Internet of Things-based Traffic and Fire Safety Risk Assessment System

Abstract

This invention discloses an Internet of Things (IoT)-based system for assessing traffic and fire safety risks, comprising the following steps: a data acquisition module for acquiring and preprocessing standardized time-series datasets; a regional time-series data generation module for obtaining a regional time-series data set; a risk action state generation module for constructing a risk action state sequence; an action chain state generation module for generating a risk action chain state sequence; a closed-loop result generation module for obtaining causal closed-loop approximation results and causal closed-loop closure results through an improved recursive fuzzy neural network; a linkage interruption result generation module for obtaining linkage interruption results and writing back to correct the risk action chain state sequence; an accident pre-state result generation module for determining the accident pre-state result; and an assessment result output module for generating traffic and fire safety risk assessment results. This invention employs an improved recursive fuzzy neural network to assess traffic and fire risks, possessing advantages such as strong foresight and high accuracy.

Description

Technical Field

[0001] This invention relates to the field of risk assessment technology, and in particular to an Internet of Things-based system for assessing traffic and fire safety risks. Background Technology

[0002] With the development of IoT sensing technology, intelligent traffic management technology, and fire-fighting linkage control technology, temperature sensors, smoke sensors, video sensing equipment, vehicle detection equipment, and fire-fighting facility monitoring devices are typically deployed in traffic tunnels, underground passages, integrated transportation hubs, and underground parking areas to collect real-time data on traffic operation status, environmental status, and fire-fighting facility status, and to realize alarm display, linkage control, and risk management through the monitoring platform.

[0003] Most existing technologies rely on single-point threshold alarms, multi-indicator static weighting, or ordinary time-series models for risk assessment. While they can identify local anomalies, they struggle to reveal the continuous transmission relationships between heat accumulation, smoke retention, traffic obstruction, weakening of linkages, and evacuation suppression. In particular, they are unable to identify the causal closed loops formed by the connection of the beginning and end of multiple risk action chains, resulting in insufficient identification of risks that accumulate continuously before an accident. Furthermore, the interruption effect after linkage execution is difficult to provide feedback to correct subsequent assessment results.

[0004] Therefore, how to provide an IoT-based assessment system for traffic and fire safety risks is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose an Internet of Things-based traffic and fire safety risk assessment system. This invention uses an improved recursive fuzzy neural network to assess traffic and fire risks, which has the advantages of strong foresight and high accuracy.

[0006] The traffic and fire safety risk assessment system based on the Internet of Things according to an embodiment of the present invention includes the following steps: The data acquisition module is used to collect multi-source sensing data in traffic scenarios and preprocess it to obtain a standardized time-series dataset. The regional time series data generation module is used to divide the collection locations corresponding to the standardized time series dataset into regions and map them to the corresponding regions to obtain a regional time series data set. The risk impact state generation module is used to extract risk features from the regional time series data set and construct a risk impact state sequence. The action chain state generation module is used to construct a risk action chain based on the risk action state sequence and generate a risk action chain state sequence arranged in chronological order. The closed-loop result generation module is used to input the current risk action chain state sequence, the region identifier of the corresponding region at the current time, and the risk action chain state sequence at the previous time into the improved recurrent fuzzy neural network to obtain the causal closed-loop approximation result and the causal closed-loop closure result at the current time. The linkage interruption result generation module is used to obtain the linkage interruption result and write the linkage interruption result back to the improved recurrent fuzzy neural network to correct the risk action chain state sequence at the next moment. The pre-accident result generation module is used to determine the causal closed loop locking result, the closed loop breakable result, and the pre-accident approximation result; The assessment results output module is used to generate traffic and fire safety risk assessment results.

[0007] Optionally, the data acquisition module includes: The multi-source sensing data includes temperature data, smoke concentration data, combustible gas concentration data, wind speed data, wind direction data, video sensing data, vehicle detection data, fire protection facility operation data, alarm execution data, traffic guidance execution data, and evacuation status data. The preprocessing includes time alignment, spatial location matching, outlier removal, missing value completion, and format unification.

[0008] Optionally, the regional time-series data generation module includes: Read the spatial distribution and facility layout relationships of each acquisition location corresponding to the standardized time-series dataset; Based on the spatial distribution and facility layout of each data collection location, the regions of each data collection location are divided and determined, and the region division results corresponding to each data collection location are obtained. Based on the area division results corresponding to each collection location, each collection location is divided into entrance area, main passage area, intersection area, equipment adjacency area and evacuation passage area; For a data collection location that corresponds to two or more areas at the same time, the area identification is determined in the order of adjacent area of ​​the device, intersection area, evacuation channel area, entrance area and main passage area to obtain the area identification corresponding to each data collection location. According to the area identifier corresponding to each collection location, the standardized time series dataset is mapped to the corresponding area to obtain the regional time series data set.

[0009] Optionally, the risk action status generation module includes: Read temperature data, smoke concentration data, video perception data, vehicle detection data, fire protection facility operation data, alarm execution data, traffic guidance execution data, and evacuation status data from the regional time series data set, and arrange the various types of data in the same area according to time order to form a regional multi-source time series record; Based on the data change results of adjacent and consecutive moments in the multi-source time series records of the region, the following features are extracted: temperature change features, hot spot persistence features, smoke growth features, smoke retention features, vehicle density features, vehicle stagnation features, area occupancy features, ventilation response features, sprinkler response features, alarm feedback features, traffic guidance feedback features, and evacuation restriction features. Based on temperature change characteristics and hot spot persistence characteristics, a heat accumulation effect state is constructed; based on smoke growth characteristics and smoke retention characteristics, a smoke retention effect state is constructed; based on vehicle density characteristics, vehicle stagnation characteristics, and area occupancy characteristics, a traffic obstruction effect state is constructed; based on ventilation response characteristics, sprinkler response characteristics, and alarm feedback characteristics, a linkage weakening effect state is constructed; based on traffic guidance feedback characteristics and evacuation restriction characteristics, an evacuation suppression effect state is constructed; and based on the corresponding changes in fire protection facility operation data, a facility failure effect state is constructed. Arrange the states of heat accumulation, smoke retention, traffic obstruction, linkage weakening, evacuation suppression, and facility failure at each time point in chronological order to obtain the risk action state sequence.

[0010] Optionally, the action chain state generation module includes: Read the risk action status, including heat accumulation action status, smoke retention action status, traffic obstruction action status, linkage weakening action status, evacuation and suppression action status, and facility failure action status. Construct a risk action chain based on the influence relationship between different risk action states at the same time; For each risk action chain at each time moment, read the risk action states at both ends of the corresponding risk action chain, and determine the chain direction of the corresponding risk action chain at that time moment based on the order of action of the risk action states at both ends at that time moment. For the same risk action chain at adjacent time points, read the risk action states at both ends of the previous time point and the risk action states at both ends of the next time point respectively, and determine the chain strength of the risk action chain at the next time point based on the consistency and continuity of the changes in the risk action states at both ends of the previous time point and the next time point. Arrange the heat accumulation enhancement chain, smoke retention enhancement chain, retardation amplification chain, linkage weakening chain, evacuation suppression chain and facility failure chain corresponding to each moment in chronological order to obtain the risk action chain state sequence.

[0011] Optionally, the closed-loop result generation module includes: Read the current risk action chain state sequence, the region identifier of the corresponding region at the current time, and the risk action chain state sequence at the previous time. The current risk action chain state sequence, the region identifier of the corresponding region at the current time, and the risk action chain state sequence at the previous time are input into an improved recursive fuzzy neural network. The improved recursive fuzzy neural network includes a risk action chain state recursive input structure, a closed-loop judgment fuzzy inference structure, and a closed-loop result output structure. Input the risk action chain state sequence of the previous moment into the risk action chain state recursive input structure, and combine the risk action chain state sequence of the previous moment and the risk action chain state sequence of the current moment according to the correspondence of the same risk action chain to obtain the recursive input result; Based on the area identifier of the corresponding area at the current time, the closed-loop judgment rule set of the corresponding area is called, and the recursive input result is input into the closed-loop judgment fuzzy inference structure. The enhancement degree of the heat accumulation enhancement chain, smoke retention enhancement chain, stagnation amplification chain, linkage weakening chain, evacuation suppression chain and facility failure chain are judged respectively, and the enhancement degree judgment result is obtained. Based on the closed-loop judgment rule set, the degree of closure of the first and last ends of the heat accumulation enhancement chain and the smoke retention enhancement chain, the smoke retention enhancement chain and the evacuation and suppression chain, the evacuation and suppression chain and the retardation and amplification chain, the retardation and amplification chain and the heat accumulation enhancement chain, the facility failure chain and the linkage weakening chain, and the linkage weakening chain and the smoke retention enhancement chain are judged respectively, and the degree of closure of the first and last ends is judged. Based on the closed-loop judgment rule set, the degree of persistence of the heat accumulation enhancement chain, smoke retention enhancement chain, retardation amplification chain, linkage weakening chain, evacuation suppression chain and facility failure chain is judged respectively, and the degree of persistence judgment results are obtained. The enhancement degree determination result, the beginning and end closure degree determination result, and the persistence degree determination result are input into the closed loop result output structure to obtain the causal closed loop approach result and the causal closed loop closure result at the current moment.

[0012] Optionally, the linkage interruption result generation module includes: The current linkage execution status is determined based on fire protection facility operation data, alarm execution data, traffic guidance execution data, and broadcast evacuation execution data. By analyzing the correlation between the current linkage execution state, the current causal closed loop approach result, and the current causal closed loop closure result, the result of the risk action chain state sequence change corresponding to the linkage execution state is obtained. The linkage interruption result is determined based on the change results of the risk action chain state sequence corresponding to the linkage execution state, and the linkage interruption result is obtained by determining the successful closed-loop interruption result and the insufficient closed-loop interruption result. The results of the linkage interruption are written back to the improved recursive fuzzy neural network to correct the risk action chain state sequence at the next moment.

[0013] Optionally, the pre-accident result generation module includes: Read the current moment's causal closed loop approach result, causal closed loop closure result, linkage interruption result, and the corrected risk action chain state sequence; Based on the causal loop approach result at the current moment, the causal loop closure result at the current moment, and the corrected risk action chain state sequence, the risk action chain participating in the causal loop closure result at the current moment in the corrected risk action chain state sequence is subjected to a loop closure lock determination to obtain the causal loop closure lock result. Based on the current moment's linkage interruption result, the current moment's causal closed loop closure result, and the corrected risk action chain state sequence, the interruption state of the risk action chain participating in the current moment's causal closed loop closure result is determined, and the closed loop can be interrupted result is obtained. Based on the current causal closed loop approach result, the current causal closed loop closure result, the causal closed loop locking result, and the closed loop interruptibility result, a joint judgment is made. When the causal closed loop approach result representation remains close, the causal closed loop closure result representation remains closed, the causal closed loop locking result representation remains locked, and the closed loop interruptibility result representation is limited in its interruptibility, the pre-accident state approximation result is determined. When the causal closed loop approach result representation weakens, the causal closed loop closure result representation weakens, the causal closed loop locking result representation weakens in its locking, and the closed loop interruptibility result representation can be interrupted, the pre-accident state approximation result is determined.

[0014] Optionally, the evaluation result output module includes: Read the causal closed loop approach results, causal closed loop closure results, causal closed loop locking results, closed loop interruptibility results, and accident pre-state approximation results; Based on the causal closed loop approach results, causal closed loop closure results, causal closed loop locking results, closed loop interruptibility results, and accident pre-state approximation results, the results are aggregated to obtain the result aggregation state. Based on the convergence status of the results, the traffic fire safety risk assessment results are determined, the risk maintenance status corresponding to the traffic fire safety risk assessment results is determined, the risk increase status corresponding to the traffic fire safety risk assessment results is determined, and the risk decrease status corresponding to the traffic fire safety risk assessment results is determined, and the traffic fire safety risk assessment results are obtained and output.

[0015] The beneficial effects of this invention are: This invention utilizes an improved recursive fuzzy neural network to jointly process the current risk action chain state sequence, the corresponding region identifier, and the previous risk action chain state sequence. Based on a closed-loop determination rule set, it performs recursive fuzzy inference on the enhancement degree, head-to-tail closure degree, and persistence degree of the risk action chain, generating causal closed-loop approach results and causal closed-loop closure results. Therefore, this invention can identify whether there are head-to-tail connections, continuous enhancement, and closure trends between risk action chains, thus elevating the traditional judgment of risk existence to a judgment of whether a risk is forming a causal closed loop. This technical solution effectively solves the problem of existing technologies' difficulty in identifying self-reinforcing risk closed loops formed by the interaction between environmental anomalies, traffic anomalies, and fire-fighting linkage anomalies. It enables the system to identify the continuously accumulating state before an accident occurs, improving the foresight of traffic and fire safety risk assessment.

[0016] This invention also generates linkage interruption results through correspondence analysis between the linkage execution state and the causal closed-loop approach result and the causal closed-loop closure result. These results are then written back to an improved recursive fuzzy neural network to correct the risk action chain state sequence for the next moment. Based on this, causal closed-loop locking results, closed-loop interruptibility results, and accident pre-state approximation results are further generated, ultimately outputting traffic and fire safety risk assessment results. This invention not only assesses whether a risk has formed but also evaluates the interruption effect of linkage execution on the risk closed loop and whether the closed loop remains locked, transforming the risk assessment result from a purely static judgment to a dynamic judgment encompassing the entire process of formation, development, interruption, and resolution. Therefore, this invention improves the accuracy, continuity, and interpretability of traffic and fire safety risk identification and provides a more reliable basis for traffic and fire linkage control, risk classification and early warning, and pre-accident intervention. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is an overall flowchart of the Internet of Things-based traffic and fire safety risk assessment system proposed in this invention; Figure 2 This is a schematic diagram illustrating the recursive fuzzy reasoning of the risk action chain using the closed-loop judgment rule set in the traffic and fire safety risk assessment system based on the Internet of Things proposed in this invention. Figure 3 This is a schematic diagram showing the output of traffic fire safety risk assessment results in the Internet of Things-based traffic fire safety risk assessment system proposed in this invention. Detailed Implementation

[0018] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0019] refer to Figures 1-3 The Internet of Things-based traffic and fire safety risk assessment system includes the following steps: The data acquisition module is used to collect multi-source sensing data in traffic scenarios and preprocess the multi-source sensing data to obtain a standardized time-series dataset. The regional time series data generation module is used to divide the collection location corresponding to the standardized time series dataset into regions based on the spatial distribution relationship and facility layout relationship in the traffic scenario, to obtain the entrance region, main passage region, intersection region, equipment adjacency region and evacuation channel region, and to map the standardized time series dataset to the corresponding region according to the region identifier to obtain the regional time series data set. The risk impact state generation module is used to extract risk features from the regional time series data set. Risk features include temperature change features, hot spot persistence features, smoke growth features, smoke retention features, vehicle density features, vehicle stagnation features, area occupancy features, ventilation response features, sprinkler response features, alarm feedback features, traffic guidance feedback features, and evacuation restriction features. Based on the direction of each feature, a risk impact state sequence arranged in chronological order is constructed. Risk impact states include heat accumulation effect state, smoke retention effect state, traffic obstruction effect state, linkage weakening effect state, evacuation suppression effect state, and facility failure effect state. The action chain state generation module is used to construct risk action chains based on the influence relationships between the states in the risk action state. The risk action chains include thermal accumulation enhancement chain, smoke retention enhancement chain, retardation amplification chain, linkage weakening chain, evacuation suppression chain, and facility failure chain. Based on the action transmission relationship between the risk action states at the same time and the continuity change relationship between the risk action states at adjacent times, the module determines the chain strength and chain direction of each risk action chain at each time and generates a risk action chain state sequence arranged in chronological order. The closed-loop result generation module is used to input the current risk action chain state sequence, the region identifier of the corresponding region at the current time, and the risk action chain state sequence of the previous time into the improved recursive fuzzy neural network. Based on the region identifier, it calls the closed-loop judgment rule set of the corresponding region and performs recursive fuzzy inference on the enhancement degree of each risk action chain, the degree of closure between the beginning and end of each risk action chain, and the degree of persistence of each risk action chain, to obtain the causal closed-loop approach result and the causal closed-loop closure result at the current time. The improved recursive fuzzy neural network includes a risk action chain state recursive input structure, a closed-loop judgment fuzzy inference structure, and a closed-loop result output structure. The risk action chain state recursive input structure is used to use the risk action chain state sequence of the previous time as the recursive input at the current time. The closed-loop judgment fuzzy inference structure is used to perform joint inference on the enhancement degree, the degree of closure between the beginning and end, and the degree of persistence of the risk action chain based on the closed-loop judgment rule set. The closed-loop result output structure is used to output the causal closed-loop approach result and the causal closed-loop closure result. The linkage interruption result generation module is used to obtain the linkage execution status at the current moment based on the regional time series data set. The linkage execution status includes the fan execution status, sprinkler execution status, alarm execution status, traffic guidance execution status, and broadcast evacuation execution status. The module performs corresponding analysis on each linkage execution status with the causal closed loop approach result and causal closed loop closure result at the current moment to obtain the linkage interruption result. The linkage interruption result is then written back to the improved recursive fuzzy neural network to correct the risk action chain state sequence at the next moment. The linkage interruption result includes the closed loop interruption success result, the closed loop interruption insufficiency result, and the closed loop interruption failure result. The closed loop interruption success result, the closed loop interruption insufficiency result, and the closed loop interruption failure result are determined by the change result of the risk action chain state sequence corresponding to the linkage execution status. The pre-accident state result generation module is used to determine the causal closed loop locking result, the closed loop interruptible result, and the pre-accident state approximation result based on the current causal closed loop approach result, causal closed loop closure result, linkage interruption result, and the corrected risk action chain state sequence. The pre-accident state approximation result is jointly determined by the causal closed loop approach result, causal closed loop closure result, causal closed loop locking result, and closed loop interruptible result. The assessment results output module is used to output traffic and fire safety risk assessment results based on causal closed loop approach results, causal closed loop closure results, causal closed loop locking results, closed loop interruptibility results, and accident pre-state approximation results.

[0020] In this embodiment, the data acquisition module includes: The multi-source sensing data includes temperature data, smoke concentration data, combustible gas concentration data, wind speed data, wind direction data, video sensing data, vehicle detection data, fire protection facility operation data, alarm execution data, traffic guidance execution data, and evacuation status data. Preprocessing includes time alignment, spatial location matching, outlier removal, missing value completion, and format unification.

[0021] In this embodiment, the regional time-series data generation module includes: Read the spatial distribution relationship and facility layout relationship of each collection location corresponding to the standardized time series dataset. The spatial distribution relationship includes the positional relationship between each collection location and the side boundary of traffic flow entry, the positional relationship with the main traffic path, the positional relationship with the intersection of the path and the evacuation path. The facility layout relationship includes the positional relationship between each collection location and the fans, sprinklers, alarms, traffic guidance devices and broadcasting devices. Based on the spatial distribution and facility layout of each collection location, the area division of each collection location is determined, and the area division results corresponding to each collection location are obtained. The entrance area division result is determined for the collection location located on the traffic flow entry side and connected to the outside of the scene. The main traffic area division result is determined for the collection location located within the coverage of the continuous passage path. The intersection area division result is determined for the collection location located at the path connection point or path divergence point. The equipment adjacent area division result is determined for the collection location located within the adjacent range of the fan, sprinkler, alarm, traffic guidance device or broadcasting device. The evacuation passage area division result is determined for the collection location located within the coverage of the evacuation path. Based on the area division results corresponding to each collection location, each collection location is divided into entrance area, main passage area, intersection area, equipment adjacency area and evacuation passage area; For a data collection location that corresponds to two or more areas at the same time, the area identification is determined in the order of adjacent area of ​​the device, intersection area, evacuation channel area, entrance area and main passage area to obtain the area identification corresponding to each data collection location. According to the area identifier corresponding to each collection location, the standardized time series dataset is mapped to the corresponding area to obtain the regional time series data set.

[0022] In this embodiment, the risk action status generation module includes: Read temperature data, smoke concentration data, video perception data, vehicle detection data, fire protection facility operation data, alarm execution data, traffic guidance execution data, and evacuation status data from the regional time series data set, and arrange the various types of data in the same area according to time order to form a regional multi-source time series record; Based on the data changes of adjacent and consecutive moments in the multi-source time-series records of the region, the following features are extracted: temperature change features, hot spot persistence features, smoke growth features, smoke retention features, vehicle density features, vehicle stagnation features, area occupancy features, ventilation response features, sprinkler response features, alarm feedback features, traffic guidance feedback features, and evacuation restriction features. Temperature change features are formed by the changes in temperature data at adjacent moments; hot spot persistence features are formed by the duration of continuous temperature increases or high levels at consecutive moments; smoke growth features are formed by the changes in smoke concentration data at adjacent moments; smoke retention features are formed by the duration of insufficient decrease in smoke concentration data at consecutive moments; vehicle density features are formed by the changes in the number of vehicles in a unit area; vehicle stagnation features are formed by the number of vehicles whose positions do not change sufficiently at consecutive moments; area occupancy features are formed by the changes in the proportion of the vehicle-covered area in the area; ventilation response features, sprinkler response features, and alarm feedback features are formed by the changes in fire protection facility operation data and alarm execution data; traffic guidance feedback features and evacuation restriction features are formed by the changes in traffic guidance execution data and evacuation status data. Based on temperature change characteristics and hot spot persistence characteristics, a heat accumulation effect state is constructed; based on smoke growth characteristics and smoke retention characteristics, a smoke retention effect state is constructed; based on vehicle density characteristics, vehicle stagnation characteristics, and area occupancy characteristics, a traffic obstruction effect state is constructed; based on ventilation response characteristics, sprinkler response characteristics, and alarm feedback characteristics, a linkage weakening effect state is constructed; based on traffic guidance feedback characteristics and evacuation restriction characteristics, an evacuation suppression effect state is constructed; and based on the corresponding changes in fire protection facility operation data, a facility failure effect state is constructed. Arrange the states of heat accumulation, smoke retention, traffic obstruction, linkage weakening, evacuation suppression, and facility failure at each time point in chronological order to obtain the risk action state sequence.

[0023] In this embodiment, the action chain state generation module includes: Read the risk action status, including heat accumulation action status, smoke retention action status, traffic obstruction action status, linkage weakening action status, evacuation and suppression action status, and facility failure action status. Based on the influence relationship between the various risk action states at the same time, a risk action chain is constructed. The influence relationship between the heat accumulation action state and the smoke retention action state is defined as the heat accumulation enhancement chain; the influence relationship between the smoke retention action state and the evacuation suppression action state is defined as the smoke retention enhancement chain; the influence relationship between the traffic obstruction action state and the heat accumulation action state is defined as the obstruction amplification chain; the influence relationship between the linkage weakening action state and the smoke retention action state is defined as the linkage weakening chain; the influence relationship between the evacuation suppression action state and the traffic obstruction action state is defined as the evacuation suppression chain; and the influence relationship between the facility failure action state and the linkage weakening action state is defined as the facility failure chain. For each risk action chain at each time moment, read the risk action states at both ends of the corresponding risk action chain, and determine the chain direction of the corresponding risk action chain at that time moment based on the order of action of the risk action states at both ends at that time moment. For the same risk action chain at adjacent time points, the risk action states at both ends of the previous time point and the risk action states at both ends of the next time point are read respectively. The chain strength of the risk action chain at the next time point is determined based on the consistency and continuity of the changes in the risk action states at both ends of the previous time point and the next time point. The consistency of the changes is used to characterize whether the risk action states at both ends of the previous time point and the next time point maintain the same direction of change. The continuity of the changes is used to characterize whether the changes in the risk action states at both ends of the previous time point and the next time point maintain continuous extension. Arrange the heat accumulation enhancement chain, smoke retention enhancement chain, retardation amplification chain, linkage weakening chain, evacuation suppression chain and facility failure chain corresponding to each moment in chronological order to obtain the risk action chain state sequence.

[0024] In this embodiment, the closed-loop result generation module includes: Read the current risk action chain state sequence, the region identifier of the corresponding region at the current time, and the risk action chain state sequence at the previous time. The current risk action chain state sequence, the region identifier of the corresponding region at the current time, and the risk action chain state sequence of the previous time are input into the improved recursive fuzzy neural network. The improved recursive fuzzy neural network includes a risk action chain state recursive input structure, a closed-loop judgment fuzzy inference structure, and a closed-loop result output structure. During the training of the improved recurrent fuzzy neural network, historical traffic scene samples, along with corresponding temperature data, smoke concentration data, video perception data, vehicle detection data, fire-fighting facility operation data, alarm execution data, traffic guidance execution data, and evacuation status data, are collected. These samples are then merged according to the same scene identifier, the same area identifier, and the same time sequence to form a historical training sample set. The historical training sample set undergoes preprocessing consistent with the current traffic and fire safety risk assessment business data to generate a historical standardized time-series dataset. Based on this dataset, a historical regional time-series data set, a historical risk action state sequence, and a historical risk action chain state sequence are constructed. From the historical risk action chain state sequence, extract the corresponding thermal accumulation enhancement chain, smoke retention enhancement chain, retardation amplification chain, linkage weakening chain, evacuation suppression chain, and facility failure chain for each historical moment in chronological order. Combine the corresponding area identifier, closed-loop judgment rule set, and risk action chain state sequence of the previous moment to form historical recursive input samples. Use the causal closed-loop approach result and causal closed-loop closure result corresponding to each historical moment as historical output labels. Quantize and normalize the historical recursive input samples to form model training input vectors. Input the model training input vectors into an improved recursive fuzzy neural network composed of risk action chain state recursive input structure, closed-loop judgment fuzzy inference structure, and closed-loop result output structure. Based on the prediction deviation between the causal closed loop approach prediction results and the causal closed loop approach results, the prediction deviation between the causal closed loop closure prediction results and the causal closed loop closure results, the rule deviation between the output of the fuzzy inference structure for closed loop determination and the closed loop determination rule set, and the continuous deviation between the output results at adjacent time points, the connection parameters, membership parameters, and rule parameters are iteratively corrected. When the total deviation corresponding to the validation sample set continues to decrease and tends to stabilize in multiple consecutive training rounds, or when the training rounds reach the set upper limit, the determination model reaches the convergence condition and training stops. The corresponding parameters are saved, and the improved recurrent fuzzy neural network with training completed is obtained. Input the risk action chain state sequence of the previous moment into the risk action chain state recursive input structure, and combine the risk action chain state sequence of the previous moment and the risk action chain state sequence of the current moment according to the correspondence of the same risk action chain to obtain the recursive input result; Based on the area identifier of the corresponding area at the current time, the closed-loop judgment rule set of the corresponding area is called, and the recursive input result is input into the closed-loop judgment fuzzy inference structure. The enhancement degree of the heat accumulation enhancement chain, smoke retention enhancement chain, stagnation amplification chain, linkage weakening chain, evacuation suppression chain and facility failure chain are judged respectively, and the enhancement degree judgment result is obtained. The closed-loop judgment rule set is a set of rules set according to regions, with different regions corresponding to different closed-loop judgment rule sets; each closed-loop judgment rule set includes the judgment rules for the degree of enhancement of the heat accumulation enhancement chain, smoke retention enhancement chain, retardation amplification chain, linkage weakening chain, evacuation suppression chain and facility failure chain, the judgment rules for the degree of closure of the beginning and end of the heat accumulation enhancement chain and smoke retention enhancement chain, the smoke retention enhancement chain and evacuation suppression chain, the evacuation suppression chain and retardation amplification chain, the retardation amplification chain and heat accumulation enhancement chain, the facility failure chain and linkage weakening chain, the linkage weakening chain and smoke retention enhancement chain, and the judgment rules for the degree of continuity of each risk action chain between consecutive moments; Based on the closed-loop judgment rule set, the degree of closure of the first and last ends of the heat accumulation enhancement chain and the smoke retention enhancement chain, the smoke retention enhancement chain and the evacuation suppression chain, the evacuation suppression chain and the retardation amplification chain, the retardation amplification chain and the heat accumulation enhancement chain, the facility failure chain and the linkage weakening chain, and the linkage weakening chain and the smoke retention enhancement chain are judged respectively. The degree of closure of the first and last ends is judged based on the temporal connection between the chain termination state of the previous risk action chain and the chain start state of the next risk action chain at the current time, as well as the continuous maintenance result between the previous risk action chain and the next risk action chain between the previous time and the current time. Based on the closed-loop judgment rule set, the degree of persistence of the heat accumulation enhancement chain, smoke retention enhancement chain, retardation amplification chain, linkage weakening chain, evacuation suppression chain and facility failure chain is judged respectively, and the degree of persistence judgment results are obtained. The enhancement degree determination result, the beginning and end closure degree determination result, and the persistence degree determination result are input into the closed loop result output structure to obtain the causal closed loop approach result and the causal closed loop closure result at the current moment.

[0025] In this embodiment, the linkage interruption result generation module includes: The current linkage execution status is determined based on the fire protection facility operation data, alarm execution data, traffic guidance execution data, and broadcast evacuation execution data. The linkage execution status includes the fan execution status, sprinkler execution status, alarm execution status, traffic guidance execution status, and broadcast evacuation execution status. By analyzing the correlation between the current linkage execution state, the current causal closed loop approach result, and the current causal closed loop closure result, the risk action chain state sequence change result corresponding to the linkage execution state is obtained. The fan execution state corresponds to the linkage weakening chain and the smoke retention enhancement chain. The sprinkler execution state corresponds to the heat accumulation enhancement chain and the linkage weakening chain. The alarm execution state corresponds to the linkage weakening chain. The traffic guidance execution state corresponds to the obstruction amplification chain and the evacuation suppression chain. The broadcast evacuation execution state corresponds to the evacuation suppression chain. The linkage interruption result is determined based on the change results of the risk action chain state sequence corresponding to the linkage execution state. When the change results of the risk action chain state sequence corresponding to the linkage execution state indicate that the causal closed loop approach result is weakened and the causal closed loop closure result is weakened, it is determined to be a successful loop interruption result. When the change results of the risk action chain state sequence corresponding to the linkage execution state indicate that the causal closed loop approach result is maintained and the causal closed loop closure result is maintained, it is determined to be an insufficient loop interruption result. When the change results of the risk action chain state sequence corresponding to the linkage execution state indicate that the causal closed loop approach result is strengthened or the causal closed loop closure result is strengthened, it is determined to be a failed loop interruption result, and the linkage interruption result is obtained. The results of the linkage interruption are written back to the improved recursive fuzzy neural network to correct the risk action chain state sequence at the next moment.

[0026] In this embodiment, the pre-accident result generation module includes: Read the current moment's causal closed loop approach result, causal closed loop closure result, linkage interruption result, and the corrected risk action chain state sequence; Based on the causal loop approach result at the current moment, the causal loop closure result at the current moment, and the corrected risk action chain state sequence, the risk action chains participating in the causal loop closure result at the current moment in the corrected risk action chain state sequence are subjected to loop closure locking determination to obtain the causal loop closure locking result. When the risk action chains participating in the causal loop closure result at the current moment maintain the same chain direction and maintain the continuity of chain strength in the corrected risk action chain state sequence, the causal loop closure locking result is determined. Based on the current moment's linkage interruption result, the current moment's causal closed loop closure result, and the corrected risk action chain state sequence, the interruption status of the risk action chain participating in the current moment's causal closed loop closure result is determined, and the closed loop can be interrupted result is obtained. When the linkage interruption result corresponds to the closed loop interruption success result and the corrected risk action chain state sequence indicates that the chain strength of the risk action chain participating in the current moment's causal closed loop closure result is weakened, the closed loop can be interrupted result is determined. When the linkage interruption result corresponds to the closed loop interruption insufficient result or the closed loop interruption failure result and the corrected risk action chain state sequence indicates that the chain strength of the risk action chain participating in the current moment's causal closed loop closure result is maintained or enhanced, the closed loop can be interrupted result is determined. Based on the current causal closed loop approach result, the current causal closed loop closure result, the causal closed loop locking result, and the closed loop interruptibility result, a joint judgment is made. When the causal closed loop approach result representation remains close, the causal closed loop closure result representation remains closed, the causal closed loop locking result representation remains locked, and the closed loop interruptibility result representation is limited in its interruptibility, the pre-accident state approximation result is determined. When the causal closed loop approach result representation weakens, the causal closed loop closure result representation weakens, the causal closed loop locking result representation weakens in its locking, and the closed loop interruptibility result representation can be interrupted, the pre-accident state approximation result is determined.

[0027] In this embodiment, the evaluation result output module includes: Read the causal closed loop approach results, causal closed loop closure results, causal closed loop locking results, closed loop interruptibility results, and accident pre-state approximation results; Based on the causal closed loop approach results, causal closed loop closure results, causal closed loop locking results, closed loop interruptibility results, and accident pre-state approximation results, the results are aggregated to obtain the result aggregation state. The traffic and fire safety risk assessment results are determined based on the convergence status of the results. When the convergence status indicates that the causal closed loop approaching result is maintained, the causal closed loop is maintained, the causal closed loop is locked, or the closed loop can be interrupted, and the interruption is limited while the pre-accident state is maintained, the risk maintenance status corresponding to the traffic and fire safety risk assessment results is determined. When the convergence status indicates that the causal closed loop approaching result is enhanced, the causal closed loop is enhanced, the causal closed loop is locked, or the closed loop can be interrupted, and the pre-accident state is enhanced, the risk rising status corresponding to the traffic and fire safety risk assessment results is determined. When the convergence status indicates that the causal closed loop approaching result is weakened, the causal closed loop is weakened, the causal closed loop is locked, or the closed loop can be interrupted, and the pre-accident state is weakened, the risk falling status corresponding to the traffic and fire safety risk assessment results is determined. The traffic and fire safety risk assessment results are then obtained and output.

[0028] Example 1: To verify the feasibility of this invention in practice, it was applied to an underground passageway in a city that serves as a diversion point for a main urban road. This passageway is enclosed, experiences rapid traffic flow changes, is subject to numerous environmental disturbances, and has a dense distribution of fire-fighting equipment. During operation, it simultaneously experiences multiple states, including continuous vehicle traffic, localized queuing and congestion, ventilation status switching, alarm feedback, and evacuation guidance standby. When localized areas experience heat accumulation, abnormal smoke, obstructed vehicle traffic, or delayed response, the related impacts do not remain at a single monitoring point but expand along traffic flow, airflow, and the area of ​​equipment operation, gradually forming a mutually reinforcing risk evolution process. Existing methods mainly rely on single-point alarm values ​​or isolated state judgments, which easily generate repeated alarms for short-term fluctuations and are difficult to identify in a timely manner the process of multiple risks that have not yet clearly exceeded limits but have already begun to couple and accumulate. This invention addresses this problem by identifying the self-reinforcing risk loop formed between environmental anomalies, traffic anomalies, and fire-fighting linkage anomalies.

[0029] In practical applications, the system continuously collects data on temperature, smoke, wind speed and direction, video, vehicle detection, fire safety facility operation, alarm execution, traffic guidance execution, and evacuation status at entrance locations, main passageways, intersection locations, adjacent equipment locations, and evacuation routes. The collected data is then uniformly organized to form a standardized time-series dataset that can be analyzed concurrently on the same timeline. Subsequently, the system, considering the spatial layout and facility deployment of the site, maps the data to entrance areas, main passageways, intersection areas, adjacent equipment areas, and evacuation route areas, forming regional time-series data sets. Based on this, the system extracts risk characteristics such as temperature changes, heat spot persistence, smoke growth, smoke retention, vehicle density, vehicle stagnation, area occupancy, ventilation response, sprinkler response, alarm feedback, traffic guidance feedback, and evacuation restrictions. It further generates states of heat accumulation, smoke retention, traffic obstruction, linkage weakening, evacuation suppression, and facility failure, and then generates risk action chain state sequences corresponding to these chains. In this way, the system no longer focuses on judging a single indicator or monitoring point, but can describe the transmission process of traffic and fire risks between areas, states, and links. Especially under conditions of continuous traffic flow merging at entrances, decreased capacity in main traffic areas, accumulated traffic pressure at intersections, and delayed linkage feedback in adjacent equipment areas, the system can simultaneously capture the correlation evolution between various risk characteristics. This integrates previously scattered monitoring phenomena into a continuous process of risk state changes, providing a more stable input basis for subsequent closed-loop identification.

[0030] In the recursive fuzzy inference stage, the system inputs the current risk action chain state sequence, the area identifier of the corresponding region at the current moment, and the risk action chain state sequence from the previous moment into an improved recursive fuzzy neural network. Based on the closed-loop determination rule set, the system jointly analyzes the enhancement degree, the degree of closure at the beginning and end, and the degree of persistence of each risk action chain, gradually forming causal closed-loop approach results and causal closed-loop closure results. Subsequently, the system continues to receive the execution status of fans, sprinklers, alarms, traffic guidance, and broadcast evacuation, and performs corresponding analysis on these linked execution statuses with the causal closed-loop approach results and causal closed-loop results to obtain linkage interruption results. These linkage interruption results are then written back to the improved recursive fuzzy neural network to correct the risk action chain state sequence for the next moment. Based on this, the system further forms causal closed-loop locking results, closed-loop interruptibility results, and accident pre-state approximation results, and finally outputs the traffic and fire safety risk assessment results.

[0031] During continuous operation, the regional monitoring information, linkage feedback information, and manual verification information recorded by the system demonstrate that this invention can identify the continuous connection trend between risk chains before obvious anomalies simultaneously appear in environmental, traffic, and linkage information. For local fluctuations caused by instantaneous disturbances, the system does not easily classify them as complete closed loops. Regarding the state after linkage execution, the system can distinguish between interrupted and non-interrupted states, elevating subsequent assessments from the equipment action level to whether the risk closed loop has been resolved. Therefore, traffic and fire safety management can shift from traditional single-point alarm handling to proactive handling oriented towards evolutionary processes and closed-loop structures, demonstrating the invention's excellent feasibility and practical application value.

[0032] Table 1. Comparison of the Comprehensive Performance of Traffic and Fire Safety Risk Assessment Methods

[0033] Overall, the method of this invention outperforms the fixed threshold method, the multi-index static weighting method, and the conventional recurrent neural network method. Specifically, the causal loop identification accuracy of this invention reaches 90.8%, higher than the 78.6% of the fixed threshold method, higher than the 82.9% of the multi-index static weighting method, and also higher than the 88.7% of the conventional recurrent neural network method. This indicates that this invention does not merely focus on identifying single-point anomalies or single-moment fluctuations, but rather can more accurately and structurally assess the formation process of risk loops by considering the connection between the beginning and end of the heat accumulation enhancement chain, smoke retention enhancement chain, retardation amplification chain, linkage weakening chain, evacuation suppression chain, and facility failure chain.

[0034] In terms of both false positive rate and false negative rate, the method of this invention has scores of 4.2 and 6.1 respectively, both lower than the three comparative methods. Compared with the fixed threshold method, the false positive rate decreases by 8.2 percentage points and the false negative rate decreases by 12.6 percentage points; compared with the multi-index static weighted method, the false positive rate decreases by 6.1 percentage points and the false negative rate decreases by 9.1 percentage points; compared with the conventional recurrent neural network method, the false positive rate decreases by 3.4 percentage points and the false negative rate decreases by 4.0 percentage points. The decrease in the false positive rate indicates that this invention can avoid misjudging instantaneous fluctuations, local short-term anomalies, or single-chain disturbances as overall risk loops; the decrease in the false negative rate indicates that this invention can grasp the coupling process from weak to strong between multiple risk action chains, reducing the omission of the true high-risk evolution state.

[0035] Regarding the accuracy of linkage interruption judgment, the method of this invention achieves 89.5%, which is higher than the fixed threshold method (74.9), the multi-index static weighted method (79.4), and the conventional recurrent neural network method (86.8). This result is directly related to the linkage interruption result write-back mechanism introduced in this invention. Traditional methods often use whether the fan, sprinkler, alarm, or traffic guidance has been activated as the indicator of completion. However, this invention further analyzes the risk action chain state sequence change results corresponding to the linkage execution state, and distinguishes between successful closed-loop interruption results, insufficient closed-loop interruption results, and failed closed-loop interruption results. Therefore, it can more realistically reflect whether the linkage measures have substantially weakened the risk closed loop.

[0036] In terms of overall F1 score, the method of this invention achieves 0.906, which is 0.144 higher than the fixed threshold method (0.762), 0.100 higher than the static weighted multi-index method (0.806), and 0.032 higher than the conventional recurrent neural network method (0.874). This indicates that the present invention achieves a better overall balance between accuracy and recall. The overall performance improvement is achieved through the combined effects of the risk action state sequence, the risk action chain state sequence, the closed-loop judgment rule set, and the write-back of the linked interruption results.

[0037] The above data demonstrates that the performance improvement of the method in this invention has clear technological origins. First, the generation of regional time-series data sets enables analysis based on the spatial characteristics of entrance areas, main traffic areas, intersection areas, equipment adjacency areas, and evacuation route areas. Second, the construction of risk effect state sequences and risk effect chain state sequences allows for a structured expression of the influence relationships between environmental anomalies, traffic anomalies, and linkage anomalies. Third, the improved recursive fuzzy neural network performs joint reasoning on the degree of enhancement, the degree of closure at the beginning and end, and the degree of persistence. Finally, the linkage interruption result write-back mechanism allows the system to continue feeding back the actual effects of linkage measures into the risk assessment at the next moment. It is precisely these multi-level synergistic effects that enable this invention to achieve superior results in closed-loop identification, pre-accident state identification, false alarm control, missed alarm control, and linkage effectiveness judgment.

[0038] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A traffic and fire safety risk assessment system based on the Internet of Things, characterized in that, Includes the following steps: The data acquisition module is used to collect multi-source sensing data in traffic scenarios and preprocess it to obtain a standardized time-series dataset. The regional time series data generation module is used to divide the collection locations corresponding to the standardized time series dataset into regions and map them to the corresponding regions to obtain a regional time series data set. The risk impact state generation module is used to extract risk features from the regional time series data set and construct a risk impact state sequence. The action chain state generation module is used to construct a risk action chain based on the risk action state sequence and generate a risk action chain state sequence arranged in chronological order. The closed-loop result generation module is used to input the current risk action chain state sequence, the region identifier of the corresponding region at the current time, and the risk action chain state sequence at the previous time into the improved recurrent fuzzy neural network to obtain the causal closed-loop approximation result and the causal closed-loop closure result at the current time. The linkage interruption result generation module is used to obtain the linkage interruption result and write the linkage interruption result back to the improved recurrent fuzzy neural network to correct the risk action chain state sequence at the next moment. The pre-accident result generation module is used to determine the causal closed loop locking result, the closed loop breakable result, and the pre-accident approximation result; The assessment results output module is used to generate traffic and fire safety risk assessment results.

2. The traffic and fire safety risk assessment system based on the Internet of Things according to claim 1, characterized in that, The data acquisition module includes: The multi-source sensing data includes temperature data, smoke concentration data, combustible gas concentration data, wind speed data, wind direction data, video sensing data, vehicle detection data, fire protection facility operation data, alarm execution data, traffic guidance execution data, and evacuation status data. The preprocessing includes time alignment, spatial location matching, outlier removal, missing value completion, and format unification.

3. The traffic and fire safety risk assessment system based on the Internet of Things according to claim 1, characterized in that, The regional time-series data generation module includes: Read the spatial distribution and facility layout relationships of each acquisition location corresponding to the standardized time-series dataset; Based on the spatial distribution and facility layout of each data collection location, the regions of each data collection location are divided and determined, and the region division results corresponding to each data collection location are obtained. Based on the area division results corresponding to each collection location, each collection location is divided into entrance area, main passage area, intersection area, equipment adjacency area and evacuation passage area; For a data collection location that corresponds to two or more areas at the same time, the area identification is determined in the order of adjacent area of ​​the device, intersection area, evacuation channel area, entrance area and main passage area to obtain the area identification corresponding to each data collection location. According to the area identifier corresponding to each collection location, the standardized time series dataset is mapped to the corresponding area to obtain the regional time series data set.

4. The traffic and fire safety risk assessment system based on the Internet of Things according to claim 1, characterized in that, The risk action status generation module includes: Read temperature data, smoke concentration data, video perception data, vehicle detection data, fire protection facility operation data, alarm execution data, traffic guidance execution data, and evacuation status data from the regional time series data set, and arrange the various types of data in the same area according to time order to form a regional multi-source time series record; Based on the data change results of adjacent and consecutive moments in the multi-source time series records of the region, the following features are extracted: temperature change features, hot spot persistence features, smoke growth features, smoke retention features, vehicle density features, vehicle stagnation features, area occupancy features, ventilation response features, sprinkler response features, alarm feedback features, traffic guidance feedback features, and evacuation restriction features. Based on temperature change characteristics and hot spot persistence characteristics, a heat accumulation effect state is constructed; based on smoke growth characteristics and smoke retention characteristics, a smoke retention effect state is constructed; based on vehicle density characteristics, vehicle stagnation characteristics, and area occupancy characteristics, a traffic obstruction effect state is constructed; based on ventilation response characteristics, sprinkler response characteristics, and alarm feedback characteristics, a linkage weakening effect state is constructed; based on traffic guidance feedback characteristics and evacuation restriction characteristics, an evacuation suppression effect state is constructed; and based on the corresponding changes in fire protection facility operation data, a facility failure effect state is constructed. Arrange the states of heat accumulation, smoke retention, traffic obstruction, linkage weakening, evacuation suppression, and facility failure at each time point in chronological order to obtain the risk action state sequence.

5. The traffic and fire safety risk assessment system based on the Internet of Things according to claim 1, characterized in that, The action chain state generation module includes: Read the risk action status, including heat accumulation action status, smoke retention action status, traffic obstruction action status, linkage weakening action status, evacuation and suppression action status, and facility failure action status. Construct a risk action chain based on the influence relationship between different risk action states at the same time; For each risk action chain at each time moment, read the risk action states at both ends of the corresponding risk action chain, and determine the chain direction of the corresponding risk action chain at that time moment based on the order of action of the risk action states at both ends at that time moment. For the same risk action chain at adjacent time points, read the risk action states at both ends of the previous time point and the risk action states at both ends of the next time point respectively, and determine the chain strength of the risk action chain at the next time point based on the consistency and continuity of the changes in the risk action states at both ends of the previous time point and the next time point. Arrange the heat accumulation enhancement chain, smoke retention enhancement chain, retardation amplification chain, linkage weakening chain, evacuation suppression chain and facility failure chain corresponding to each moment in chronological order to obtain the risk action chain state sequence.

6. The traffic and fire safety risk assessment system based on the Internet of Things according to claim 1, characterized in that, The closed-loop result generation module includes: Read the current risk action chain state sequence, the region identifier of the corresponding region at the current time, and the risk action chain state sequence at the previous time. The current risk action chain state sequence, the region identifier of the corresponding region at the current time, and the risk action chain state sequence at the previous time are input into the improved recursive fuzzy neural network. The improved recursive fuzzy neural network includes a risk action chain state recursive input structure, a closed-loop judgment fuzzy inference structure, and a closed-loop result output structure. Input the risk action chain state sequence of the previous moment into the risk action chain state recursive input structure, and combine the risk action chain state sequence of the previous moment and the risk action chain state sequence of the current moment according to the correspondence of the same risk action chain to obtain the recursive input result; Based on the area identifier of the corresponding area at the current time, the closed-loop judgment rule set of the corresponding area is called, and the recursive input result is input into the closed-loop judgment fuzzy inference structure. The enhancement degree of the heat accumulation enhancement chain, smoke retention enhancement chain, stagnation amplification chain, linkage weakening chain, evacuation suppression chain and facility failure chain are judged respectively, and the enhancement degree judgment result is obtained. Based on the closed-loop judgment rule set, the degree of closure of the first and last ends of the heat accumulation enhancement chain and the smoke retention enhancement chain, the smoke retention enhancement chain and the evacuation and suppression chain, the evacuation and suppression chain and the retardation and amplification chain, the retardation and amplification chain and the heat accumulation enhancement chain, the facility failure chain and the linkage weakening chain, and the linkage weakening chain and the smoke retention enhancement chain are judged respectively, and the degree of closure of the first and last ends is judged. Based on the closed-loop judgment rule set, the degree of persistence of the heat accumulation enhancement chain, smoke retention enhancement chain, retardation amplification chain, linkage weakening chain, evacuation suppression chain and facility failure chain is judged respectively, and the degree of persistence judgment results are obtained. The enhancement degree determination result, the beginning and end closure degree determination result, and the persistence degree determination result are input into the closed loop result output structure to obtain the causal closed loop approach result and the causal closed loop closure result at the current moment.

7. The traffic and fire safety risk assessment system based on the Internet of Things according to claim 1, characterized in that, The linkage interruption result generation module includes: The current linkage execution status is determined based on fire protection facility operation data, alarm execution data, traffic guidance execution data, and broadcast evacuation execution data. By analyzing the correlation between the current linkage execution state, the current causal closed loop approach result, and the current causal closed loop closure result, the result of the risk action chain state sequence change corresponding to the linkage execution state is obtained. The linkage interruption result is determined based on the change results of the risk action chain state sequence corresponding to the linkage execution state, and the linkage interruption result is obtained by determining the successful closed-loop interruption result and the insufficient closed-loop interruption result. The results of the linkage interruption are written back to the improved recursive fuzzy neural network to correct the risk action chain state sequence at the next moment.

8. The traffic and fire safety risk assessment system based on the Internet of Things according to claim 1, characterized in that, The accident pre-state result generation module includes: Read the current moment's causal closed loop approach result, causal closed loop closure result, linkage interruption result, and the corrected risk action chain state sequence; Based on the causal loop approach result at the current moment, the causal loop closure result at the current moment, and the corrected risk action chain state sequence, the risk action chain participating in the causal loop closure result at the current moment in the corrected risk action chain state sequence is subjected to a loop closure lock determination to obtain the causal loop closure lock result. Based on the current moment's linkage interruption result, the current moment's causal closed loop closure result, and the corrected risk action chain state sequence, the interruption state of the risk action chain participating in the current moment's causal closed loop closure result is determined, and the closed loop can be interrupted result is obtained. Based on the current causal closed loop approach result, the current causal closed loop closure result, the causal closed loop locking result, and the closed loop interruptibility result, a joint judgment is made. When the causal closed loop approach result representation remains close, the causal closed loop closure result representation remains closed, the causal closed loop locking result representation remains locked, and the closed loop interruptibility result representation is limited in its interruptibility, the pre-accident state approximation result is determined. When the causal closed loop approach result representation weakens, the causal closed loop closure result representation weakens, the causal closed loop locking result representation weakens in its locking, and the closed loop interruptibility result representation can be interrupted, the pre-accident state approximation result is determined.

9. The traffic and fire safety risk assessment system based on the Internet of Things according to claim 1, characterized in that, The evaluation result output module includes: Read the causal closed loop approach results, causal closed loop closure results, causal closed loop locking results, closed loop interruptibility results, and accident pre-state approximation results; Based on the causal closed loop approach results, causal closed loop closure results, causal closed loop locking results, closed loop interruptibility results, and accident pre-state approximation results, the results are aggregated to obtain the result aggregation state. Based on the convergence status of the results, the traffic fire safety risk assessment results are determined, the risk maintenance status corresponding to the traffic fire safety risk assessment results is determined, the risk increase status corresponding to the traffic fire safety risk assessment results is determined, and the risk decrease status corresponding to the traffic fire safety risk assessment results is determined, and the traffic fire safety risk assessment results are obtained and output.

Images