.g., "62.5"), the predicted risk index (e.g., "71.0"), the key node states of the evolution path (e.g., "THERMAL_RUNAWAY_INITIATED"), and precise timestamps and spatial coordinates.

[0065] The spatiotemporal identification code can be written into the electronic tag on the physical container of the chemical through RFID (Radio Frequency Identification), so that any workstation equipped with a reading device can obtain the latest risk status of the chemical in real time during the flow of the chemical; at the same time, the identification code is also bound to the corresponding chemical 3D model in the digital twin model of the plant area and presented in a visual form on the monitoring screen.

[0066] S4: Match control strategies based on risk levels and response evolution paths, and issue control strategies and dynamic classification labels. At the same time, monitor the status changes of real-time usage scenario data and environmental status data to trigger dynamic adjustments to risk levels and updates to dynamic classification labels.

[0067] This step aims to automatically match and execute targeted control strategies based on the multidimensional risk characteristics calculated above, while maintaining closed-loop monitoring to achieve dynamic updates.

[0068] In this embodiment, the matching of control strategies not only depends on the risk level, but also introduces the reaction evolution path as the matching basis. This is because the same risk index may correspond to completely different runaway mechanisms, and the required control measures are also completely different.

[0069] For example, both are extremely high risks, but those caused by violent exothermic reactions require immediate spray cooling, while those caused by the release of toxic gases require immediate closure of the sealing valve and activation of negative pressure ventilation.

[0070] The control strategies in step S4 include at least one of the following: storage isolation strategy, hierarchical requisition approval strategy, usage limit strategy, fire alarm linkage strategy, ventilation interlock strategy, and global resource game scheduling strategy. The system calls upon at least one corresponding strategy from the strategy library and executes it in parallel based on the combination of risk level and response evolution path.

[0071] Specifically, the implementation methods for each control strategy are as follows: The storage isolation strategy dynamically adjusts the safe isolation distance of a target chemical within the storage area based on its risk level and the energy field distribution derived from the spatiotemporal evolution state matrix. When the spatial distance between the target chemical and other target chemicals with incompatible relationships is lower than the safe isolation distance, a spatial rearrangement warning is triggered.

[0072] Traditional isolation distances are usually fixed values, such as at least 2 meters between all flammable materials. However, in real-world scenarios, changes in ambient temperature and ventilation conditions can alter the risk diffusion range. This embodiment uses real-time energy field distribution derived from a spatiotemporal evolution state matrix, such as the thermal radiation range or gas diffusion range, combined with the risk level, to dynamically calculate an adaptive safe isolation distance.

[0073] For example, in high-temperature and high-ventilation environments, the safe isolation distance may be increased from 2 meters to 4 meters. The system continuously calculates the real-time spatial distance between the chemical and other surrounding chemicals, especially the real-time spatial distance with incompatible materials identified from the incompatible materials list based on inherent attribute data. Once the distance falls below the dynamically calculated safe isolation distance, a spatial rearrangement warning is immediately triggered, instructing AGVs (Automated Guided Vehicles) or warehouse personnel to transfer the relevant chemicals to a safe area.

[0074] The approval hierarchy strategy for chemical requisition increases the number of approval nodes according to the risk level. When the target chemical is classified as high-risk or extremely high-risk, a dual online and offline review and approval mechanism is triggered. When requisitioning, the system reads the spatiotemporal identification code in the dynamic classification label to verify whether the current environmental status data meets the requisition conditions. If not, the requisition operation is rejected.

[0075] As the risk level dynamically changes, the approval process for requisitioning the chemical automatically adjusts. At low risk, only online approval from the team leader is required. Once the risk level jumps to high or extremely high, the system automatically freezes the regular approval process, forcibly triggering a dual online and offline review mechanism: online approval requires joint signatures from the safety director and plant manager; offline approval requires the on-site safety officer to scan the spatiotemporal identification code on the chemical container using a mobile terminal. The system backend compares the reaction evolution path in the identification code with the current environmental status data to verify whether the current environment (such as temperature, humidity, and ventilation) meets the safe requisition conditions for that reaction path. If the environmental status data exceeds the safety threshold, even if the approval is granted, the system issues a locking command to the smart lock, rejecting the requisition operation and preventing the dangerous flow out at the source.

[0076] A global resource game scheduling strategy is used in scenarios where multiple chemicals coexist and emergency resources are limited: a game scheduling model is constructed with the goal of minimizing the global dynamic risk index and the available emergency resource capacity in the real-time enterprise resource data as a constraint, to allocate the optimal ventilation volume, spray cooling water volume and emergency personnel intervention order to each target chemical.

[0077] When multiple risk events occur simultaneously within a factory area, such as two areas experiencing high risk at the same time, and emergency resources are limited, such as total cooling water flow, total ventilation fan power, and number of emergency personnel, traditional average allocation or first-come-first-served strategies often result in the inability to effectively control the overall risk.

[0078] This embodiment introduces a game-theoretic scheduling model, treating each target chemical as a player in the game. The objective function is to minimize the sum of the global dynamic risk indices, and the constraints are the available emergency resource capacity in the enterprise resource data. By solving the Nash equilibrium or using heuristic algorithms, an optimal resource allocation scheme is output.

[0079] For example, for chemical A, which reacts with intense exothermic energy and is close to thermal runaway, 80% of the cooling water volume is allocated; while for chemical B, whose risk increases more slowly, only 20% of the water volume is allocated and emergency personnel are prioritized to intervene in the process of B, thereby maximizing the overall safety benefits under resource constraints.

[0080] The usage restriction strategy dynamically adjusts the single-use limit and the cumulative usage limit for target chemicals based on the dynamic risk index and the predicted risk index; when the predicted risk index indicates a future risk shift, the single-use limit and the cumulative usage limit are automatically lowered.

[0081] Chemical usage quotas in the MES system are no longer static values. When the dynamic risk index rises, or when the predicted risk index indicates a risk shift due to environmental degradation within a future time period T, the system automatically sends instructions to the MES system to lower the single-use limit and the cumulative usage limit for that chemical (e.g., daily / weekly). For example, the single-use limit is lowered from 500 kg to 200 kg. This reduces the potential scale of reactions from a total quantity control perspective, lowering the overall risk level of the system.

[0082] The fire-fighting linkage strategy combines risk levels and reaction evolution paths with pre-stored special response plans; when the dynamic risk index reaches the threshold of extremely high risk level, the response plan, which includes the physicochemical properties of the target chemical and its reaction evolution path, will be automatically sent to the regional fire control system.

[0083] Different chemical fires or leaks require different extinguishing agents or treatment materials; for example, water-repellent substances cannot be extinguished with water. This strategy combines risk level and reaction evolution path to accurately match specific response plans from a pre-stored plan library. When the dynamic risk index reaches the extremely high risk level threshold, the system automatically sends the response plan, which includes the physicochemical properties and reaction evolution path of the target chemical, to the area fire control system (FAS) via OPC protocol or hardwiring. The FAS then automatically selects the corresponding foam, dry powder, or gaseous extinguishing agent for pre-setting and simultaneously closes the fire doors and smoke exhaust valves in the corresponding area, achieving second-level pre-fire preparation.

[0084] The ventilation interlock strategy dynamically adjusts the operating frequency of ventilation equipment in the area based on real-time temperature, humidity and gas concentration data in the environmental status data, combined with the predicted risk index; when the predicted risk index approaches the threshold boundary, the ventilation equipment is upgraded to the preset maximum power in advance.

[0085] The HVAC system no longer passively adjusts based solely on constant temperature and humidity setpoints, but instead employs proactive variable frequency control incorporating a predicted risk index. When the predicted risk index approaches a threshold boundary, indicating a potential risk due to gas accumulation or temperature rise, the system sends an instruction to the PLC control submodule in advance, directly increasing the ventilation equipment's operating speed from low to the preset maximum power. This increases the number of air exchanges, diluting the gas concentration or removing heat from the reaction, thus preventing the reaction from progressing to the next dangerous stage.

[0086] In addition to the six core control strategies mentioned above, step S4 can further introduce a proactive intervention mechanism. When the predicted risk index indicates that the target chemical will jump to an extremely high risk level within the next T period, a preparatory intervention instruction can be issued in advance at the current moment.

[0087] For example, if a predictive model indicates that a storage tank will reach its boiling point in 30 minutes, posing an extremely high risk, the system will immediately issue preemptive intervention instructions. These instructions may include reducing current usage limits, activating auxiliary ventilation equipment in advance, or sending frequency reduction requests to surrounding unrelated equipment to reduce regional power grid load and heat load. This proactive intervention provides a significant buffer to prevent accidents from occurring.

[0088] Finally, in order to maintain the closed loop and real-time tracking of the system, step S4 monitors the status changes of real-time usage scenario data and environmental status data in real time to trigger dynamic adjustment of risk level and update of dynamic classification labels.

[0089] The mechanisms for triggering dynamic adjustments include periodic triggering, event-driven triggering, and predictive triggering. Periodic triggering involves re-executing steps S1 to S3 at preset time intervals to keep the system status refreshed periodically. Event-driven triggering involves immediately reassessing the system when the magnitude / rate of change in real-time scene or environmental data exceeds a preset threshold, such as a sudden temperature rise of 5°C / min. Predictive triggering involves immediately reassessing the system when the predicted risk index deteriorates beyond the system's self-recovery capability threshold.

[0090] When the risk level of a target chemical jumps from low to high, the system automatically locks the current requisition operation and requires a re-execution of the control strategy before it can be unlocked, ensuring that the material is in an absolutely static and controlled state under high-risk conditions.

[0091] Through the cyclical execution of steps S1 to S4 above, the embodiments of this application realize the closed-loop management of chemical risks throughout the entire process, from data collection, spatiotemporal modeling, dual-drive analysis, forward-looking prediction, dynamic classification, and refined intervention, which completely changes the lag and extensiveness of traditional static management.

[0092] Based on the same inventive concept described above, this application also provides a chemical dynamic classification and control system 100 based on risk level, applied to the methods described in the above embodiments. See also... Figure 3 , Figure 3 This is a schematic diagram of an embodiment of the dynamic classification and control system for chemicals based on risk levels provided in this application.

[0093] The chemical dynamic classification and control system 100 includes a data acquisition module 10, a risk assessment and prediction module 20, a dynamic classification module 30, and a control execution module 40.

[0094] The data acquisition module 10 is used to collect inherent property data, real-time usage scenario data, and environmental status data of the target chemical. In this embodiment, the data acquisition module 10 includes an IoT sensing unit deployed on-site and a business interface unit for interfacing with the enterprise ERP system to obtain inherent property data and enterprise resource data. The IoT sensing unit includes temperature and humidity sensors, gas concentration sensors, video recognition devices, RFID readers, and UWB positioning devices. The business interface unit interfaces with industrial systems such as ERP, MES, and SCADA via API to obtain chemical ledgers, process parameters, and equipment operating status. The IoT sensor network supports edge computing. Before the data is uploaded to the cloud server, the edge nodes perform preprocessing, including preliminary filtering of the raw data, outlier removal, and spatiotemporal alignment, which significantly reduces the data processing pressure and network transmission latency in the cloud.

[0095] The risk assessment and prediction module 20 is communicatively connected to the data acquisition module 10. It constructs a spatiotemporal evolution state matrix using preprocessed real-time usage scenario data and environmental state data. Based on a mechanism- and data-driven model, it calculates the current dynamic risk index, reaction evolution path, and future predicted risk index by combining the inherent attribute data. This module incorporates a physicochemical mechanism model library and a time-series prediction network model. Furthermore, the risk assessment and prediction module 20 includes an energy overlap verification subunit, used to perform spatiotemporal overlap comparison of reaction energy fields in multi-chemical coexistence scenarios and introduce a penalty coefficient to adjust the risk index, accurately identifying superimposed risks.

[0096] The dynamic classification module 30 is connected to the risk assessment and prediction module 20. It is used to classify chemicals into corresponding risk levels based on dynamic risk indices and predicted risk indices, and generate dynamic classification labels. Internally, this module runs fuzzy logic or probabilistic statistical algorithms to perform risk level classification and forward-looking transition determination, and encodes the results into machine-readable spatiotemporal identifiers for distribution.

[0097] The control execution module 40 is connected to the dynamic classification module 30. It is used to match control strategies based on risk level and response evolution path, issue control strategies and dynamic classification labels, and monitor the status changes of real-time usage scenario data and environmental status data to trigger dynamic adjustments and label updates.

[0098] Specifically, the control and execution module 40 includes a PLC control submodule and a resource scheduling submodule that communicate directly with field equipment. Upon receiving high-risk or proactive intervention commands, it directly controls the start / stop of ventilation equipment, the opening / closing of solenoid valves, or executes global resource game-theoretic scheduling. Through the PLC control submodule, the system can directly issue low-level hardware control commands, such as increasing the inverter frequency, cutting off power, or opening sprinkler valves, ensuring that intervention actions are executed without delay. The resource scheduling submodule is responsible for running game-theoretic algorithms when multiple objectives conflict, achieving a rational allocation of global resources.

[0099] In this embodiment, the entire system can be deployed on electronic devices such as edge computing gateways or industrial control computers. These electronic devices possess multi-protocol parsing capabilities, such as Modbus, OPC UA, and MQTT, enabling direct data interaction with the DCS system at chemical storage or production sites.

[0100] Furthermore, to ensure the system's safety under extreme failure conditions, the electronic devices also include a hardware watchdog circuit. When the evaluation system software crashes or encounters a memory overflow that prevents the dynamic risk index from being refreshed, the hardware watchdog is triggered to reset, and the hardware safety output circuit is activated to issue the highest-level default safety control command, such as cutting off all feed valves and opening emergency ventilation, to ensure that the system tilts towards the safety side in a failure state.

[0101] Unlike existing technologies, this application discloses a method and system for dynamic classification and control of chemicals based on risk levels. This application constructs a spatiotemporal evolution state matrix, mapping isolated multi-source sensors and operational data to a unified spatiotemporal coordinate system, revealing the spatial propagation path and temporal coupling effect of risks. Through a dual-driven model of mechanism and data, it not only calculates theoretical hazard benchmarks and reaction evolution paths using a physicochemical mechanism model, clarifying the inherent physicochemical evolution logic of chemicals, but also uses a data-driven model to calculate spatiotemporal disturbance correction coefficients, correcting deviations caused by complex environments and scenario disturbances, and achieving accurate calculation of the dynamic risk index. Simultaneously, by using a time-series prediction network to extrapolate the evolution trend of future environmental state data, a predicted risk index is calculated, endowing the system with forward-looking early warning and intervention capabilities. Furthermore, this application classifies risk levels based on the dynamic risk index and the predicted risk index, and uses both the reaction evolution path and the risk level as the basis for matching control strategies, generating dynamic classification labels, achieving refined and differentiated control for different reaction mechanisms and risk levels. Finally, by triggering a dynamic update mechanism through real-time monitoring of state changes, the real-time tracking of risk ratings and control strategies is ensured, significantly improving the safety, refinement, and resource allocation efficiency of chemical control.

[0102] The above description is merely an embodiment of this application and does not limit the patent scope of this application. Any equivalent structural or procedural transformations made using the content of this application's specification and drawings, or direct or indirect applications in other related technical fields, are similarly included within the patent protection scope of this application.

Claims

1. A method for dynamic classification and control of chemicals based on risk levels, characterized in that, Includes the following steps: S1: Collect data on the inherent properties of the target chemical, real-time usage scenario data, and environmental status data, and preprocess the collected data; S2: Construct a spatiotemporal evolution state matrix using the preprocessed real-time usage scenario data and environmental state data. Based on the preset mechanism and data dual-drive model, analyze the spatiotemporal evolution state matrix in combination with the inherent attribute data, and calculate the current dynamic risk index, reaction evolution path, and predicted risk index of the target chemical in the future preset period. S3: Based on the dynamic risk index and the predicted risk index, classify the target chemical into the corresponding risk level and generate a dynamic classification label containing a timestamp; S4: Based on the risk level and the reaction evolution path, match the control strategy and issue the control strategy and the dynamic classification label. At the same time, monitor the status changes of the real-time usage scenario data and the environmental status data in real time to trigger the dynamic adjustment of the risk level and the update of the dynamic classification label.

2. The method for dynamic classification and control of chemicals according to claim 1, characterized in that, In step S2, the calculation of the dynamic risk index, reaction evolution path, and predicted risk index based on the preset mechanism and data dual-driven model includes: Based on the inherent attribute data, the theoretical hazard baseline value and the reaction evolution path are calculated using a physicochemical mechanism model; Based on the spatiotemporal evolution state matrix, the spatiotemporal disturbance correction coefficient is analyzed and calculated using a data-driven model; The theoretical risk benchmark value is fused with the spatiotemporal disturbance correction coefficient to obtain the current dynamic risk index; The spatiotemporal evolution state matrix is ​​input into the time series prediction network, and combined with the reaction evolution path, the evolution trend of environmental state data within a preset time period is deduced, and the predicted risk index is calculated.

3. The method for dynamic classification and control of chemicals according to claim 1, characterized in that, Step S2 also includes: When two or more target chemicals are detected in the same physical space, the energy field distribution of each target chemical's exothermic reaction or gas diffusion is calculated based on the reaction evolution path and usage of each target chemical. If the energy fields of the target chemicals overlap in space and time, a penalty coefficient is introduced to adjust the calculated dynamic risk index and predicted risk index in a stepwise manner.

4. The method for dynamic classification and control of chemicals according to claim 1, characterized in that, The risk levels mentioned in step S3 include four levels: low risk, medium risk, high risk, and extremely high risk. The classification is based not only on the threshold range of the current dynamic risk index, but also on the probability that the predicted risk index will exceed the threshold of the next higher level. When the probability exceeds the preset confidence level, the target chemical is classified to the next higher level in advance.

5. The method for dynamic classification and control of chemicals according to claim 1, characterized in that, The control strategy includes a hierarchical requisition and approval strategy, which includes: The approval process is accelerated according to the risk level. When a target chemical is classified as high-risk or extremely high-risk, a dual online and offline review and approval mechanism is triggered. When the device is issued, the spatiotemporal identifier code in the dynamic classification label is read to verify whether the current environmental status data meets the issuance conditions. If it does not meet the conditions, the issuance operation is rejected.

6. The method for dynamic classification and control of chemicals according to claim 1, characterized in that, The control strategy includes a global resource game-theoretic scheduling strategy, which is used in scenarios where multiple chemicals coexist and emergency resources are limited. A game-theoretic scheduling model is constructed with the goal of minimizing the global dynamic risk index and the available emergency resource capacity in the real-time enterprise resource data as a constraint. This model allocates the optimal ventilation volume, spray cooling water volume, and emergency personnel intervention sequence to each target chemical.

7. The method for dynamic classification and control of chemicals according to claim 1, characterized in that, The control strategy includes a storage isolation strategy, which includes: Based on the risk level of the target chemical and the energy field distribution derived from the spatiotemporal evolution state matrix, the safe isolation distance of the chemical within the storage area is dynamically adjusted. When the spatial distance between a target chemical that has an incompatible relationship and another target chemical is less than the safe isolation distance, a spatial rearrangement warning is triggered.

8. The method for dynamic classification and control of chemicals according to claim 1, characterized in that, The control strategy includes a fire-fighting linkage strategy, which includes: Based on the risk level and the response evolution path, match the pre-stored special response plans; When the dynamic risk index reaches the threshold of extremely high risk level, the emergency response plan, which includes the physicochemical properties of the target chemical and the reaction evolution path, will be automatically sent to the regional fire control system.

9. The method for dynamic classification and control of chemicals according to claim 1, characterized in that, The control strategy includes a ventilation interlock strategy, which includes: Based on the real-time temperature, humidity and gas concentration in the environmental status data, and combined with the predicted risk index, the operating frequency of ventilation equipment in the area is dynamically adjusted. When the predicted risk index approaches the threshold boundary, the ventilation equipment should be upgraded to the preset maximum power in advance.

10. A dynamic classification and control system for chemicals based on risk levels, used to implement the method as described in any one of claims 1-9, characterized in that, The dynamic classification and control system for chemicals includes: The data acquisition module is used to collect data on the inherent properties of the target chemical, real-time usage scenario data, and environmental status data. The risk assessment and prediction module is used to construct a spatiotemporal evolution state matrix using the preprocessed real-time usage scenario data and the environmental state data. Based on the mechanism and data dual-driven model, it calculates the current dynamic risk index, reaction evolution path and future predicted risk index by combining the inherent attribute data. The dynamic classification module is used to classify chemicals into corresponding risk levels based on dynamic risk indices and predicted risk indices, and generate dynamic classification labels. The control and execution module is used to match control strategies based on risk level and the reaction evolution path, issue control strategies and dynamic classification tags, and monitor the status changes of the real-time usage scenario data and the environmental status data to trigger dynamic adjustments and tag updates.

Images