An intelligent fire door management method and system based on the Internet of Things

Abstract

This application relates to the field of fire management technology and discloses an intelligent fire door management method and system based on the Internet of Things. The method includes the following steps: Step S1, collecting multi-environment data of the fire door; Step S2, standardizing the data to obtain multi-environment data in a unified format and establishing a feature matrix; Step S3, determining the importance of each data item in the feature matrix and deriving a comprehensive weight; Step S4, constructing a membership function and outputting the fire risk level; Step S5, determining the current fire stage; Step S6, selecting a final risk response strategy using a fuzzy game method; Step S7, controlling the fire door's action according to the selected final risk response strategy. This invention uses information entropy and hierarchical analysis algorithms to jointly determine the importance of multi-environment data and calculate the comprehensive weight of fire risk, making the assessment more objective and comprehensive, avoiding excessive bias in a single model, and solving problems of biased assessment and poor applicability.

Description

Technical Field

[0001] This invention relates to the field of fire management technology, specifically to an intelligent fire door management method and system based on the Internet of Things. Background Technology

[0002] In modern building safety systems, fire doors, as crucial passive fire protection components, play a vital role in preventing the spread of fire, slowing its spread, and ensuring unobstructed evacuation routes. Especially in large shopping malls, hospitals, rail transit systems, and warehouses, fire doors often work in conjunction with fire alarm systems and emergency broadcast systems to provide multiple safety barriers in the event of a fire. With the increasing prevalence of the Internet of Things (IoT) and sensing technologies, fire doors are being endowed with more and more intelligent functions, enabling real-time perception and dynamic response to environmental conditions, thus becoming an important part of intelligent fire protection systems.

[0003] Current technologies still employ a relatively crude approach. A common practice is to set fixed thresholds: triggering an alarm when smoke concentration exceeds a certain level, or controlling door opening when temperature exceeds a certain value. Therefore, the logic of judging based on a single factor is too idealistic and cannot cover the diverse paths of fire development in reality. Often, carbon monoxide concentration has just increased while other parameters are still within normal ranges, and traditional systems cannot detect the potential risk, leading to excessive bias, inaccurate assessments, and poor applicability. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an intelligent fire door management method and system based on the Internet of Things, which solves the problem that the judgment logic of fire doors based on a single factor in existing technologies is too idealistic, leading to excessive deviation, biased evaluation, and poor applicability.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a smart fire door management method based on the Internet of Things, comprising the following steps:

[0006] Step S1: Collect multi-environmental data for the fire door;

[0007] Step S2: Standardize the multi-environment data to obtain multi-environment data in a unified format, and establish a feature matrix;

[0008] Step S3: The importance of each data item in the feature matrix is ​​determined by the information entropy algorithm and the hierarchical analysis algorithm, respectively, and the comprehensive weight is calculated by calculating the calculation results of the information entropy algorithm and the hierarchical analysis algorithm.

[0009] Step S4: Construct a membership function based on the unified format multi-environment data and the comprehensive weight as input. The membership function is used to assess the current environmental state and output the fire risk level.

[0010] Step S5: Based on the assessed fire risk level, determine the current fire stage of the environment using Bayesian inference methods.

[0011] Step S6: Based on the current fire stage of the environment, select the final risk response strategy through fuzzy game theory and issue an alarm message, while uploading the alarm message to the cloud.

[0012] Step S7: Control the movement of the fire door according to the selected final risk response strategy, and record the fire door data and upload it to the cloud.

[0013] Preferably, in step S1, the multiple environmental data include fire door temperature, smoke concentration around the fire door, carbon monoxide concentration around the fire door, fire door vibration, and the rate of change of fire door surface temperature.

[0014] The environmental data is acquired through a variety of sensors installed in the fire door area, including temperature sensors, smoke sensors, carbon monoxide sensors, vibration sensors, and infrared thermal imaging sensors.

[0015] Preferably, in step S2, the standardization process uses an interval scaling method to linearly transform each environmental data item according to its corresponding historical maximum and minimum values, so that the values ​​of multiple environmental data in a unified format are all between 0 and 1.

[0016] Preferably, in step S3:

[0017] The information entropy algorithm is used to calculate the objective weights of various environmental data.

[0018] The hierarchical analysis algorithm calculates subjective weights through a judgment matrix;

[0019] The objective weights and the subjective weights are weighted and synthesized according to a preset ratio to form a comprehensive weight for assessing the fire status.

[0020] Preferably, in step S4, the fire risk level includes: safe status, early warning status, initial fire status, and developing fire status.

[0021] Preferably, in the process of assessing the fire risk level, a membership function is constructed based on various standardized multi-environmental data. The membership function is used to represent the degree of membership of the multi-environmental data to different fire risk levels. Based on the membership degree of the multi-environmental data in each fire risk level, combined with the comprehensive weight, a fuzzy comprehensive evaluation method is used to calculate the score value of each fire risk level, and finally determine the fire risk level corresponding to the current environmental state.

[0022] Preferably, in step S5, the Bayesian inference method calculates the posterior probability of the fire state based on the fire risk level output by the evaluation model after assessing the current environmental state and the prior probability of the fire state in historical statistics. The state with the highest posterior probability is the fire stage of the current environment.

[0023] Preferably, in step S6, the risk response strategy includes opening the fire door, keeping it closed, and activating the alarm device;

[0024] The fuzzy game theory method uses the fire stage as input to establish a risk assessment value for the risk response strategy, and selects the strategy with the lowest risk value as the final response strategy based on the current fire stage of the environment.

[0025] Preferably, in step S7, the fire door data includes: fire door multi-environment data, fire risk level, fire stage, selected response strategy, and fire door actions.

[0026] An IoT-based intelligent fire door management system includes:

[0027] The multi-environment data acquisition module is used to collect various environmental data of the fire door and its surroundings, including temperature, smoke concentration, carbon monoxide concentration, door vibration, and surface temperature change rate.

[0028] The data processing module is used to standardize the multi-environment data, generate processed data in a unified format, and establish a feature matrix.

[0029] The weight calculation module is used to determine the importance of each environmental data item in the multi-environment data based on the feature matrix using the information entropy algorithm and the hierarchical analysis algorithm, and to calculate the comprehensive weight.

[0030] The status assessment module is used to construct an assessment model based on the unified format of multi-environment data and the comprehensive weights, assess the current environmental status, and output the fire risk level.

[0031] The fire stage reasoning module is used to determine the current fire stage of the environment based on the fire risk level using Bayesian reasoning methods.

[0032] The response strategy decision module is used to select the final risk response strategy based on the current fire stage using a fuzzy game method, and generate corresponding alarm information.

[0033] The control execution module is used to control the opening or closing of fire doors according to the final risk response strategy, issue alarm information, and record fire door data;

[0034] The communication module is used to upload the alarm information and the recorded fire door data to the cloud server.

[0035] This invention provides an intelligent fire door management method and system based on the Internet of Things (IoT). It has the following beneficial effects:

[0036] 1. This invention uses information entropy and hierarchical analysis algorithms to determine the importance of multi-environment data, ultimately calculating the comprehensive weight for fire risk assessment. This results in a more objective and comprehensive assessment, avoiding excessive bias in assessments based on a single model. Compared to most existing technologies that use fixed thresholds or single-factor judgments, this invention significantly improves the accuracy and reliability of risk identification, solving the problems of biased and impractical judgments in the past.

[0037] 2. This invention constructs a feature matrix of multi-environmental data and uses fuzzy comprehensive evaluation and Bayesian inference to jointly assess the fire level and its development stage. It can dynamically perceive and proactively identify state evolution in the actual environment, avoiding the risks of response lag and misjudgment.

[0038] 3. This invention provides a decision-making basis for the response strategy of fire doors by adopting fuzzy game theory and dynamically selecting the optimal action in real-time fire stages. This allows the device to flexibly adjust the opening and closing status of fire doors, solving the problem of rigid response strategies and inability to cope with changing fire conditions in the original technology, and significantly enhancing the level of intelligence. Attached Figure Description

[0039] Figure 1 This is a schematic diagram of the method flow of the present invention;

[0040] Figure 2 This is a schematic diagram of the system architecture of the present invention. Detailed Implementation

[0041] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0042] To better understand the present invention, the above content will be described in detail below with reference to specific embodiments.

[0043] Please see the appendix Figure 1 This invention provides an IoT-based intelligent fire door management method, comprising the following steps:

[0044] Step S1: Collect multi-environmental data for the fire door;

[0045] In this embodiment, step S1 is used to complete the multi-dimensional perception of the fire door and its surrounding environment, and is the starting point of the entire IoT-based smart fire door management method. This step provides key input data for subsequent processes such as data standardization, fire risk assessment, and strategy response, and its completeness and accuracy directly affect the overall performance of the system.

[0046] Therefore, in step S1, the multi-environmental data collected by the present invention includes, but is not limited to, the following types:

[0047] Surface temperature of the fire door body;

[0048] The smoke concentration in the air around the fire door;

[0049] The concentration of carbon monoxide (CO) in the air around the fire door;

[0050] The degree of vibration experienced by the fire door;

[0051] The rate at which the surface temperature of the fire door changes over time.

[0052] In one specific implementation, the above environmental parameters are all collected in real time by multiple sensors installed on and near the fire door. Sensor types include, but are not limited to:

[0053] Temperature sensor: Used to sense the real-time temperature of the door surface or the surrounding environment;

[0054] Smoke sensor: Used to detect the concentration of smoke particles in the air, reflecting possible combustion behavior;

[0055] Carbon monoxide sensor: Used to monitor changes in CO concentration in the air to identify potential accumulations of harmful gases;

[0056] Vibration sensor: used to detect whether the door is vibrating, combined with external impacts or abnormal structural changes;

[0057] Infrared thermal imaging sensor: used to capture images of the thermal distribution of the door and further used to calculate the rate of temperature change.

[0058] Generally, sensor deployment should cover different heights of the door and multiple areas of adjacent walls to improve data integrity and system robustness. For example, a set of multi-functional sensor units can be deployed at the top, middle, and bottom of the door to ensure coverage of dimensions such as heat distribution, gas concentration, and mechanical disturbance.

[0059] In some embodiments, the data acquisition cycle is set to no more than 1 second to meet the rapid response requirements in the early stages of a fire. Alternatively, the acquisition cycle can also be dynamically adjusted based on real-time data fluctuations, system load status, or set safety policies. For example, when an abnormal temperature rise trend is detected, the sampling interval can be automatically shortened to 0.2 seconds to enhance response sensitivity.

[0060] To reduce the impact of noise interference or data anomalies on system judgment, this invention also introduces filtering algorithms or preprocessing mechanisms during the data acquisition phase. For example:

[0061] Vibration data are processed by a moving average filter after acquisition to suppress occasional high-frequency jitter;

[0062] Before being used to calculate the rate of temperature change, infrared thermal image data undergoes grayscale enhancement and edge sharpening to improve the contrast of the time series.

[0063] In one specific implementation, the rate of temperature change on the surface of the fire door can be calculated using the following formula:

[0064] ;

[0065] in, This represents the rate of temperature change per unit time (unit: °C / s).

[0066] The current time point represents the surface temperature of the door as obtained by the infrared thermal imaging sensor.

[0067] The temperature value recorded in the previous sampling period;

[0068] , These are the current and previous timestamps, respectively, in seconds.

[0069] The rate of temperature change is an important dynamic characteristic indicator, especially suitable for identifying "rapid heating" scenarios, which are common in sudden fire scenarios such as electrical short circuits and oil combustion. In contrast, "slow heating" may occur in environments with heat source penetration or shielding. This indicator can effectively assist in distinguishing fire risk levels and judging evolution paths.

[0070] In some embodiments, the outputs of the aforementioned multiple sensors can be transmitted to the local edge computing unit via standard communication interfaces (such as RS485, I2C, or CAN bus) for preliminary data integration and caching, thereby reducing network dependence and enhancing the distributed processing capabilities of the system.

[0071] Furthermore, in the extended implementation, the system can also be configured with an event-driven sampling mechanism. When the value collected by a certain sensor exceeds a preset threshold (such as smoke concentration exceeding a certain limit), the entire acquisition subsystem will be triggered to enter "high-frequency mode" to monitor parameter changes with higher time resolution and improve the accuracy of early warning for emergencies.

[0072] In summary, step S1 achieves high-frequency acquisition of multi-source environmental elements. Through the fusion deployment of heterogeneous sensors, dynamic sampling control, redundant layout and pre-processing algorithm collaboration, the system’s sensitivity and real-time performance in identifying early signs of fire are greatly improved, laying a solid perceptual foundation for subsequent intelligent judgment and response.

[0073] Step S2: Standardize the multi-environment data to obtain multi-environment data in a unified format, and establish a feature matrix;

[0074] In this embodiment, step S2 is an important processing step that follows multi-environment data acquisition (step S1). Its main purpose is to standardize and transform the raw environmental data from different sources and with different dimensions so that they can be uniformly analyzed and calculated in the future.

[0075] Specifically, since environmental parameters such as temperature, smoke concentration, carbon monoxide concentration, vibration intensity, and temperature change rate have different physical dimensions and numerical ranges, directly using the raw data will lead to inconsistent feature scales, thus affecting the accuracy of subsequent feature weight calculations and risk modeling. Therefore, in step S2 of this invention, all collected environmental data are subjected to interval scaling processing, so that all feature values ​​are mapped to the same standardized interval, i.e., within the range of [0,1], forming a multi-environment dataset with a unified format.

[0076] Specifically, the interval scaling standardization method uses the following transformation formula:

[0077] ;

[0078] in, Indicates the first The standardized values ​​of the environmental data are in the range of [0,1].

[0079] This represents the actual raw environmental data values ​​collected.

[0080] Indicates the first The historical minimum value of the environmental data is based on long-term observation data or system initialization settings.

[0081] Indicates the first The historical maximum value of an environmental data item is usually a preset limit value or an empirical statistical value.

[0082] In general, the maximum and minimum values ​​of all features are obtained by statistical analysis of historical operating datasets and stored in the system database. In the case of initial system deployment or lack of historical data, the limit values ​​can also be preset by experts based on their experience and used as the initial range.

[0083] In one specific implementation, standardization can dynamically adjust the maximum and minimum values ​​based on a sliding time window to improve adaptability to environmental changes. For example, the system can set a 24-hour sliding window period to update the historical extreme values ​​of each feature to reflect short-term trends.

[0084] Alternatively, this invention also supports multi-level management of maximum and minimum values ​​based on context. For example, for temperature data, the extreme value ranges differ significantly between winter and summer operating environments. The system can automatically switch the interval setting based on the current seasonal label or external meteorological interface to improve standardization accuracy.

[0085] After standardization, the system synchronizes and integrates the data according to timestamps, forming a unified multi-environment data vector. Data from multiple time points can be further combined sequentially to construct a feature matrix in the following form:

[0086] ;

[0087] in,

[0088] This represents the standardized feature matrix with dimension 1. ;

[0089] Indicates the number of samples in the time series;

[0090] The number of dimensions representing environmental characteristics (in this embodiment, the number is 5, corresponding to temperature, smoke concentration, CO concentration, vibration, and temperature change rate).

[0091] Indicates the first At the 1st time point, the 1st Standardized values ​​of feature items ( , ).

[0092] In some embodiments, this feature matrix serves as the input data structure for subsequent weight calculation modules, and can also be used for training or online inference of the fire identification model. In this way, data from different sensor sources are unified and structured, greatly improving the efficiency of cross-modal feature fusion.

[0093] To ensure data stability and model convergence, some outlier or abnormal data points can be removed or smoothed before entering the standardization module. For example, when a feature value experiences an instantaneous jump exceeding the historical mean ± 3 standard deviations, data filtering logic can be triggered to automatically postpone its inclusion in the standardization process.

[0094] In one possible system deployment approach, standardized modules are deployed on edge node processing platforms to ensure low-latency response and localized processing capabilities, which is particularly suitable for scenarios with dense sensor deployments or edge environments with unstable network connections.

[0095] In summary, step S2 not only unifies the scale and structures the dimensions of various environmental features, but also ensures the stability of data input and the compatibility of subsequent processing links through flexible interval scaling strategies, multi-source historical data fusion, dynamic window mechanisms, and abnormal data processing logic, thereby further guaranteeing the practicality and reliability of the fire identification system.

[0096] Step S3: Use the information entropy algorithm and the hierarchical analysis algorithm to determine the importance of each data item in the feature matrix, and calculate the comprehensive weight based on the calculation results of the information entropy algorithm and the hierarchical analysis algorithm.

[0097] In this embodiment, step S3 follows the generation of the standardized multi-environment feature data matrix (step S2). Its core purpose is to establish a multi-index weight evaluation mechanism that combines data-driven and expert knowledge, thereby providing a comprehensive weight coefficient with discriminative ability for subsequent fire status identification.

[0098] Specifically, since different environmental characteristics exhibit varying degrees of differentiation in different fire scenarios, relying solely on objective data statistics is insufficient to fully reflect their importance in actual judgment. Therefore, this invention combines the information entropy algorithm with the Analytic Hierarchy-Process (AHP) algorithm to calculate objective weights and subjective weights respectively, and further forms a unified comprehensive weight vector through weighted fusion, which serves as one of the inputs to the subsequent risk identification model.

[0099] Specifically, the information entropy algorithm is used to calculate the objective weights of each environmental feature in the feature matrix, and the process is as follows:

[0100] Based on the aforementioned feature matrix It can be seen that , indicating the first Time of the first Standardized values ​​of the features.

[0101] Based on this, firstly, calculate the... probability distribution values ​​of feature items The calculation formula is:

[0102] ;

[0103] in, Indicates the number of samples in the time series;

[0104] The number of dimensions representing environmental characteristics (in this embodiment, the number is 5, corresponding to temperature, smoke concentration, CO concentration, vibration, and temperature change rate).

[0105] Then, calculate the first... Information entropy of a feature , the specific calculation formula is:

[0106] ;

[0107] in, , is the normalization coefficient, to ensure the information entropy value ;

[0108] when When, define .

[0109] Furthermore, we obtain the first Entropy weight of feature , the specific calculation formula is:

[0110] ;

[0111] in, Indicates the first Objective weight values ​​for environmental characteristics;

[0112] Therefore, the entropy weight mentioned above reflects the degree of data variation. The greater the variation, the higher the contribution to the system state judgment, and thus the greater its objective weight.

[0113] In one specific implementation, the Analytic Hierarchy Process (AHP) is used to calculate the subjective weights of each feature term, mainly including the following steps:

[0114] ;

[0115] in, Indicates the first Subjective weight values ​​for environmental features; Represents the th element in the characteristic matrix The largest eigenvector of an environmental feature; This represents the sum of each largest eigenvector in the feature matrix;

[0116] In one possible implementation, the subjective and objective weights are weighted and fused in the following manner to obtain the final comprehensive weight:

[0117] ;

[0118] in, Indicates the first The comprehensive weight of each environmental characteristic;

[0119] The proportion of objective weight. The proportion of subjective weight, and satisfying .

[0120] In summary, step S3 introduces two weighting strategies, information entropy and hierarchical analysis, to quantify the importance of features from the perspectives of data statistical distribution and expert experience, respectively. It also establishes a comprehensive weighting system for fire status assessment through weighted fusion. This step maintains the sensitivity of the model while improving the rationality and adaptability of the weighting configuration, providing a scientific basis for subsequent status modeling, alarm threshold judgment, and dynamic response strategies.

[0121] Step S4: Construct a membership function based on multi-environment data in a unified format and comprehensive weights as input. The membership function is used to assess the current environmental state and output the fire risk level.

[0122] In this embodiment, step S4 follows immediately after the generation of the comprehensive weight vector calculated based on information entropy and the hierarchical analysis method (step S3). The comprehensive weight is used to perform fuzzy evaluation on the standardized multi-environment data to construct a fuzzy recognition model, which is used to realize intelligent identification of the fire risk level under the current environmental conditions.

[0123] Specifically, the core of this step is to establish a fuzzy mapping relationship between multiple environmental characteristics and fire risk levels, so as to achieve quantitative assessment of different risk states (such as safety, early warning, initial fire, and developing fire).

[0124] To achieve this function, this invention introduces a fuzzy membership function model, which constructs the membership degree expression form of each type of environmental feature under each fire level, and finally combines the comprehensive weights to obtain the fire risk level score result of the current state through the fuzzy comprehensive evaluation algorithm.

[0125] Specifically, the membership function is constructed as follows:

[0126] For standardized environmental characteristic values Its membership degree under each fire risk level is expressed as ,in, Indicates the fire risk level, with a value of [value missing]. These correspond to "safe status", "early warning status", "initial fire status" and "developing fire status" respectively;

[0127] This indicates that the feature value belongs to a level. The degree of membership.

[0128] In one specific implementation, membership functions can be modeled using triangular, trapezoidal, or Gaussian functions to improve the system's adaptability to uncertainties. For example, when using a triangular membership function, its general expression is:

[0129] ;

[0130] in, Indicates the first Item features at the level The left endpoint of the membership function;

[0131] This indicates the position with a membership degree of 1, corresponding to the maximum membership point;

[0132] Indicates the right endpoint;

[0133] The graph of the function shows a peak located at The triangle at that location.

[0134] And the above parameters, , It can be set based on historical monitoring samples and domain knowledge, or trained through sample fitting algorithms.

[0135] In the fuzzy comprehensive evaluation stage, this embodiment adopts the following calculation process:

[0136] Fire rating The overall membership score is as follows The calculation method is as follows:

[0137] ;

[0138] in, The first one calculated in step S3 The comprehensive weight of each environmental characteristic;

[0139] For the current eigenvalue In level Membership degree;

[0140] For the current input in the th Fuzzy rating values ​​for each fire level.

[0141] Finally, the ratings for all levels are combined into a fuzzy output vector:

[0142] ;

[0143] In one possible implementation, a threshold enhancement mechanism can be introduced to suppress the sensitivity to false positives.

[0144] That is when When the threshold is exceeded (for example, the threshold is set to 0.5), a flag value of "state uncertain" or "determination required" can be output to improve the robustness of the discrimination system.

[0145] Generally, the membership functions at each level of fuzzy comprehensive evaluation should have overlapping regions to ensure the identifiability of the state in the fuzzy boundary region. For example, a certain overlapping region can be set between "early warning state" and "initial fire state" for transitional identification.

[0146] In some embodiments, the present invention can construct multiple sets of membership function templates to meet the fire response requirements of different application scenarios.

[0147] For example, residential environments and industrial warehousing environments have significant differences in their environmental characteristics and fire evolution characteristics. The system can automatically switch the membership function configuration according to the equipment deployment scenario to improve adaptability.

[0148] As an alternative, an adaptive membership function adjustment strategy can be adopted, which involves periodically optimizing the membership function parameters using methods such as cluster analysis based on continuous data collection, thereby maintaining the discriminative ability in sync with environmental evolution.

[0149] To achieve a clear output of fire risk classification, this invention defines fire risk levels. It includes the following four categories:

[0150] Safety Status (Level-0): Indicates that the current environment is in a normal and safe operating state;

[0151] Warning status (Level-1): Some characteristic values ​​exceed the normal range but do not yet pose a fire threat;

[0152] Initial fire status (Level-2): There are clear signs of abnormality, and the characteristics of an early fire are present.

[0153] Developing fire status (Level-3): Multiple characteristic values ​​deviate significantly from normal levels, indicating obvious signs of fire.

[0154] The system uses the comprehensive membership score vector corresponding to the current input data. The highest scorer corresponds to the current fire status level, which can be used to drive subsequent alarm responses, fire linkage, or control strategies.

[0155] In summary, step S4 constructs a membership mapping between multiple environmental features and fire levels, combines it with a comprehensive weight vector, and introduces a fuzzy comprehensive evaluation method to achieve effective fusion and accurate identification of multi-source data in different fire level discrimination tasks. This mechanism exhibits good robustness in handling environmental uncertainty, provides the system with a clear and progressive risk level output strategy, and significantly enhances the engineering adaptability and practical performance of the intelligent fire identification system.

[0156] Step S5: Based on the assessed fire risk level, determine the current fire stage of the environment using Bayesian inference methods.

[0157] In this embodiment, the Bayesian inference method bases the fire risk level output by the evaluation model after assessing the current environmental state and the prior probability of the fire state from historical statistics to calculate the posterior probability of the fire state. The state with the highest posterior probability is the fire stage of the current environment.

[0158] In general, a single fire level output is insufficient to reflect the continuity and diversity of fire evolution. Therefore, in this invention, a Bayesian probability model is used to establish a mapping reasoning framework of "risk level - fire stage", enabling the system to further identify the stage based on existing risk level information, such as the stage evolution process from "fire source generation" to "spread and development" and then to "full combustion".

[0159] Specifically, in a Bayesian model, the required parameters include the following two categories:

[0160] Prior probability : Indicates the probability of occurrence of each fire stage in historical statistics, which can be obtained by frequency statistics of stage labels in historical monitoring datasets;

[0161] Likelihood function : Indicates the stage of a fire The system determines the current fire risk level as follows: The probability can be constructed based on the co-occurrence frequency in the training samples or empirical rules. For example, the probability of a fire risk level occurring in step 4 can be statistically calculated.

[0162] ;

[0163] in, For posterior probability, This indicates the total number of fire stages divided by the system, such as: no fire, initial fire, ongoing combustion, and spread. Each stage corresponds to one... ;

[0164] The final output stage judgment results are as follows:

[0165] ;

[0166] in, This represents the most likely fire stage at present, which is the stage with the highest posterior probability value.

[0167] The results of this stage will serve as key inputs for subsequent linkage modules (such as alarm triggering and fire response decisions).

[0168] Alternatively, the present invention can further optimize the method of obtaining the prior probability and the likelihood function.

[0169] In some implementations, the system can periodically update the prior probabilities. This is to reflect the dynamic changes in the evolution trend of fires in different scenarios or seasons.

[0170] In addition, the likelihood function Alternatively, it can be established based on Bayesian network structure learning or conditional probability table estimation methods, and the inference accuracy can be improved through continuous learning mechanisms. For example, in an industrial environment, the level distribution corresponding to stage 3 may be biased towards "initial fire state", while in a residential environment it may be biased towards "early warning state", so a scenario-specific modeling mechanism needs to be introduced.

[0171] In some embodiments, to enhance model interpretability and robustness, this invention supports outputting a complete posterior probability vector. :

[0172] ;

[0173] This a priori probability distribution can then be further used by higher-level policy modules, for example:

[0174] If the highest probability is only slightly higher than the second highest value, the "transition period" state can be output.

[0175] Alternatively, a risk range can be set to control the uncertainty of posterior distribution fluctuations.

[0176] Specifically, this is based on the risk level output from step S4 above. This step inputs the value as an observation variable into the Bayesian inference module and links it with the system's built-in fire stage probability model to build an inference link from level to stage.

[0177] For example:

[0178] If an input is determined to be "initial fire state" (level 2) by step S4, and historical statistics show that this level has the highest probability of occurring in stage 3 (continuous fire state), then the system outputs that the current stage is 3.

[0179] This reasoning process, while maintaining low complexity of the input data, enhances the mapping from risk level to fire stage, forming a complete and hierarchical fire judgment system.

[0180] In summary, step S5, by introducing a Bayesian inference model, constructs a risk level-driven probabilistic inversion mechanism for fire stages. Based on fully utilizing historical statistical knowledge and current state information, this step achieves dynamic estimation of the fire state evolution path, providing scientific support for the accurate matching of subsequent response strategies and improving the overall intelligence level and judgment reliability of the system.

[0181] Step S6: Based on the current fire stage, select the final risk response strategy through fuzzy game theory and issue an alarm message, while uploading the alarm message to the cloud.

[0182] In this embodiment, step S6 follows immediately after the fire stage determination result completed in step S5. Based on the current fire development stage, a fuzzy game analysis mechanism is further introduced to select the optimal response path from multiple possible risk response strategies, and alarm information is generated and uploaded in real time. This step aims to achieve strategy adaptation and risk avoidance in different environmental situations while ensuring timely response, based on the response priority and risk-reward relationship corresponding to different fire stages. Together with the aforementioned fuzzy comprehensive evaluation (step S4) and Bayesian inference stage identification (step S5), it forms a closed-loop intelligent response system, jointly constructing an intelligent fire response logic link of "level identification - stage judgment - strategy linkage".

[0183] Generally, the assessment of a fire stage cannot be directly mapped to a single response action, especially in situations where multiple strategies coexist and their influences are mutually exclusive. A single decision criterion is insufficient to effectively weigh response priorities in complex scenarios. Therefore, this invention further constructs a fuzzy game model, using the current fire stage as the decision input variable and combining it with risk assessment functions for different strategies at the current stage, selecting a response strategy based on the principle of minimum risk.

[0184] Specifically, the response strategy selection method driven by fuzzy game theory mainly includes the following:

[0185] Define the fire response strategy set as follows:

[0186] ;

[0187] in: Indicates the first One optional response strategy ;

[0188] In one possible implementation, the response strategy includes:

[0189] Maintain the current state (do not react);

[0190] : Activate the alarm device;

[0191] Automatically opens fire doors or switches fire extinguishing modes;

[0192] Define the fire stage set as:

[0193] ;

[0194] The current stage of the environment is determined by step S5;

[0195] Furthermore, based on this, a fuzzy risk assessment function is introduced. This indicates the fire stage. Execution strategy The intensity of the risks involved.

[0196] At this point, the core formula for fuzzy game strategy selection is:

[0197] ;

[0198] in, The final response strategy selected;

[0199] In the current stage Execution strategy The corresponding risk value.

[0200] Regarding risk assessment functions The construction method is as follows:

[0201] In some embodiments, the risk function can be constructed by a fuzzy rule inference system, setting the following fuzzy rules:

[0202] If the fire is in its initial stage, the risk of triggering an alarm is low, the risk of opening the fire door is medium, and the risk of maintaining the current status is high.

[0203] If the situation is in a "high-risk spread state", then the risk of "maintaining the status quo" is extremely high, while the risks of "alarming" and "opening the fire door" are relatively low.

[0204] If in a "no-fire state", the risk values ​​of all strategies are relatively high, but "staying current" has the lowest risk.

[0205] The degree of fuzziness in the risk intensity of a strategy can be described by using triangular membership functions or Gaussian membership functions. A fuzzy rule base can be constructed, and fuzzy reasoning can be performed to obtain the fuzzy risk value of each strategy. Then, the centroid method or the maximum membership method can be used to defuzzify the risk and obtain the specific risk quantification value.

[0206] Specifically, in one possible implementation, the fuzzy game system includes the following elements:

[0207] Input item: The current fire stage as determined ;

[0208] Rule base: A fuzzy set of rules for strategy risks at each stage;

[0209] Inference engine: Employs fuzzy synthesis method for risk assessment between strategies;

[0210] Output: Selected optimal response strategy .

[0211] For example:

[0212] If the current determination is a stage (Continuous fire state), the fuzzy rule system evaluation yields the following results:

[0213] ;

[0214] ;

[0215] ;

[0216] The final response strategy at this point is:

[0217] ;

[0218] This means executing the "activate alarm device" strategy.

[0219] As an option, the response strategy execution results will simultaneously generate alarm information and upload it to the cloud management platform.

[0220] In some implementations, alarm information includes, but is not limited to:

[0221] Current fire stage ;

[0222] Response strategy number ;

[0223] System timestamp, monitoring location coordinates, fire level Auxiliary information;

[0224] Alarm information is organized in a structured data format (such as JSON or XML) and uploaded to the cloud platform via standard communication protocols (such as MQTT or HTTPS) for remote response deployment or platform-level data archiving.

[0225] In some embodiments, to enhance the flexibility and scalability of response strategy selection, the system supports dynamically adjusting the strategy set. and risk assessment function The parameter set.

[0226] For example:

[0227] For large industrial sites, additional strategy items can be added, such as "trigger sprinkler system" and "isolate power module".

[0228] For residential buildings, the risk function structure can be dynamically adjusted according to scenarios such as night / day or population density, making the response strategy more adaptable to different scenarios.

[0229] In summary, step S6, based on the fire stage determination, introduces a fuzzy game theory approach and constructs a multi-strategy response risk model to achieve response strategy selection driven by the minimum risk criterion. This mechanism balances response efficiency and risk control, ensuring the most appropriate response actions are taken at different fire stages. Simultaneously, it enables the systematic uploading of alarm information, providing timely and accurate data for platform-level response deployment and significantly enhancing the system's intelligent linkage control capabilities.

[0230] Step S7: Control the movement of the fire door according to the selected final risk response strategy, and record the fire door data and upload it to the cloud.

[0231] In this embodiment, step S7 follows immediately after the risk response strategy selection in step S6. It further executes the response strategy through the intelligent control system to automate the operation of the fire door and record and upload the corresponding data to the cloud. The core of this step is to precisely control the opening and closing status of the fire door according to the selected final risk response strategy, and to ensure that each operation record and environmental data is uploaded to the cloud platform in a timely and accurate manner for data archiving and subsequent remote monitoring.

[0232] Generally, response strategies for fire scenarios must consider not only the real-time development of the fire but also the flexibility and stability of the system. Especially in high-risk environments, the opening and closing of fire doors must be precisely controlled according to the specific fire stage and the requirements of the response strategy. Therefore, this invention addresses the risk response strategy selected in step S6. The interpretation, combined with the fire door control system, enables precise execution and real-time feedback mechanisms, ensuring that every response meets predetermined safety standards and system requirements.

[0233] Specifically, the execution of fire door actions needs to be coordinated with the system's built-in sensors and execution modules to ensure that the fire door actions are accurate.

[0234] In some embodiments, the system can also dynamically adjust the operation of fire doors according to the specific fire scenario and its changes. For example, during the spread of a fire, some areas may automatically adjust to close fire doors to prevent the fire from spreading, while at other stages, fire doors may automatically open to expel smoke or gases. Through precise control, each operation can react to the needs of the real-time environment, improving the timeliness and safety of the response.

[0235] Technical implementation of fire door data recording and uploading:

[0236] In this embodiment, each action of the fire door generates corresponding data records, including but not limited to:

[0237] Operation timestamp;

[0238] Current stage of the fire;

[0239] Operation type (e.g., on or off);

[0240] Current environmental fire level;

[0241] Operation execution status (whether it was executed successfully), etc.

[0242] This data will be organized in a structured data format (such as JSON or XML) using secure encryption and uploaded to the cloud management platform. The specific data upload process includes:

[0243] Package the data into a standard format;

[0244] Send to the cloud platform using standard communication protocols (such as MQTT, HTTPS, etc.);

[0245] Data storage and real-time monitoring are performed in the cloud to provide foundational data for subsequent remote response and decision-making.

[0246] Specifically, in one possible implementation, the technical solution for the data upload process is as follows:

[0247] Data collection: The fire door control system records details of each operation, such as timestamps, current fire stage, and operation type;

[0248] Data formatting: Convert all data into a uniform format (such as JSON) to ensure data consistency and parsability during the upload process;

[0249] Data encryption: Encrypting data to ensure its privacy and security;

[0250] Data upload: Data is uploaded to the remote monitoring system in real time through communication protocols supported by the cloud platform (such as MQTT, HTTPS, etc.);

[0251] Cloud platform processing: The cloud platform stores and analyzes the uploaded data, supporting functions such as real-time display, trend prediction, and alarm linkage.

[0252] As an alternative, the system can simultaneously transmit real-time fire stage and response strategy information during data upload, facilitating big data analysis and intelligent decision-making on the cloud platform.

[0253] For example, when a fire door opens, the cloud platform can simultaneously receive information about the door's opening status and the current fire stage (such as "high-risk spread status"). Through historical data analysis, the platform can dynamically optimize and adjust response strategies to ensure the system is more efficient and intelligent in the next response.

[0254] In some embodiments, to improve the adaptability of the system, the system can adjust the fire door control strategy according to changes in the fire stage.

[0255] For example, when the fire is determined to be in a "continuous fire state", the system may automatically close some fire doors to prevent the fire from spreading, but in the "initial fire state", it may open the fire doors to allow smoke to be vented in time.

[0256] Through real-time monitoring and dynamic adjustments, the system can be ensured to respond flexibly to changes in specific fire scenarios.

[0257] In summary, step S7, by introducing a fire door control and data upload mechanism, ensures the accurate operation of fire doors while promptly recording and uploading relevant data to the cloud. This mechanism not only improves the efficiency of the response system but also provides valuable data support for subsequent decision analysis, further enhancing the system's intelligence and adaptability.

[0258] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A smart fire door management method based on the Internet of Things, characterized in that, Includes the following steps: Step S1: Collect multi-environmental data for the fire door; Step S2: Standardize the multi-environment data to obtain multi-environment data in a unified format, and establish a feature matrix; Step S3: The importance of each data item in the feature matrix is ​​determined by the information entropy algorithm and the hierarchical analysis algorithm, respectively, and the comprehensive weight is calculated by calculating the calculation results of the information entropy algorithm and the hierarchical analysis algorithm. Step S4: Construct a membership function based on the unified format multi-environment data and the comprehensive weight as input. The membership function is used to assess the current environmental state and output the fire risk level. Step S5: Based on the assessed fire risk level, determine the current fire stage of the environment using Bayesian inference methods. Step S6: Based on the current fire stage of the environment, select the final risk response strategy through fuzzy game theory and issue an alarm message, while uploading the alarm message to the cloud. Step S7: Control the movement of the fire door according to the selected final risk response strategy, and record the fire door data and upload it to the cloud; In step S3: The information entropy algorithm is used to calculate the objective weights of various environmental data. The hierarchical analysis algorithm calculates subjective weights through a judgment matrix; The objective weight and the subjective weight are weighted and synthesized according to a preset ratio to form a comprehensive weight for assessing the fire status. In the process of assessing the fire risk level, a membership function is constructed based on various standardized multi-environment data. The membership function is used to represent the degree of membership of the multi-environment data to different fire risk levels. Based on the membership degree of multiple environmental data in each fire risk level, combined with the comprehensive weight, the fuzzy comprehensive evaluation method is used to calculate the score value of each fire risk level, and finally determine the fire risk level corresponding to the current environmental state. In step S5, the Bayesian inference method calculates the posterior probability of the fire state based on the fire risk level output by the evaluation model after assessing the current environmental state and the prior probability of the fire state in historical statistics. The state with the highest posterior probability is the fire stage of the current environment. In step S6, the risk response strategy includes opening the fire door, keeping it closed, and activating the alarm device; The fuzzy game theory method uses the fire stage as input to establish a risk assessment value for the risk response strategy, and selects the strategy with the lowest risk value as the final response strategy based on the current fire stage of the environment.

2. The intelligent fire door management method based on the Internet of Things according to claim 1, characterized in that, In step S1, the multiple environmental data include fire door temperature, smoke concentration around the fire door, carbon monoxide concentration around the fire door, fire door vibration, and the rate of change of fire door surface temperature. The environmental data is acquired through a variety of sensors installed in the fire door area, including temperature sensors, smoke sensors, carbon monoxide sensors, vibration sensors, and infrared thermal imaging sensors.

3. The intelligent fire door management method based on the Internet of Things according to claim 1, characterized in that, In step S2, the standardization process uses an interval scaling method to linearly transform each environmental data item according to its corresponding historical maximum and minimum values, so that the values ​​of multiple environmental data in a unified format are all between 0 and 1.

4. The intelligent fire door management method based on the Internet of Things according to claim 1, characterized in that, In step S4, the fire risk level includes: safe status, early warning status, initial fire status, and developing fire status.

5. The intelligent fire door management method based on the Internet of Things according to claim 1, characterized in that, In step S7, the fire door data includes: fire door multi-environment data, fire risk level, fire stage, selected response strategy, and fire door actions.

6. An IoT-based intelligent fire door management system, based on the IoT-based intelligent fire door management method according to any one of claims 1-5, characterized in that, include: The multi-environment data acquisition module is used to collect various environmental data of the fire door and its surroundings, including temperature, smoke concentration, carbon monoxide concentration, door vibration, and surface temperature change rate. The data processing module is used to standardize the multi-environment data, generate processed data in a unified format, and establish a feature matrix. The weight calculation module is used to determine the importance of each environmental data item in the multi-environment data based on the feature matrix using the information entropy algorithm and the hierarchical analysis algorithm, and to calculate the comprehensive weight. The status assessment module is used to construct an assessment model based on the unified format of multi-environment data and the comprehensive weights, assess the current environmental status, and output the fire risk level. The fire stage reasoning module is used to determine the current fire stage of the environment based on the fire risk level using Bayesian reasoning methods. The response strategy decision module is used to select the final risk response strategy based on the current fire stage using a fuzzy game method, and generate corresponding alarm information. The control execution module is used to control the opening or closing of fire doors according to the final risk response strategy, issue alarm information, and record fire door data; The communication module is used to upload the alarm information and the recorded fire door data to the cloud server.

Images