Intelligent fire monitoring system based on Internet of Things

Abstract

The invention relates to the technical field of monitoring, and particularly discloses an intelligent fire monitoring system based on the Internet of Things, which comprises a data acquisition module for acquiring environmental data of a monitoring area through a multi-source environmental perception unit by adopting a mixed acquisition mode; the data transmission module is used for transmitting the data acquired by the data acquisition module to a cloud or an edge server according to a transmission triggering condition; the data analysis module is used for carrying out anomaly detection on the data received by the data transmission module based on a preset rule and judging a fire early warning signal; and the early warning and response module is used for triggering an alarm according to an analysis result of the data analysis module and executing corresponding fire emergency response measures. According to the invention, a mixed mode architecture of steady-state timing sampling and abnormal event triggering is adopted, and basic monitoring is maintained by low-frequency sampling in a conventional state, so that resources can be saved; once a potential risk feature is detected, a high-frequency acquisition mode is immediately switched to ensure that the data is complete and correct.

Description

Technical Field

[0001] The present invention relates to the field of monitoring technologies, and particularly to an intelligent fire monitoring system based on the Internet of Things. Background Art

[0002] Existing industrial fire monitoring systems often face challenges such as slow response, frequent false alarms, and insufficient adaptability. Most systems rely on fixed-frequency sampling and threshold judgment of a single sensor, which makes it difficult for them to detect the rapid change characteristics in the early stage of a fire, such as a sharp rise in gas concentration or an abnormal increase in temperature. At the same time, these systems are vulnerable to interference in the industrial environment, such as dust and steam, which can cause false alarms. In addition, data transmission mainly relies on a single communication protocol, such as TCP. In a complex industrial network environment, such as electromagnetic interference and multi-node competition, packet loss or delay often occurs, which will make the fire warning not timely enough.

[0003] In an intelligent fire monitoring system, the data acquisition method usually adopts fixed-period sampling or event-triggered sampling based only on events. Both of these methods have deficiencies: If the sampling frequency is set too high in the fixed-period sampling method, it will increase redundant data and consume more storage and computing resources; if the frequency is set too low, it may miss key transient fire characteristics, such as a sharp rise in gas concentration. And the event-triggered sampling method based only on events relies on the threshold judgment of a single sensor and is easily interfered by noise, which may lead to false triggering or missed detection.

[0004] In the patent "An Indoor Intelligent Fire Alarm System Based on ZigBee Communication" (application number: CN201711109664.1, hereinafter referred to as the prior art 1), an indoor intelligent fire alarm system is disclosed. The present invention uses ZigBee communication to realize the star network layout of sensors, expand the system coverage, and use a ZigBee coordinator and a main controller to centrally process data signals, improving the accuracy and reliability of fire alarms; however, the prior art 1 still adopts a single detection method and cannot balance data integrity and resource efficiency. Summary of the Invention

[0005] In view of this, the embodiments of the present invention provide an intelligent fire monitoring system based on the Internet of Things to solve the problem that traditional fire monitoring systems cannot balance data integrity and resource efficiency due to using a single fixed sampling or event-triggered mode for monitoring.

[0006] The embodiments of the present invention provide an intelligent fire monitoring system based on the Internet of Things, including

[0007] A data acquisition module, which acquires environmental data of the monitoring area through a multi-source environmental perception unit using a hybrid acquisition mode;

[0008] A data transmission module that transmits the data acquired by the data acquisition module to the cloud or an edge server according to a transmission trigger condition;

[0009] A data analysis module that performs anomaly detection on the data received by the data transmission module based on preset rules and judges a fire warning signal;

[0010] A warning and response module that triggers an alarm according to the analysis result of the data analysis module and executes corresponding fire emergency response measures.

[0011] Preferably, the hybrid acquisition mode includes:

[0012] The data acquisition module adopts a timing sampling mode in a steady-state environment to acquire environmental data at a first sampling frequency;

[0013] When the data detected by any module in the multi-source environment perception unit exceeds a first threshold, it switches to an event-driven mode to acquire environmental data at a second sampling frequency;

[0014] If multiple modules of the multi-source environment perception unit jointly determine that they meet the fire characteristics, the sampling rate is further increased to a third sampling frequency;

[0015] When the data in the environmental data returns to the safe range, the sampling rate is gradually decreased until it returns to the timing sampling mode.

[0016] Preferably, the first threshold includes:

[0017] The temperature sensor detects that the temperature rise rate is greater than 3°C / s, or the temperature is greater than 50°C;

[0018] The smoke sensor detects that the PM2.5 concentration is greater than 100mg / m 3 , and lasts for 5 seconds;

[0019] The combustible gas sensor detects that the CO concentration is greater than 30ppm, or the CH4 concentration is greater than 5% LEL.

[0020] Preferably, the data transmission module includes:

[0021] The data preprocessing unit performs compression, denoising, and sharding encapsulation on the data acquired by the data acquisition module;

[0022] The protocol adaptive selection unit selects the MQTT-SN, TCP, or UDP protocol for transmission according to the data type;

[0023] The heterogeneous network fusion unit dynamically switches the communication link among LoRa, 5G, and fiber optic networks;

[0024] Transmit the processed data to the cloud or edge server through the selected protocol and communication link.

[0025] Preferably, the compression of the data collected by the data collection module includes numerical data compression and image data compression;

[0026] Among them, for numerical data compression, adaptive Huffman coding or floating-point quantization compression algorithm is used to compress the original data to 30% - 50% of the original size;

[0027] Among them, for image data compression, the JPEG2000 coding algorithm based on discrete wavelet transform is used, and the compression ratio is dynamically adjusted and the lossless storage of key frames is retained.

[0028] Preferably, when the data preprocessing unit performs JPEG2000 compression on image data, lossless coding is used for the thermal imaging area and lossy compression is used for the background area.

[0029] Preferably, the transmission trigger conditions include a timing transmission mode and an event-triggered transmission mode;

[0030] The timing transmission mode sends heartbeat data packets at a preset frequency and stores them;

[0031] The event-triggered transmission mode immediately starts data transmission when the multi-source environment perception unit collects environmental data at the second sampling frequency or the third sampling frequency.

[0032] Preferably, the data analysis module includes:

[0033] Real-time monitor the data in the data collection module and compare it with the first threshold;

[0034] When the data detected by any module in the multi-source environment perception unit exceeds the first threshold, allocate corresponding dynamic weights according to the module type in the multi-source environment perception unit;

[0035] Calculate the weighted comprehensive fire risk value. If the comprehensive fire risk value exceeds the preset risk value, generate a corresponding fire warning signal.

[0036] Preferably, the fire warning signal includes:

[0037] A first-level warning signal. When the comprehensive fire risk value exceeds the second risk threshold and is less than the third risk value, start local alarm;

[0038] A second-level warning signal. When the comprehensive fire risk value exceeds the third risk threshold, send an emergency signal to the fire warning platform in the cloud.

[0039] Preferably, the dynamic weights include:

[0040] In the first type of scenario, the modules in the multi-source environment perception unit operate using the first dynamic weight;

[0041] In the second type of scenario, the modules in the multi-source environment perception unit operate using the second dynamic weight.

[0042] The intelligent fire monitoring system based on the Internet of Things provided by the present invention has the following beneficial effects:

[0043] In the present invention, a hybrid mode architecture of "steady-state timed sampling + abnormal event triggering" is adopted. In the normal state, low-frequency sampling is used to maintain basic monitoring, which can save resources; once potential risk characteristics are detected, it immediately switches to the high-frequency acquisition mode to ensure the integrity and correctness of the data. This dynamic adjustment method not only avoids the waste of resources caused by continuous high-load operation but also can capture all sudden fire characteristics without omission. The essence of this solution is to transform the static resource allocation mode into a dynamic adaptive intelligent allocation system, and through the coordinated linkage of each functional module, an optimal balance is achieved between resource consumption and monitoring effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required to be used in the embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings, and all of these are within the protection scope of the present invention.

[0045] Figure 1 It is a flowchart of the intelligent fire monitoring system based on the Internet of Things. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0046] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. It should be noted that in this document, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. In the description of the present invention, it should be understood that the orientation or positional relationships indicated by the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. are based on the orientation or positional relationships shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as a limitation of the present invention. Moreover, the terms "include", "comprise" or any other variant thereof are intended to cover non-exclusive inclusion, such that a process, method, article or device including a series of elements not only includes those elements but also includes other elements not expressly listed, or further includes elements inherent to such process, method, article or device. Without further limitation, elements defined by the statement "including..." do not exclude the existence of additional identical elements in the process, method, article or device including the said elements. If there is no conflict, the embodiments of the present invention and the various features in the embodiments may be combined with each other, and all are within the protection scope of the present invention.

[0047] Embodiment 1

[0048] Please refer to Figure 1 , the embodiments of the present invention provide an Internet of Things-based intelligent fire monitoring system; in the monitoring systems for industrial factories, warehouses, and other industrial environments, the currently commonly used methods are fixed-frequency sampling methods or single-sensor threshold judgment methods. However, such monitoring systems often face problems such as slow response, frequent false alarms, and insufficient adaptability. The existence of these problems has led to unsatisfactory fire prevention and control effects. At the same time, the monitoring system also struggles in balancing data integrity and resource efficiency.

[0049] In the embodiments of the present invention, an Internet of Things-based intelligent fire monitoring system is provided to solve the problem of imbalance between data integrity and resource efficiency in the existing monitoring systems.

[0050] In this embodiment, the Internet of Things-based intelligent fire monitoring system includes:

[0051] A data acquisition module that uses a multi-source environment perception unit to obtain environmental data of the monitoring area in a hybrid acquisition mode.

[0052] The multi-source environment perception unit consists of the following sensor arrays:

[0053] A temperature sensing module for real-time monitoring of the environmental temperature rise rate and absolute value.

[0054] A smoke sensing module for measuring the PM2.5 and PM10 concentrations based on the principle of laser scattering.

[0055] A gas sensing module that uses electrochemical elements to detect the gas concentrations of combustible gases (CO, CH4).

[0056] A multi-spectral vision acquisition module, which is a combined device including a visible light camera and an infrared thermal imager, for capturing the flame spectral characteristics and smoke diffusion images.

[0057] Furthermore, the system adopts a hybrid acquisition mode. The hybrid acquisition mode includes a timing sampling mode and an event-driven mode. Each module of the multi-source environment perception unit realizes microsecond-level data synchronization through a hardware synchronization signal.

[0058] The application of the hybrid acquisition mode enables the system to perform timing sampling at a lower frequency during normal times, saving resources. When an abnormal event is detected, such as a sharp rise in temperature, a sudden increase in smoke concentration, or the appearance of a flame, etc., it immediately switches to the event-driven mode, increasing the sampling frequency, promptly capturing and responding to fire risks, and achieving a balance between resource efficiency and data integrity.

[0059] Data synchronization adopts a distributed clock synchronization mechanism. Each sensor module (temperature / smoke / infrared, etc.) is built-in with a high-precision clock chip (such as a PHY layer clock compatible with the IEEE1588 protocol), and transmits the reference clock pulse through a dedicated synchronization signal line (SYNC). The main control unit broadcasts a time calibration signal periodically (such as once per second), and all slave modules uniformly correct the local clock offset under the trigger of a hardware interrupt to achieve an absolute time synchronization of ±1us level. The advantage of this setting is that it can ensure that the data collected among the modules of the multi-source environment perception unit has a high degree of time consistency, thereby improving the accuracy and reliability of data processing.

[0060] The hybrid acquisition mode includes:

[0061] The data acquisition module adopts a timing sampling mode in a steady-state environment to collect environmental data at a first sampling frequency. Generally, the first sampling frequency can be adaptively adjusted and set within the range of 0.1 - 1Hz, and collect environmental data according to the set frequency to avoid unnecessary resource waste caused by the system running all the time. When the environmental parameters are stable and there are no significant changes, a lower sampling frequency is sufficient to maintain effective monitoring of the environmental state, while reducing the system power consumption and data processing burden.

[0062] When the data detected by any module in the multi-source environment perception unit exceeds a first threshold, the mode is switched to an event-driven mode to collect environment data at a second sampling frequency.

[0063] The first threshold includes:

[0064] The temperature sensor module detects that the temperature rise rate is greater than 3°C / s or the temperature is greater than 50°C.

[0065] The smoke sensor module detects that the PM2.5 concentration is greater than 100mg / m 3 , and lasts for 5 seconds.

[0066] The gas sensing module detects that the CO concentration is greater than 30 ppm, or the CH4 concentration is greater than 5% LEL.

[0067] When the infrared thermal imaging in the multispectral vision acquisition module identifies a high-temperature area with a temperature greater than 300°C and meets the flame morphological characteristics or the flame pixels detected according to the HSV color space of the visible light image [0-10, 100-255, 100-255] account for more than 15%.

[0068] This setup enables rapid response to emergencies, promptly capturing dramatic changes in environmental parameters and improving the sensitivity and real-time nature of environmental monitoring. In event-driven mode, the second sampling frequency is typically much higher than the first sampling frequency in scheduled sampling mode, ensuring sufficient data points are captured when an anomaly occurs for subsequent data analysis, fault diagnosis, and incident handling.

[0069] Furthermore, the second sampling frequency can be adaptively adjusted within the range of 5-20 Hz, and the second frequency is graded according to the event type: 5-10 Hz when the combustible gas concentration exceeds the threshold, and 10-20 Hz when the temperature rise rate exceeds the threshold.

[0070] This allows for more effective allocation of sampling resources based on the urgency of different types of events, ensuring more detailed data information is obtained in emergency situations. While close attention is required when combustible gas concentrations exceed the threshold, the rate of change may be slower than that of temperature changes. Therefore, a sampling frequency of 5-10 Hz can capture critical data while avoiding data redundancy. When the temperature rise rate exceeds the threshold, it often means an emergency such as a fire. At this time, a higher sampling frequency, i.e., 10-20 Hz, is required. Using 20 Hz sampling, a temperature rise mutation of 0.5°C / ms can be captured to capture every detail of the temperature change, providing accurate data support for subsequent fire warnings and emergency response. This hierarchical setting strategy not only improves the efficiency of data processing, but also ensures the response speed and accuracy of the environmental monitoring system.

[0071] If multiple modules of the multi-source environment perception unit jointly determine that the fire characteristics are met, the sampling rate is further increased to the third sampling frequency.

[0072] The third sampling frequency is the highest-level emergency sampling mode of the system and is designed specifically for the fire confirmation stage.

[0073] The frequency range of the third sampling frequency is 50 - 100 Hz (sampling once every 10 - 20 ms); the trigger conditions need to be met simultaneously:

[0074] Data detected by at least two types of modules in the multi-source environment perception unit are greater than the first threshold, the comprehensive fire risk value is greater than the preset value, and the multi-spectral vision acquisition module identifies the dynamic characteristics of the flame (such as when the flame flicker frequency is greater than 3 Hz).

[0075] For example, when there is a smoldering fire due to cable overload, its temperature rise rate is 0.2 °C / ms and the smoke particle diffusion speed is 10 cm / s; at the sampling verification of 50 Hz sampling frequency, the time resolution of the temperature data can track the change of 0.4 °C / ms, and compared with the 10 Hz sampling, the arc feature recognition rate is increased from 45% to 92%.

[0076] Furthermore, the setting of the sampling frequency meets the following conditions:

[0077]

[0078] where f sample is the third sampling frequency, in Hz; dT / dt is the temperature rise rate, in °C / ms; f flame is the flame flicker frequency; the constant 0.5 °C / ms is the "critical temperature rise rate of deflagration" defined in fire science.

[0079] The temperature rise rate term To quantify the ratio of the heat release rate of the current fire to the reference deflagration threshold. For example, when the temperature rise rate is 0.5℃ / ms, it corresponds to the basic sampling value at 50Hz; when a violent explosion occurs, the temperature rise rate may be 1℃ / ms, and at this time the temperature rise rate rises to twice the basic sampling value. According to the NFPA921 standard, 0.5℃ / ms is the critical value for the out-of-control of liquid fuel fires; and the typical value range of the brightness fluctuation frequency caused by flame turbulence is 3 - 15Hz. According to the sampling theorem, sampling at the Nyquist frequency (more than 2 times) is required to completely record the process, ensure distortion-free signal reconstruction, and achieve efficient and accurate fire feature recognition. On this theoretical basis, the system can dynamically adjust the sampling frequency to adapt to the monitoring requirements under different fire conditions. For example, in the initial stage of a fire, the changes in smoke and temperature are relatively slow, and the system can use a lower sampling frequency to reduce the amount of data processing; while when the fire intensifies, the smoke spreads rapidly and the temperature rises sharply, the system automatically increases the sampling frequency to 50Hz or even higher to ensure that every subtle change in fire features is captured.

[0080] When the data in the environmental data returns to the safe range, gradually reduce the sampling rate until returning to the regular sampling mode.

[0081] During the process of environmental data monitoring, once it is found that the data has returned to a safe level and range, measures should be taken to gradually reduce the sampling frequency. This is done to reduce resource consumption and improve system efficiency. Through such adjustments, the continuity and accuracy of data collection can be ensured, while avoiding resource waste caused by over-collection. Finally, when the environmental data is completely stable and within the safety threshold, the sampling mode can be switched back to the normal regular sampling mode, which not only ensures the real-time nature of data monitoring but also ensures the efficient operation of the entire system.

[0082] The data transmission module transmits the data obtained by the data acquisition module to the cloud or edge server according to the transmission trigger conditions.

[0083] The transmission trigger conditions include the regular transmission mode and the event-triggered transmission mode.

[0084] The timed transmission mode sends heartbeat data packets (periodic status signals used by IoT devices to maintain the survival of communication links) at a preset frequency, such as once every 30s, 60s, etc.; this can ensure continuous monitoring of the system status, and maintain the activity and responsiveness of the system even in the absence of changes in environmental data. In addition, the timed transmission mode helps the cloud or edge server receive data regularly. The data transmitted at a fixed time is not judged, but needs to be stored for data analysis and processing to provide a reliable basis for subsequent decision-making. When the stored timed transmission data exceeds the preset storage duration, it is automatically deleted to ensure the effective use of storage space and the timeliness of data.

[0085] When the multi-source environmental perception unit uses the second sampling frequency to collect environmental data, the event-triggered transmission mode immediately starts data transmission; at this time, the data transmission module will quickly send the collected environmental data to the cloud or edge server to promptly respond to environmental changes. This transmission mode can ensure that when significant changes occur in environmental data, the system can quickly respond and transmit key data to the processing center in real time, thus effectively shortening the response time and improving the emergency handling ability of the system.

[0086] At the same time, the event-triggered transmission mode can also reduce unnecessary data transmission, reduce network load and storage costs, and improve the overall efficiency of the system. By combining the timed transmission mode and the event-triggered transmission mode, the data transmission module can flexibly adjust the transmission strategy according to actual needs to ensure the timeliness and accuracy of data, providing strong support for the stable operation of the environmental monitoring system.

[0087] Furthermore, when data is transmitted, it is preferentially transmitted to the edge server, and in the edge server, it is judged according to the analysis of the data analysis module whether it needs to be transmitted to the cloud-based early warning platform; this setting can reduce the delay of data transmission and improve the timeliness of data processing. As the "front line" of data processing, the edge server can quickly conduct preliminary analysis and judgment on the collected environmental data. For urgent or important data, it can be immediately uploaded to the cloud-based early warning platform so that relevant departments can take measures promptly. At the same time, this setting can also effectively reduce the processing pressure on the cloud, optimize resource allocation, and make the environmental monitoring system operate more efficiently and stably.

[0088] Furthermore, when the information sender outputs information, if the receiver successfully obtains the data, it returns an ACK (acknowledgment signal). If the sender does not receive the ACK and is within the timeout window, data retransmission is triggered. The timeout window is the maximum waiting time value between the completion of packet transmission and the receipt of the ACK, for example, 3 seconds. If the maximum waiting time value is exhausted and the ACK is still not received, it is determined that the transmission has failed; after the transmission fails, the communication link can be switched for retransmission.

[0089] In this embodiment, the described data transmission method includes a series of steps aimed at ensuring that data packets can be reliably transmitted from the sender to the receiver.

[0090] First, when the sender is ready to send a data packet, it starts a timer, which is used to track the time interval from when the data packet is sent until the receiver acknowledges its receipt.

[0091] Next, the sender monitors this timeout window, the length of which is set to a specific value, usually no more than 3 seconds, to wait for the receiver's acknowledgment signal (ACK).

[0092] If within this timeout window, the sender does not receive the expected acknowledgment signal, indicating that the data packet may not have successfully reached the receiver, the sender will then initiate a retransmission mechanism to re-send the data packet.

[0093] In addition, if the number of retransmissions reaches a predetermined upper limit value N (for example, N is set to 3 times), or the retransmission attempts of the data packet exceed the predetermined communication window time range, the sender will take further measures, namely, switch to an alternative communication link to attempt to complete data transmission through a different path.

[0094] Such a setting can ensure the reliability and stability of data transmission. In a fire environment monitoring system, the integrity and accuracy of data are of crucial importance. Any loss or error of data may lead to deviations in monitoring results, which in turn affect relevant decisions and result in immeasurable losses. By introducing the ACK acknowledgment mechanism and the timeout window, the data transmission module can promptly switch the communication link and re-transmit data in case of transmission failure, thus effectively avoiding data loss problems caused by reasons such as network instability or equipment failures. At the same time, this setting can also improve the success rate of data transmission, reduce resource waste and time costs brought about by transmission failures, and provide a more solid technical guarantee for the stable operation of the environmental monitoring system.

[0095] The data preprocessing unit performs compression, denoising, and fragmentation encapsulation on the data collected by the data acquisition module.

[0096] Performing compression on the data collected by the data acquisition module includes numerical data compression and image data compression.

[0097] Adaptive Huffman coding or floating-point quantization compression algorithm is used for numerical data compression, compressing the original data to 30% - 50% of its original size;

[0098] Before data transmission, the system preprocesses the collected environmental data to reduce redundancy and noise and improve transmission efficiency. First, for numerical data such as temperature, smoke, and gas, Huffman coding or floating-point compression algorithms are used for compression; for infrared images or AI vision data, JPEG2000 or WebP formats are used for efficient compression. At the same time, the system performs packetization and fragmentation processing according to the data type. Small data packets are directly sent using MQTT or COAP, and large data packets such as video streams are transmitted in a streaming manner using TCP fragmentation, UDP, or RTSP. This can effectively reduce the bandwidth occupancy of data transmission and improve the stability and real-time performance of data transmission. Especially in the case of poor network conditions, fragmented transmission can reduce transmission failures caused by packet loss and ensure the integrity and reliability of data. In addition, by compressing and denoising the original data through the data preprocessing unit, the quality of the data can be further improved, providing a more accurate and reliable data source for subsequent data analysis and processing.

[0099] In this embodiment, for the compression processing of image data, the JPEG2000 coding algorithm based on discrete wavelet transform is adopted. This algorithm can not only effectively compress image data but also dynamically adjust the compression ratio according to actual needs to achieve the best compression effect. At the same time, to ensure image quality, the lossless storage of key frames is particularly reserved to ensure that the complete information of the original image can be restored when needed.

[0100] For the data preprocessing unit, when it performs JPEG2000 compression on image data, different coding strategies are adopted according to different regions of the image. Specifically, for the thermal imaging region in the image, lossless coding technology is used to ensure the integrity and accuracy of this part of the image data without loss. For the background region of the image, lossy compression technology is adopted to achieve a higher compression ratio by sacrificing a certain amount of image quality, thereby effectively reducing the storage space requirements.

[0101] The protocol adaptive selection unit selects MQTT-SN (Message Queuing Telemetry Transport), TCP (Transmission Control Protocol), or UDP protocol (User Datagram Protocol) for transmission according to the data type;

[0102] The protocol adaptive selection unit dynamically switches the transmission protocol according to the data type and network conditions; low-frequency data such as temperature and humidity, gas concentration, etc. (≤1Hz) is transmitted through MQTT-SN or CoAP+DTLS to support QoS classification (levels 0-2); fire alarm signals are forced to use the TCP+retransmission timeout (RTO) adaptive algorithm to ensure 100% delivery; warning data with high real-time requirements (such as flame positioning coordinates) enables the QUIC protocol to achieve 0-RTT connection; fixed monitoring points use RTSP or UDP and support H.264 / H.265 encoding; in the wide-area deployment scenario, cloud storage adopts the HLS sharding strategy (generating ts slices every 2 seconds) and embeds flame detection metadata.

[0103] The protocol adaptive selection unit can intelligently select the most suitable transmission protocol according to the data type and its transmission requirements, as well as the current network conditions. For low-frequency data such as temperature and humidity, gas concentration, etc., since its changes are slow and the real-time requirement is not high, the MQTT-SN or CoAP+DTLS protocol is used for transmission. Both of these protocols can support QoS classification to ensure the reliable transmission of data under different network conditions. At the same time, MQTT-SN is particularly suitable for resource-constrained devices such as Internet of Things sensors, while CoAP+DTLS provides security for data transmission.

[0104] For key data such as fire alarm signals, the accuracy and reliability of its transmission are crucial. Therefore, the system forces the use of the TCP+retransmission timeout (RTO) adaptive algorithm. Through the reliable connection of TCP and the adaptive adjustment of the RTO algorithm, it is ensured that the fire alarm signal can be delivered 100%, and the reliable transmission of data can be guaranteed even in the case of unstable network.

[0105] For warning data with high real-time requirements, such as flame positioning coordinates, the system enables the QUIC protocol. The QUIC protocol combines the reliability of TCP and the low-latency characteristics of UDP, and can achieve 0-RTT connection, that is, data transmission can start without going through the complete TCP three-way handshake process, thus greatly improving the data transmission speed. This is crucial for warning data that requires rapid response.

[0106] For the video data of fixed monitoring points, the system uses the RTSP or UDP protocol for transmission. The RTSP protocol is specifically used for the control, transmission, and playback of streaming media data, and can support high-definition video coding formats such as H.264 / H.265, providing high-quality video monitoring services. The UDP protocol, due to its low-latency characteristics, is suitable for video monitoring scenarios with high requirements for real-time performance. In the wide-area deployment scenario, the cloud storage adopts the HLS sharding strategy. By cutting the video data into ts slices every 2 seconds and embedding the flame detection metadata, it not only facilitates the storage and management of data but also is conducive to subsequent data analysis and processing. At the same time, the HLS sharding strategy can also improve the transmission efficiency and playback fluency of video data.

[0107] The heterogeneous network fusion unit dynamically switches the communication link among LoRa, 5G, and fiber optic networks; this can ensure the stability and efficiency of data transmission. In remote or poorly network-covered areas, the LoRa network can provide long-distance and low-power communication capabilities to ensure the stable transmission of data. In areas with good network coverage and high requirements for data transmission speed, the 5G network can give play to its advantages of high speed and low latency to achieve the rapid transmission of data. The fiber optic network serves as the backbone network for high-speed data transmission to ensure that a large amount of data can be efficiently and stably transmitted between the cloud and each monitoring point. Through the dynamic switching of the heterogeneous network fusion unit, the system can select the most suitable communication link according to the actual scenario and requirements, thus ensuring the stability and efficiency of the entire monitoring system.

[0108] The processed data is transmitted to the cloud or edge server through the selected protocol and communication link.

[0109] The data analysis module performs anomaly detection on the data received by the data transmission module based on preset rules and judges the fire warning signal.

[0110] The data in the real-time monitoring data acquisition module is monitored and compared with the first threshold.

[0111] When the data detected by any module in the multi-source environment perception unit exceeds the first threshold, corresponding dynamic weights are assigned according to the module type in the multi-source environment perception unit; the dynamic weights can be set according to the working conditions of the monitored area. When in the first type of scenario (working period or working scenario), the modules in the multi-source environment perception unit operate with the first dynamic weight; in the second type of scenario (non-working period or non-working scenario), the modules in the multi-source environment perception unit operate with the second dynamic weight. The above settings can adjust the monitoring sensitivity according to the actual working conditions to avoid false alarms and missed alarms.

[0112] During working hours or in a working scenario, due to frequent personnel activities, there may be more interference factors. Therefore, the first dynamic weight is adopted to increase the monitoring threshold and reduce the possibility of false alarms.

[0113] While during non - working hours or in a non - working scenario, due to the reduction of personnel activities, the potential fire risk may increase. Therefore, the second dynamic weight is adopted to lower the monitoring threshold and improve the monitoring sensitivity to ensure that potential fire risks can be detected in a timely manner. Through this setting of dynamic weights, the data analysis module can more accurately judge the fire warning signal, improving the accuracy and reliability of the entire monitoring system.

[0114] Calculate the weighted comprehensive fire risk value. If the comprehensive fire risk value exceeds the preset risk value, then generate the corresponding fire warning signal.

[0115] Furthermore, set the comprehensive fire risk value based on the weighted voting method:

[0116] F I = aT + bS + cC + dF + eV

[0117] Among them, F I is the comprehensive fire risk value; a, b, c, d, e are the dynamic weights of the corresponding modules respectively; T is the temperature, usually provided by the temperature sensing module; S is the smoke concentration, usually provided by the smoke sensing module; C is the concentration of combustible gas, usually provided by the gas sensing module; F is the probability of the presence of flame directly detected by the infrared sensor or AI vision, taking a probability value between 0 - 1, that is, F = P(presence of flame); V is the smoke coverage analyzed by AI computer vision, expressed as a value between 0 and 1 (such as the proportion of smoke pixels, smoke concentration).

[0118] Different physical quantities have different units. For example, the temperature is °C, the smoke is mg / m 3 , the combustible gas is ppm, and the flame detection and vision detection are probability values. Direct addition will lead to unreasonable calculations. Therefore, all values need to be normalized to between 0 - 1 before the weight calculation can be carried out, making different data sources comparable.

[0119] Temperature normalization: For example, assuming that the fire phase change temperature range is between 20 - 300 °C, if the current temperature is 100 °C, then after normalization, T = 0.2857.

[0120] Smoke concentration normalization: The original unit is mg / m 3 (PM2.5) or ppm (CO, smoke particles); for example, if the maximum set value of S is 500 mg / m 3 , and the current concentration is 250 mg / m 3, then S = 0.5.

[0121] The normalization principle of combustible gas is the same as that of smoke concentration.

[0122] In the first type of scenario, it is in a high - pedestrian - flow scenario. Therefore, increase the weights of temperature (T) and smoke (S) because there may be other interference factors in the environment (such as cooking, smoking, vehicle exhaust, etc.), and a higher threshold is required to reduce false alarms.

[0123] Therefore, F I = 0.35T + 0.3S + 0.15C + 0.1F + 0.1V.

[0124] Enhance the weight of the temperature factor to increase the attention to real high - temperature fires; due to the possible smoke interference caused by human activities, appropriately increase the smoke weight but not be overly sensitive; keep the gas weight with a certain influence, but lower than that of temperature and smoke to avoid false alarms caused by minor gas leaks; due to possible light or heat source interference in the working scenario, moderately reduce the influence of the flame weight; AI vision may be misjudged due to interference factors such as pedestrian flow and lighting, so reduce the proportion.

[0125] In the second type of scenario, since there are fewer people, it is necessary to improve the fire monitoring sensitivity to ensure that fires can be detected in high - risk environments such as at night and unoccupied areas in a timely manner.

[0126] Therefore, F I = 0.25T + 0.2S + 0.2C + 0.2F + 0.15V.

[0127] In the second type of scenario, the generation of smoke is more likely to be a fire signal. Therefore, reduce the threshold but control the weight; since there is no pedestrian flow interference at night, the accuracy of AI vision analysis is improved, so increase its influence; therefore, mainly focus on the flame (F) and visual smoke (V) to improve the sensitivity of the system and ensure that the fire risk in the unoccupied state can be detected in a timely manner.

[0128] Setting dynamic weights can flexibly adjust the sensitivity of the monitoring system according to different environments and scenarios, achieving more accurate and reliable fire warnings. During working hours or in high - pedestrian - flow scenarios, by increasing the weights of temperature and smoke, false alarms caused by other interference factors can be effectively reduced, improving the stability of the system. While during non - working hours or in low - pedestrian - flow scenarios, by reducing the monitoring threshold and increasing the weights of the flame and visual smoke, it can ensure that the system can detect fires in high - risk environments in a timely manner, improving safety. This setting of dynamic weights not only improves the accuracy and reliability of the monitoring system, but also makes the system more intelligent and adaptive, and can better meet the fire monitoring requirements in different environments and scenarios.

[0129] In this embodiment, according to the comprehensive fire risk value (FI ) can be used to set different alarm levels so as to take corresponding response measures. Generally speaking, it can be divided into multiple levels according to the value range of F I , such as low risk, medium risk, high risk, and extremely high risk.

[0130] When the risk is low, F I ≤ 0.3; the environmental parameters of the area are within the normal range, and the fire risk is relatively low. It may be just slight temperature or smoke fluctuations, such as normal monitoring of kitchen operations, vehicle exhaust, heat generated by heating equipment, etc. Continue to collect data without triggering an alarm.

[0131] When the risk is medium, 0.3 < F I ≤ 0.5; there may be a fire risk, but it has not reached the clear fire determination standard. It may be that the sensor detects abnormal fluctuations in temperature or smoke concentration, but has not reached the fire threshold, such as local equipment overheating, slight gas leakage, etc.

[0132] When the risk is high, 0.5 < F I ≤ 0.7; the monitoring data indicates a relatively high fire risk, and intervention measures need to be taken. It may be detected that the smoke concentration has increased significantly, the leakage of combustible gas exceeds the standard, the temperature has risen abnormally, but no obvious flame or large-scale fire signal has been detected; if F I continues to rise within 10 minutes, it will be upgraded to the extremely high risk state.

[0133] When the risk is extremely high, F I > 0.7; it is highly suspected that a fire has occurred, and immediate emergency response measures must be taken. It may be detected that the temperature has risen sharply, the smoke concentration has increased sharply, the flame detection signal is confirmed, or even open fire is detected by AI vision.

[0134] Furthermore, in this embodiment, F I = 0.3 is set as the first risk threshold, F I = 0.5 is set as the second risk threshold, and F I = 0.7 is set as the third risk threshold.

[0135] Furthermore, the fire warning signal includes:

[0136] No alarm is triggered. When no module in the multi-source environment perception unit is triggered or the comprehensive fire risk value does not exceed the first risk threshold, the system determines that the current environment is in a safe state, and only continuously monitors and records data for subsequent analysis. At this time, the system will not send any warning information to the management personnel to avoid unnecessary panic and resource waste. At the same time, the system will also automatically adjust the acquisition frequency to minimize energy consumption and data storage pressure on the premise of ensuring the monitoring effect.

[0137] Level 1 warning signal: When one of the modules in the multi-source environmental perception unit is triggered and the comprehensive fire risk value exceeds the second risk threshold and is less than the third risk value, local alarm is activated, a low-level warning is sent on the management system interface to remind the management personnel to pay attention and eliminate the risk;

[0138] Level 2 warning signal: When at least two modules in the multi-source environmental perception unit are triggered and the comprehensive fire risk value exceeds the third risk threshold, an emergency signal is sent to the fire warning platform in the cloud.

[0139] In the case of a Level 1 warning signal, if any single module in the multi-source environmental perception unit is triggered, or when the comprehensive fire risk value does not exceed the first risk threshold set by us, the system will activate the local alarm mechanism. At the same time, a low-level warning signal will be issued on the interface of the management system to remind the local on-duty management personnel (such as security guards) to pay attention to the current situation and take corresponding measures to eliminate potential risks.

[0140] When the Level 2 warning signal is triggered, it means that at least two different modules in the multi-source environmental perception unit are triggered, or the comprehensive fire risk value has exceeded the first risk threshold preset by us. In this case, the system will not only stay at the local alarm, but will send an emergency signal to the fire warning platform in the cloud so that the fire department can receive the alarm in time; and automatically activate the fire extinguishing system (such as sprinkler, gas fire extinguishing, isolating the fire source), provide real-time data (fire location, smoke concentration, flame detection situation, etc.), continuously record and analyze the development of the fire to ensure the safety of personnel and property protection.

[0141] Warning and response module: It triggers the corresponding alarm mechanism based on the analysis results provided by the data analysis module and is responsible for executing a series of fire emergency response measures. The main function of this module is to issue alarm signals to ensure that relevant personnel can be notified in time to take actions when a fire occurs, thereby minimizing the losses and injuries that may be caused by the fire.

[0142] Embodiment 2

[0143] The embodiment of the present invention provides an Internet of Things-based intelligent fire monitoring system. In Embodiment 1, two scenarios for the allocation of dynamic weights of the comprehensive fire risk value are provided, including the first type of scenario and the second type of scenario. However, the first type of scenario and the second scenario are only applicable to the working environment and the non-working environment. Therefore, in this embodiment, a dynamic weight adjustment strategy for chemical factory buildings is provided;

[0144] Chemical factory buildings contain flammable and explosive chemicals (such as organic solvents, hydrogen). When a fire breaks out, it will quickly release and cause flash fires / explosions; there is no obvious smoke or flame at the initial stage of gas leakage in pipelines / valves; local fires may trigger chain explosions (such as the rupture of pressure vessels due to heat) interference factors; typical fire risk characteristics such as steam and welding sparks during the production process are prone to false alarms;

[0145] Therefore, based on these fire risk characteristics, the dynamic weights of the corresponding modules in the comprehensive fire risk value can be adjusted accordingly.

[0146] In chemical factory buildings, the comprehensive fire risk value is set according to the leakage priority principle; 80% of major accidents in chemical fires are caused by gas leakage (refer to NFPA86 standard), and the highest monitoring priority needs to be given.

[0147] According to the characteristics of methane, the lower explosion limit of CH4 is 5% LEL. When the CH4 concentration is greater than 5% LEL, the explosion risk will increase exponentially; therefore, in this embodiment, the basic weight of the combustible gas concentration is set to 0.45 (covering conventional monitoring); when the CH4 concentration > 5% VOL, +0.2 (leakage scenario weight 0.65); when the CH4 concentration > 15% LEL, an additional +0.25 (explosion compensation weight 0.9); further, unit VOL is used for gas leakage scenarios, and LEL is used for explosion scenarios.

[0148] Due to the large temperature fluctuations in the chemical factory environment (reactor heating / cooling), the normal sensitivity needs to be reduced, and the basic weight of temperature is set to 0.25; when there is a slow temperature rise in the scenario of gas leakage, it may be accompanied by leakage (such as pipeline rupture), and the monitoring needs to be enhanced, so the weight of temperature increases by 0.05; when there is an explosion risk, the temperature rise rate > 10°C / s is a sign that the pressure vessel is about to fail (refer to ASME BPV Code), and the weight jumps to 0.5.

[0149] In chemical plants, the weight of infrared thermal imaging changes non-linearly. According to its optical characteristics, the basic weight of F is set to 0.35 to match conventional thermal radiation monitoring and cover equipment overheating detection. In the open fire scenario, when the infrared characteristics of the flame (35μm band) appear, it is increased by 0.2 because the flicker frequency of chemical flames is 515Hz, and high-frequency sampling is required to capture; when the heat diffusion rate > 20cm 2 / s, the weight increases to 0.65, corresponding to the expansion speed of the deflagration fireball (NFPA68 standard).

[0150] In the embodiment of the present invention, the full cycle coverage of "slow leakage → open fire → explosion" is achieved through dynamic weights.

[0151] In chemical plants, dust / steam can cause a false alarm rate of PM2.5 as high as 60%, so anti-interference design is required; its basic weight is set to 0.15 (lower than other sensors); for the explosion scenario, the weight is 0.1 (reduced to 0.05), because shock waves can raise dust and generate false smoke signals. Among them, if CO>100ppm (a sign of incomplete combustion) is detected at the same time, the smoke weight is restored to 0.3.

[0152] Since there is less smoke in chemical plant rooms and there is a possibility of explosion, the smoke coverage rate (V) in F I is replaced by the explosion risk compensation factor (K). In high-risk scenarios such as chemical plants, traditional fire monitoring systems have a risk of missing explosion alarms. Traditional fire monitoring systems only rely on a single gas concentration or temperature threshold (such as CH4>10%VOL or temperature rise>5°C / s), and cannot capture the synergistic effect of the two.

[0153] For example: when methane leaks (CH4 = 15%LEL) accompanied by a slow temperature rise (2°C / s), the traditional system may determine it as a low risk, but in fact it is already in the explosion critical state. The linear weighted model is insensitive to the non-linear risk evolution such as explosion, resulting in a lag in emergency response.

[0154] In this embodiment, by coupling the gas concentration and the temperature rise rate, the K factor realizes dynamic risk amplification

[0155] When the methane concentration and the temperature rise rate are both close to the explosion threshold, the K value approaches 1, making the comprehensive risk value (F I ) increase by up to 50% (multiplied by 1 + 0.5K).

[0156] Among them, the calculation formula of the said K is:

[0157]

[0158] The comprehensive fire risk value is:

[0159] F I =(aT + bS + cC + dF)×(1 + 0.5K)

[0160] In explosion high-risk scenarios such as chemical plant rooms, the comprehensive risk value needs to be multiplied by the explosion compensation term (1 + 0.5K). Among them, K is determined by the product of the methane concentration and the temperature rise rate. When K = 1 (that is, CH≥20%LEL and the daily temperature rise rate>10°C / s), the total risk value is amplified by 1.5 times, directly triggering the highest-level emergency response.

[0161] 20%LEL of the molecule is the upper explosion limit of methane, 10°C / s is the critical temperature rise rate for the failure of pressure vessels (refer to the ASME standard), and the min function is used to limit K≤1 to avoid excessive amplification of noise interference.

[0162] When K = 1, the three elements of explosion are simultaneously satisfied: the concentration of combustible gas ≥ 20% LEL (upper explosion limit), 2. the temperature rise rate > 10 °C / s (the critical value for causing overpressure); there is an ignition source (confirmed by the infrared flame detection module).

[0163] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: they can still modify the technical solutions recorded in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. An intelligent fire monitoring system based on the Internet of Things, characterized in that, including a data acquisition module that acquires environmental data of a monitoring area through a multi-source environment perception unit using a hybrid acquisition mode; a data transmission module that transmits the data acquired by the data acquisition module to a cloud or edge server according to a transmission trigger condition; a data analysis module that performs anomaly detection on the data received by the data transmission module based on preset rules and judges a fire warning signal; a warning and response module that triggers an alarm according to the analysis result of the data analysis module and executes corresponding fire emergency response measures.

2. The intelligent fire monitoring system based on the Internet of Things according to claim 1, wherein The hybrid acquisition mode includes: The data acquisition module adopts a timed sampling mode in a steady-state environment and acquires environmental data at a first sampling frequency; When the data detected by any module in the multi-source environment perception unit exceeds a first threshold, it switches to an event-driven mode and acquires environmental data at a second sampling frequency; If multiple modules of the multi-source environment perception unit jointly determine that they conform to fire characteristics, the sampling rate is further increased to a third sampling frequency; When the data in the environmental data returns to the safe range, the sampling rate is gradually decreased until it returns to the timed sampling mode.

3. The intelligent fire monitoring system based on the Internet of Things according to claim 2, characterized in that The first threshold includes: The temperature sensor detects that the temperature rise rate is greater than 3°C / s or the temperature is greater than 50°C; The smoke sensor detects that the PM2.5 concentration is greater than 100 mg / m 3 , and it lasts for 5 seconds; The combustible gas sensor detects that the CO concentration is greater than 30 ppm or the CH4 concentration is greater than 5% LEL.

4. The intelligent fire monitoring system based on the Internet of Things according to claim 2, characterized in that, The data transmission module includes: Performing compression, denoising, and fragment encapsulation on the data acquired by the data acquisition module through a data preprocessing unit; The protocol adaptive selection unit selects the MQTT-SN, TCP, or UDP protocol for transmission according to the data type; Dynamically switching communication links in LoRa, 5G, and fiber optic networks through a heterogeneous network fusion unit; Transmitting the processed data to a cloud or edge server through the selected protocol and communication link.

5. The intelligent fire monitoring system based on the Internet of Things according to claim 4, characterized in that, The compression of the data acquired by the data acquisition module includes numerical data compression and image data compression; Among them, adaptive Huffman coding or floating-point quantization compression algorithm is used for numerical data compression, and the original data is compressed to 30% - 50% of the original size; Among them, the JPEG2000 coding algorithm based on discrete wavelet transform is used for image data compression, and the compression ratio is dynamically adjusted and the lossless storage of key frames is retained.

6. The intelligent fire monitoring system based on the Internet of Things according to claim 5, characterized in that, When the data preprocessing unit performs JPEG2000 compression on image data, the thermal imaging area uses lossless coding and the background area uses lossy compression.

7. The intelligent fire monitoring system based on the Internet of Things according to claim 2, characterized in that, The transmission trigger conditions include a timed transmission mode and an event-triggered transmission mode; The timed transmission mode sends heartbeat data packets at a preset frequency and stores them; The event-triggered transmission mode immediately starts data transmission when the multi-source environment perception unit acquires environmental data at the second sampling frequency or the third sampling frequency.

8. The intelligent fire monitoring system based on the Internet of Things according to claim 2, characterized in that, The data analysis module includes: Real-time monitoring of the data in the data acquisition module and comparing it with the first threshold; When the data detected by any module in the multi-source environment perception unit exceeds the first threshold, corresponding dynamic weights are assigned according to the module type in the multi-source environment perception unit; Calculate the weighted comprehensive fire risk value. If the comprehensive fire risk value exceeds the preset risk value, generate corresponding fire warning signals.

9. The intelligent fire monitoring system based on the Internet of Things according to claim 8, characterized in that, The fire warning signals include: A first-level warning signal. When the comprehensive fire risk value exceeds the second risk threshold and is less than the third risk value, initiate local alarm. A second-level warning signal. When the comprehensive fire risk value exceeds the third risk threshold, send an emergency signal to the fire warning platform in the cloud.

10. The intelligent fire monitoring system based on the Internet of Things according to claim 8, characterized in that, The dynamic weights include: In the first type of scenario, the modules in the multi-source environment perception unit operate with the first dynamic weight. In the second type of scenario, the modules in the multi-source environment perception unit operate with the second dynamic weight.

Images