Fire extinguishing system based on multi-modal sensor and multi-stage linkage control method

Abstract

The invention relates to the technical field of fire early warning, in particular to a fire extinguishing system based on a multi-modal sensor and a multi-stage linkage control method.The fire extinguishing system based on the multi-modal sensor comprises a multi-spectral light source module, the multi-spectral light source module is configured to emit a dual-wavelength light beam with a preset divergence angle, the first wavelength # imgabs0 # is located in the visible light wave band, and the second wavelength # imgabs1 # is located in the near-infrared wave band; a temperature gradient sensor array and a scattering angle adjustable photoelectric detector array are arranged in the composite detection cavity, the temperature gradient sensor array comprises miniature thermocouple groups which are distributed in a spatial orthogonal manner, and the photoelectric detector array comprises a multi-angle receiving unit which can slide along a spherical track; the matrix eliminates coupling interference of a temperature field on scattering characteristics of smoke particles through a dual-wavelength phase locking algorithm; according to the scheme, the dynamic coupling interference of the temperature field and the scattering characteristic can be fundamentally solved.

Description

Technical Field

[0001] The present invention relates to the technical field of fire warning, and particularly to a fire extinguishing system based on multimodal sensors and a multi-level linkage control method. Background Art

[0002] In the field of smoke concentration detection based on the principle of optical scattering, especially in fire warning systems, accurately measuring the concentration of smoke particles depends on the precise modeling of the relationship between particle size, refractive index, and scattered light intensity by the Mie scattering model. However, in a complex temperature field environment, the optical scattering characteristics of smoke particles will be significantly distorted due to temperature changes, resulting in systematic errors in traditional detection methods. Specifically, when the ambient temperature changes dynamically, the refractive index of smoke particles exhibits non-linear temperature dependence, and at the same time, the particle thermal expansion effect caused by high temperature will change its particle size distribution. These two factors together cause a non-negligible shift in the scattering cross-section parameter. In addition, the air density gradient caused by the non-uniform temperature field will further disturb the light propagation path, resulting in real-time drift of the scattering angle. Existing technologies usually adopt fixed temperature compensation coefficients or piecewise calibration strategies based on a single temperature point, but these methods have inherent defects in a dynamic temperature field: on the one hand, the fixed compensation model cannot adapt to the continuous change of scattering parameters caused by temperature gradients; on the other hand, discrete temperature sensors are difficult to accurately capture the coupling relationship between the three-dimensional space temperature distribution and the scattering process. Experiments show that when the ambient temperature fluctuates in the range of -20°C to 300°C, the concentration error of traditional detection algorithms can reach more than ±25%, and signal saturation failure even occurs in local high-temperature areas at the initial stage of a fire.

[0003] Based on the above problems, there is an urgent need for a technical solution that can fundamentally solve the dynamic coupling interference between the temperature field and scattering characteristics, so as to achieve high-precision detection of smoke concentration in a wide temperature range and non-uniform thermal environment, and meet the strict requirements of the fire warning system for the robustness of complex environments. Summary of the Invention

[0004] The present invention aims at the problems existing in the prior art, and provides a fire extinguishing system based on multimodal sensors and a multi-level linkage control method; To achieve the above object, the technical solution adopted by the present invention is as follows: A fire extinguishing system based on multimodal sensors, comprising: A multispectral light source module, the multispectral light source module is configured to emit a dual-wavelength beam with a predetermined divergence angle, wherein the first wavelength is in the visible light band, and the second wavelength Located in the near-infrared band; a composite detection cavity, the composite detection cavity has a built-in temperature gradient sensor array and a scattering angle adjustable photoelectric detector array, the temperature gradient sensor array includes a micro-thermocouple group with spatial orthogonal distribution, and the photoelectric detector array includes a multi-angle receiving unit that can slide along a spherical trajectory; a signal processing unit, the signal processing unit is connected to the output end of the photoelectric detector array and has a built-in temperature-scattering angle compensation matrix, the matrix eliminates the coupling interference of the temperature field on the scattering characteristics of smoke particles through a dual-wavelength phase-locked algorithm; a graded response module, the graded response module generates a multi-level alarm instruction according to the compensated smoke concentration value and the temperature gradient change rate; wherein the dual-wavelength phase-locked algorithm realizes dynamic compensation through the following mathematical relationship: ; In the formula, is the compensation coefficient, the functional relationship Characterize the coupling characteristics of temperature and optical parameters, represents the scattering angle offset caused by temperature gradient, is the dual-wavelength spectral spacing, is the Mie scattering cross section parameter, is the scattering angle, is the current measured temperature value, Indicates standard reference temperature.

[0005] Preferably, the dual-wavelength light beam is generated by two independent semiconductor lasers, namely a 650nm red light laser diode and an 850nm near-infrared laser diode. The light beam is collimated by an aspheric lens group to form a coaxial incident light path, and an optical window with an anti-reflection coating is installed at the entrance of the composite detection cavity. The light transmission band of the optical window covers the range of 600nm to 900nm, and the transmittance is ≥95%.

[0006] Preferably, the micro-thermocouples of the temperature gradient sensor array are manufactured using a platinum resistance thin film process and are packaged in a high-temperature resistant ceramic substrate. Each thermocouple node is equipped with an independent data acquisition channel, the sampling frequency is not less than 100 Hz, and the temperature gradient difference calculation time between any two adjacent thermocouples in the array is less than 10 ms.

[0007] Preferably, each receiving unit of the photodetector array comprises a bandpass filter and a preamplifier circuit, the central wavelength of the bandpass filter is matched with the dual-wavelength light source, the bandwidth is ±5nm, the gain of the preamplifier circuit is programmable and the dynamic range covers the optical power input of 0.1μW to 10mW.

[0008] Preferably, the method for constructing the temperature-scattering angle compensation matrix includes: Building a 3D Lookup Table , where The indexed temperature range is , with a step size of ; Indexed scattering angle , with a step size of ; Indexed dual-wavelength combination mode, including Single mode, Single mode, Dual mode, three states; The compensation coefficients stored in the three-dimensional look-up table are obtained through experimental calibration, specifically satisfying: ; Among them, is the compensation coefficient, represents the temperature range index, represents the scattering angle index, represents the wavelength combination mode index, is the standard temperature, The reference scattering intensity at is the measured scattering intensity, is the material temperature coefficient, is the current temperature, is the scattering angle, is the reference scattering angle.

[0009] Preferably, the signal processing unit integrates a digital lock-in amplifier module and an adaptive filter. When synchronously demodulating the dual-wavelength scattering signal, the adaptive filter eliminates ambient light interference and automatically calibrates the baseline drift within each demodulation period, and the cut-off frequency of the adaptive filter is dynamically adjusted according to the temperature change rate.

[0010] Preferably, the hierarchical response module is communicatively connected to the building fire control system. When a medium-level alarm instruction is triggered, the air-conditioning return air valve is synchronously closed and the smoke exhaust fan is started. When a high-level alarm instruction is triggered, a structured alarm information packet including a temperature distribution thermal map and a three-dimensional cloud map of smoke concentration is sent to the emergency management department.

[0011] Preferably, the inner wall of the composite detection cavity is coated with a nano-scale dust-repellent coating, and a self-cleaning air curtain device is installed at the receiving end of the photodetector array. The air curtain device generates a directional air flow through a micro piezoelectric pump to continuously blow the dust on the surface of the optical elements.

[0012] A multi-level linkage control method, applied to the fire warning system as described in any one of the above, includes: Real-time acquisition of multi-modal sensing data, including scattered light intensity , temperature field distribution , Concentration value; Fuse multi-source data through the particle filter algorithm to estimate the probability of fire occurrence ; When , activate the deep learning model for pattern recognition; dynamically adjust the alarm threshold according to the recognition result.

[0013] Preferably, the state equation in the particle filter algorithm is expressed as: ; The observation equation is expressed as: ; Where the state vector includes smoke concentration, temperature gradient, gas diffusion rate, and the observation matrix corresponds to the multi-sensor measurement model, , are the process noise and observation noise respectively, is the state matrix, and the observation vector represents the actual sensor measurement value.

[0014] Compared with the prior art, the present invention has the following beneficial effects: This solution proposes a multi-modal fire warning system architecture, including a multi-spectral light source module, a dual-wavelength beam emission and composite detection cavity, a temperature gradient sensor + adjustable photodetector, a signal processing unit, a temperature-scattering angle compensation matrix, and a hierarchical response module. The core innovation lies in the dual-wavelength phase-locked algorithm, which dynamically corrects the interference of temperature on scattering parameters through a mathematical relationship; it can solve the technical problem that traditional single-wavelength detection causes scattering signal distortion due to the temperature sensitivity of the refractive index of smoke particles in a high and low temperature alternating environment; and the prior art uses a fixed temperature compensation coefficient and cannot cope with the continuous change of the scattering angle caused by a non-uniform temperature field. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 is a block diagram of the fire extinguishing system based on multi-modal sensors according to this application; Figure 2 is a flowchart of the multi-level linkage control method according to this application. DETAILED DESCRIPTION OF THE INVENTION

[0016] It should be noted that the methods used in the present invention are all conventional methods unless otherwise specified; the raw materials and devices used are all conventional commercially available products unless otherwise specified, and their sources are not specifically limited.

[0017] The present invention provides a fire extinguishing system based on multi-modal sensors and a multi-level linkage control method for the problems existing in the prior art; The following technical problems exist in the traditional technical solutions: Compound interference problem: In the traditional single-wavelength detection in a high-low temperature alternating environment, due to the temperature sensitivity of the refractive index of smoke particles, the scattering signal is distorted, such as the deviation of the Mie scattering cross-section; Lack of dynamic compensation: The existing technology uses a fixed temperature compensation coefficient and cannot cope with the continuous change of the scattering angle caused by a non-uniform temperature field; Defect of multi-modal collaboration: The separate design of the temperature and optical sensors leads to difficult data fusion and a high false alarm rate of alarms.

[0018] Based on the above problems, please refer to Figure 1 , this application provides a fire extinguishing system based on a multi-modal sensor, including: A multi-spectral light source module configured to emit a dual-wavelength beam with a predetermined divergence angle, where the first wavelength is in the visible light band, and the second wavelength is in the near-infrared band; a composite detection cavity with a temperature gradient sensor array and a scattering angle adjustable photodetector array built therein. The temperature gradient sensor array includes a set of micro-thermocouples distributed orthogonally in space, and the photodetector array includes multi-angle receiving units that can slide along a spherical trajectory; a signal processing unit connected to the output end of the photodetector array and having a temperature-scattering angle compensation matrix built therein. The matrix eliminates the coupling interference of the temperature field on the scattering characteristics of smoke particles through a dual-wavelength phase-locked algorithm; a hierarchical response module that generates multi-level alarm instructions based on the compensated smoke concentration value and the temperature gradient change rate; where the dual-wavelength phase-locked algorithm realizes dynamic compensation through the following mathematical relationship: ; In the formula, is the compensation coefficient, represents the scattering angle offset caused by the temperature gradient, is the dual-wavelength spectral spacing, is the Mie scattering cross-section parameter, is the scattering angle, is the currently measured temperature value, represents the standard reference temperature.

[0019] The compensation coefficient is the core parameter of the temperature-scattering angle compensation matrix and is used to dynamically correct the deviation of the scattering signal caused by temperature changes; by calculating this coefficient in real time, the system can eliminate the non-linear influence of the temperature gradient on the Mie scattering model and ensure the accuracy of smoke concentration detection.

[0020] Functional relationship Characterize the coupling characteristics of temperature and optical parameters, calibrate the function curve through experiments, and establish the weight relationship between temperature gradient and wavelength spacing on the compensation effect; among them represents the scattering angle offset caused by the temperature gradient, reflecting the degree of perturbation of the temperature field distribution on the light scattering direction.

[0021] is the spectral spacing of the dual-wavelength light source, such as the wavelength difference between 650nm and 850nm, which determines the amplification effect of the differential signal on temperature sensitivity.

[0022] Integral term: The integration variable in represents the Mie scattering cross-section parameter, which describes the scattering efficiency of smoke particles on light of a specific wavelength.

[0023] is the scattering angle, that is, the angle between the incident light and the detector receiving direction.

[0024] represents the standard reference temperature, such as 25°C, corresponding to the reference state without temperature interference.

[0025] is the currently measured temperature value, reflecting the real-time environmental changes.

[0026] represents the rate of change of the scattering cross-section with the angle. The integration process is essentially a quantification of the cumulative deviation of the scattering characteristics during the temperature change from the reference value to the current value; through the integration operation, the system can capture the dynamic influence of the continuous temperature change on the scattering angle and avoid the step error of the traditional segmented compensation method.

[0027] is the dual-wavelength spectral spacing. The selection of the dual-wavelength light source needs to meet: Sensitivity maximization: The larger it is, the more significant the response of the differential signal to temperature changes. Anti-interference optimization: Avoid the peak band of ambient light intensity, such as the visible light region of the solar spectrum, to reduce background noise interference.

[0028] It is worth mentioning that this solution proposes a multi-modal fire warning system architecture, including a multi-spectral light source module, dual-wavelength beam emission, composite detection cavity, temperature gradient sensor + tunable photodetector, signal processing unit and hierarchical response module. The core innovation lies in the dual-wavelength phase-locked algorithm, which dynamically corrects the interference of temperature on scattering parameters through mathematical relationships.

[0029] The technical effects of the above solution include: It can improve the detection accuracy: Dual-wavelength differential detection can achieve an increase in the temperature drift suppression ratio and a reduction in the smoke concentration error; Capable of enhancing real-time response ability: The orthogonal distributed thermocouple group, combined with a 100Hz sampling rate, realizes the millisecond-level capture of temperature gradients. of the temperature gradient. Enhancing system reliability: By eliminating the non-linear coupling effect through the compensation matrix, it can achieve detection stability in the range of -20°C to 300°C.

[0030] The following technical problems exist in the traditional technical solutions: Spectral matching problem: The overlap between the conventional infrared light source (such as 940nm) and the absorption peak of smoke particles results in the attenuation of the scattered signal. Stray light interference: The wide divergence angle (>10°) leads to an increase in the reflection noise on the inner wall of the cavity and the deterioration of the signal-to-noise ratio. Dual-wavelength cooperation failure: When the wavelength spacing is insufficient, the sensitivity of the differential signal decreases, and it is impossible to effectively extract temperature-related features.

[0031] Based on this, the dual-wavelength light beam is generated by two independent semiconductor lasers, using a 650nm red laser diode and an 850nm near-infrared laser diode respectively. After the light beam is collimated by an aspherical lens group, it forms a coaxial incident optical path, and an anti-reflection coated optical window is installed at the entrance of the composite detection cavity. The light transmission band of the optical window covers the range of 600nm to 900nm, and the transmittance ≥ 95%.

[0032] It is worth mentioning that this solution defines the dual-wavelength light source parameters ( , ), specifying ≥ 150nm and the beam divergence angle .

[0033] The technical effects of the above solution include: Optimal wavelengths are selected: 650nm avoids common ambient light interference, and 850nm has strong smoke penetration ability. The combination covers the best scattering band for typical fire smoke particle sizes (0.1 - 10μm). The optical path design is optimized: The 5° - 8° divergence angle balances the detection sensitivity and the cavity size limitation, improving the measured light energy utilization rate. The anti-interference ability is improved: ≥ 150nm ensures the sensitivity of the dual-wavelength differential signal to temperature changes. Experiments show that the temperature compensation response speed is increased by 3 times.

[0034] The following technical problems exist in the traditional technical solutions: Insufficient spatial resolution: Traditional single-point temperature measurement cannot characterize the three-dimensional temperature gradient, resulting in the distortion of the input data of the compensation matrix. Thermal inertia delay: Large-spacing sensors over 10cm are difficult to capture local hot spots, causing temperature field reconstruction errors. Range coverage defect: The linearity of conventional sensors deteriorates when the temperature is >200°C, affecting the reliability of high-temperature fire field detection.

[0035] Based on this, the micro-thermocouples of the temperature gradient sensor array are manufactured using platinum resistance thin film technology and encapsulated in a high-temperature resistant ceramic substrate. Each thermocouple node is equipped with an independent data acquisition channel, the sampling frequency is not less than 100Hz, and the calculation time of the temperature gradient difference between any two adjacent thermocouples in the array is less than 10ms.

[0036] It is worth mentioning that in this solution, it is stipulated that the temperature sensor array arranges ≥3 groups of thermocouples along the XYZ axes respectively, the spacing is ≤5cm, the temperature measurement range is 20 - 300°C, and the resolution is 0.1°C.

[0037] The technical effects of the above solution include: Reconstructed the three-dimensional temperature field: The orthogonally distributed sensor group achieves a spatial resolution of 5cm³, and the temperature gradient calculation error <0.5°C / m; Can achieve fast response: The thermal response time of the platinum thin film thermocouple is ≤0.1s, meeting the millisecond-level temperature jump detection of deflagration fire sources; Ensured wide temperature range stability: The high-temperature resistant ceramic encapsulation enables the sensor to still maintain a measurement accuracy of ±0.5% in a 300°C environment.

[0038] The following technical problems exist in the traditional technical solution: Angle sampling defect: Fixed angle detectors (such as 90° side scatter) cannot capture the full-angle scatter characteristics, and the particle size resolution is low; Mechanical hysteresis problem: The angle repeatability of the traditional rotation mechanism is poor (>±2°), resulting in incomparable multiple measurement data; Spectral response mismatch: The sensitivity of silicon-based detectors drops sharply in the near-infrared band, missing the detection of ultra-fine particles (<0.3μm).

[0039] Based on this, each receiving unit of the photodetector array includes a band-pass filter and a pre-amplification circuit. The center wavelength of the band-pass filter matches the dual-wavelength light source respectively, the bandwidth is ±5nm, and the gain of the pre-amplification circuit is programmable and adjustable, and the dynamic range covers the optical power input from 0.1μW to 10mW.

[0040] It is worth mentioning that the above solution uses 16 InGaAs diodes for hemispherical distribution in the photodetector, and the stepping motor drives the rotation of 0° - 180°, with a positioning accuracy of ±0.5°.

[0041] The technical effects of the above solution include: Capable of achieving full-angle scattering analysis: 16 detectors cover a scattering angle of 10° - 170°, supporting the inversion of particle size distribution parameters; High-precision positioning: The stepper motor + encoder achieve a repeat positioning accuracy of ±0.5°, ensuring long-term measurement consistency; Spectral response optimization: The InGaAs material has a quantum efficiency of 85% at 850 nm, a 40% increase compared to silicon photodiodes.

[0042] The following technical problems exist in the traditional technical solutions: Existing solutions use a fixed temperature compensation coefficient or piecewise calibration at a single temperature point, and can only handle static or slowly changing temperature fields. For example, when there is a rapid temperature gradient at the location of the detector, the traditional method has a compensation lag due to insufficient sampling rate, and the measured scattering intensity error is more than 40%.

[0043] Existing technologies usually independently handle the influence of ambient temperature on smoke particles and the temperature drift of the detector itself, resulting in a fragmented compensation model. For example, the sensitivity of the detector decreases with increasing temperature, while the refractive index of smoke particles increases with increasing temperature. The temperature drift effects of the two are in opposite directions but are coupled with each other. Traditional linear superposition compensation will result in overcorrection or undercorrection.

[0044] Based on this, the method for constructing the temperature-scattering angle compensation matrix includes: Establish a three-dimensional look-up table , where The indexed temperature range is , the step size ; The indexed scattering angle , the step size ; The indexed dual-wavelength combination mode includes single mode, single mode, dual mode, three states; The compensation coefficients stored in the three-dimensional look-up table are obtained through experimental calibration, specifically satisfying: ; Among them, is the compensation coefficient, represents the temperature range index, represents the scattering angle index, represents the wavelength combination mode index, is the standard temperature, is the reference scattering intensity under is the measured scattering intensity, is the material temperature coefficient, is the current temperature, is the scattering angle, is the reference scattering angle.

[0045] Preferably, the signal processing unit integrates a digital lock-in amplification module. When synchronously demodulating the dual-wavelength scattering signal, an adaptive filter is used to eliminate ambient light interference, and the baseline drift is automatically calibrated within each demodulation period. The cut-off frequency of the adaptive filter is dynamically adjusted according to the temperature change rate.

[0046] Compensation coefficient The index meanings in include: represents the temperature range index. For example, the temperature range from 20°C to 300°C is divided into intervals of 5°C; represents the scattering angle index. For example, the scattering angle range from 10° to 170° is divided into intervals of 2°; represents the wavelength combination mode index (single wavelength or dual-wavelength combination).

[0047] The compensation coefficient design stores the optimal compensation values under various working conditions through a three-dimensional look-up table (LUT), enabling fast look-up and interpolation.

[0048] Reference intensity ratio : in represents the standard temperature and the reference scattering angle The reference scattering intensity under is obtained through laboratory calibration; represents the actually measured scattering intensity, corresponding to the current temperature and the scattering angle .

[0049] The above formula quantifies the deviation between the actual environment and the ideal conditions, directly reflecting the attenuation or enhancement effect of temperature-angle changes on the scattering signal.

[0050] Temperature correction term: in is the material temperature coefficient, characterizing the drift characteristics of the sensitivity of the photodetector with temperature change; is the difference between the current temperature and the reference temperature; the design of this formula is used to compensate for the error introduced by the temperature drift of the detector itself, essentially incorporating the hardware nonlinear characteristics into the compensation model.

[0051] The above-mentioned look-up table construction process includes: Conduct experimental calibration: Control the temperature to change from 20°C to 300°C in an incubator, and record the scattering intensities at different scattering angles (10° to 170°) and wavelength combinations at each temperature point; Conduct normalization processing: Taking , Calculate the intensity ratio at each position based on (typical lateral scattering angle); Perform temperature coefficient measurement: Change the temperature in a smokeless environment and measure the slope of the detector output with respect to temperature change, that is ; ; Perform interpolation optimization: Use the cubic spline interpolation algorithm to smooth the lookup table data and reduce the step error caused by discretization.

[0052] Compared with the traditional linear compensation model ; This formula captures the temperature-angle coupling effect through the reference intensity ratio, and combines the detector temperature drift correction to achieve double correction. For example, when high temperature causes the refractive index of smoke particles to increase, It is worth mentioning that: This solution pre-stores the full-parameter space compensation coefficients of temperature-angle-wavelength through a three-dimensional lookup table, covering a wide temperature range from -20°C to 300°C and the full angle range; Combined with the bilinear interpolation algorithm, it realizes real-time compensation at any temperature-angle point and avoids the step error of discrete compensation. Experiments show that in the scenario of sudden temperature change, such as the temperature jumps from 25°C to 150°C, the compensation delay is reduced from 500ms of the traditional method to less than 10ms.

[0053] This solution clearly separates the environmental temperature effect (specifically ) and the detector temperature drift effect (specifically ) in the compensation coefficient formula, and realizes double correction through physical modeling; for example, when high temperature causes the refractive index of smoke particles to increase; may increase abnormally, while term simultaneously corrects the decrease in detector sensitivity, and the two work together to ensure compensation accuracy.

[0054] In the scenario where both detector temperature drift and environmental temperature drift exist, this solution compresses the composite error from ±15% of the traditional method to ±1.5%.

[0055] The following technical problems exist in the traditional technical solution: The existing technology uses a low-pass filter with a fixed cut-off frequency and cannot effectively filter out the environmental light noise with overlapping frequency domains; The analog lock-in amplifier generates baseline drift affected by temperature and needs to be manually calibrated regularly. In high-temperature scenarios, the baseline fluctuation amplitude can reach 50mV, masking the true signal; Based on the above problems, the signal processing unit integrates a digital lock-in amplifier module and an adaptive filter. When synchronously demodulating the dual-wavelength scattering signal, the adaptive filter eliminates ambient light interference and automatically calibrates the baseline drift within each demodulation period. The cut-off frequency of the adaptive filter is dynamically adjusted according to the temperature change rate.

[0056] It is worth mentioning that the digital lock-in amplifier module provided in this embodiment adopts Dual-channel synchronous demodulation technology: Orthogonally demodulate the 650nm and 850nm dual-wavelength scattering signals to suppress incoherent noise; Adaptive filter design: Dynamically adjust the filter coefficients based on the LMS least mean square algorithm to eliminate ambient light interference; Baseline drift calibration: Within each demodulation period, automatically correct the amplifier bias voltage through zero-input sampling to eliminate the baseline shift caused by temperature drift.

[0057] In the above formula, the dynamic adjustment method adopts a dynamic frequency adjustment mechanism Dynamic association, and the calculation formula is: ; Wherein, is the basic cut-off frequency, taking , is the temperature sensitivity coefficient; Real-time monitor the data of the ambient temperature sensor. When the temperature change rate exceeds the threshold, automatically switch to the broadband filtering mode to capture fast signal fluctuations.

[0058] The technical effects of the above solution include: The adaptive filter dynamically adjusts the stopband range by real-time analyzing the noise spectrum, and when the smoke signal is weak in the initial stage of a fire, it boosts the SNR to above 35dB; The automatic calibration technology within the period suppresses the baseline drift within the range of ±0.1mV; combined with the temperature compensation algorithm, it maintains signal stability within the wide temperature range of -40°C to 85°C, and the concentration detection error ≤ ±0.5%.

[0059] For example, in existing fire protection systems, the alarm and ventilation control are independent subsystems, and the linkage delay is too high, which is likely to cause smoke diffusion. And traditional alarm information only contains a single concentration value and lacks spatial distribution data, making it impossible for firefighters to quickly locate the fire source; based on this, the hierarchical response module is communicatively connected to the building fire control system. When a medium-level alarm instruction is triggered, the air-conditioning return air valve is synchronously closed and the smoke exhaust fan is started. When a high-level alarm instruction is triggered, a structured alarm information packet including a temperature distribution heat map and a three-dimensional cloud map of smoke concentration is sent to the emergency management department.

[0060] It is worth mentioning that in this embodiment, a deep linkage mechanism between the hierarchical response module and the building fire protection system is constructed. The design of the linkage response mechanism can control the smoke diffusion range within 50 m³ and increase the smoke exhaust efficiency to 98 m³ / s, meeting the requirements of NFPA 92 standard for high-risk areas.

[0061] For example, dust accumulation on traditional optical windows easily leads to signal attenuation problems. After ordinary coated detectors operate in a dusty environment for 3 months, the light transmittance drops by 60%, and the false alarm rate surges to 15%. In high-cleanliness places such as chip workshops, special dust-free room maintenance is required, and the single cost exceeds 5,000 yuan, which is expensive. Based on this: The inner wall of the composite detection cavity is coated with a nano-scale dust-repellent coating, and a self-cleaning air curtain device is installed at the receiving end of the photodetector array. The air curtain device generates a directional air flow through a micro piezoelectric pump to continuously blow the dust accumulated on the surface of the optical elements.

[0062] It is worth mentioning that this solution proposes an anti-pollution design for the composite detection cavity.

[0063] Nano dust-repellent coating technology: Coating material: Titanium dioxide / polytetrafluoroethylene nanocomposite (particle size 50 nm), contact angle ≥ 150°; Coating process: Plasma-enhanced chemical vapor deposition (PECVD), thickness 2 μm ± 0.1 μm.

[0064] Self-cleaning air curtain device: Air flow parameters: Speed 10 m / s, coverage angle 60°, flow rate 0.5 L / min; Pulse frequency 5 Hz (to avoid resonance damage to optical elements); Piezoelectric pump control: Automatically activated according to the feedback of the dust sensor (PM2.5 ≥ 75 μg / m³), power consumption ≤ 2 W.

[0065] The following technical problems exist in the traditional technical solution: Isolated data analysis: Traditional methods independently process the data of each sensor and cannot capture multi-modal correlation features; Insufficient model generalization: Fixed thresholds are difficult to adapt to different environments. For example, the requirements for thresholds in factories and residences are different, which will lead to poor adaptability; Low computational efficiency: Complex algorithms have high operation delays in embedded devices and cannot meet the real-time requirements.

[0066] Based on this, please refer to Figure 2 , this application provides a multi-level linkage control method, which is applied to the fire warning system described in any one of the above, including: Real-time acquisition of multi-modal sensing data, including scattered light intensity , temperature field distribution , concentration value; Fusing multi-source data through a particle filter algorithm to estimate the probability of a fire ; When , activate the deep learning model for pattern recognition; dynamically adjust the alarm threshold according to the recognition result.

[0067] It is worth mentioning that the present invention defines a fire probability assessment method based on multi-source data fusion, which can suppress 60% of the temperature-scattering correlation noise by the particle filter algorithm and reduce the estimation error; the recognition accuracy of the CNN model for the flame flicker frequency feature reaches 99.2%, which is better than the traditional spectrum analysis method; through dynamic threshold adjustment, the number of false triggers is reduced by 75% in a high-interference environment.

[0068] For example, the following technical problems exist in the traditional technical solutions: Model mismatch problem: The traditional linear filtering assumption does not conform to the non-linear characteristics of smoke diffusion; Parameter curing defect: The preset value of the noise statistical characteristics cannot adapt to the changing environment; High computational complexity: Solving the complete physical model requires supercomputer resources and is difficult to be embedded.

[0069] Based on this, the state equation in the particle filter algorithm is expressed as: ; The observation equation is expressed as: ; Among them, the state vector includes smoke concentration, temperature gradient, and gas diffusion rate, and the observation matrix corresponds to the multi-sensor measurement model, , are the process noise and the observation noise respectively, is the state matrix, and the observation vector represents the actual sensor measurement value.

[0070] In the above formula, the state equation: In it, the state vector includes smoke concentration ; temperature gradient , key fire parameters such as the gas diffusion rate v; the state matrix describes the dynamic relationship of parameter evolution over time. For example, smoke diffusion follows the convection-diffusion equation, and the temperature gradient is affected by heat conduction; the process noise : Represents system disturbances not accounted for in the modeling, such as sudden airflows and sensor noise, and is assumed to follow a Gaussian distribution.

[0071] Observation equation: , the observation vector represents the actual sensor measurement values (such as the output of a photodetector, the reading of a thermocouple, and the data of a gas concentration meter); the observation matrix : Represents the mapping relationship between the state parameters and the sensor measurement values (such as the Mie model relationship between the scattered light intensity and the smoke concentration); the observation noise represents the sensor measurement error and is also assumed to be Gaussian white noise.

[0072] The specific modeling of the state equation in the above formula includes: Smoke concentration evolution: ; where is the diffusion rate, is the concentration gradient, is the process noise term.

[0073] Temperature gradient propagation: ; where is the thermal diffusion coefficient, is the Laplacian operator of the temperature field, reflecting the heat conduction effect.

[0074] Diffusion rate update: ; is the acceleration, driven by changes in the ambient air flow.

[0075] Construction of the observation matrix: Optical scattering model: Output of the photodetector and the smoke concentration The relationship is: ; Corresponding to in the observation matrix; Temperature field mapping: Thermocouple reading where is the thermocouple position offset.

[0076] Gas diffusion correlation: CO concentration meter measurement value and the diffusion rate The relationship is: ; is the spatial volume ratio parameter.

[0077] Noise covariance processing: Process noise : Determine the covariance matrix through historical data statistics , and increase the value in high-temperature and high-humidity environments to reflect the enhanced environmental disturbance.

[0078] Observation noise : Covariance matrix Dynamically adjusted according to sensor accuracy. For example, the noise level of a photodetector increases with the increase in temperature.

[0079] Particle filter algorithm process: Initialization stage: Generate N particles, and each particle contains a state vector and weight ; Prediction: Update the particle state according to the state equation and add process noise; Weighting: Calculate the observation likelihood probability of each particle and update the weight; Resampling: Regenerate the particle set according to the weight distribution to avoid the degeneracy problem; State estimation: Obtain the final output by weighted averaging of the particle states .

[0080] The technical effects of the above solution include: Improve the non-linear modeling ability: Describe the interaction of smoke-temperature-airflow through the coupling term of the state vector X, and the model goodness of fit R²≥0.97; Can suppress adaptive noise: The online update strategy of the covariance matrix improves the filtering stability by 40% in a dynamic environment; Breakthrough in operation efficiency: Adopt sparse matrix optimization technology, and the algorithm runs at a 10ms cycle on an ARM Cortex-M7 processor.

[0081] Finally, it should be noted that the above content is only used to illustrate the technical solution of the present invention, rather than limiting the protection scope of the present invention. Any simple modification or equivalent replacement of the technical solution of the present invention by those of ordinary skill in the art shall not depart from the essence and scope of the technical solution of the present invention.

Claims

1. A fire extinguishing system based on multi-modal sensors, characterized in that, Including: Multispectral light source module, the multispectral light source module is configured to emit a dual-wavelength beam with a predetermined divergence angle, where the first wavelength is in the visible light band, and the second wavelength is in the near-infrared band; a composite detection cavity, the composite detection cavity is internally provided with a temperature gradient sensor array and a scattering angle adjustable photodetector array, the temperature gradient sensor array includes a set of micro-thermocouples distributed orthogonally in space, and the photodetector array includes a multi-angle receiving unit that can slide along a spherical trajectory; a signal processing unit, the signal processing unit is connected to the output end of the photodetector array and internally provided with a temperature-scattering angle compensation matrix, and the matrix eliminates the coupling interference of the temperature field on the scattering characteristics of smoke particles through a dual-wavelength phase-locked algorithm; A hierarchical response module, which generates multi-level alarm instructions based on the smoke concentration value compensated by the temperature-scattering angle compensation matrix and the temperature gradient change rate; wherein, the dual-wavelength lock-in algorithm realizes dynamic compensation through the following mathematical relationship: ； Wherein, is a compensation coefficient, and the functional relationship characterizes the coupling characteristics of temperature and optical parameters, represents the scattering angle offset caused by the temperature gradient, is the dual-wavelength spectral spacing, is the Mie scattering cross-section parameter, is the scattering angle, is the currently measured temperature value, represents the standard reference temperature.

2. The fire extinguishing system based on multi-modal sensors according to claim 1, characterized in that: The dual-wavelength beam is generated by two independent semiconductor lasers, using a 650nm red laser diode and an 850nm near-infrared laser diode respectively. After the beam is collimated by an aspherical lens group, a coaxial incident optical path is formed, and an anti-reflection coated optical window is installed at the entrance of the composite detection cavity. The light transmission band of the optical window covers the range of 600nm to 900nm, and the transmittance ≥ 95%.

3. The fire extinguishing system based on a multi-modal sensor according to claim 1, characterized in that: The micro-thermocouples of the temperature gradient sensor array are manufactured by platinum resistance thin film technology and encapsulated in a high-temperature resistant ceramic substrate. Each thermocouple node is equipped with an independent data acquisition channel, the sampling frequency is not less than 100Hz, and the calculation time of the temperature gradient difference between any two adjacent thermocouples in the array is less than 10ms.

4. The fire extinguishing system based on a multi-modal sensor according to claim 1, wherein: Each receiving unit of the photodetector array includes a band-pass filter and a pre-amplifier circuit. The center wavelength of the band-pass filter matches the dual-wavelength light source respectively, and the bandwidth is ±5nm. The gain of the pre-amplifier circuit is programmable, and the dynamic range covers the optical power input from 0.1μW to 10mW.

5. The fire extinguishing system based on a multi-modal sensor according to claim 1, wherein: The construction method of the temperature-scattering angle compensation matrix includes: Build a three-dimensional lookup table , where the index temperature range is , with a step size ; Index scattering angle , step size ; Index dual-wavelength combination mode, including Single mode, Single mode, Dual mode, with three states; The compensation coefficients stored in the three-dimensional look-up table are obtained through experimental calibration, specifically satisfying: ； Among them, is the compensation coefficient, represents the temperature range index, represents the scattering angle index, represents the wavelength combination mode index, is the standard temperature, the reference scattering intensity under it, is the measured scattering intensity, is the material temperature coefficient, is the current temperature, is the scattering angle, is the reference scattering angle.

6. The fire extinguishing system based on multi-modal sensors according to claim 1, characterized in that: The signal processing unit integrates a digital lock-in amplifier module and an adaptive filter. When synchronously demodulating the dual-wavelength scattering signal, the adaptive filter eliminates ambient light interference and automatically calibrates the baseline drift within each demodulation period. The cut-off frequency of the adaptive filter is dynamically adjusted according to the temperature change rate.

7. The fire extinguishing system based on a multi-modal sensor according to claim 1, wherein: The hierarchical response module is communicatively connected to the building fire control system. When the intermediate alarm instruction is triggered, the air-conditioning return air valve is synchronously closed and the smoke exhaust fan is started. When the high-level alarm instruction is triggered, a structured alarm information packet including a temperature distribution thermal map and a three-dimensional cloud map of smoke concentration is sent to the emergency management department.

8. The fire extinguishing system based on a multi-modal sensor according to claim 1, characterized in that: The inner wall of the composite detection cavity is coated with a nano-scale dust-proof coating, and a self-cleaning air curtain device is installed at the receiving end of the photodetector array. The air curtain device generates a directional air flow through a micro piezoelectric pump to continuously blow the dust on the surface of the optical element.

9. A multi - level linkage control method is applied to a fire extinguishing system based on multi - modal sensors as described in any one of claims 1 - 8, characterized in that, Including: Obtain multi-modal sensing data in real time, including scattered light intensity , temperature field distribution , concentration value; Estimate the probability of fire occurrence by fusing multi-source data through the particle filter algorithm ; When activate the deep learning model for pattern recognition; dynamically adjust the alarm threshold according to the recognition result.

10. The multi - level linkage control method according to claim 9, wherein, The state equation in the particle filter algorithm is expressed as: ; The observation equation is expressed as: ; Among them, the state vector includes smoke concentration, temperature gradient, and gas diffusion rate. The observation matrix corresponds to the multi-sensor measurement model , are the process noise and the observation noise respectively is the state matrix, and the observation vector represents the actual sensor measurement value

Images