Smoke alarm device and alarm control method

Abstract

The invention discloses a smoke alarm device and an alarm control method, and particularly relates to the technical field of fire-fighting alarm devices.The device comprises a shell, and a multi-mode sensor module is arranged in the shell and used for collecting parameters in the environment in real time and generating detection signals; the data processing unit is electrically connected with the multi-mode sensor module through an I2C bus and is used for performing multi-dimensional analysis on the detection signal and generating an alarm decision instruction; the grading alarm module is used for receiving the alarm decision instruction through a GPIO (General Purpose Input / Output) interface; wherein the grading alarm module comprises a visible light warning assembly which comprises an RGBLED array and a laser projector and is used for generating a visual alarm and escape guidance; the sound wave generating device is used for transmitting a directional frequency-adjustable sound wave signal; the wireless communication module is electrically connected with the data processing unit; the invention aims to solve the problem that a smoke alarm device in the prior art can only play a role in warning and cannot effectively guide evacuation of users.

Description

Technical Field

[0001] The present invention relates to the technical field of fire alarm devices, and particularly to a smoke alarm device and an alarm control method. Background Art

[0002] As the core device for fire early warning, the technology of smoke alarm devices began in the mid-20th century. Early products were mainly based on ionization or photoelectric sensor principles, triggering audible and visual alarms by detecting the concentration of smoke particles. With the acceleration of urbanization and the improvement of fire safety standards, the limitations of traditional devices in terms of false alarm rate, response sensitivity, and environmental adaptability have become increasingly prominent. For example, non-fire particles such as cooking fumes and water vapor are likely to cause false alarms, while the risk of missed alarms for low-smoke concentration fires such as smoldering fires is relatively high; in addition, the single-threshold triggering mechanism is difficult to meet the multi-scenario requirements under complex building structures. In recent years, the integration of Internet of Things and intelligent algorithm technologies has promoted the upgrade of alarm control methods. For example, through multi-sensor data fusion (temperature, CO concentration, smoke particle size analysis) combined with machine learning models, dynamic discrimination of fire types can be achieved. However, there is still room for optimization in the edge computing ability, multi-device cooperation strategy, and cross-platform compatibility of existing systems, and low-power design and real-time requirements need to be taken into account to meet the strict specifications of international standards such as UL 217 for reliability and response time.

[0003] The invention patent with the application number: CN201710919773.3 and the publication number: CN107767614A (hereinafter referred to as "Prior Art 1") discloses a smoke alarm method and a smoke alarm device, which includes the steps: S1. Real-time obtain a smoke signal and detect the smoke concentration value, and record the time of this acquisition; S2. Judge whether the smoke concentration value exceeds a preset smoke concentration threshold. If it does not exceed, return to S1, otherwise enter S3; S3. Generate a smoke alarm trigger signal and send out an alarm signal, monitor the time interval value between the current time and the time when the smoke signal was obtained in S1, and enter S4; S4. When the time interval from the most recent acquisition of the smoke signal reaches 5 minutes, obtain the smoke signal again and detect the smoke concentration value; S5. Judge whether it exceeds the preset smoke concentration threshold. If it does not exceed, return to S1; otherwise, send out an alarm signal and return to S4.

[0004] The specification of Prior Art 1 discloses a smoke alarm method. When in use, there are two modes: real-time monitoring and interval monitoring. After the alarm mechanism is triggered, the smoke detection in the air adopts an interval detection method, which can avoid repeated alarms all the time. However, in actual applications, it is impossible to guide the evacuation of users, and it can only play a warning role and cannot effectively guide the evacuation of users. Summary of the Invention

[0005] The present invention provides a smoke alarm device and an alarm control method, aiming to solve the problem that the existing smoke alarm devices can only play a warning role and cannot effectively guide the evacuation of users.

[0006] To solve the above technical problems, the technical solution adopted by the present invention is: A smoke alarm device, including a housing, inside which are provided: A multimodal sensor module, configured to collect parameters in the environment in real time and generate detection signals; A data processing unit, electrically connected to the multimodal sensor module through an I2C bus, and configured to perform multi-dimensional analysis on the detection signals and generate alarm decision instructions; A hierarchical alarm module, configured to receive the alarm decision instructions through a GPIO interface; Among them, the hierarchical alarm module includes: A visible light warning component, including an RGB LED array and a laser projector, and configured to generate visual alarms and escape guides; An acoustic wave generating device, configured to emit a directionally adjustable frequency acoustic wave signal; A wireless communication module, electrically connected to the data processing unit, and configured to transmit alarm information and environmental data to a user terminal and a cloud server; An emergency power supply unit, which supplies power to the multimodal sensor module, the data processing unit, the hierarchical alarm module, the visible light warning component, the acoustic wave generating device, and the wireless communication module through a power management circuit.

[0007] Further, the multimodal sensor module includes a photoelectric smoke detector, a wide-spectrum infrared thermal imaging sensor, a VOC gas concentration sensor, and an ultrasonic airflow monitoring probe; the photoelectric smoke detector is arranged below the air inlet hole at the top of the housing and is configured to detect the concentration of smoke particles through the scattered light intensity; the wide-spectrum infrared thermal imaging sensor is installed inside the heat conduction plate on the side wall of the housing and is configured to generate a temperature gradient distribution map; the VOC gas concentration sensor is communicated with external air through an air pump and is configured to detect the content of volatile organic compounds; the ultrasonic airflow monitoring probe is arranged in the convection channel at the bottom of the housing and is configured to measure the air flow rate and turbulence characteristics.

[0008] Further, the data processing unit includes a microcontroller and a dynamic threshold algorithm memory, the microcontroller is connected to the multimodal sensor module, the hierarchical alarm module, the wireless communication module, and the dynamic threshold algorithm memory; the dynamic threshold algorithm memory is configured to obtain data through the microcontroller and is used to store and execute algorithms; Among them, the algorithms stored in the dynamic threshold algorithm memory include an environmental baseline generation module, a time-frequency joint analysis module, a sensitivity compensation module, and an interference pattern recognition module. The environmental baseline generation module is used to establish a dynamic baseline based on historical data and perform a differential comparison with real-time monitoring values; the time-frequency joint analysis module is used to simultaneously execute on optoelectronic signals; the sensitivity compensation module is used to adjust the alarm threshold according to the input values of temperature and humidity sensors according to a pre-stored compensation matrix; the interference pattern recognition module is used to perform similarity matching between sensor data and a feature library.

[0009] Furthermore, the visible light warning component includes: a rotary scanning drive circuit, a laser projection control unit, and a stroboscopic synchronization module; among them, the rotary scanning drive circuit is connected to the RGB LED array and is used to generate a spiral warning light effect; the laser projection control unit is used to generate an escape path projection according to building floor plan data; the stroboscopic synchronization module is used to coordinate the alarm rhythm of the RGB LED array and the sound wave generating device, and the stroboscopic frequency has a 1:4 harmonic relationship with the sound wave frequency.

[0010] Furthermore, the wireless communication module includes a dual-mode RF front end, a Mesh networking unit, and a linkage instruction generator; among them, the Mesh networking unit is used to automatically form a self-healing network with adjacent devices when emergency power supply is detected; the linkage instruction generator is used to trigger operations according to the alarm level; the operations include closing the intelligent gas valve through a Zigbee gateway, sending a positioning beacon to a fire-fighting drone, and activating the emergency start circuit of the building smoke exhaust system.

[0011] Among them, the alarm control method of the smoke alarm device includes the following steps: S1: Environmental baseline modeling stage: S101: Continuously collect environmental parameters through a multi-modal sensor module; S102: Use a sliding window algorithm to calculate the dynamic baseline value of each parameter, and the window size is adjustable from 1 to 24 hours; S2: Real-time monitoring stage; S201: Perform a differential comparison between the current detection value and the dynamic baseline value to generate a deviation index; S202: When the deviation index exceeds the first threshold, activate the time-frequency joint analysis; S3: Multi-level verification stage; S301: Primary verification, judge whether the single sensor over-threshold duration is greater than T1 (T1 = 10 - 60 seconds); S302: Secondary verification, calculate the multi-sensor data correlation coefficient, and trigger false alarm filtering when the value is less than R (R = 0.6 - 0.8); S303: Final verification, analyzing the spatial distribution pattern through a convolutional neural network; S4: Hierarchical response stage, selecting the response level according to the verification result, and synchronously adjusting the working modes of each alarm component through the control bus.

[0012] Further, in S4, the hierarchical response stage further includes the following steps: S401: Primary response mode; S4011: Activate the breathing flash mode of the RGBLED array, making the flashing frequency of the RGBLED array 1 - 2 Hz; S4012: Send an APP push notification through the wireless communication module; S402: Secondary response mode; S4021: Start the directional alarm of the sound wave generating device, with the sound pressure level adjustable from 75 - 90 dB; S4022: Control the laser projector to display the shortest escape path; S403: Tertiary response mode; S4031: Trigger a full - power audible and visual alarm, and synchronously upload the alarm coordinates to the fire cloud platform; S4032: Cut off the power supply of modules unrelated to the communication link, and give priority to ensuring the operation of the communication link.

[0013] Further, in S302, false alarm filtering includes: S3021: Feature comparison step, dynamically time - warping and matching the current sensor data with the interference feature library; S3022: Adversarial verification step, when the matching degree exceeds 85%, send a confirmation request to the user, and delay the alarm trigger for 30 - 120 s; S3023: Self - learning step, add the features of the confirmed false alarm events to the local feature library, and update the cloud model through federated learning.

[0014] Further, in S1, the environmental baseline modeling stage further includes: S103: Environmental compensation sub - step, adjusting the baseline value according to the input values of the temperature and humidity sensors according to the following formula: Corrected baseline = Original baseline × [1 + α(T - 25) + β(H - 50)] (α = 0.02 / °C, β = 0.005 / %RH) S104: Sudden fluctuation handling sub - step, when a parameter mutation exceeding ±30% is detected, start the buffer period mechanism and pause the baseline update for 5 - 15 min.

[0015] Further, the emergency power supply unit includes the following steps when in use: S501: Power monitoring loop to detect the voltage of the lithium battery and the charge of the supercapacitor in real time; S502: Power consumption adjustment strategy. When the remaining battery power < 40%, the continuous scanning of the thermal imaging sensor is turned off; when the remaining battery power < 20%, switch to the geofence broadcast mode and send a beacon every 60 seconds; S503: Network mutual assistance mechanism to coordinate the power supply balance of adjacent devices through the Mesh network and extend the battery life of the overall system.

[0016] Compared with the prior art, the present invention has the following beneficial effects: Through the multi-modal sensor module, the present invention realizes the four-dimensional collaborative perception of environmental parameters. Compared with the traditional single optoelectronic detection, the false alarm rate of this solution is reduced and the fire recognition accuracy is increased in the kitchen cooking interference scenario. At the same time, the data processing unit can adjust the alarm threshold in real time according to the changes in environmental temperature and humidity; the hierarchical alarm module outputs through visible light and sound waves in a dual-modal manner. It breathes and blinks at 1Hz through the annular RGB LED to achieve the first-level response for primary early warning. The laser projector generates an escape path (positioning accuracy ±0.5m), and at the same time, the sound wave generating device emits a 95dB directional sound wave (effective coverage radius 15m) to form a second-level response, thereby achieving the effect of guiding the escape path of the user; finally, the emergency power supply unit ensures the reliability of the system. Specific embodiments

[0017] The present invention will be further described below in conjunction with embodiments. The described embodiments are only a part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative work fall within the protection scope of the present invention.

[0018] This embodiment discloses a smoke alarm device, including a housing, and the following are arranged inside the housing: A multi-modal sensor module for collecting parameters in the environment in real time and generating detection signals; A data processing unit electrically connected to the multi-modal sensor module through the I2C bus and used for multi-dimensional analysis of the detection signals and generating alarm decision instructions; A hierarchical alarm module for receiving the alarm decision instructions through the GPIO interface; Among them, the hierarchical alarm module includes: A visible light warning component, including an RGB LED array and a laser projector, and used for generating visual alarms and escape guidance; A sound wave generating device for emitting a directional adjustable frequency sound wave signal; A wireless communication module, electrically connected to the data processing unit, and used for transmitting alarm information and environmental data to a user terminal and a cloud server; An emergency power supply unit, which supplies power to the multi-modal sensor module, the data processing unit, the hierarchical alarm module, the visible light warning component, the sound wave generating device, and the wireless communication module through a power management circuit.

[0019] Through the multi-modal sensor module, the present invention realizes the four-dimensional collaborative perception of environmental parameters. Compared with the traditional single optoelectronic detection, the false alarm rate of this solution is reduced and the fire recognition accuracy is increased in the kitchen cooking interference scenario. At the same time, the data processing unit can adjust the alarm threshold in real time according to the changes in environmental temperature and humidity; the hierarchical alarm module outputs through visible light and sound wave dual modalities. The annular RGB LED breathes and blinks at 1Hz to achieve the first-level response for primary early warning. The laser projector generates an escape route (positioning accuracy ±0.5m), and at the same time, the sound wave generating device emits a 95dB directional sound wave (effective coverage radius 15m) to form a second-level response, thereby achieving the effect of guiding the escape route of users; finally, the emergency power supply unit ensures the reliability of the system.

[0020] In some embodiments, the multi-modal sensor module includes a photoelectric smoke detector, a wide-spectrum infrared thermal imaging sensor, a VOC gas concentration sensor, and an ultrasonic air flow monitoring probe; the photoelectric smoke detector is arranged below the air inlet hole at the top of the housing and is used for detecting the concentration of smoke particles through the scattered light intensity; the wide-spectrum infrared thermal imaging sensor is installed inside the heat conduction plate on the side wall of the housing and is used for generating a temperature gradient distribution map; the VOC gas concentration sensor is communicated with the external air through an air pump and is used for detecting the content of volatile organic compounds; the ultrasonic air flow monitoring probe is arranged in the convection channel at the bottom of the housing and is used for measuring the air flow rate and turbulence characteristics.

[0021] As an alternative implementation manner, in this embodiment, the housing is injection-molded with a flame-retardant ABS material. The internal circuit board is fixed to the bottom surface of the housing through a heat-conducting silica gel pad. The multi-modal sensor module communicates with the data processing unit through a 4-wire I2C bus at a rate of 400kHz. The bus impedance matching resistor (120Ω) is arranged near both sides of the connector. The data processing unit uses an STM32H743 microcontroller. Its GPIO34-37 pins are connected to the SPI interface of the dynamic threshold algorithm memory (model W25Q128JVSIQ), and the clock is configured in a 50MHz dual-line mode. The GPIO interface of the hierarchical alarm module drives the MOSFET switch tube (AO3400) of the RGBLED array through an optocoupler isolation circuit.

[0022] The optoelectronic smoke detector uses an OSD-5ME particulate sensor, which is installed 15 mm directly below the 6×Φ3 mm intake hole at the top, and the sampling period is adjustable (100 - 500 ms).

[0023] The wide-spectrum infrared thermal imaging sensor (MLX90640) is closely attached to the 2-mm-thick aluminum alloy heat conduction plate on the side wall, and thermal contact is ensured through thermal paste (GD900).

[0024] The VOC sensor (SGP30) is connected to a micro air pump (FSP0504) to achieve an air exchange of 300 ml per minute.

[0025] The ultrasonic probe (MB7060) is set in the 8-mm-high convection channel at the bottom, and the measurement angle is 30°.

[0026] In some embodiments, the data processing unit includes a microcontroller and a dynamic threshold algorithm memory. The microcontroller is connected to the multimodal sensor module, the hierarchical alarm module, the wireless communication module, and the dynamic threshold algorithm memory. The dynamic threshold algorithm memory is used to obtain data through the microcontroller and is used to store and execute algorithms; Among them, the algorithms stored in the dynamic threshold algorithm memory include an environmental baseline generation module, a time-frequency joint analysis module, a sensitivity compensation module, and an interference pattern recognition module. The environmental baseline generation module is used to establish a dynamic baseline based on historical data and perform a differential comparison with the real-time monitoring value. The time-frequency joint analysis module is used to simultaneously execute on optoelectronic signals. The sensitivity compensation module is used to adjust the alarm threshold according to the input value of the temperature and humidity sensor according to a pre-stored compensation matrix. The interference pattern recognition module is used to perform a similarity match between the sensor data and the feature library.

[0027] In the actual use process, data collection and preprocessing need to be carried out first. In this process, the microcontroller polls the multimodal sensor module through the I²C bus at a clock frequency of 400 kHz and collects data in the following order: 1. Optoelectronic smoke concentration: Read once every 200 ms, 12-bit ADC quantization value (0 - 4095); 2. Infrared temperature distribution: Obtain a 32×24 pixel matrix every 5 seconds (accuracy ±0.5°C); 3. VOC gas concentration: Read the equivalent CO2 value every 10 seconds (range 400 - 60000 ppm) 4. Airflow velocity: Measure the turbulence intensity once every 1 second (0 - 5 m / s) Then, signal normalization processing is required. Next, the dynamic baseline needs to be generated through the environmental baseline generation module. In this process, historical data modeling is first performed through the following formula: The microcontroller executes a sliding window algorithm (the window size can be configured from 1 to 24 hours): calculate the exponentially weighted moving average (EWMA): Wherein, : current smoke concentration; : baseline value of the previous hour; : new baseline value; ( = 0.05) After calculating the exponentially weighted moving average, it is stored, and the 72-hour historical baseline curve is stored in the FRAM (anti-power-loss).

[0028] As an alternative implementation, in this embodiment, the currently detected = 2.8% / m, the original baseline = 1.2% / m, therefore, after substituting into the formula, .

[0029] Then calculate the real-time differential comparison and generate the deviation index (DI) formula as follows: Wherein, when DI > 15% for 30 seconds, trigger time-frequency joint analysis; The new baseline value is stored in the dynamic threshold algorithm memory for the next cycle of differential comparison. When the real-time value continuously exceeds the baseline by 15% for 5 minutes (i.e., 2.8 > 1.28×1.15 = 1.47), trigger time-frequency analysis.

[0030] After triggering the time-domain analysis module, it is necessary to use the time-frequency joint analysis module to calculate the time-domain integration and extract the frequency-domain features; When calculating the time-domain integration, window integration is performed on the optoelectronic signal (the window length is adjustable from 200 ms to 5 min), and the formula is as follows: Where N is the number of points collected within the window; is the smoke value at the kth sampling point; corresponds to the above baseline value; As an alternative implementation, in this embodiment, 100 sampling points are set, and the average deviation is 0.5% / m: after substituting into the formula: 100×(0.5) 2 = 25% 2 ·s; When exceeding the first-level threshold of 3.5%2·s (preset value), it is marked as a suspicious signal.

[0031] Then, it is necessary to extract the frequency-domain features. During this process, it is necessary to calculate the energy proportion of the frequency band of the fire characteristics. The formula is as follows: When > 65%, it is marked as a suspected fire.

[0032] As an alternative implementation, in this embodiment, if the total energy is 2000 and the energy of the characteristic frequency band is 1560, after substituting into the formula, Since it exceeds 65% of the threshold, an activation signal is sent to the sensitivity compensation module.

[0033] When the sensitivity compensation module receives the signal from the time-frequency joint analysis module; The compensation coefficient for reading the current temperature and humidity look-up table is calculated through the following formula: Temperature index: ; Humidity index: ; Then, the compensation coefficient is read from the dynamic threshold algorithm memory ; Then, the alarm threshold is trimmed. The formula is as follows: Among them, is the basic threshold; T is the current temperature; H is the current humidity; As an alternative implementation, in this embodiment, when the current temperature is 40 °C and the humidity is 70%, = 1.3; after substituting into the formula = 3.0 × [1 + 0.02(40 - 25) + 0.005(70 - 50)] = 3.0 × 1.4 = 4.2% / m; from the calculation results, it can be seen that the threshold is increased to 4.2% / m due to the high temperature and high humidity environment to avoid false alarms caused by environmental interference.

[0034] During this process, the interference pattern recognition module also needs to perform feature library matching and false alarm suppression strategies. Among them, it is necessary to use the dynamic time warping (DTW) algorithm to calculate the similarity between the sensor data and the feature library: the feature template is: store the 60-second time series data of typical interference scenarios (cooking, steam, dust); The similarity calculation formula is as follows: Among them, is the preset maximum distance; When > 85%, it is determined as interference; When implementing the false alarm suppression strategy, it is first necessary to trigger a delayed alarm for 30 - 120 seconds. During this period, the trend of the data is continuously verified. When the user confirms a false alarm, the feature is added to the local feature library.

[0035] As an alternative implementation method, in this embodiment, The preset value is 100, and the between the current data and the steam interference template = 12; After substituting into the formula,

[0036] After triggering the delayed alarm, it is necessary to perform multi - level verification and decision - making on the alarm; In this process, primary verification is first performed. The condition is that the smoke concentration > 4.2% / m lasts for 30 seconds, and the response action is to activate the RGBLED breathing flash. The secondary judgment condition is R < 0.7, and the response action is to turn off non - essential sensors and increase the acquisition frequency. Finally, final verification is required. The judgment condition is that the fire probability output by the CNN is approximately equal to 90%, and the response action is to trigger a full - system alarm and upload to the fire protection platform; In the above actual measurement, when performing primary verification, the smoke concentration is 4.5% / m for 45 seconds, and the verification is passed; Then secondary verification is performed. It is necessary to calculate the coefficient relationship between delay and temperature. The calculation formula is as follows: After substituting specific values: R = 0.96; Since R > 0.7, therefore, the verification is passed; Finally, final verification is performed. Through the thermal imaging matrix, the fire probability is judged to be 89.7%, which is greater than 80%. Therefore, a three - level alarm is triggered.

[0037] After triggering the three - level alarm, audible and visual alarms are given. The visible light warning component performs stroboscopic flashing with a frequency of 4Hz and a period of 250ms.

[0038] In summary, the serial logic of the data processing unit is as follows: Generate a real - time dynamic reference benchmark through the environmental baseline generation module, then capture transient and steady - state features through the time - frequency joint analysis module, then adjust the threshold according to the specific environmental conditions through the sensitivity compensation module, then filter out false information through the interference pattern recognition module, and finally determine the level of the fire situation through multi - level verification and perform corresponding responses through the hierarchical alarm module; The advantage of such a setting is that different types of alarms can be given according to the fire situation in different situations, enabling users to judge the severity of the fire situation based on specific sounds, facilitating user evacuation. At the same time, the number of false alarms can also be reduced through the interference recognition module.

[0039] In some embodiments, the visible light warning component includes: a rotating scanning drive circuit, a laser projection control unit and a strobe synchronization module; wherein the rotating scanning drive circuit is connected to the RGBLED array and is used to generate a spiral warning light effect; the laser projection control unit is used to generate an escape path projection based on the building plan data; the strobe synchronization module is used to coordinate the alarm rhythm of the RGBLED array and the sound wave generating device, and the strobe frequency and the sound wave frequency are in a 1:4 harmonic relationship.

[0040] As an optional implementation, in this embodiment, when the visible light warning component is in standby mode, the RGBLED array is observed to see if it is operating normally, the laser projector is in standby mode, the microcontroller continuously monitors the data of the multimodal sensor, and the dynamic baseline is updated every 5 minutes; When the primary alarm is triggered, the RGBLED array flashes and the light switches to green, then drives the laser projector to rotate and scan 180°. The wireless communication module sends positioning information to the remote control terminal until the alarm is lifted. When the secondary alarm is triggered, the RBG LED array lights switch to red with a brightness of 800cd / m 2 ; The laser projector projects a pre-set escape route, and several laser projectors project a continuous arrow pattern to facilitate the user's escape instructions.

[0041] When the third-level alarm is triggered, the light of the RBGLED array switches to blue and red alternating flashes, with a brightness of 1000cd / m 2 ; The width of the path projected by the laser projector is increased to 8 pixels, and the frequency of the flashing of the arrow pattern is increased to 4 Hz; the emergency power supply unit intervenes to release 6A current to ensure full power operation of the laser projection.

[0042] The advantage of this setting is that since there are several smoke alarm devices installed on the ceiling of the building, the escape route can be projected through a laser projector to guide users to escape. Different smoke concentrations can be adapted according to different light colors and brightness, so that users can see the escape arrows more clearly and the guiding effect is better.

[0043] In some embodiments, the wireless communication module includes a dual-mode radio frequency front end, a Mesh networking unit, and a linkage instruction generator; wherein, the Mesh networking unit is configured to automatically form a self-healing network with adjacent devices when emergency power supply is detected; the linkage instruction generator is configured to trigger operations according to the alarm level; the operations include closing the intelligent gas valve through the Zigbee gateway, sending a positioning beacon to the fire-fighting drone, and activating the emergency start circuit of the building smoke exhaust system.

[0044] As an alternative implementation, in this embodiment, the dual-mode radio frequency front end has a wifi mode and a Bluetooth mode. The acoustic wave generating device is connected to the user's smartphone through the dual-mode radio frequency front end. When the acoustic wave generating device alarms, it transmits to the user's smartphone, enabling the user to be aware of the fire in the first time for evacuation.

[0045] In some different embodiments, this embodiment also discloses an alarm control method for a smoke alarm device, including the following steps: S1: Environmental baseline modeling stage: S101: Continuously collect environmental parameters through the multi-modal sensor module; S102: Calculate the dynamic baseline value of each parameter using the sliding window algorithm, and the window size is adjustable from 1 to 24 hours; S2: Real-time monitoring stage; S201: Perform differential comparison between the current detected value and the dynamic baseline value to generate a deviation index; S202: When the deviation index exceeds the first threshold, activate time-frequency joint analysis; S3: Multi-level verification stage; S301: Primary verification, determine whether the single sensor over-threshold duration is greater than T1 (T1 = 10 - 60 seconds); S302: Secondary verification, calculate the multi-sensor data correlation coefficient, and trigger false alarm filtering when the value is less than R (R = 0.6 - 0.8); S303: Final verification, analyze the spatial distribution pattern through a convolutional neural network; S4: Hierarchical response stage, select the response level according to the verification result, and synchronously adjust the working mode of each alarm component through the control bus.

[0046] In some embodiments, in S4, the hierarchical response stage further includes the following steps: S401: Primary response mode; S4011: Activate the breathing flash mode of the RGBLED array, such that the flashing frequency of the RGBLED array is 1 - 2Hz; S4012: Send an APP push notification through the wireless communication module; S402: Secondary response mode; S4021: Activate the directional alarm of the acoustic wave generating device, with the sound pressure level adjustable between 75 - 90 dB; S4022: Control the laser projector to display the shortest escape route; S403: Tertiary response mode; S4031: Trigger the full - power acoustic - optical alarm and synchronously upload the alarm coordinates to the fire cloud platform; S4032: Cut off the power supply of modules irrelevant to the communication link and prioritize ensuring the operation of the communication link.

[0047] In some embodiments, in S302, false alarm filtering includes: S3021: Feature comparison step, dynamically time - warping match the current sensor data with the interference feature library; S3022: Adversarial verification step, when the matching degree exceeds 85%, send a confirmation request to the user and delay the alarm trigger for 30 - 120 s; S3023: Self - learning step, add the features of the confirmed false alarm events to the local feature library and update the cloud model through federated learning.

[0048] In some embodiments, in S1, the environmental baseline modeling stage further includes: S103: Environmental compensation sub - step, according to the input values of the temperature and humidity sensors; S104: Sudden fluctuation handling sub - step, when it is detected that the parameter mutation exceeds ±30%, start the buffer period mechanism and pause the baseline update for 5 - 15 min.

[0049] In some embodiments, the supercapacitor bank (117, 2.7 V / 10 F × 3 in series) inside the emergency power supply unit is connected in parallel to the positive pole of the main power line through a Schottky diode (SS34). The lithium - battery slot adopts a spring - pin contact structure, and the power management circuit uses a BQ25895 chip to achieve automatic switching, with the switching response time < 200 μs.

[0050] The usage includes the following steps: S501: Power monitoring loop, real - time detect the lithium - battery voltage and the supercapacitor charge; S502: Power consumption regulation strategy, when the remaining power < 40%, turn off the continuous scanning of the thermal imaging sensor; when the remaining power < 20%, switch to the geofence broadcast mode and send a beacon every 60 seconds; S503: Network mutual - assistance mechanism, coordinate the power supply balance of adjacent devices through the Mesh network to extend the overall system endurance.

[0051] The specific usage method of the present invention is: First, physically install the housing in the center of the ceiling, less than 30 cm away from the wall, ensuring that the air inlet holes at the top are not blocked. Then, use expansion bolts to install the heat conduction plate on the wall to ensure the efficiency of infrared sensor heat conduction.

[0052] Then, configure the power supply. Connect to the 220V AC main power supply, observe the power indicator light, and then install the replaceable lithium battery. The final stage of physical installation requires wireless pairing to connect the smartphone to the wireless communication module for easy receipt of alarms.

[0053] Next, model the environmental baseline. The multi-modal sensors continuously collect environmental parameters. Among them, the optoelectronic smoke concentration is sampled every 10 s, the distribution of infrared temperature is monitored every 5 min, and the VOC concentration is averaged hourly. Then, the microcontroller executes the above sliding window algorithm to generate a dynamic baseline. Then, learn the compensation parameters. Continuously record for 48 hours through the temperature and humidity sensors to construct a compensation matrix, and then store typical interference scenarios in the feature library to complete the identification of interference items.

[0054] Then, regularly organize users and trigger first-level, second-level, and third-level alarms for users to identify and conduct fire warning drills.

[0055] In addition, the terms "first", "second", "third", and "fourth" are only used for descriptive purposes and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, features defined with "first", "second", "third", and "fourth" may explicitly or implicitly include at least one of such features.

[0056] In the present invention, unless otherwise clearly specified and defined, terms such as "install", "set", "connect", "fix", "swivel connection", etc. should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or integrated; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the internal communication of two components or the interaction relationship between two components. Unless otherwise clearly defined, for those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0057] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. Smoke alarm device, comprising a housing, characterized in that, Inside the housing are provided: A multimodal sensor module for collecting parameters in the environment in real time and generating detection signals; A data processing unit electrically connected to the multimodal sensor module via an I2C bus and used for performing multi-dimensional analysis on the detection signals and generating alarm decision instructions; A hierarchical alarm module for receiving the alarm decision instructions via a GPIO interface; Wherein, the hierarchical alarm module includes: A visible light warning component, including an RGBLED array and a laser projector, and used for generating visual alarms and escape guidance; An acoustic wave generating device for emitting directionally adjustable frequency acoustic wave signals; A wireless communication module electrically connected to the data processing unit and used for transmitting alarm information and environmental data to a user terminal and a cloud server; An emergency power supply unit for supplying power to the multimodal sensor module, the data processing unit, the hierarchical alarm module, the visible light warning component, the acoustic wave generating device, and the wireless communication module through a power management circuit.

2. The smoke alarm device according to claim 1, wherein: The multimodal sensor module includes a photoelectric smoke detector, a broadband infrared thermal imaging sensor, a VOC gas concentration sensor, and an ultrasonic airflow monitoring probe; the photoelectric smoke detector is arranged below the air inlet hole at the top of the housing and used for detecting the concentration of smoke particles through the intensity of scattered light; the broadband infrared thermal imaging sensor is installed inside the heat conducting plate on the side wall of the housing and used for generating a temperature gradient distribution map; the VOC gas concentration sensor is communicated with external air through an air pump and used for detecting the content of volatile organic compounds; the ultrasonic airflow monitoring probe is arranged inside the convection channel at the bottom of the housing and used for measuring the air flow rate and turbulence characteristics.

3. The smoke alarm device according to claim 1, characterized in that: The data processing unit includes a microcontroller and a dynamic threshold algorithm memory, the microcontroller is connected to the multimodal sensor module, the hierarchical alarm module, the wireless communication module, and the dynamic threshold algorithm memory; the dynamic threshold algorithm memory is used for obtaining data through the microcontroller and used for storing and executing algorithms; Wherein, the algorithms stored in the dynamic threshold algorithm memory include an environmental baseline generation module, a time-frequency joint analysis module, a sensitivity compensation module, and an interference pattern recognition module. The environmental baseline generation module is used for establishing a dynamic baseline based on historical data and performing differential comparison with real-time monitoring values; the time-frequency joint analysis module is used for simultaneously executing on photoelectric signals; the sensitivity compensation module is used for adjusting the alarm threshold according to the input value of a temperature and humidity sensor according to a pre-stored compensation matrix; the interference pattern recognition module is used for performing similarity matching between sensor data and a feature library.

4. The smoke alarm device according to claim 1, characterized in that: The visible light warning component includes: a rotary scanning drive circuit, a laser projection control unit, and a stroboscopic synchronization module; wherein, the rotary scanning drive circuit is connected to the RGBLED array and used for generating a spiral warning light effect; the laser projection control unit is used for generating an escape path projection according to building floor plan data; the stroboscopic synchronization module is used for coordinating the alarm rhythm of the RGBLED array and the acoustic wave generating device, and the stroboscopic frequency has a 1:4 harmonic relationship with the acoustic wave frequency.

5. The smoke alarm device according to claim 1, characterized in that: The wireless communication module includes a dual-mode RF front-end, a Mesh networking unit, and a linkage instruction generator. Among them, the Mesh networking unit is used to automatically form a self-healing network with adjacent devices when emergency power supply is detected. The linkage instruction generator is used to trigger operations according to the alarm level. The operations include closing the intelligent gas valve through the Zigbee gateway, sending a positioning beacon to the fire drone, and activating the emergency start circuit of the building smoke exhaust system.

6. A method for alarm control of a smoke alarm device, based on the smoke alarm device according to any one of claims 1-5, characterized in that, It includes the following steps: S1: Environmental baseline modeling stage: S101: Continuously collect environmental parameters through the multi-modal sensor module; S102: Use the sliding window algorithm to calculate the dynamic baseline values of each parameter, and the window size is adjustable from 1 to 24 hours; S2: Real-time monitoring stage; S201: Perform differential comparison between the current detection value and the dynamic baseline value to generate a deviation index; S202: When the deviation index exceeds the first threshold, activate time-frequency joint analysis; S3: Multi-level verification stage; S301: Primary verification, determine whether the duration of a single sensor exceeding the threshold is greater than T1 (T1 = 10 - 60 seconds); S302: Secondary verification, calculate the correlation coefficient of multi-sensor data, and trigger false alarm filtering when the value is less than R (R = 0.6 - 0.8); S303: Final verification, analyze the spatial distribution pattern through a convolutional neural network; S4: Hierarchical response stage, select the response level according to the verification result, and synchronously adjust the working modes of each alarm component through the control bus.

7. The alarm control method of the smoke alarm device according to claim 6, characterized in that: In S4, the hierarchical response stage further includes the following steps: S401: First-level response mode; S4011: Activate the breathing flashing mode of the RGBLED array, so that the flashing frequency of the RGBLED array is 1 - 2 Hz; S4012: Send an APP push notification through the wireless communication module; S402: Second-level response mode; S4021: Start the directional alarm of the sound wave generating device, and the sound pressure level is adjustable from 75 to 90 dB; S4022: Control the laser projector to display the shortest escape path; S403: Third-level response mode; S4031: Trigger a full-power audible and visual alarm, and synchronously upload the alarm coordinates to the fire cloud platform; S4032: Cut off the power supply of modules irrelevant to the communication link, and give priority to ensuring the operation of the communication link.

8. The alarm control method of the smoke alarm device according to claim 6, characterized in that: In S302, false alarm filtering includes: S3021: Feature comparison step, perform dynamic time warping matching between the current sensor data and the interference feature library; S3022: Adversarial verification step, when the matching degree exceeds 85%, send a confirmation request to the user and delay the alarm trigger for 30 - 120 s; S3023: Self-learning step, add the features of the confirmed false alarm events to the local feature library, and update the cloud model through federated learning.

9. The alarm control method of the smoke alarm device according to claim 6, characterized in that: In S1, the environmental baseline modeling stage further includes: S103: Environmental compensation sub-step, adjust the baseline value according to the input values of the temperature and humidity sensors according to the following formula: Corrected baseline = Original baseline × [1 + α(T - 25) + β(H - 50)] (α = 0.02 / ℃, β = 0.005 / %RH); S104: Sub-step for handling sudden fluctuations. When it is detected that the parameter mutation exceeds ±30%, start the buffer period mechanism and pause the baseline update for 5 - 15 minutes.

10. The alarm control method of the smoke alarm device according to claim 6, characterized in that: When the emergency power supply unit is in use, it includes the following steps: S501: Power supply monitoring loop, continuously detect the voltage of the lithium battery and the charge quantity of the supercapacitor in real time; S502: Power consumption adjustment strategy. When the remaining power < 40%, turn off the continuous scanning of the thermal imaging sensor; when the remaining power < 20%, switch to the geofence broadcast mode and send a beacon every 60 seconds; S503: Network mutual assistance mechanism, coordinate the power supply balance of adjacent devices through the Mesh network to extend the battery life of the overall system.

Images