A dangerous goods intelligent monitoring and emergency deduction method and system, and an electronic device

Abstract

The specification discloses a dangerous goods intelligent monitoring and emergency deduction method and system and electronic equipment, and relates to the technical field of dangerous goods storage and logistics safety monitoring. The method comprises the following steps: acquiring static basic attribute data and dynamic multi-source sensing data of dangerous goods, and fusing and encoding the data into structured state data according to the semantic structure of dangerous goods field knowledge; updating the state data time sequence in a preset period; inputting the time sequence into a risk assessment model to obtain a risk level determined based on field semantics and comprehensive change trend; and constructing an accident scene based on the structured state data and generating a target disposal scheme when the early warning condition is met. The specification fuses and encodes the multi-source heterogeneous data to carry the dangerous goods physical and chemical characteristic semantics through the field semantics, captures the multi-dimensional complex evolution trend through the time sequence model to realize early warning, and realizes the transformation of the disposal scheme from the fixed template to the dynamic customization through the scene construction based on the real-time state.

Description

Technical Field

[0001] This specification relates to the field of hazardous materials storage and logistics safety monitoring technology, specifically to a method, system, and electronic equipment for intelligent monitoring and emergency simulation of hazardous materials. Background Technology

[0002] During the storage and logistics transfer of hazardous materials, due to their inherent flammable, explosive, and toxic physical and chemical properties, extremely high requirements are placed on the timeliness and accuracy of monitoring and emergency response.

[0003] Currently, hazardous materials safety monitoring primarily relies on sensor networks deployed in storage areas or on transport vehicles. At the risk assessment level, a single-dimensional threshold alarm mechanism is typically used, where upper and lower safety limits are set for each sensor parameter, and an alarm is triggered when a parameter exceeds the threshold. The judgment logic can be expressed as:

[0004] in, express The risk status at all times, Indicates a single sensor in The measured value at time, This indicates a preset safety threshold.

[0005] However, hazardous materials accidents are often gradual and complex, and their evolution is usually the result of the coupling of multiple state parameters. For example, a slow rise in ambient temperature may not have reached the alarm threshold on its own, but if this temperature change is combined with the low flash point of a specific hazardous material and the ongoing hot work, the actual risk has increased significantly. Because the aforementioned threshold judgment mechanism only focuses on the instantaneous exceeding of limits by a single sensor value, it cannot perceive the coordinated changes in environmental parameters, operational status, and the physical and chemical properties of the hazardous material itself over time. This leads to a significant lag in the perception of early signs of complex risks, easily missing the optimal intervention window.

[0006] In emergency response, existing technologies typically retrieve fixed-text emergency response procedures from a pre-set contingency plan database using keyword matching upon receiving an alarm signal. Because these plans are pre-compiled, the assumed scenarios are often fixed and singular, making dynamic adaptation impossible based on real-time, multi-dimensional conditions at the time of the incident (such as wind direction and speed, distribution of nearby goods, and current operational status). This static matching approach is detached from the actual physical evolution of the incident, resulting in response plans that often lack specificity and feasibility.

[0007] How to achieve dynamic and accurate risk assessment and real-time generation of response plans under the premise of multi-source heterogeneous data fusion for hazardous materials is a technical problem that urgently needs to be solved. Summary of the Invention

[0008] The purpose of this specification is to provide a method, system, and electronic device for intelligent monitoring and emergency simulation of hazardous materials, in order to solve the technical problems in existing hazardous materials safety monitoring technologies, such as the difficulty of identifying multi-dimensional complex evolution trends due to single-dimensional threshold alarms, resulting in delayed early warnings, and the lack of targeted response plans due to the disconnect between fixed text plans and real-time on-site conditions.

[0009] According to the first aspect of this specification, a method for intelligent monitoring and emergency simulation of hazardous materials is provided, including: The process involves acquiring static basic attribute data and dynamic multi-source sensing data of hazardous materials, and then fusing and encoding them according to the semantic structure of hazardous materials domain knowledge to obtain structured state data for the hazardous materials. Static basic attribute data refers to the inherent attribute information of hazardous materials obtained through the declaration system that does not change with the operational process, such as hazardous material identification (e.g., UN number), hazard category, packaging category, and corresponding physicochemical properties such as flash point, explosion limits, and toxicity parameters. Dynamic multi-source sensing data refers to the observational information that changes over time and is collected in real-time by IoT sensing devices throughout the entire hazardous materials operation process, such as location sensing data, environmental sensing data, video sensing data, and operational sensing data. The semantic structure of hazardous materials domain knowledge refers to the semantic organization method that uses the hazardous material classification number as an anchor point to associate knowledge in areas such as physicochemical properties, isolation rules, and emergency response procedures with the corresponding hazardous materials. The fusion encoding process is not simply concatenating data from different sources into an array, but rather semantically mapping the original observation values ​​in the dynamic sensing data with the associated physicochemical properties in the static basic attributes, so that the encoded structured state data naturally carries the domain semantic information of the hazardous material. For example, when encoding temperature values ​​collected by temperature sensors, they are associated with the flash point attribute corresponding to the hazardous material classification number, enabling subsequent processing steps to understand the actual risk meaning of the temperature value for that specific hazardous material at the domain semantic level.

[0010] The structured status data is updated at a preset period to form a time series of status data. The specific value of the preset period can be set according to actual business needs, such as a fixed value between 1 and 10 seconds, or it can be dynamically adjusted according to the risk situation. Through periodic updates, the structured status data forms a continuous sequence in the time dimension, enabling subsequent risk assessments to make comprehensive judgments based on the changing trends of multiple parameters over time, rather than just based on the instantaneous value at a single moment.

[0011] By inputting time series data of state conditions into a pre-trained risk assessment model, a risk assessment result including a risk level is obtained. This risk level is determined based on the semantics of the hazardous materials domain carried by the structured state data and its comprehensive changing trend over time. Specifically, when processing time series data of state conditions, the risk assessment model considers not only the numerical changes of parameters in each dimension, but also the correlation between these parameters and the semantics of the hazardous materials domain. For example, the model can comprehensively judge the superimposed effect between "temperature rising trend," "gradual change in gas concentration," "flash point attribute of the hazardous material," and "current operating status," thereby outputting a risk level that reflects the comprehensive changing trend of multiple dimensions of state conditions, rather than making a binary judgment based solely on whether a single parameter exceeds a fixed threshold.

[0012] Once the risk assessment results meet the preset early warning conditions, an accident scenario is constructed based on the structured state data corresponding to the hazardous materials, generating a target response plan. The preset early warning conditions can be set to specific risk level thresholds according to actual safety management needs. The accident scenario is constructed based on structured state data, which includes multi-dimensional on-site information such as the spatial location of the hazardous materials, environmental parameters, and information on nearby goods. Therefore, the constructed accident scenario reflects the specific on-site conditions rather than being based on a preset fixed template. The target response plan generated on this basis is a customized solution for the specific scenario, capable of adapting to differentiated response needs under different on-site conditions.

[0013] According to a second aspect of this specification, a hazardous materials intelligent monitoring and emergency simulation system is provided, comprising: The data aggregation unit is used to acquire static basic attribute data and dynamic multi-source sensing data of hazardous materials, and to fuse and encode them according to the semantic structure of hazardous materials domain knowledge to obtain the structured state data of the hazardous materials; the structured state data is updated at a preset period to form a state data time series.

[0014] The risk assessment unit is used to input the time series of state data into a pre-trained risk assessment model to obtain a risk assessment result containing a risk level. This risk level is determined based on the semantics of the hazardous materials domain carried by the structured state data and its comprehensive changing trend over time.

[0015] The emergency simulation unit is used to construct accident scenarios and generate target response plans based on the structured status data corresponding to hazardous materials, provided that the risk assessment results meet the preset early warning conditions.

[0016] According to a third aspect of this specification, an electronic device is provided, including at least one processor and at least one memory, the memory being used to store instructions that, when executed by the processor, cause the processor to perform the steps of the method as described in the first aspect of this specification.

[0017] The technical solutions provided in the embodiments of this specification have at least the following beneficial effects: By acquiring static basic attribute data and dynamic multi-source sensing data of hazardous materials, and fusing and encoding them according to the semantic structure of hazardous materials domain knowledge, structured state data is obtained. Since the fusing and encoding process associates dynamic observation data with the inherent physical and chemical properties of hazardous materials at the semantic level, the structured state data is no longer an isolated set of sensor readings, but a comprehensive data representation carrying domain semantics, providing a rich semantic foundation for subsequent risk assessment.

[0018] By updating structured state data at a preset period to form a state data time series, and inputting this time series into a pre-trained risk assessment model, the model can determine the risk level based on the semantics of the hazardous materials being carried and their comprehensive changing trends over time. This allows for the accurate capture of the complex evolution patterns of hazardous material states, overcoming the shortcomings of single-dimensional threshold alarm mechanisms in perceiving early complex risk signs, and achieving dynamic and accurate risk assessment.

[0019] By constructing accident scenarios and generating target disposal plans based on structured state data corresponding to hazardous materials when the risk assessment results meet the preset early warning conditions, the static matching of fixed text plans is transformed into dynamic deduction based on real-time multi-dimensional states. This makes the generated disposal plans closely fit the actual physical evolution process on site, overcomes the defect of fixed plans being divorced from real-time on-site conditions, and realizes a complete closed loop from early warning to disposal. Attached Figure Description

[0020] Figure 1 This is a schematic diagram of the physical environment and hardware deployment of a port dangerous goods storage yard provided in the embodiments of this specification.

[0021] Figure 2 This is a flowchart illustrating a method for intelligent monitoring and emergency simulation of hazardous materials provided in the embodiments of this specification.

[0022] Figure 3 This is a schematic diagram of a structured state data fusion encoding process provided in the embodiments of this specification.

[0023] Figure 4 This is a schematic diagram of a risk assessment model architecture and training phase provided in the embodiments of this specification.

[0024] Figure 5 This is a schematic diagram of an accident scenario construction and target disposal plan generation process provided in the embodiments of this specification.

[0025] Figure 6 This is a schematic diagram of a material diffusion simulation calculation scenario provided in the embodiments of this specification.

[0026] Figure 7 This is a schematic diagram of the structure of a hazardous materials intelligent monitoring and emergency simulation system provided in the embodiments of this specification.

[0027] Figure label explanations: 101—Temperature and humidity sensor, 102—Combustible gas detector, 103—Smoke sensor, 104—RFID reader / writer, 105—AI camera, 106—Edge gateway, 107—Declaration terminal; 301—Static basic attribute data, 302—Dynamic multi-source sensing data, 303—Semantic anchor point, 304—Fusion coding, 305—Structured state data; 401—Input layer, 402—Self-attention coding layer, 403—Output layer, 404—Pre- Training phase: 405—Supervised fine-tuning phase; 406—Alignment optimization phase; 501—Scenario construction; 502—Material diffusion simulation calculation; 503—Knowledge base retrieval and compatibility filtering; 504—Simulation simulation evaluation; 601—Hazardous impact range; 602—Location of leak source; 603—Wind direction arrow; 604—Location of nearby goods; 605—Suggested evacuation boundary line; 700—System; 701—Data aggregation unit; 702—Risk assessment unit; 703—Emergency simulation unit. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the embodiments in this specification clearer, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. The described embodiments are only some, not all, of the embodiments in this specification. All other embodiments obtained by those skilled in the art based on the embodiments in this specification without creative effort are within the scope of protection of this specification.

[0029] The intelligent monitoring and emergency simulation solution for hazardous materials provided in the embodiments of this specification can be deployed in scenarios involving the full-process management of hazardous materials, such as port hazardous goods yards, hazardous chemical storage parks, and logistics transfer centers. The following description uses port hazardous goods yards as the main application scenario, but those skilled in the art will understand that this solution is also applicable to other scenarios requiring hazardous materials safety monitoring and emergency response.

[0030] The following is an explanation of some of the terms used in this specification.

[0031] UN numbers (United Nations Numbers) are internationally recognized dangerous goods identification codes. Each UN number corresponds to a specific set of physical and chemical properties and emergency response procedures.

[0032] The IMDG Code (International Maritime Dangerous Goods Code) is a standard for classifying and segregating dangerous goods developed by the International Maritime Organization. It specifies the minimum segregation distance between different categories of dangerous goods and the prohibitions on co-existence.

[0033] TOS (Terminal Operating System) is the core business system used by ports to manage ship loading and unloading, yard operations, and cargo turnover.

[0034] LEL (Lower Explosive Limit) is the lowest concentration of a flammable gas in air at which it can explode, and is usually expressed as a volume percentage.

[0035] Structured state data, in this specification, can be understood as a data set that represents the current multidimensional state of a hazardous material by uniformly encoding the static inherent properties of the hazardous material and the dynamic real-time observation data according to the semantic relationship of domain knowledge. It differs from simple sensor reading splicing in that the encoded data naturally carries semantic information in the fields of hazardous material category, physical and chemical properties, etc.

[0036] In this specification, state data time series can be understood as an ordered data set formed in the time dimension after structured state data is continuously updated at fixed or dynamic periods. This allows subsequent processing to make comprehensive judgments based on the changing trends of multiple parameters over time.

[0037] In this manual, an accident scenario can be understood as a structured scenario description based on the current multidimensional state information of the hazardous materials, describing the on-site conditions when an accident is assumed to occur. It differs from a fixed contingency plan template in that an accident scenario reflects real-time on-site information such as the current specific spatial location, environmental parameters, and nearby goods.

[0038] like Figure 1 As shown in the embodiments of this specification, the intelligent monitoring and emergency simulation scheme for dangerous goods operates in the physical environment of a port dangerous goods yard. The yard is spatially divided into multiple dangerous goods container areas (such as Yard 1, Yard 3, Yard 5, etc.), each containing several dangerous goods containers. The container areas are arranged according to the isolation distances specified in the IMDG Code. A gate is installed at the entrance of the yard to control the entry and exit of goods.

[0039] At the hardware deployment level, the yard is equipped with various IoT sensing devices to continuously monitor the hazardous materials storage environment and operational status. Temperature and humidity sensors 101 are installed on the yard pillars and the surfaces of hazardous materials containers to collect ambient temperature and humidity data, with a collection cycle set to every 5 minutes. Combustible gas detectors 102 are deployed around the hazardous materials container area to monitor the concentration of combustible gases in the air in real time (expressed as a percentage of LEL). Smoke sensors 103 are used to detect smoke particles produced by open flames or smoldering. RFID readers 104 are installed at the yard gates and key passageways to read RFID tags on containers to obtain real-time cargo location information. AI cameras 105 are deployed in key areas of the yard to identify abnormal behaviors such as unauthorized hot work operations and personnel not wearing protective equipment. The data collected by the above sensing devices is converted and initially aggregated through an edge gateway 106 before being transmitted via wired or wireless networks to a server cluster deployed in the port data center.

[0040] At the business system level, cargo owners or freight forwarders input basic attribute information of dangerous goods through the declaration terminal 107, including UN number, hazard class, packaging category, quantity, flash point, explosion limits, toxicity parameters, and other physicochemical properties. The declaration terminal 107 interfaces with external maritime declaration systems and customs supervision systems via network interfaces to obtain approval status. In addition, the Terminal Operating System (TOS) records operational plan information such as the arrival time, planned departure time, and loading / unloading arrangements for each shipment. The meteorological service interface provides current and 24-48 hour weather forecast data (temperature, wind direction, wind speed, precipitation probability, etc.).

[0041] At the computing infrastructure level, the server cluster runs a data aggregation service, a risk assessment inference service, and an emergency simulation service. The data aggregation service receives dynamic sensing data from the edge gateway 106 and static attribute data from the reporting terminal 107, and performs fusion encoding processing. The risk assessment inference service loads a pre-trained risk assessment model and performs inference calculations on the time series of the fused encoded state data. Upon receiving an early warning signal, the emergency simulation service invokes the material diffusion simulation calculation module and the accident handling knowledge base to construct accident scenarios and generate handling plans. These three services communicate with each other via an internal message queue for data transfer and event notification.

[0042] However, the aforementioned hardware environment and business systems face the following prominent problems in actual operation. First, the physicochemical properties of hazardous materials stored in the declaration terminal 107 (such as flash point and explosion limits) and the real-time environmental data collected by IoT sensing devices (such as temperature and gas concentration) belong to different data systems. They are not aligned in terms of data format, time base, and semantic level, making it impossible to correlate environmental observations with the inherent properties of hazardous materials during risk assessment. Second, the collection frequency and data triggering methods of each sensing device vary (some collect data every 5 minutes, while others only generate records when an event is triggered), lacking a unified time alignment mechanism, making it difficult to form a continuous multi-dimensional state sequence for trend analysis. Third, existing risk judgments rely on fixed threshold alarms for a single parameter, failing to perceive the gradual superposition and linkage changes of multiple parameters over time, resulting in a significant lag in identifying early signs of complex risks. Fourth, emergency response relies on pre-prepared fixed text plans, unable to dynamically generate targeted response plans based on real-time on-site conditions (wind direction, nearby goods, operational status, etc.) at the time of the accident.

[0043] To address the aforementioned issues, this specification provides an intelligent monitoring and emergency simulation method for hazardous materials. Its overall processing logic is as follows: First, static attribute data and dynamic sensing data are fused and encoded according to the semantic structure of hazardous materials domain knowledge to form structured state data carrying domain semantics, which is continuously updated at a preset period to form a time series. Then, this time series is input into a risk assessment model for multi-dimensional composite trend analysis, outputting a risk level determined based on domain semantics and comprehensive change trends. When the risk level meets the warning conditions, an accident scenario is constructed based on the current state data, and a target response plan is generated. The following, in conjunction with… Figure 2 Each step is explained in detail. For example... Figure 2 As shown, the method includes steps S201 to S204.

[0044] Step S201: Obtain the static basic attribute data and dynamic multi-source perception data of the hazardous materials, and fuse and encode them according to the semantic structure of the hazardous materials domain knowledge to obtain the structured state data of the hazardous materials.

[0045] Specifically, the acquisition of static basic attribute data is achieved by connecting to an external declaration system. Taking UN1203 gasoline as an example, the information entered by the cargo owner at the declaration terminal includes: UN number UN1203, hazard class Class 3 flammable liquid, packaging category II, quantity 500 tons, flash point -43℃, and explosion limits 1.4%-7.6% (by volume). The system automatically connects to the maritime declaration system to obtain the approval status of this batch of goods and stores the above information as the static basic attribute data of this batch of gasoline.

[0046] The collection of dynamic multi-source sensing data is achieved through IoT devices deployed in the yard, specifically including four categories: location sensing data, which obtains the real-time location of containers in the yard through RFID tags and readers (e.g., "Container No. 12 in Area A of Yard 3"); environmental sensing data, which obtains current environmental parameters through temperature and humidity sensors, combustible gas detectors, and smoke sensors (e.g., temperature 35℃, combustible gas concentration 8% LEL); video sensing data, which obtains abnormal behavior identification results through video analysis using AI cameras (e.g., "Illegal hot work detected" or "No abnormality"); and operation sensing data, which obtains the operation status of the batch of goods through the TOS system (e.g., "Arrived in the yard, loading and unloading operations scheduled for 14:00").

[0047] In engineering implementation, the four types of sensing data mentioned above have different collection frequencies and data formats: location sensing data is updated when goods pass by an RFID reader; environmental sensing data is automatically collected at fixed intervals (e.g., every 5 minutes); video sensing data is event-driven, only generating records when AI identifies an anomaly; and operation sensing data is updated when the operation status changes in the TOS system. After receiving the above multi-source data, the data aggregation unit 701 needs to perform time alignment and format unification processing. For data sources with different collection frequencies, nearest neighbor interpolation or linear interpolation can be used to align data of different frequencies to a unified time base.

[0048] The core mechanism of fusion coding lies in using the hazardous materials classification number (i.e., UN number) as the semantic anchor 303, and semantically mapping the original observations in the dynamic sensing data with the associated physicochemical properties in the static basic attribute data 301. For example... Figure 3 As shown, taking UN1203 gasoline as an example, the left side contains static basic attribute data 301 (UN1203 → Class 3 flammable liquid → flash point -43℃ → explosion limits 1.4%-7.6%), and the right side contains dynamic multi-source sensing data 302 (temperature 35℃, combustible gas concentration 8%LEL, location No. 3, area A, operating status: awaiting loading and unloading). When the fusion coding 304 is used, the value of 35℃ collected by the temperature sensor is not stored as an isolated number in the structured state data 305, but is semantically associated with the flash point -43℃ corresponding to UN1203—35℃ is much higher than the flash point of this gasoline 78℃ (35-(-43)=78), which means that combustible vapors already exist on the surface of the gasoline at the current temperature. Similarly, the combustible gas concentration 8%LEL is associated with the lower explosive limit 1.4%, indicating that although the current concentration has not reached the lower explosive limit, it is within the range of concern.

[0049] Through the aforementioned fusion encoding, the generated structured state data is no longer a simple compilation of various sensor readings, but a comprehensive data representation carrying semantics from the hazardous materials domain. This structured state data includes at least the following dimensions: physicochemical property data associated with UN numbers (flash point, explosion limits, toxicity, etc.), operational status data representing the current lifecycle stage (declared, arrived, stored, being loaded / unloaded, already dispatched, etc.), environmental sensing data (temperature, humidity, gas concentration, smoke, etc.), and spatial location data (storeyard number, container coordinates, etc.). In one implementation, the structured state data can be organized in key-value pairs, with each key corresponding to a semantic dimension, and the value range determined by the data type of that dimension (numerical, enumerated, or textual). In another implementation, the structured state data can also be encoded as a fixed-length numerical vector, where the values ​​of each dimension are normalized and arranged in a predetermined order for easy processing by subsequent models.

[0050] Step S202: Update the structured state data at a preset period to form a state data time series.

[0051] Unlike the single-parameter threshold judgment logic shown in formula (1) in the background art, the embodiments of this specification do not make a limit judgment on the value of a single sensor at a certain moment, but organize the multi-dimensional state data into a sequence in the time dimension for subsequent model to perform comprehensive trend analysis. Specifically, let If the structured state data obtained after performing a fusion encoding on a certain hazardous material at any given time is S(t), then the preset period is used to determine the structured state data. After continuous updates, the time series of state data formed within the time window [t_0, t_n] can be represented as:

[0052] in, , Each For structured data containing multidimensional attributes, its dimensions must at least cover the dimensions of physical and chemical properties, environmental perception, spatial location, and operational status.

[0053] The structured state data obtained from a single update only reflects the instantaneous state of hazardous materials at a certain moment and cannot support the judgment of risk evolution trends. It is by continuously updating the structured state data at a preset period to form the above time series that the subsequent risk assessment model can make a comprehensive judgment based on the changing trends of multiple parameters over time, rather than making a binary judgment based on the instantaneous value of a single parameter as in formula (1).

[0054] The specific value of the preset cycle can be set according to actual business needs. In one embodiment, the preset cycle is set to 5 minutes, that is, a full update of the structured status data of all hazardous materials in stock is performed every 5 minutes. In another embodiment, the preset cycle can be dynamically adjusted according to the current risk level of the hazardous materials: when the risk assessment results indicate a low risk level (such as below blue alert), a longer update cycle (such as 10 minutes) is used to save computing resources; when the risk level increases (such as reaching yellow or orange alert), the update cycle is shortened (such as 1 minute) to increase the sampling frequency and ensure sufficient perception resolution for rapidly changing risk situations.

[0055] Taking UN1203 gasoline as an example, the system updates its structured status data every 5 minutes, and after continuous operation, it forms a time series of status data for this batch of gasoline. This time series records multi-dimensional status snapshots at each point in time over a period of time (such as the past 30 days), including temperature change curves, gas concentration change curves, and operation status switching records.

[0056] Step S203: Input the time series of state data into the pre-trained risk assessment model to obtain a risk assessment result containing a risk level. The risk level is determined based on the semantics of the hazardous materials domain carried by the structured state data and its comprehensive change trend in the time series.

[0057] In contrast to the threshold judgment logic of formula (1) in the background art, the risk assessment model in this specification for determining the risk level can be summarized as follows:

[0058] in, for The risk level is output in real time. For parameters Risk assessment model, For the state data time series defined by formula (2), This refers to the semantic information about hazardous materials carried in structured state data (including physicochemical properties, isolation rules, etc.). This differs from formula (1), which relies solely on the instantaneous value of a single sensor. and fixed threshold The binary judgments differ, and the risk level in formula (3) is different. It is based on the comprehensive change trend of multidimensional state data over time series, combined with semantics in the field of hazardous materials. A jointly determined continuous quantization output.

[0059] Specifically, when determining the risk level, the model relies not only on the absolute values ​​of the parameters in each dimension, but also on the rate of change of these parameters over time, the direction of change, and the correlation between different parameters. Moreover, this analysis is based on data that carries the semantics of the hazardous materials domain—the model can understand what 35°C means for gasoline with a flash point of -43°C, but for diesel with a flash point of 60°C, it means a completely different level of risk.

[0060] Taking the assessment of UN1203 gasoline as an example, the risk assessment model comprehensively analyzes the following information: current environmental data (temperature 35℃, close to the summer high temperature threshold; combustible gas concentration 8% LEL, within the normal range but close to the concern value); historical trend data (the area has recorded 3 high temperature warnings in the past 30 days, and the temperature is showing an upward trend); meteorological forecast data (the highest temperature in the next 24 hours is 38℃, with no precipitation); and operational plan data (there is a loading and unloading operation plan at 14:00 on the same day). The model integrates the above multi-dimensional information and its temporal trends, outputting a risk score of 72 points.

[0061] In one embodiment, the risk assessment model employs a sequence modeling network based on a self-attention mechanism. For example... Figure 4 As shown, the model includes an input layer 401, a self-attention encoding layer 402, and an output layer 403. The input layer 401 receives the time series of state data. The self-attention encoding layer 402 establishes correlations between multiple attribute dimensions (temperature, gas concentration, location, operation status, etc.) and multiple time steps of the state data, thereby extracting composite evolutionary features across dimensions and time. The output layer 403 outputs a risk assessment result containing risk levels.

[0062] The training process of this model includes three progressive stages (such as...). Figure 4 As shown in the second half, there are fundamental differences in training data, optimization objectives, and learned capabilities at each stage.

[0063] Pre-training Phase 404: Self-supervised training using time series of state data from historical operational data in the hazardous materials sector. The training data consists of time series of state data accumulated during normal port operations over several years (i.e., routine monitoring data during accident-free periods), typically much larger than accident samples. The training task can be designed as mask prediction (randomly masking certain dimensions at certain time steps in the time series, causing the model to predict the masked values ​​based on the context) or prediction for the next time step (based on previous...). State prediction at time step # (The state at each time step). Through training in this stage, the model learns the temporal dependencies of state data across multiple time steps, establishing a baseline representation of the "normal operating state"—that is, the model internalizes the prior knowledge of "how each dimension parameter should change over time under normal conditions".

[0064] Supervised Fine-Tuning Phase 405: Supervised training based on labeled risk event samples. Labeled samples are derived from historical accident records (such as leaks and fires) and post-event annotations of anomalous events by safety experts (e.g., "Although no accident occurred during this period, obvious abnormal signs appeared"). Each sample contains a time series of state data and its corresponding risk label (risk level, risk type). Through training in this phase, the self-attention mechanism can establish correlations across multiple dimensions of the state data to extract composite evolutionary features—for example, learning the correspondence between the combination of "temperature increase + gradual change in gas concentration + switch to hot work status" and risk events. The core value of this phase lies in the model's evolution from "knowing what is normal" to "knowing what combinations constitute risk."

[0065] Alignment Optimization Phase 406: Training based on safety expert decision-making preferences to adjust the mapping strategy from composite evolutionary features to risk levels. In this phase, preference pairs can be constructed: safety experts rank multiple sets of assessment results output by the model for the same risk scenario. The top-ranked assessments are used as preferred samples, and the bottom-ranked ones as unpreferred samples. Through preference learning, the risk levels output by the model are made more consistent with the decision-making logic of safety experts. For example, for the scenario of "temperature 38℃ + gas concentration 12% LEL + hot work," a conservative assessment (outputting an 85-point red alert) might be ranked higher by experts than an aggressive assessment (outputting a 62-point yellow alert). By learning this ranking preference, the model's output is made more aligned with the "better to falsely report than to miss a report" decision-making tendency in safety management practice. The core value of this phase is that the model evolves from "knowing what constitutes risk" to "knowing how the degree of risk should be quantified," ensuring that the output risk level is consistent with the judgment scale of safety experts.

[0066] The progressive relationship of the three stages can be summarized as follows: pre-training establishes a temporal baseline (what is normal) → supervised fine-tuning extracts risk patterns (what constitutes risk) → alignment optimization calibrates the quantification scale (how to quantify the degree of risk). Each stage adds a capability based on the previous stage, and all three are indispensable.

[0067] Furthermore, the risk assessment results can also include the risk level of mixed storage. Before inputting the state data time series into the risk assessment model, the system retrieves a pre-defined isolation rule knowledge base based on the hazardous material category information in the structured state data to obtain matching isolation constraints. For example, for the mixed storage scenario of UN2037 aerosols (Class 2.1 flammable gas) and UN1405 potassium amino compounds (Class 4.3 flammable gas released upon contact with water), the system retrieves the IMDG Code isolation table to obtain the isolation constraint "Class 2.1 and Class 4.3 substances must be isolated 2 (away from each other)". This isolation constraint, together with the state data time series, serves as the input to the risk assessment model, which determines the risk level of mixed storage accordingly.

[0068] Furthermore, the risk assessment model can also output risk evolution trend predictions, representing the expected direction and rate of change in risk level within a preset future time window. For example, the model can output a trend prediction that "the risk level is expected to rise from 72 points to 80 points within the next 2 hours," providing a basis for decision-making in advance deployment of emergency resources.

[0069] Step S204: If the risk assessment results meet the preset early warning conditions, construct an accident scenario based on the structured state data corresponding to the hazardous materials and generate a target disposal plan.

[0070] Preset warning conditions can be set according to actual safety management needs. In one embodiment, the warning levels are divided into four levels: a risk score of 50-59 is a blue warning (attention), 60-69 is a yellow warning (caution), 70-84 is an orange warning (warning), and 85 and above is a red warning (danger). When the risk score reaches an orange or red warning, the emergency simulation process is triggered.

[0071] Taking UN1203 gasoline risk score of 72 (orange alert) as an example, emergency simulation unit 703 initiated the accident scenario construction and response plan generation process. For example... Figure 5 As shown, the process specifically includes the following sub-steps.

[0072] Sub-step 501: Extract the spatial location of the hazardous material (Container 12, Area A, Yard 3), environmental parameters (temperature 35°C, combustible gas concentration 8% LEL, southeast wind 3 m / s), and information on adjacent cargo (UN1405 potassium amino compound is stored in Yard 5) from the structured state data. Combine this with the acquired meteorological data (maximum temperature 38°C in the next 24 hours, no precipitation) to generate a structured accident scenario. This accident scenario assumes a gasoline leak occurs at 14:30, with a leak volume of approximately 200 liters.

[0073] Sub-step 502: Based on the physicochemical properties of the hazardous materials and meteorological data, determine the scope of the accident's impact through material diffusion simulation calculations. The material diffusion simulation calculations select the appropriate physical diffusion model based on the material state and diffusion characteristics of the hazardous materials. For the flammable vapor formed after the UN1203 gasoline leak, since its density is slightly greater than air and it diffuses in open spaces, a Gaussian plume diffusion model is used for calculations. For example... Figure 6 As shown, under the meteorological condition of a southeast wind of 3 m / s, the calculated concentration of combustible gas within 150 meters downwind may reach the lower explosive limit, and this area is marked as the hazardous influence zone 601. Figure 6 The document also shows the location of the leak source 602 (Area A, Yard 3), wind direction arrow 603, location of adjacent cargo 604 (UN1405, Yard 5), and suggested evacuation boundary line 605.

[0074] Different physical diffusion models can be selected for hazardous materials with different states of matter and diffusion characteristics: for scenarios where liquid leakage leads to the formation of a liquid pool and evaporation (such as volatile liquids like benzene and toluene), a pool-fire evaporation model can be used; for scenarios where pressurized gas is released instantaneously (such as liquefied petroleum gas), a heavy gas diffusion model (such as the SLAB model) can be used; and for explosion scenarios, an explosion shock wave overpressure model can be used. The selection of the model is automatically determined by parameters such as the state of matter, vapor pressure, and density in the physicochemical properties of the hazardous material.

[0075] Sub-step 503: Based on the accident scenario and the scope of its impact, retrieve matching procedures from the accident response knowledge base. Filter the feasibility of the solutions by considering the chemical compatibility constraints of the hazardous materials, generating a set of candidate solutions. For the UN1203 gasoline leak scenario, matching procedures retrieved from the knowledge base include: foam covering, spray dilution, and dry powder fire extinguishing. Chemical compatibility constraint filtering ensures that the candidate solutions do not include any measures that adversely react with gasoline (e.g., do not use direct water jets to impact the leaking liquid surface to prevent static electricity ignition). After filtering, a set of candidate solutions is generated, including: Solution A (evacuate personnel within 200 meters downwind, shut off nearby power, and cover the leaked material with foam) and Solution B (adding the activation of the spray system to dilute the vapors, based on Solution A).

[0076] Sub-step 504: For each option in the candidate treatment plan set, simulate the accident evolution trajectory after implementing that option. Based on preset evaluation indicators (such as the rate of decrease in combustible gas concentration, the speed of reduction in the affected area, and the probability of secondary risks), quantify and rank the expected treatment effects of each option. The simulation results show that option B can reduce the combustible gas concentration more quickly, but attention should be paid to preventing sewage outflow. Finally, option B with the highest score is determined as the target treatment plan, and dike measures are added to prevent sewage diffusion.

[0077] When the risk level of mixed loading in the risk assessment results meets the preset blocking conditions, the generated target disposal plan will include a spatial transfer strategy for hazardous materials. Taking the mixed loading scenario in Example 2 as an example: UN2037 aerosol (Class 2.1) and UN1405 potassium amino compound (Class 4.3) are arranged in the same storage area. The system retrieves the IMDG Code isolation table and finds that the two need to be "isolated 2 (away from each other)". Considering the aging drainage facilities in the area and the forecast of heavy rain in the next 48 hours (potassium amino compound produces hydrogen gas when it comes into contact with water), the model outputs a mixed loading risk score of 88 points (red alert). The target disposal plan generated at this time includes a spatial transfer strategy: immediately suspend the mixed storage plan of the two shipments, transfer the potassium amino compound to an indoor dry warehouse, and keep it at least 12 meters away from the aerosol. In addition, the disposal plan may also include long-term recommendations, such as repairing drainage facilities and adding waterproof dikes.

[0078] In another application scenario, the aforementioned simulation chain can also be used for virtual drills and training. Unlike the automatic simulation mode, in drill mode, the system initializes the drill scenario based on a structured accident scenario. It then receives response decisions from multiple drill roles (such as dispatchers, safety officers, and rescue team leaders) at preset time points. Utilizing the aforementioned material diffusion simulation calculations and accident evolution trajectory simulation rules, the system dynamically updates the accident's impact range and evolution trajectory based on each role's response decisions. For example, if the dispatcher decides to "initiate evacuation" at T+5 minutes, the system simulates the evacuation progress and accident evolution based on this decision to determine if the evacuation is timely. If the decision is incorrect (such as using a water-based fire extinguisher for a sodium leak), the system simulates the escalation of the situation. After the drill, the system quantifies and scores each role's response decisions based on preset evaluation indicators, outputting a drill evaluation report that includes response time, decision accuracy, and an estimate of the response effect. During the drill evaluation, the simulation engine can be called to generate the optimal solution for the current scenario as a comparison benchmark, indicating the deviation between the trainees' decisions and the optimal solution, and providing suggestions for improvement.

[0079] The following three specific embodiments further illustrate the practical application of the solutions in this specification.

[0080] Example 1: Hazardous Goods Leakage Risk Warning and Emergency Response. A batch of UN1203 gasoline (500 tons, Class 3 flammable liquid, Packaging Category II, flash point -43℃, explosive limits 1.4%-7.6%) is stored in a hazardous goods yard at a port. The system executes the following process: the cargo owner enters basic information about the gasoline, and the system connects to the maritime declaration system to obtain the approval status. IoT devices continuously collect environmental data (temperature 35℃, flammable gas concentration 8% LEL), location data (RFID tracking to area A of yard 3), and operational data (TOS shows planned loading and unloading at 14:00). The risk assessment model integrates current environmental data, three high-temperature warning records within the past 30 days, weather forecast (maximum 38℃ in the next 24 hours), and operational plans, outputting a risk score of 72, triggering an orange alert. The system then initiates an emergency simulation: assuming a leak of 200 liters occurs at 14:30, with a southeast wind of 3 m / s, the Gaussian diffusion model calculates that the lower explosive limit may be reached within 150 meters downwind. Option A (evacuation + power outage + foam covering) and Option B (Option A + spray dilution) are generated. After simulation evaluation, Option B is recommended, with the addition of a cofferdam. The dispatcher adopts the recommended option, and the system records the handling process and results.

[0081] Example 2: Risk Identification of Mixed Hazardous Goods. A port plans to store two batches of hazardous goods in the same area: Goods A is UN2037 aerosol (Class 2.1 flammable gas, 10 tons), and Goods B is UN1405 potassium amino compound (Class 4.3, flammable gas released upon contact with water, 5 tons). The system searches the IMDG Code segregation table and finds that Class 2.1 and Class 4.3 require "Segregation 2 (away from each other)". The risk assessment model comprehensively considers: both are flammable substances and Class 4.3 reacts with water, the drainage facilities in the area are aging and have experienced water accumulation, and a heavy rain is forecast for the next 48 hours, outputting a mixed storage risk score of 88 points (red alert). The system generates a disposal plan: immediately suspend the mixed storage plan, transfer the potassium amino compound to an indoor dry warehouse, maintain a distance of at least 12 meters from the aerosol, and recommend repairing the drainage facilities and adding a waterproof dike. This mixed storage prohibition is updated as a new rule in the segregation rule knowledge base, and similar situations will be automatically identified and intercepted in the future.

[0082] Example 3: Virtual Drill Training. The system provides virtual drill functionality for training port safety management personnel and emergency rescue personnel. After logging into the system, trainees select a drill scenario (such as liquefied petroleum gas leak, fire caused by nitric acid leak, dangerous goods container falling into water, etc.) and choose their role (dispatcher, safety officer, rescue team leader, etc.). The system advances the scenario chronologically (such as leak discovery at T+0, activation of the emergency plan at T+5 minutes, and rescue arrival at T+15 minutes), requiring trainees to make a decision at each time point. Based on the trainees' decisions, the system dynamically updates the situation using material diffusion simulation calculations—correct decisions reduce risk, while incorrect decisions escalate risk. After the drill, an evaluation report is generated, including response time, decision accuracy rate, estimated casualties and property damage, and comparisons with the optimal response plan, pointing out deviations and providing improvement suggestions.

[0083] During continuous system operation, actual risk events, emergency response processes, and outcomes can serve as feedback data to optimize the risk assessment model and emergency simulation engine. Specifically, after each actual response, the system acquires a snapshot of structured state data at the time of the response, the actual implemented response plan, and multi-dimensional evaluations of the response effectiveness (such as whether the leak was controlled, whether evacuation was timely, and whether secondary reactions occurred). This information is then correlated with the historical evaluation results of the risk assessment model to form structured training samples for incremental optimization and updates to the risk assessment model. Simultaneously, the actual response plan and its effectiveness evaluation are added as new cases to the accident response knowledge base, and risk identification events and corresponding response recommendations are added as new rules to the isolation rule knowledge base, enabling the automatic identification and interception of subsequent risk events of the same type. Through this data feedback mechanism, the system's risk assessment accuracy and simulation plan quality can continuously improve with the accumulation of actual events. In one implementation, incremental fine-tuning of the model can be triggered when the accumulated feedback data reaches a preset threshold; in another implementation, model updates can be triggered immediately after a major risk event to ensure the model quickly absorbs experience from extreme cases.

[0084] This specification also provides an intelligent monitoring and emergency simulation system for hazardous materials, which can be deployed in the aforementioned electronic equipment as the main body for executing the various method steps. For example... Figure 7 As shown, the system 700 includes the following functional units.

[0085] The data aggregation unit 701 is used to acquire static basic attribute data and dynamic multi-source perception data of hazardous materials, and to fuse and encode them according to the semantic structure of hazardous materials domain knowledge to obtain the structured state data of the hazardous materials; the structured state data is updated at a preset period to form a state data time series.

[0086] Risk assessment unit 702 is used to input the time series of state data into a pre-trained risk assessment model to obtain a risk assessment result containing a risk level. The risk level is determined based on the semantics of the hazardous materials domain carried by the structured state data and its comprehensive change trend in the time series.

[0087] Emergency simulation unit 703 is used to construct accident scenarios and generate target disposal plans based on the structured status data corresponding to hazardous materials, provided that the risk assessment results meet the preset early warning conditions.

[0088] Optionally, the data aggregation unit 701 is specifically used to: obtain static basic attribute data including hazardous material identification, category and physicochemical properties by connecting to an external declaration system; and collect dynamic multi-source sensing data including location sensing data, environmental sensing data, video sensing data and operation sensing data by IoT devices.

[0089] Optionally, the system also includes an isolation rule retrieval unit, used to retrieve a preset isolation rule knowledge base based on the hazardous material category information in the structured state data to obtain matching isolation constraints. The risk assessment unit 702 is also used to output a mixed-packing risk level, which is determined by the risk assessment model based on the hazardous material category information in the structured state data and its corresponding isolation constraints. The emergency simulation unit 703 is specifically used to: when the mixed-packing risk level meets preset blocking conditions, generate a target disposal plan including a hazardous material spatial transfer strategy, which is used to redistribute the hazardous materials to storage areas that meet the isolation constraints.

[0090] Optionally, the risk assessment model in risk assessment unit 702 is a sequence modeling network based on a self-attention mechanism, which is trained through three stages: self-supervised pre-training, supervised fine-tuning, and alignment optimization.

[0091] Optionally, the emergency simulation unit 703 is specifically used for: extracting the spatial location, environmental parameters, and information on nearby goods of hazardous materials from structured state data, and generating structured accident scenarios by combining meteorological data; determining the scope of the accident's impact through material diffusion simulation calculations based on the physicochemical properties of hazardous materials and meteorological data, wherein the material diffusion simulation calculations select the corresponding physical diffusion model according to the material state and diffusion characteristics of hazardous materials; retrieving matching disposal procedures from the accident disposal knowledge base based on the accident scenario and the scope of the accident's impact, filtering the feasibility of the solutions by combining the chemical compatibility constraints of hazardous materials, and generating a set of candidate disposal solutions; simulating the accident evolution trajectory after executing each solution in the set of candidate disposal solutions, quantitatively scoring and ranking the expected disposal effects of each solution based on preset evaluation indicators, and determining the candidate solution with the highest score as the target disposal solution.

[0092] Optionally, the system also includes a virtual drill unit, which is used to initialize drill scenarios based on structured accident scenarios, receive decision inputs from multiple roles at preset time nodes, update the scope of accident impact and evolution trajectory using material diffusion simulation calculations and accident evolution trajectory simulation rules, and output a drill evaluation report by quantifying the decision-making process.

[0093] The specific functional implementation, interaction logic, and technical effects of each unit in the system provided in this embodiment are consistent with the descriptions in the aforementioned method embodiments, and will not be repeated here.

[0094] This specification also provides an electronic device including at least one processor and at least one memory. The memory stores computer programs or instructions that, when executed by the processor, cause the processor to implement the steps of the aforementioned intelligent monitoring and emergency simulation method for hazardous materials. The processor may be a central processing unit (CPU), a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices. The memory may include volatile memory (such as random access memory, RAM) and non-volatile memory (such as read-only memory, flash memory, disk storage, etc.). The electronic device may also include a communication interface for data interaction with external reporting systems, IoT sensing devices, and user terminals.

[0095] This specification also provides a computer-readable storage medium storing a computer program or instructions that, when executed by a processor, implement the steps of the aforementioned intelligent monitoring and emergency simulation method for hazardous materials. The storage medium can be non-transitory, such as a disk, optical disk, solid-state drive, or flash memory.

[0096] This specification also provides a computer program product, including a computer program or instructions, which, when executed by a processor, implement the steps of the aforementioned intelligent monitoring and emergency simulation method for hazardous materials.

[0097] The above description is merely a preferred embodiment of this specification and is not intended to limit this specification. Various modifications and variations can be made to this specification by those skilled in the art. Any modifications, equivalent substitutions, or improvements made within the spirit and principles of this specification should be included within the scope of protection of this specification.

Claims

1. A method for intelligent monitoring and emergency simulation of hazardous materials, characterized in that, include: Acquire static basic attribute data and dynamic multi-source sensing data of hazardous materials, and fuse and encode them according to the semantic structure of hazardous materials domain knowledge to obtain the structured state data of the hazardous materials. The structured state data is updated at a preset period to form a state data time series; The time series of the state data is input into a pre-trained risk assessment model to obtain a risk assessment result containing risk levels. The risk levels are determined based on the hazardous materials domain semantics carried by the structured state data and their comprehensive change trends over time. If the risk assessment results meet the preset early warning conditions, an accident scenario is constructed based on the structured state data corresponding to the hazardous materials, and a target disposal plan is generated.

2. The method according to claim 1, characterized in that, The acquisition of static basic attribute data and dynamic multi-source sensing data of hazardous materials includes: By connecting to external declaration systems, static basic attribute data including hazardous material identification, category, and physical and chemical properties can be obtained; Dynamic multi-source sensing data, including location sensing data, environmental sensing data, video sensing data, and operation sensing data, is collected through IoT devices.

3. The method according to claim 1, characterized in that, Before inputting the time series of the state data into the pre-trained risk assessment model, the following steps are also included: Based on the hazardous material category information in the structured state data, a preset isolation rule knowledge base is retrieved to obtain matching isolation constraints; wherein, the risk assessment result also includes a mixed loading risk level, which is determined by the risk assessment model based on the hazardous material category information in the structured state data and its corresponding isolation constraints; The generated target handling plan includes: When the risk level of the mixed loading meets the preset blocking conditions, a target disposal plan including a spatial transfer strategy for hazardous materials is generated. The spatial transfer strategy is used to redistribute the hazardous materials to a storage area that meets the isolation constraints.

4. The method according to claim 1, characterized in that, The risk assessment model is a sequence modeling network based on a self-attention mechanism, and the training process of the risk assessment model includes: The self-attention mechanism is used to perform self-supervised pre-training on historical operational data in the hazardous materials field to learn the temporal dependencies of the state data across multiple time steps; Supervised fine-tuning is performed based on labeled risk event samples, enabling the self-attention mechanism to establish correlations between multiple dimensions of the state data to extract composite evolutionary features; Alignment optimization is performed based on the decision preferences of security experts to adjust the mapping strategy from the composite evolutionary features to the risk level.

5. The method according to claim 1, characterized in that, The process of constructing an accident scenario based on the structured state data corresponding to the hazardous materials and generating a target response plan includes: The spatial location, environmental parameters, and information on nearby goods of hazardous materials are extracted from the structured state data, and a structured accident scenario is generated by combining it with the acquired meteorological data. Based on the physicochemical properties of hazardous materials and the meteorological data, the scope of the accident's impact is determined through material diffusion simulation calculations; the material diffusion simulation calculations select the corresponding physical diffusion model according to the material state and diffusion characteristics of the hazardous materials. Based on the accident scenario and the scope of the accident impact, matching disposal procedures are retrieved from the accident disposal knowledge base. The feasibility of the solutions is filtered in combination with the chemical compatibility constraints of the hazardous materials to generate a set of candidate disposal solutions. For each of the candidate disposal schemes in the set, the accident evolution trajectory after the scheme is executed is simulated; the expected disposal effect of each scheme is quantitatively scored and ranked based on preset evaluation indicators; and the candidate scheme with the highest score is determined as the target disposal scheme.

6. The method according to claim 5, characterized in that, The method further includes: The drill scenario is initialized based on the structured accident scenario; Receive decision-making input from multiple roles at preset time points; Using the aforementioned material diffusion simulation calculations and accident evolution trajectory simulation rules, the accident's impact range and evolution trajectory are updated based on the aforementioned response decisions; The effectiveness of the proposed decision-making process is quantitatively scored based on preset evaluation indicators, and an exercise evaluation report is output.

7. A hazardous materials intelligent monitoring and emergency simulation system, characterized in that, include: The data aggregation unit is used to acquire static basic attribute data and dynamic multi-source sensing data of hazardous materials, and to fuse and encode them according to the semantic structure of hazardous materials domain knowledge to obtain the structured state data of the hazardous materials. The structured state data is updated at a preset period to form a state data time series; A risk assessment unit is used to input the time series of the state data into a pre-trained risk assessment model to obtain a risk assessment result containing a risk level. The risk level is determined based on the hazardous materials domain semantics carried by the structured state data and its comprehensive change trend in the time series. The emergency simulation unit is used to construct accident scenarios and generate target response plans based on the structured status data corresponding to the hazardous materials, when the risk assessment results are determined to meet the preset early warning conditions.

8. An electronic device, characterized in that, include: At least one processor; At least one memory for storing instructions; When the instructions are executed by the at least one processor, the at least one processor performs the method as described in any one of claims 1 to 6.

9. A storage medium storing a computer program or instructions, characterized in that, When the computer program or instructions are executed by a processor, they implement the method as described in any one of claims 1 to 6.

10. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the steps of the method as described in any one of claims 1 to 6.

Images