Fire-fighting remote monitoring and early warning method based on Internet of Things

Abstract

This invention discloses a fire remote monitoring and early warning method based on the Internet of Things (IoT), belonging to the technical field of IoT. The method includes deploying IoT monitoring terminals at key monitoring points of large equipment in a factory to collect real-time fire-related parameters such as temperature, combustible gas, smoke, and flames, as well as equipment image data. After preprocessing, the data is transmitted to a remote monitoring platform via an IoT communication module. By calculating the probability of open flame, smoldering fire, and no-flame, and simultaneously identifying fire characteristics in the images, the method jointly determines the fire occurrence and fire level. Based on the fire level, a graded early warning is triggered, and the factory's fire control system is activated to execute emergency response operations. Simultaneously, all data is stored and traced. This invention solves the problems of delayed fire monitoring, high false alarm and missed alarm rates, and untimely response for large equipment in factories, achieving real-time remote monitoring, accurate early warning, and rapid coordinated response of equipment fire status.

Description

Technical Field

[0001] This invention relates to the field of Internet of Things (IoT) technology, and more particularly to a fire remote monitoring and early warning method based on IoT. Background Technology

[0002] Large equipment in factories, such as machine tools, boilers, lifting equipment, and chemical reactors, are the core carriers of industrial production. During their operation, there are many fire hazards: electrical wiring of the equipment is prone to high-temperature sparks due to aging and overload; hydraulic system interfaces may have oil leaks that may cause flammability risks; fuel / lubricating oil storage and transportation parts are prone to flammable medium leaks; and high-temperature working areas are prone to fires due to temperature runaway. Moreover, such equipment is large in size and complex in structure, with many monitoring points and wide distribution, making it difficult for traditional fire monitoring methods to meet safety management needs.

[0003] Currently, fire monitoring of large equipment in factories mostly relies on a combination of manual inspections and traditional fire alarms, which has significant drawbacks: manual inspections are inefficient and costly, prone to overlooking potential hazards due to human negligence, and cannot achieve 24-hour uninterrupted monitoring; traditional fire alarms are mostly single-point monitoring devices, capable of detecting only a single fire parameter, such as smoke or temperature, with a limited monitoring range, and are easily affected by the high temperature, dust, and flammable and explosive environment of factories, leading to frequent false alarms and missed alarms; at the same time, existing monitoring systems lack remote linkage capabilities, and even if fire hazards or fires are detected, they cannot quickly trigger early warnings and link fire-fighting equipment for response, which can easily lead to the spread of fire and cause significant casualties and property damage.

[0004] With the rapid development of IoT technology, its application in the field of fire monitoring has gradually become widespread. However, existing IoT-based fire monitoring methods are mostly applicable to the overall fire monitoring of buildings, with few specialized monitoring solutions for large equipment in factories. Furthermore, these methods suffer from problems such as low data fusion accuracy, inaccurate fire detection, and poor linkage between early warning and response, failing to fully adapt to the special operating environment and fire monitoring needs of large equipment in factories.

[0005] Therefore, developing a targeted, highly intelligent, accurate, and efficient IoT-based remote monitoring and early warning method for fire protection of large factory equipment has become an urgent technical problem to be solved. Summary of the Invention

[0006] To address the shortcomings of the aforementioned background technologies, this invention proposes a fire remote monitoring and early warning method based on the Internet of Things (IoT) to overcome the deficiencies in existing technologies. The technical solution adopted by this invention is as follows: A fire remote monitoring and early warning method based on the Internet of Things is provided, comprising the following steps: S1. Real-time collection of fire-related parameters and corresponding image data of key monitoring parts of large equipment in the factory through IoT monitoring terminals. Then, noise reduction and calibration of the collected sensor data are performed through data preprocessing, and grayscale and defogging preprocessing of the image data are performed to remove invalid data and obtain standardized monitoring data and standardized image data. S2. Through the Internet of Things (IoT) communication module, standardized monitoring data and standardized image data are transmitted to the remote monitoring platform in real time using wireless communication. S3. Perform fusion calculation on standardized monitoring data, and obtain the probability of open flame, smoldering fire and no fire by combining confidence distance theory. At the same time, analyze standardized image data, identify fire features in the images, and make a joint judgment with the data fusion results and the image fire feature identification results. S4. When the data fusion result exceeds the dynamic risk threshold, a fire is determined to have occurred, triggering a graded early warning instruction. The remote monitoring platform triggers the corresponding level of early warning based on the fire level, and simultaneously links the factory's fire control system and on-site response equipment to issue early warning information and execute emergency response operations.

[0007] Preferably, the IoT monitoring terminal includes multiple sensor modules, an image acquisition module, a data preprocessing module, and an IoT communication module. The key monitoring locations include electrical wiring points, hydraulic system interfaces, fuel / lubricating oil storage and transportation locations, and high-temperature operating areas. The multiple sensor modules include temperature sensors, combustible gas sensors, smoke sensors, and flame sensors. The image acquisition module uses a CCD image sensor to acquire image information of the monitored locations in real time.

[0008] Preferably, the data is encrypted using an encryption algorithm during the wireless communication transmission process.

[0009] Preferably, the calculation steps for the fusion calculation are as follows: Let the data collected by the i-th sensor be... Where i = 1, 2, ..., n, and n is the number of sensors. Follows a normal distribution. for Given the probability density function, calculate the confidence distance between any two sensors i and j. and The formula is as follows: ; ; in , Let y and x be the standard deviations of the k-th and j-th sensor data, respectively, and p(y|x) be the conditional probability density function. Then, the confidence distance threshold was set. According to confidence distance and Determine the mutual support relationship between the data from two sensors, denoted as ,when ≤ hour, =1 indicates mutual support; when > hour, =0 indicates that it is not supported.

[0010] Even better, a binary relationship matrix is ​​constructed based on the mutual support relationship. Combining the weight coefficients of each sensor and the fire characteristics in the identified image, the probability of open flame, the probability of smoldering fire, and the probability of no fire are calculated by weighted summation. When the probability of open flame is ≥80%, it is determined to be an open flame fire. When the probability of smoldering fire is ≥70% and the probability of open flame is <80%, it is determined to be a smoldering fire. When the probability of no fire is ≥90%, it is determined to be no fire.

[0011] More preferably, the image conditions for determining no fire are that no fire features are identified in the image, including no flame, no abnormal smoke, and no obvious high-temperature grayscale anomaly; the image conditions for determining smoldering fire are that smoldering-related features are identified in the image, including light gray / black smoke or abnormally high grayscale values ​​in local areas but no obvious flame outline; and the image conditions for determining open flame fire are that an open flame outline is identified in the image.

[0012] Preferably, the determination of the dynamic risk threshold includes: The current time period's factory equipment scenario type is obtained through IoT monitoring terminals, and the scenario type includes normal operation scenario, maintenance scenario, and shutdown scenario; Preset basic thresholds are matched according to the scenario type. The basic thresholds correspond to normal operation scenarios, maintenance scenarios dynamically reduce the basic thresholds based on the risk of personnel gathering, and shutdown scenarios dynamically further reduce the basic thresholds based on the risk of response delay. Furthermore, by overlaying real-time weather condition impact assessments, when meteorological conditions are detected to exacerbate the risk of fire spread, the basic threshold is adaptively adjusted to obtain a dynamic risk threshold.

[0013] As a preferred approach, fire severity is divided into three levels: Level 1 is a minor hazard, with no open flame and sensor data exceeding the normal range but not reaching the alarm threshold; Level 2 is a general fire, a smoldering fire or a small open flame, not spreading to the core parts of the equipment; and Level 3 is a major fire, with open flame spreading to the core parts of the equipment or the concentration of combustible gas reaching the high alarm threshold. Different levels correspond to different warning and linkage logics. Level 1 hazards only trigger on-site sound warnings and platform recordings; Level 2 fires trigger on-site sound and light warnings, platform pop-up warnings, and the pre-activation of fire extinguishing equipment in the corresponding area; and Level 3 fires trigger plant-wide sound and light warnings, SMS warnings, and emergency pop-ups on the platform. At the same time, the fire extinguishing system is immediately activated, the main power supply to the equipment and surrounding non-fire-fighting power supplies are cut off, the smoke exhaust system is turned on, and alarm information is simultaneously pushed to the factory's fire management department and emergency management platform. By acquiring real-time personnel location data and fire spread data, evacuation routes are obtained based on the factory evacuation map. A neural network model is used to predict the optimal evacuation route based on personnel location data and evacuation routes, and evacuation guidance is dynamically generated.

[0014] Even better, the remote monitoring platform has a manual intervention function, allowing fire management personnel to manually trigger warnings, start or stop linked response equipment through the platform. Manual operation permissions have higher priority than automatic linkage operations.

[0015] Preferably, the remote monitoring platform classifies and stores the collected raw monitoring data, standardized data, image data, fire determination results, early warning records, and linkage response records, and establishes a historical database.

[0016] Compared with the prior art, the present invention has the following significant advantages: 1. By deploying dedicated IoT monitoring terminals at key monitoring points of large equipment in the factory, such as electrical wiring points and hydraulic system interfaces, multiple types of sensors and image acquisition modules are integrated to simultaneously collect multi-dimensional fire protection parameters such as temperature, combustible gas, smoke, and flame, as well as real-time images of the equipment. This covers areas with high risk of fire hazards. Combined with explosion-proof sensor design, it is suitable for the special operating environment of factories with high temperature, high dust, and flammable and explosive conditions, avoiding monitoring deviations caused by environmental interference.

[0017] 2. A data fusion and image recognition joint judgment mode is adopted. The confidence distance theory is used to fuse and calculate the data of multiple sensors to obtain accurate open flame probability, smoldering fire probability and no fire probability. At the same time, the results of image fire feature recognition are mutually corroborated, which solves the problems of false alarm and missed alarm caused by single parameter monitoring or single judgment method, improves the accuracy and reliability of fire judgment, and can accurately distinguish between no fire, smoldering fire and open flame fire.

[0018] 3. Based on the degree of fire hazard and the fire situation, the fire level is divided into three levels, corresponding to different early warning methods and linkage response logic, so as to achieve "early detection of hazards, early warning triggering, and early response". At the same time, it supports the combination of automatic linkage and manual intervention, with manual operation authority having higher priority than automatic linkage. This ensures rapid response in the event of a sudden fire while taking into account the flexibility of fire management, which can effectively curb the spread of fire and reduce casualties and property losses.

[0019] 4. By leveraging IoT technology, real-time remote transmission and monitoring of fire protection data can be achieved, allowing fire management personnel to remotely monitor the fire status of equipment without on-site supervision, thus reducing management costs. Simultaneously, all monitoring data, judgment results, early warning records, and handling records are categorized and stored to establish a historical database, supporting data query, traceability, and statistical analysis. This provides reliable data support for fire hazard investigation and early warning model optimization for large-scale equipment in factories. Attached Figure Description

[0020] Figure 1 This is a flowchart of a fire remote monitoring and early warning method based on the Internet of Things in a specific embodiment of the present invention; Figure 2 This is a flowchart of the wireless transmission process in a specific embodiment of the present invention; Figure 3 This is a flowchart illustrating the dynamic risk threshold determination process in a specific embodiment of the present invention. Detailed Implementation

[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] This embodiment focuses on a large chemical reactor in a chemical plant as the monitoring object. This reactor is used to synthesize organic chemical raw materials and operates under high temperature and high pressure conditions. Its reactor body interface, electrical control cabinet, fuel delivery pipeline, and heat transfer oil circulation system are high-risk fire hazard areas. (Refer to...) Figure 1 As shown, this invention provides a fire remote monitoring and early warning method based on the Internet of Things, and the specific implementation steps are as follows: S1: Real-time collection of fire-related parameters and corresponding image data of key monitoring locations of large equipment in the factory via IoT monitoring terminals. Subsequently, data preprocessing is performed to reduce noise and calibrate the collected sensor data, and image data undergoes grayscale conversion and dehazing preprocessing to remove invalid data, resulting in standardized monitoring data and standardized image data. Specifically, this includes: The first step is to deploy IoT monitoring terminals at key monitoring locations in the chemical plant. These IoT monitoring terminals include multiple types of sensor modules, image acquisition modules, data preprocessing modules, and IoT communication modules. The key monitoring locations include electrical wiring points, hydraulic system interfaces, fuel / lubricating oil storage and transportation areas, and high-temperature operating areas. The multiple types of sensor modules include at least temperature sensors, combustible gas sensors, smoke sensors, and flame sensors. The image acquisition module uses a CCD image sensor to acquire image information of the monitored equipment locations in real time.

[0023] The temperature sensor uses an NTC thermistor type temperature detector with a measurement range of -10℃ to 150℃ and an accuracy of ≤±0.5℃. The combustible gas sensor uses an MQ-7 gas detector with a detection range of 0 to 100% LEL and alarm thresholds set to 20% LEL for low alarm and 50% LEL for high alarm. The flame sensor uses an R2868 ultraviolet detector with a response time of ≤5 seconds, a detection angle of ≥90°, and anti-sunlight interference function. All sensors adopt an explosion-proof design with an explosion-proof rating of not less than Ex d IIBT4, suitable for high-temperature, dusty, flammable and explosive operating environments of large factory equipment.

[0024] One IoT monitoring terminal is installed in the electrical control cabinet (electrical wiring point), which integrates an NTC thermistor temperature sensor, a smoke sensor and a CCD image sensor to monitor temperature changes, smoke generation and equipment operating status at the electrical wiring point. An IoT monitoring terminal is installed at the reactor body interface (hydraulic / sealing interface), which integrates a temperature sensor, a combustible gas sensor, and a CCD image sensor to monitor the temperature at the interface, the concentration of leaked combustible medium, and the sealing status of the interface. One IoT monitoring terminal is installed in the fuel delivery pipeline (fuel storage and delivery section), which integrates a combustible gas sensor, a flame sensor and a CCD image sensor to monitor whether there is a combustible gas leak or open flame in the pipeline. Heat transfer oil circulation system (high temperature operation area): Deploy one IoT monitoring terminal, which integrates temperature sensor, flame sensor and CCD image sensor to monitor the temperature of heat transfer oil and whether there is a risk of high temperature runaway or open flame.

[0025] All IoT monitoring terminals adopt an explosion-proof design with an explosion-proof rating of Ex d IIB T4, making them suitable for the high-temperature, dusty, flammable and explosive working environment of chemical plants. Each monitoring terminal integrates a data preprocessing module, an NB-IoT+5G dual-mode IoT communication module, and dual power supply and UPS backup power module (UPS continuous power supply time ≥180 minutes) to ensure stable operation of the terminal.

[0026] Multiple sensor modules collect fire protection parameters for the corresponding monitoring locations every 10 seconds. Among them, the temperature sensor collects the real-time temperature of the electrical wiring, vessel interface, and heat transfer oil circulation system; the combustible gas sensor collects the combustible gas concentration at the vessel interface and fuel delivery pipeline; the smoke sensor collects the smoke concentration at the electrical control cabinet; and the flame sensor collects the flame signal at the fuel delivery pipeline and heat transfer oil circulation system. The CCD image sensor collects image information for the corresponding location every 30 seconds, capturing the surface condition of the equipment and the presence of features such as smoke / flame. After the IoT monitoring terminal is started, it runs continuously for 24 hours. Subsequently, the collected sensor data is processed by wavelet transform algorithm to remove noise data caused by dust and electromagnetic interference in the factory environment. Then, the sensor data is calibrated by linear regression calibration method to ensure data accuracy. The image data is grayscaled and dehazed to remove invalid image data with a resolution of less than 300 dpi. The preprocessed sensor data and image data are uniformly converted into JSON format. The sensor data is labeled with the acquisition time, monitoring location, parameter type and value, and the image data is labeled with the acquisition time, monitoring location and image resolution, which is convenient for the remote monitoring platform to parse and process.

[0027] S2: Through the IoT communication module, standardized monitoring data and standardized image data are transmitted to the remote monitoring platform in real time using wireless communication, specifically including: Please see Figure 2 In this embodiment, the IoT monitoring terminal transmits standardized monitoring data and image data to the remote monitoring platform in real time through the NB-IoT+5G dual-mode communication module. The specific transmission method is as follows: Communication mode selection: Under normal operating conditions, NB-IoT communication mode is used to transmit data (low power consumption, wide coverage, and suitable for complex factory environments); when the sensor data is detected to be close to the alarm threshold or when suspected fire features appear in the image, it automatically switches to 5G communication mode to achieve high-speed data transmission and ensure rapid feedback of early warning information. Data encryption: During transmission, the AES-128 encryption algorithm is used to encrypt the data to prevent data from being tampered with or stolen, ensuring the security of data transmission; at the same time, a data retransmission mechanism is set up. When data transmission fails, it will automatically retry 3 times. If it still fails, a local alarm will be triggered on the terminal to remind on-site personnel to check the communication status. Data upload frequency: Under normal operating conditions, comprehensive monitoring data and image data are uploaded once every 30 seconds; when sensor data exceeds the normal range, the upload frequency is increased to once every 5 seconds to ensure that the remote monitoring platform can keep track of the equipment's fire protection status in real time.

[0028] S3: Standardized monitoring data are fused and calculated to obtain the probability of open flame, smoldering fire, and no-flame, using confidence distance theory. Simultaneously, standardized image data is analyzed to identify fire features in the images. The data fusion results are then jointly judged with the image fire feature identification results, specifically including: The remote monitoring platform is deployed in the factory's fire control center. After receiving standardized data transmitted from various IoT monitoring terminals, it performs joint analysis. The specific judgment process is as follows: The confidence distance algorithm is used to fuse data from multiple sensors at the same monitoring location. Taking a fuel pipeline monitoring terminal as an example, let the data from the combustible gas sensor be x1, the data from the flame sensor be x2, and the data from the temperature sensor be x3. The confidence distance between any two sensors is calculated. Set confidence distance threshold =0.3, when When the value is ≤0.3, the data from the two sensors are considered to be mutually compatible. =1), otherwise it is not supported (=1), =0); The calculation steps for fusion computing are as follows: Let the data collected by the i-th sensor be... Where i = 1, 2, ..., n, and n is the number of sensors. Follows a normal distribution. for Given the probability density function, calculate the confidence distance between any two sensors i and j. and The formula is as follows: ; ; in , Let y and x be the standard deviations of the k-th and j-th sensor data, respectively, and p(y|x) be the conditional probability density function. Then, a binary relation matrix is ​​constructed based on the mutual support relationship. Combining the weight coefficients of each sensor and the fire characteristics in the identified image, the probability of open flame, the probability of smoldering fire, and the probability of no fire are calculated by weighted summation. When the probability of open flame is ≥80%, it is determined to be an open flame fire. When the probability of smoldering fire is ≥70% and the probability of open flame is <80%, it is determined to be a smoldering fire. When the probability of no fire is ≥90%, it is determined to be no fire.

[0029] In this embodiment, the combustible gas sensor has a weight of 0.4, the flame sensor has a weight of 0.3, and the temperature sensor has a weight of 0.3. The probability of open flame, the probability of smoldering flame, and the probability of no flame are calculated by weighted summation.

[0030] Simultaneously, by analyzing standardized image data, image segmentation algorithms are used to extract flame outlines and smoke areas from the images. Combined with grayscale analysis, the presence of fire characteristics is determined: open flames are characterized by irregular bright yellow / orange outlines with a grayscale value ≥ 200; smoldering fires are characterized by light gray / black smoke areas with grayscale values ​​between 50 and 100, and no obvious bright spots. Specifically, the image condition for determining no fire is that no fire characteristics are identified in the image, including no flames, no abnormal smoke, and no obvious high-temperature grayscale anomalies; the image condition for determining smoldering fires is that smoldering-related characteristics are identified in the image, including light gray / black smoke or abnormally high grayscale values ​​in local areas without obvious flame outlines; the image condition for determining open flame fires is that an open flame outline is identified in the image.

[0031] Based on the data fusion results and image feature recognition results, the fire situation and level are determined according to the following rules: (1) No fire: The probability of no fire is ≥90%, and the probability of open flame is <80% and the probability of smoldering fire is <70%. No flame or smoke features are detected in the image. It is determined that there is no fire. The platform continuously monitors and records the data. (2) Level II general fire (smoldering): Smoldering fire probability ≥70%, open flame probability <80%, no fire probability <90%, light gray smoke area is identified in the image, and it is judged as a level II smoldering fire; (3) Level II general fire (open flame): The probability of open flame is ≥80%, the probability of no fire is <90%, a small open flame outline (area <0.5㎡) is identified in the image, and it has not spread to the core part of the reactor. It is judged as a Level II open flame fire. (4) Level III Major Fire: The probability of open flame is ≥80%, the probability of no fire is <90%, and the concentration of combustible gas reaches the high reporting threshold. The open flame outline is detected in the image and spreads to the reactor body interface, which is greater than the preset threshold. Accompanied by a large amount of smoke, it is judged as a Level III major fire.

[0032] S4: When the data fusion result exceeds the dynamic risk threshold, a fire is determined to have occurred, triggering a tiered early warning command. The remote monitoring platform triggers the corresponding level of early warning based on the fire severity, simultaneously coordinating with the factory's fire control system and on-site response equipment to issue warning information and execute emergency response operations. The specific operations are as follows: This invention uses a dynamic risk threshold method to determine the occurrence of a fire and trigger a tiered early warning command. The determination of the dynamic risk threshold is as follows: Figure 3 As shown, it includes: The current time period's factory equipment scenario type is obtained through IoT monitoring terminals, and the scenario type includes normal operation scenario, maintenance scenario, and shutdown scenario; Preset basic thresholds are matched according to the scenario type. The basic thresholds correspond to normal operation scenarios, maintenance scenarios dynamically reduce the basic thresholds based on the risk of personnel gathering, and shutdown scenarios dynamically further reduce the basic thresholds based on the risk of response delay. Furthermore, by overlaying real-time weather condition impact assessments, when meteorological conditions are detected to exacerbate the risk of fire spread, the basic threshold is adaptively adjusted to obtain a dynamic risk threshold.

[0033] First, the CCD image sensor is used to collect and identify the factory scene type for the current time period. Depending on the time period, the activities and behaviors of people in the factory vary. Common scene types include normal operation scene, maintenance scene, and shutdown scene. Among them, the normal operation scene is generally during the daytime, and people are relatively dispersed near various equipment and workstations, making evacuation relatively easy. The maintenance scene is such as the repair of a large piece of equipment, where people are concentrated near a certain equipment area, and the evacuation pressure is greater. During the shutdown scene, most people are not in the factory area, and there is less personnel movement (maybe only inspection or nearby staff), and the response delay after a fire usually is longer.

[0034] Subsequently, a preset base threshold is matched according to the scenario type. For normal operation scenarios, due to the dispersed personnel and regular use of large facilities, the fire risk is relatively low, so a lower base threshold is set to reflect the fire risk level during that period. For maintenance scenarios, considering the concentration of personnel and the greater difficulty and potential risk of evacuation in the event of a fire, the base threshold will be adjusted, specifically according to actual needs. For shutdown scenarios, since the equipment does not need to be used and is basically unattended, the response delay after a fire occurs is longer, so a more sensitive early warning is required. In this case, the base threshold needs to be further lowered. Then, weather changes are monitored in real time, especially meteorological conditions such as temperature, humidity, and wind speed. When meteorological conditions exacerbate the risk of fire spread, i.e., when one or more of the temperature, humidity, and wind speed exceed the meteorological impact threshold, the base threshold will be adaptively adjusted according to weather factors. For example, high wind speed or excessively high temperature may accelerate the spread of fire and increase the fire risk. In this case, the difference between these abnormal parameters and the threshold will be calculated, and the ratio will be calculated with the threshold. The calculation results will then be weighted and summed to obtain the impact coefficient. Then, the impact coefficient is multiplied by the base threshold, and the product is subtracted from the base threshold to obtain a dynamic risk threshold. This dynamic risk threshold reflects the real-time fire risk level, providing a more accurate reference for subsequent early warning and emergency response, ensuring timely and reasonable responses in different scenarios and environments, and maximizing the safety of the plant area.

[0035] When the data fusion result and image feature recognition result exceed a preset dynamic risk threshold, it means that the current fire risk has reached a certain dangerous level and may threaten factory safety. At this time, a tiered early warning mechanism will be automatically triggered, issuing corresponding early warning instructions based on different risk levels. Each early warning level corresponds to different response measures, helping relevant personnel to take timely preventive or emergency measures to mitigate potential fire risks. Through this tiered early warning system, different situations can be dynamically addressed, ensuring the fire safety of the factory.

[0036] Based on the joint judgment results, the remote monitoring platform automatically triggers corresponding level of early warning and coordinated response actions, while also supporting manual intervention by fire management personnel (manual operation privileges take precedence over automatic linkage). The specific implementation is as follows: Level 1 Hazard, No Fire, Only Abnormal Parameters: When sensor data exceeds the normal range but does not reach the alarm threshold, such as when the temperature of the electrical control cabinet reaches 55℃ (normal range 0~50℃), and the probability of no fire is ≥90%, only an on-site sound alarm is triggered. The buzzer built into the monitoring terminal will sound an alarm, and the platform will record the hazard information to remind on-site personnel to investigate. Level II general fire (smoldering / small open flame): Triggers on-site audible and visual warning, buzzer continuously alarms, warning light flashes red, remote monitoring platform pop-up warning, displays fire location, monitoring data, real-time image, and links the corresponding area sprinkler system to prepare for start-up. Fire management personnel can view the real-time situation through the platform and manually confirm whether to start fire extinguishing operation. Level 3 Major Fire: Triggers a plant-wide audible and visual alarm, an emergency pop-up on the platform, and an SMS alert for management personnel. Simultaneously, it pushes alerts to the mobile phones of the factory's fire safety officer and emergency management specialist. At the same time, it automatically activates the corresponding area's foam extinguishing system, cuts off the main power supply to the reactor and surrounding non-fire-fighting power supplies, turns on the smoke exhaust fan, and closes the fire damper. Simultaneously, it pushes alarm information to the local emergency management platform, including the location of the fire, the fire situation, and monitoring data. Fire safety management personnel can manually adjust the coordinated response plan through the platform to control the spread of the fire.

[0037] In addition, the remote monitoring platform adopts a distributed storage architecture to classify and store the following data: Raw data: Unprocessed temperature, combustible gas, smoke, and flame data collected by each sensor, and raw image data collected by the CCD image sensor; Processed data: preprocessed standardized monitoring data and standardized image data; The platform has a built-in deep learning module that optimizes the confidence distance threshold and sensor weights of the data fusion algorithm every quarter based on historical monitoring data and fire case data, thereby improving the accuracy of fire detection and reducing the false alarm rate. It also supports data query, traceability and statistical analysis. Fire management personnel can use the platform to query monitoring data and handling records for any time period and any monitoring location, providing data support for fire hazard investigation and daily maintenance of reactors.

[0038] The above embodiments are merely descriptions of preferred embodiments of the present invention and are not intended to limit the scope of the present invention. Various modifications and improvements made by those skilled in the art to the technical solutions of the present invention without departing from the spirit of the present invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A fire remote monitoring and early warning method based on the Internet of Things, characterized in that, Includes the following steps: S1. Real-time collection of fire-related parameters and corresponding image data of key monitoring parts of large equipment in the factory through IoT monitoring terminals. Then, noise reduction and calibration of the collected sensor data are performed through data preprocessing, and grayscale and defogging preprocessing of the image data are performed to remove invalid data and obtain standardized monitoring data and standardized image data. S2. Through the Internet of Things (IoT) communication module, standardized monitoring data and standardized image data are transmitted to the remote monitoring platform in real time using wireless communication. S3. Perform fusion calculation on standardized monitoring data, and obtain the probability of open flame, smoldering fire and no fire by combining confidence distance theory. At the same time, analyze standardized image data, identify fire features in the images, and make a joint judgment with the data fusion results and the image fire feature identification results. S4. When the data fusion result exceeds the dynamic risk threshold, a fire is determined to have occurred, triggering a graded early warning instruction. The remote monitoring platform triggers the corresponding level of early warning based on the fire level, and simultaneously links the factory's fire control system and on-site response equipment to issue early warning information and execute emergency response operations.

2. The fire remote monitoring and early warning method based on the Internet of Things according to claim 1, characterized in that, The IoT monitoring terminal includes multiple sensor modules, an image acquisition module, a data preprocessing module, and an IoT communication module. The key monitoring locations include electrical wiring points, hydraulic system interfaces, fuel / lubricating oil storage and transportation locations, and high-temperature operating areas. The multiple sensor modules include temperature sensors, combustible gas sensors, smoke sensors, and flame sensors. The image acquisition module uses a CCD image sensor to acquire image information of the monitored locations in real time.

3. The fire remote monitoring and early warning method based on the Internet of Things according to claim 1, characterized in that, The wireless communication process employs an encryption algorithm to encrypt the data.

4. The fire remote monitoring and early warning method based on the Internet of Things according to claim 1, characterized in that, The calculation steps for the fusion calculation are as follows: Let the data collected by the i-th sensor be... Where i = 1, 2, ..., n, and n is the number of sensors. Follows a normal distribution. for Given the probability density function, calculate the confidence distance between any two sensors i and j. The formula is as follows: ； ； in , Let y and x be the standard deviations of the k-th and j-th sensor data, respectively, and p(y|x) be the conditional probability density function. Then, the confidence distance threshold was set. According to confidence distance and Determine the mutual support relationship between the data from two sensors, denoted as ,when ≤ hour, =1 indicates mutual support; when > hour, =0 indicates that it is not supported.

5. A fire remote monitoring and early warning method based on the Internet of Things according to claim 4, characterized in that, A binary relationship matrix is ​​constructed based on the mutual support relationship. Combining the weight coefficients of each sensor and the fire characteristics in the identified image, the probability of open flame, the probability of smoldering fire, and the probability of no fire are calculated by weighted summation. When the probability of open flame is ≥80%, it is determined to be an open flame fire. When the probability of smoldering fire is ≥70% and the probability of open flame is <80%, it is determined to be a smoldering fire. When the probability of no fire is ≥90%, it is determined to be no fire.

6. A fire remote monitoring and early warning method based on the Internet of Things according to claim 5, characterized in that, The image conditions for determining no fire are that no fire features are identified in the image, including no flame, no abnormal smoke, and no obvious high-temperature grayscale anomaly; the image conditions for determining smoldering fire are that smoldering-related features are identified in the image, including light gray / black smoke or abnormally high grayscale values ​​in local areas but no obvious flame outline; the image conditions for determining open flame fire are that an open flame outline is identified in the image.

7. A fire remote monitoring and early warning method based on the Internet of Things according to claim 1, characterized in that, The determination of the dynamic risk threshold includes: The current time period's factory equipment scenario type is obtained through IoT monitoring terminals, and the scenario type includes normal operation scenario, maintenance scenario, and shutdown scenario; Preset basic thresholds are matched according to the scenario type. The basic thresholds correspond to normal operation scenarios, maintenance scenarios dynamically reduce the basic thresholds based on the risk of personnel gathering, and shutdown scenarios dynamically further reduce the basic thresholds based on the risk of response delay. Furthermore, by overlaying real-time weather condition impact assessments, when meteorological conditions are detected to exacerbate the risk of fire spread, the basic threshold is adaptively adjusted to obtain a dynamic risk threshold.

8. A fire remote monitoring and early warning method based on the Internet of Things according to claim 1, characterized in that, Fire severity is classified into three levels: Level 1 is a minor hazard, with no open flame and sensor data exceeding the normal range but not reaching the alarm threshold; Level 2 is a general fire, a smoldering fire or a small open flame, not spreading to the core parts of the equipment; and Level 3 is a major fire, with open flames spreading to the core parts of the equipment or the concentration of combustible gas reaching the high alarm threshold. Different levels correspond to different warning and linkage logics. Level 1 hazards only trigger on-site audible warnings and platform recordings; Level 2 fires trigger on-site audible and visual warnings, platform pop-up warnings, and the pre-activation of fire extinguishing equipment in the corresponding area; and Level 3 fires trigger plant-wide audible and visual warnings, SMS warnings, and emergency pop-ups on the platform. At the same time, the fire extinguishing system is immediately activated, the main power supply to the equipment and surrounding non-fire-fighting power supplies are cut off, the smoke exhaust system is turned on, and alarm information is simultaneously pushed to the factory's fire management department and emergency management platform. By acquiring real-time personnel location data and fire spread data, evacuation routes are obtained based on the factory evacuation map. A neural network model is used to predict the optimal evacuation route based on personnel location data and evacuation routes, and evacuation guidance is dynamically generated.

9. A fire remote monitoring and early warning method based on the Internet of Things according to claim 8, characterized in that, The remote monitoring platform has a manual intervention function, allowing fire management personnel to manually trigger warnings, start or stop linked response equipment through the platform. Manual operation permissions have higher priority than automatic linkage operations.

10. A fire remote monitoring and early warning method based on the Internet of Things according to claim 1, characterized in that, The remote monitoring platform classifies and stores the collected raw monitoring data, standardized data, image data, fire judgment results, early warning records, and linkage response records, and establishes a historical database.

Images