Smoke methane composite detection method and device

Abstract

The invention discloses a smog methane composite detection method and device, relates to the technical field of smog methane detection, and solves the technical problems of weak anti-interference capability in a data processing link and lack of a multi-parameter compensation and calibration system. The effective acquisition range is dynamically calculated in combination with a factory calibration interval, the signal amplification factor is adaptively adjusted by using a digital potentiometer, the electric signal strength is ensured to be stabilized in the optimal interval, the influence of laser loss and environment temperature on basic data is eliminated from the source, and the filtering coefficient is dynamically adjusted in combination with the signal-to-noise ratio by using a layered filtering strategy. Random noise is filtered out, and key features of an absorption peak are reserved; through segmented weighted polynomial fitting, accurate extraction of core parameters of the absorption peak is realized; and a unified multi-round outlier elimination mechanism eliminates the interference of abnormal data on the methane concentration, the absorption peak position and the smoke sensing value, and improves the credibility of the data.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of smoke methane detection, in particular to a smoke methane composite detection method and device. BACKGROUND

[0002] In the existing household kitchen fire safety monitoring scene, users usually need to install smoke fire detection alarm and household combustible gas detector respectively to meet the basic protection needs, but this independent setting mode has significant defects: not only the installation operation is cumbersome, the later maintenance cost is high, it is difficult to realize unified and regular self-checking, but also the multiple devices are scattered and arranged, which destroys the overall appearance of the kitchen space; more importantly, the functions of the two types of detectors are independent and cannot be associated, making it difficult to form a collaborative monitoring mechanism and easily causing a safety warning blind area.

[0003] CN109949534A proposes a smoke detector based on double light path scattering, which realizes the effects of reducing false positives and mutual calibration through redundancy design, but this scheme increases system cost and complexity. CN120801186A proposes a scheme and implementation method based on laser methane detection, which describes each functional module, but the data processing part is not sufficient. CN120446050A mentions part of the data processing method of laser methane detection, but the processing of abnormal data is too simple and may not achieve the desired effect. At present, there is no clear and feasible high-quality solution for infrared smoke and laser methane dual-light-source composite detection, making it difficult to achieve the synergistic effect of "1+1>2". SUMMARY

[0004] In view of the deficiencies of the prior art, the present application provides a smoke methane composite detection method and device, which solves the problems of weak anti-interference ability in data processing link and lack of multi-parameter compensation and calibration system.

[0005] To achieve the above purpose, the present application realizes the following technical scheme: a smoke methane composite detection method, which specifically includes the following steps: Step one, a laser detector is used to receive the reflection signal of the target detection area and complete data acquisition, extract the maximum value of the electric signal intensity parameter in the collected data and the preset best data acquisition interval, and record the initial laser collection data under the condition that the former meets the latter; Step two, the initial laser collection data is filtered and denoised by using the exponential weighted moving average algorithm to obtain filtered data, and the filtered data is weighted and fitted by using the least square method polynomial fitting algorithm to obtain fitted data; Step three, correct the post-fitting data drift with a slow EWMA baseline vector, extract the peak height, peak area, peak width, optical depth four characteristics, together with temperature, humidity input the 3-layer minimalist neural network cross term model which has been offline trained, output the final methane concentration after aging-pollution self-checking; Step four, use the infrared receiving tube to receive the infrared scattering signal of the target detection area in real time and complete data acquisition, and perform random filtering and environmental temperature compensation on the smoke sensing effective sampling value data, and the processing steps of the methane concentration data are the same, to obtain the final smoke sensing detection value; Step five, preset the methane concentration alarm threshold, if the methane concentration exceeds the threshold and triggers the methane alarm for multiple times, dynamically adjust the smoke sensing threshold according to the laser intensity attenuation rate, if the smoke sensing detection value exceeds the threshold and triggers the smoke sensing alarm for multiple times.

[0006] As a further scheme of the present application, the acquisition method of the preset optimal data acquisition interval is: Read the factory calibration interval [L0 min , L0 max ] from EEPROM, read the real-time temperature t and the cumulative working hours h of the laser, correct according to the aging-temperature drift model, = , h is the cumulative working time of the laser, t is the real-time temperature value, is the aging coefficient, the value is 0.3% / kh, is the temperature coefficient, the value is 0.15% / ℃, is the relative attenuation ratio, further according to the formula L min =L0 min ·(1−), L max =L0 max ·(1−), the acquisition interval [L min , L max ] is calculated.

[0007] As a further scheme of the present application, the specific way to obtain the filtered data is: According to the formula =median(S0[n-1],S0[n],S0[n+1]) filtering processing is carried out, the current input sample after median pre-filtering is obtained, wherein n is the sampling times, then according to the exponential weighted moving average algorithm formula y S [n]=a s × +(1-a s )×y S [n-1], wherein a is a fixed filtering coefficient, the value is 0.05, y S [n] represents the slow channel output value after the n th sampling, y S[n-1] represents the last time base estimate value, and the signal-to-noise ratio in the window is calculated, and the calculation formula is , if SNR> 20dB, d f = 0.3, if SNR> 10dB, d f = 0.15, and the formula y f [n] = a f × + (1-a f ) × y f [n-1] is used to calculate the fast channel output of the nth point, and the filtered data is obtained by comprehensively considering the fast channel output and the slow channel output, and the specific S1[n] = y f [n] - y s [n] is obtained.

[0008] As a further scheme of the application, the specific way to obtain the fitted data is: The 512-point vector S1 is equally divided into four segments to obtain A segment [0, 63], B segment [64, 191], C segment [192, 319], and D segment [320, 511], and the SNR of each segment is calculated, according to the formula SNR sub = RMS (S1 sub ) / , wherein is the noise standard deviation, and = , SNR sub = is the signal-to-noise ratio of the current sub-interval, RMS (S1 sub ) represents the root mean square value of 128-point data in the sub-interval, which represents the effective amplitude of the signal, then the maximum value index i peak is searched in the B segment and the C segment, and a 64-point Hann window is generated with i peak as the center, and the sharpness S = |S1[i peak ] - 0.5· (S1[i peak - 4] + S1[i peak + 4]) | / RMS local is calculated, wherein RMS local represents the noise effective value of the same local interval, S1[i peak ] represents the intensity of the current peak, and S1[i peak + 4] represents the intensity of each of the four sampling points on the left and right of the peak top. If S<0.5, it represents a blunt peak, and is not enhanced, if S≥0.5, 64 points are selected as the reference width w base = 64, and the boundary safety half-width is calculated, half = min(i peak -B start , B end -ipeak , w base / 2), then the Hann window with area = 1 is generated, that is, H[i]=0.5-0.5*cos(2pi*k / w), k=0...w-1, and normalization processing is performed H norm [i]=H[i] / , and finally the peak area weight superposition W peak [i]=W0[i]+1.2*S*H norm [i], i∈[i peak -half, i peak +half-1].

[0009] As a further scheme of the application, the specific way of correcting the drift of the fitted data with a slow EWMA baseline vector is as follows: The value of the slow EWMA channel y s [n] in the same scanning period is used to establish the background vector B line , and the corrected absorption curve S2 corr =S2-B line , and four characteristics of peak height Q, peak area A, peak width W and optical depth tau are extracted at the same time, wherein the peak height Q=max(S2 corr ), the peak area A=∑(S2 corr >0.2*Q), the peak width W=right intersection point-left intersection point, and the depth tau=-ln(1-Q / I 0） , wherein I0 represents the real-time zero gas light intensity.

[0010] As a further scheme of the application, the way of obtaining the final methane concentration is as follows: The peak height, peak area, peak width, optical depth, temperature t and humidity RH are input into the cross-term model of the minimalist neural network which has been trained offline, the relative humidity RH is additionally input into the model, and D comp =D raw *(1+k1*(t-25)+k2*(RH-50)) is calculated by the model, wherein k1 and k2 are regression coefficients, to obtain the original concentration D raw , D comp represents the compensated concentration, I0 is updated in real time by using the saturated pulse in the period, the aging-pollution self-checking is performed on D raw , D=round(D comp *100) / 100, and the final concentration D is output.

[0011] As a further aspect of the present invention, v sets of methane concentration and absorption peak position data {(D1, P1), (D2, P2), ..., (Dv, Pv)} are continuously collected v times. For these v sets of data, outlier values ​​for concentration D and absorption peak position P are removed to eliminate interference. The mean M of the v sets of concentration data is then calculated. D With standard deviation S D For each group of concentration data Di, calculate the deviation coefficient |D i -M D | / S D A preset deviation coefficient threshold K is set. Data with a deviation coefficient > K are identified as outliers and the corresponding concentration D and absorption peak position P data are removed as a whole.

[0012] As a further aspect of the present invention, the final smoke detection value is obtained as follows: An infrared receiver tube is used to receive the infrared scattering signal of the target detection area in real time and complete the data acquisition. For the raw data of a single acquisition, random noise interference is filtered out by an average filtering algorithm to obtain the intermediate sampling value H. Then, the intermediate sampling value H is temperature compensated in combination with the ambient temperature parameter to correct the detection deviation caused by temperature. Finally, the effective sampling value data G of the smoke detector in a single acquisition is obtained. Repeat the above steps to continuously collect m smoke sensor samples, obtaining a dataset {G1, G2, ..., Gm}. Process this dataset using the same outlier removal method as the methane concentration data, first calculating the mean M of the m data sets. G With standard deviation S G Then calculate the sample value G for each group. i deviation coefficient |G i -M G | / S G Outliers with a deviation coefficient greater than a preset threshold are removed; the outlier removal operation is repeated for the remaining data until the amount of remaining data is ≤m / 3 or no new outliers are generated; finally, the mean of the valid data after outlier removal is taken to obtain the final smoke detection value G. S .

[0013] As a further aspect of the present invention, the specific method for triggering the smoke alarm is as follows: Based on methane concentration detection value D S With smoke detection value G S It achieves accurate alarms by combining dynamic threshold adjustment strategies and sends notifications through multiple channels, with a preset methane concentration alarm threshold T. D It can be tiered according to application scenarios and compares D in real time. S With T D If D S >T D If this state is maintained for a preset number of consecutive times, a methane concentration alarm will be triggered immediately. preset smoke sensing initial alarm threshold T G , first call the cache module of laser intensity C history monitoring data, analysis of the fluctuation amplitude of C in the preset time window, according to the attenuation of C according to the preset rules dynamic adjustment of detection limit, such as C attenuation 20% then T G Down 15%), get dynamic adjustment alarm threshold T G ; Then determine whether the smoke detection value G S Exceeds T G : if G S ≤T G , it is determined that there is no smoke signal; if G S >T G , and the state is maintained for many times, the smoke alarm is triggered.

[0014] The smoke methane composite detection device comprises a detection light path and a signal acquisition component, a laser driving and temperature control component, a signal calculation processing component, an acousto-optic linkage alarm component and a wireless networking communication component. The detection light path and the signal acquisition component comprise a smoke infrared emitter tube, an infrared receiver tube, a laser emitter and a laser receiver. The infrared emitter tube emits an infrared light beam, which is scattered by smoke particles in the light path and then received by the infrared receiver tube and converted into an electrical signal. The laser emitter emits a specific wavelength of light signal, which is received by the laser receiver and converted into an electrical signal. The laser driving and temperature control component uses PWM adjustment combined with precise PID control. The signal calculation processing component processes the collected electrical signals and outputs the indication results of the smoke state, the flammable gas concentration state and the alarm state through fusion algorithm. The acousto-optic linkage alarm component synchronously sends out clear sound and light signals when the device detects smoke alarm or gas leakage alarm. The wireless networking communication component reports heartbeat data and alarm data in real time, and platform monitoring personnel can view and process the data in real time.

[0015] The present application provides a smoke methane composite detection method and device. Compared with the prior art, the following beneficial effects are achieved: The present application realizes smoke detection and methane leakage monitoring simultaneously through two-in-one function integration design, reduces installation difficulty and improves kitchen space aesthetics, and realizes flexible and accurate alarm response through laser methane light intensity inversion optimization of infrared scattered light intensity, dynamic threshold parameter adjustment, and the like.

[0016] The present application establishes an aging-temperature drift correction model, dynamically calculates effective collection range in combination with factory calibration interval, and uses a digital potentiometer to adaptively adjust signal amplification multiple, so as to ensure that the electric signal intensity is stable in an optimal interval, eliminate the influence of laser loss and environmental temperature on basic data from the source, lay a precise data foundation for subsequent concentration detection, adopt a hierarchical filtering strategy, dynamically adjust the filtering coefficient in combination with the signal-to-noise ratio, filter out random noise and retain the key characteristics of the absorption peak, realize accurate extraction of the core parameters of the absorption peak through piecewise weighted polynomial fitting, and effectively eliminate the interference of abnormal data on methane concentration, absorption peak position and smoke value through a unified multi-round outlier elimination mechanism, thereby improving the reliability of the data.

[0017] The present application establishes a multi-factor concentration compensation model, dynamically updates zero gas light intensity in combination with periodic saturation pulse and completes aging-pollution self-checking, greatly offsets the influence of device loss, environmental temperature and humidity on methane concentration detection, makes the error of the final output concentration value controllable, adapts to long-term stable detection under complex working conditions, dynamically adjusts the smoke alarm threshold according to the fluctuation range of laser intensity, avoids false positives and false negatives caused by light path attenuation, realizes linkage determination of methane and smoke detection in the case of light path shielding failure, prevents false alarm removal due to smoke accumulation and light path shielding, and realizes differential sound-light alarm + multi-channel notification mechanism, which is convenient for on-site personnel to quickly distinguish alarm types, ensures timely access to alarm information, and reduces safety hazards. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 The present application establishes an aging-temperature drift correction model, dynamically calculates effective collection range in combination with factory calibration interval, and uses a digital potentiometer to adaptively adjust signal amplification multiple, so as to ensure that the electric signal intensity is stable in an optimal interval, eliminate the influence of laser loss and environmental temperature on basic data from the source, lay a precise data foundation for subsequent concentration detection, adopt a hierarchical filtering strategy, dynamically adjust the filtering coefficient in combination with the signal-to-noise ratio, filter out random noise and retain the key characteristics of the absorption peak, realize accurate extraction of the core parameters of the absorption peak through piecewise weighted polynomial fitting, and effectively eliminate the interference of abnormal data on methane concentration, absorption peak position and smoke value through a unified multi-round outlier elimination mechanism, thereby improving the reliability of the data. Figure 2 The present application establishes an aging-temperature drift correction model, dynamically calculates effective collection range in combination with factory calibration interval, and uses a digital potentiometer to adaptively adjust signal amplification multiple, so as to ensure that the electric signal intensity is stable in an optimal interval, eliminate the influence of laser loss and environmental temperature on basic data from the source, lay a precise data foundation for subsequent concentration detection, adopt a hierarchical filtering strategy, dynamically adjust the filtering coefficient in combination with the signal-to-noise ratio, filter out random noise and retain the key characteristics of the absorption peak, realize accurate extraction of the core parameters of the absorption peak through piecewise weighted polynomial fitting, and effectively eliminate the interference of abnormal data on methane concentration, absorption peak position and smoke value through a unified multi-round outlier elimination mechanism, thereby improving the reliability of the data. DETAILED DESCRIPTION

[0019] With reference to the drawings of the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of the present application.

[0020] First embodiment Please refer to Figure 1 The present application provides a smoke methane composite detection method, which specifically comprises the following steps: Step one, a laser detector is used to receive the reflection signal of a target detection area and complete data acquisition. Specifically, a laser emits a 512-point triangular wave, the sampling rate fs=20kS / s, the original vector W[0…511] is obtained, then the first-order difference dW[i]=W[i+1]-W[i] is calculated, i is the number of sampling points, a continuous 128-point region is searched, and the maximum value is found, and the maximum value represents the most significant segment of the absorption feature, and the starting point idx is recorded. A linear background fitting is performed on the segment to obtain a background straight line B[i], and the background is deducted and averaged to obtain the electrical signal intensity L= , wherein W[i] is the value of the i-th sampling point in the original sampling waveform, B[i] is the value of the linear background fitting straight line at the i-th point on the segment, W[i]-B[i] represents the net signal after deducting the background, and idx represents the starting index of the 128-point interval with the most significant absorption feature found from the 512-point waveform. Read the factory calibration interval [L0 min , L0 max ] from the EEPROM, read the real-time temperature t and the cumulative working hours h of the laser, correct according to the aging-temperature drift model, = , h is the cumulative working time of the laser, t is the real-time temperature value, is the aging coefficient, and the value is 0.3% / kh, is the temperature coefficient, and the value is 0.15% / ℃, is the relative attenuation ratio, and further according to the formula L min =L0 min ·(1−), L max =L0 max ·(1−), the collection interval [L min , L max ] is calculated, and the current electrical signal intensity L is interval determined. If L does not belong to [L min , L max ], the signal amplification multiple F is adjusted through a digital potentiometer, and the electrical signal intensity L is adaptively adjusted. ​After adjustment, the laser is re-emitted to emit a triangular wave detection signal, and the collection, determination and adjustment process is repeated until the electric signal intensity L falls into the interval [L min , L max ], according to the adjusted stable electric signal intensity L and the amplification factor F, the current device laser intensity C is calculated according to the formula C=L / F, and C is stored in the cache module and real-time monitoring of its mutation state, when the electric signal intensity meets the interval requirement, the initial laser collection data S0 is recorded synchronously.

[0021] Step two, the initial laser collection data S0 is filtered and denoised by using the exponential weighted moving average algorithm to obtain the filtered data S1, and the specific processing method is as follows: According to the formula =median(S0[n-1],S0[n],S0[n+1]) filtering processing is carried out, and the current input sample after median pre-filtering is obtained, wherein n is the sampling number, then according to the formula of exponential weighted moving average algorithm y S [n]=a s × +(1-a s )×y S [n-1],wherein a is a fixed filter coefficient, and the value is 0.05, y S [n] represents the slow channel output value after the n th sampling, y S [n-1] represents the last basal estimate value, and the signal-to-noise ratio in the window is calculated, and the calculation formula is , if SNR>20dB, d f value is 0.3, if SNR>10dB, d f value is 0.15, and the formula y f [n]=a f × +(1-a f )×y f [n-1] is combined to calculate the fast channel output of the n th point, and the filtered data is obtained by comprehensively considering the fast channel output and the slow channel output, and specifically S1[n]=y f [n]-y s [n].

[0022] Step three, the filtered data S1 is weighted and fitted by using the least square method polynomial fitting algorithm to obtain the fitted data S2, and a segmented weighting strategy is adopted in the fitting process, and the specific processing method is as follows: The 512-point vector S1 is equally divided into four segments to obtain A segment [0, 63], B segment [64, 191], C segment [192, 319] and D segment [320, 511], and the SNR of each segment is calculated, according to the formula SNRsub = RMS(S1 sub ) where is the standard deviation of the noise, and = RMS(S1 ) sub is the signal-to-noise ratio of the current sub-interval, RMS(S1 sub ) represents the root mean square value of 128 points of data in the sub-interval, representing the effective amplitude of the signal, then search for the maximum value index i in section B and section C peak , and generate a 64-point Hann window centered on i peak , while calculating the sharpness S = |S1[i peak ] - 0.5 · (S1[i peak - 4] + S1[i peak + 4]) | / RMS local , where RMS local represents the effective value of the noise in the same local interval, S1[i peak ] represents the intensity of the current peak, and S1[i peak + 4] represents the intensity of the 4 sampling points on the left and right of the peak; If S < 0.5, it represents a blunt peak, then it is not enhanced, if S ≥ 0.5, select 64 points as the reference width w base = 64, and calculate the boundary safety half-width, half = min(i peak - B start , B end - i peak , w base / 2), then generate a Hann window with an area of 1, then H[i] = 0.5 - 0.5 · cos(2π·k / w), k = 0…w-1, and perform normalization processing H norm [i] = H[i] / sum(H) , and finally perform peak area weight superposition W peak [i] = W0[i] + 1.2 · S · H norm [i], i ∈ [i peak - half, i peak + half - 1].

[0023] Step four, drift correction of the fitting data S2 using the slow EWMA baseline vector, to obtain S2 corr , specifically using the value of the slow EWMA channel y s [n] in the same scan period to establish the background vector B line , the corrected absorption curve S2 corr = S2 - B lineSimultaneously, four features are extracted: peak height Q, peak area A, peak width W, and optical depth τ. Among them, peak height Q = max(S2) corr Peak area A = Σ(S2) corr >0.2·Q), peak width W = right intersection point – left intersection point, depth τ = –ln(1 – Q / I) 0） Where I0 represents real-time zero atmospheric light intensity, the above four features, along with temperature t and humidity RH, are input into a minimal neural network cross-term model that has been trained offline. Specifically, a three-layer minimal network is used, including five nodes for the four input features + temperature t; six hidden nodes; and one output node. Relative humidity RH is then added to the model input, and the model calculates D. comp =D raw ·(1+k1·(t–25)+k2·(RH–50)), where k1 and k2 are regression coefficients, to obtain the original concentration D. raw D comp This represents the compensated concentration, with I0 updated in real-time using saturation pulses within the period, for D. raw Perform aging-contamination self-check, D=round(D comp ·100) / 100, output the final concentration D.

[0024] V sets of methane concentration and absorption peak position data {(D1, P1), (D2, P2), ..., (Dv, Pv)} were obtained by continuously collecting data v times. Outliers in the concentration D and absorption peak position P of these v sets of data were removed to eliminate interference. The mean M of the concentration data for the v sets was calculated. D With standard deviation S D For each group of concentration data Di, calculate the deviation coefficient |D i -M D | / S D A preset deviation coefficient threshold K is set, and data with a deviation coefficient > K are identified as outliers and the corresponding concentration D and absorption peak position P data of that group are removed as a whole. Repeat the calculation of mean, standard deviation, and coefficient of deviation, and outlier removal for the remaining data until the remaining data volume is ≤v / 3 or no new outliers are generated. After removing outliers from the concentration data, use the same method to process the data from concentration data D. i Replace with absorption peak position data P i The remaining data were used to screen for outliers in the absorption peak positions, and the concentration D and absorption peak position P data of the outlier groups were removed simultaneously. The average of the concentration data after removing outliers was then taken to obtain the final methane concentration detection value D. S The mean value of the peak position data after removing outliers is used to obtain the final absorption peak position P. S .

[0025] Step 5: Use an infrared receiver tube to receive the infrared scattering signal of the target detection area in real time and complete the data acquisition. For the raw data of a single acquisition, use an average filtering algorithm to filter out random noise interference and obtain the intermediate sampling value H. Then, combine the ambient temperature parameter to perform temperature compensation on the intermediate sampling value H to correct the detection deviation caused by temperature. Finally, obtain the effective smoke sensor sampling value data G of a single acquisition. Repeat the above steps to continuously collect m smoke sensor samples, obtaining a dataset {G1, G2, ..., Gm}. Process this dataset using the same outlier removal method as the methane concentration data, first calculating the mean M of the m data sets. G With standard deviation S G Then calculate the sample value G for each group. i deviation coefficient |G i -M G | / S G Outliers with a deviation coefficient greater than a preset threshold are removed; the outlier removal operation is repeated for the remaining data until the amount of remaining data is ≤m / 3 or no new outliers are generated; finally, the mean of the valid data after outlier removal is taken to obtain the final smoke detection value G. S .

[0026] Step 6: Based on the methane concentration detection value D S With smoke detection value G S It achieves accurate alarms by combining dynamic threshold adjustment strategies and sends notifications through multiple channels, with a preset methane concentration alarm threshold T. D It can be tiered according to application scenarios and compares D in real time. S With T D If D S >T D If this state is maintained for a preset number of consecutive times, a methane concentration alarm will be triggered immediately. Preset initial alarm threshold T for smoke detector G First, historical monitoring data of laser intensity C in the cache module is retrieved, and the fluctuation range of C within a preset time window is analyzed. Based on the attenuation of C, the detection lower limit is dynamically adjusted according to preset rules. For example, if C attenuates by 20%, then T... G (Reduced by 15%), resulting in a dynamically adjusted alarm threshold T. G Subsequently, the smoke detection value G was determined. S Does it exceed T? G If G S ≤T G If G is determined to be a smoke-free signal; S >T G If this state is maintained for multiple consecutive times, a smoke alarm will be triggered.

[0027] When the laser intensity C is below the minimum critical value, it can be determined that the optical path is severely blocked, and the equipment will immediately enter a fault state. At this time, it is necessary to indicate that the laser optical path is blocked and clarify the impact of this situation on methane detection. If the smoke detector was in normal condition before the optical path was blocked, it is determined that there is a fault in the infrared optical path synchronization of the smoke detector, which needs to be checked manually. If the smoke detector has already triggered an alarm before the blockage, it will not actively deactivate the alarm state even if the smoke sensor does not detect abnormal data, in order to prevent false alarms or other misjudgments caused by smoke and dust accumulation.

[0028] Methane concentration alarms and smoke alarms use differentiated audible and visual methods. For example, the methane alarm is a "long sound + slow red flash," while the smoke alarm is a "short sound + fast red light flash," making it easy for on-site personnel to quickly distinguish between them. Both alarms trigger a linkage output mechanism: on the one hand, the alarm data and alarm time are stored locally, triggering the shutdown of the gas valve and the activation of the exhaust fan; on the other hand, the alarm data is uploaded to the cloud platform via the network communication module (4G / 5G / Wi-Fi), simultaneously triggering the telephone and SMS notification modules to push the alarm information to preset user terminals and maintenance team terminals, achieving multi-channel synchronous notification of alarm information.

[0029] Second Embodiment As a second embodiment of the present invention, please refer to Figure 2 This application provides a smoke methane composite detection device, which includes a detection optical path and signal acquisition component, a laser driving and temperature control component, a signal calculation and processing component, an audible and visual alarm component, and a wireless network communication component; The detection optical path and signal acquisition components include a smoke-detecting infrared emitting tube, an infrared receiving tube, a laser emitter (LD), and a laser receiver (PD). The infrared emitting tube emits an infrared beam, which is scattered by smoke particles in the optical path and then received by the infrared receiving tube, converting the light signal into an electrical signal. The laser emitter (LD) emits a light signal of a specific wavelength, which is received by the laser receiver (PD) and also converted into an electrical signal. The laser drive and temperature control components use PWM regulation combined with precise PID control to ensure that the laser operates within the set temperature range to maintain wavelength stability. At the same time, by applying a fixed excitation signal, the laser can stably emit triangular wave signals. The signal processing component processes the collected electrical signals and outputs indications of smoke detector status, combustible gas concentration status, and alarm status through a fusion algorithm. When the device detects a smoke alarm or a gas leak alarm, the sound and light linkage alarm component will simultaneously emit a clear sound and light signal to alert the user; if a fire hazard is detected, it can also trigger a linkage action of gas circuit cut-off and exhaust fan activation. The wireless network communication component enables real-time monitoring of device status by periodically reporting heartbeat data and real-time reporting alarm data. It also supports instant reminders via SMS and telephone, allowing platform monitoring personnel to view and process the data in real time.

[0030] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.

[0031] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. A method for detecting methane in smoke, characterized in that, The method specifically includes the following steps: Step 1: Use a laser detector to receive the reflected signal from the target detection area and complete data acquisition. Extract the electrical signal intensity parameter with the maximum value in the acquired data and the preset optimal data acquisition range. Record the initial laser acquisition data under the condition that the former satisfies the latter. Step 2: The initial laser acquisition data is filtered and denoised using an exponential weighted moving average algorithm to obtain filtered data. At the same time, the filtered data is weighted and fitted using a least squares polynomial fitting algorithm to obtain fitted data. Step 3: Correct the drift of the fitted data using the slow EWMA baseline vector, extract four features: peak height, peak area, peak width, and optical depth, and input them along with temperature and humidity into the offline-trained 3-layer minimal neural network cross-term model. After aging-contamination self-verification, the final methane concentration is output. Step 4: Use an infrared receiver tube to receive the infrared scattering signal of the target detection area in real time and complete the data acquisition. Then, perform random filtering and ambient temperature compensation on the effective smoke sensor sampling data. Similarly, process the methane concentration data to obtain the final smoke sensor detection value. Step 5: Preset the methane concentration alarm threshold. If the methane concentration exceeds the threshold and the methane alarm is triggered multiple times in a row, dynamically lower the smoke detection threshold according to the laser intensity attenuation rate. If the smoke detection value exceeds the threshold and the smoke detection alarm is triggered multiple times in a row.

2. The method for detecting combined smoke and methane according to claim 1, characterized in that, The method for obtaining the preset optimal data collection range is as follows: Read the factory calibration range [L0] from EEPROM min L0 max Read the real-time temperature t and the cumulative working hours h of the laser, and correct according to the aging-temperature drift model. = h represents the cumulative operating time of the laser, and t represents the real-time temperature measurement value. The aging factor is 0.3% / kh. This is the temperature coefficient, with a value of 0.15% / ℃. As the relative attenuation ratio, further according to the formula L min =L0 min ·(1−), L max =L0 max ·(1−), the acquisition interval [L] is calculated. min L max ].

3. The method for detecting combined smoke and methane according to claim 1, characterized in that, The specific method for obtaining the filtered data is as follows: According to the formula =median(S0[n-1],S0[n],S0[n+1]) performs filtering to obtain the current input sample after median pre-filtering, where n is the number of samples. Then, the exponentially weighted moving average algorithm formula y is applied. S [n]=a s × +(1-a s )×y S [n-1], where a is a fixed filter coefficient with a value of 0.05, and y S [n] represents the slow channel output value after the nth sample, y S [n-1] represents the previous basis estimate, and the in-window signal-to-noise ratio is calculated using the following formula: If SNR > 20dB, d f The value is 0.

3. If SNR > 10dB, d f The value is 0.15, and it is combined with the formula y. f [n]=a f × +(1-a f )×y f [n-1], calculate the fast channel output at point n, and combine the fast channel output and the slow channel output to obtain the filtered data. Specifically, S1[n]=y f [n]-y s [n].

4. The method for detecting combined smoke and methane according to claim 1, characterized in that, The specific method for obtaining the fitted data is as follows: Divide the 512-point vector S1 into four equal segments: segment A [0, 63], segment B [64, 191], segment C [192, 319], and segment D [320, 511]. Calculate the SNR of each segment. sub =RMS(S1) sub ) / ,in The noise standard deviation, and = SNR sub = represents the signal-to-noise ratio of the current sub-interval, RMS(S1) sub The value represented by ) indicates the root mean square value of the 128 data points within this sub-interval, representing the effective amplitude of the signal. Then, the maximum value index i is searched in segments B and C. peak and with i peak Generate a 64-point Hann window centered on the target area, and simultaneously calculate the sharpness S = |S1[i peak ]−0.5·(S1[i peak -4]+S1[i peak +4])| / RMS local RMS local S1[i] represents the effective value of noise in the same local interval. peak ] represents the intensity of the current peak, S1[i peak +4] indicates the intensity of 4 sampling points on each side of the peak; If S < 0.5, it indicates a blunt peak and no enhancement is performed. If S ≥ 0.5, 64 points are selected as the reference width w. base =64, and calculate the boundary safety half-width, half=min(i peak -B start B end -i peak w base / 2), then generate a Hann window with an area of ​​1, then H[i] = 0.5 − 0.5·cos(2π·k / w), k = 0 … w - 1, and perform normalization H norm [i]=H[i] / Finally, the peak region weights are superimposed (W). peak [i] = W0[i] + 1.2·S·H norm [i], i∈[i peak -half, i peak +half−1].

5. The method for detecting combined smoke and methane according to claim 1, characterized in that, The specific method for correcting the drift of the fitted data using the slow EWMA baseline vector is as follows: Slow EWMA channel y s [n] values ​​within the same scan period, establishing the base vector B. line Corrected absorption curve S2 corr =S2–B line Simultaneously, four features are extracted: peak height Q, peak area A, peak width W, and optical depth τ. Among them, peak height Q = max(S2) corr Peak area A = Σ(S2) corr >0.2·Q), peak width W = right intersection point – left intersection point, depth τ = –ln(1 – Q / I) 0） Where I0 represents real-time zero atmospheric light intensity.

6. The method for detecting combined smoke and methane according to claim 1, characterized in that, The final methane concentration was obtained as follows: Peak height, peak area, peak width, optical depth, and temperature (t) and humidity (RH) are input together into a minimally oriented neural network cross-term model trained offline. Relative humidity (RH) is then added to the model input, and the model calculates D. comp =D raw ·(1+k1·(t–25)+k2·(RH–50)), where k1 and k2 are regression coefficients, to obtain the original concentration D. raw D comp This represents the compensated concentration, with I0 updated in real-time using saturation pulses within the period, for D. raw Perform aging-contamination self-check, D=round(D comp ·100) / 100, output the final concentration D.

7. The method for detecting combined smoke and methane according to claim 6, characterized in that, V sets of methane concentration and absorption peak position data {(D1, P1), (D2, P2), ..., (Dv, Pv)} were obtained by continuously collecting data v times. Outliers in the concentration D and absorption peak position P of these v sets of data were removed to eliminate interference. The mean M of the concentration data for the v sets was calculated. D With standard deviation S D For each group of concentration data Di, calculate the deviation coefficient |D i -M D | / S D A preset deviation coefficient threshold K is set. Data with a deviation coefficient > K are identified as outliers and the corresponding concentration D and absorption peak position P data are removed as a whole.

8. The method for detecting combined smoke and methane according to claim 1, characterized in that, The final smoke detection value is obtained as follows: An infrared receiver tube is used to receive the infrared scattering signal of the target detection area in real time and complete the data acquisition. For the raw data of a single acquisition, random noise interference is filtered out by an average filtering algorithm to obtain the intermediate sampling value H. Then, the intermediate sampling value H is temperature compensated in combination with the ambient temperature parameter to correct the detection deviation caused by temperature. Finally, the effective sampling value data G of the smoke detector in a single acquisition is obtained. Repeat the above steps to continuously collect m smoke sensor samples, obtaining a dataset {G1, G2, ..., Gm}. Process this dataset using the same outlier removal method as the methane concentration data, first calculating the mean M of the m data sets. G With standard deviation S G Then calculate the sample value G for each group. i deviation coefficient |G i -M G | / S G Outliers with a deviation coefficient greater than a preset threshold are removed; the outlier removal operation is repeated for the remaining data until the amount of remaining data is ≤m / 3 or no new outliers are generated; finally, the mean of the valid data after outlier removal is taken to obtain the final smoke detection value G. S .

9. The method for detecting combined smoke and methane according to claim 1, characterized in that, The specific way to trigger a smoke alarm is as follows: Based on methane concentration detection value D S With smoke detection value G S It achieves accurate alarms by combining dynamic threshold adjustment strategies and sends notifications through multiple channels, with a preset methane concentration alarm threshold T. D It can be tiered according to application scenarios and compares D in real time. S With T D If D S >T D If this state is maintained for a preset number of consecutive times, a methane concentration alarm will be triggered immediately. Preset initial alarm threshold T for smoke detector G First, historical monitoring data of laser intensity C in the cache module is retrieved, and the fluctuation range of C within a preset time window is analyzed. Based on the attenuation of C, the detection lower limit is dynamically adjusted according to preset rules. For example, if C attenuates by 20%, then T... G (Reduced by 15%), resulting in a dynamically adjusted alarm threshold T. G Subsequently, the smoke detection value G was determined. S Does it exceed T? G If G S ≤T G If G is determined to be a smoke-free signal; S >T G If this state is maintained for multiple consecutive times, a smoke alarm will be triggered.

10. A smoke-methane composite detection device, used to perform the smoke-methane composite detection method according to any one of claims 1-9, characterized in that the smoke... The methane composite detection device includes a detection optical path and signal acquisition component, a laser driving and temperature control component, a signal calculation and processing component, an audible and visual alarm component, and a wireless network communication component. The detection optical path and signal acquisition components include a smoke-detecting infrared emitting tube, an infrared receiving tube, a laser emitter, and a laser receiver. The infrared emitting tube emits an infrared beam, which is scattered by smoke particles in the optical path and then received by the infrared receiving tube, converting the optical signal into an electrical signal. The laser emitter emits a light signal of a specific wavelength, which is received by the laser receiver and also converted into an electrical signal. The laser drive and temperature control components use PWM regulation combined with precise PID control. The signal processing component processes the collected electrical signals and outputs indications of smoke detector status, combustible gas concentration status, and alarm status through a fusion algorithm. When the device detects a smoke alarm or a gas leak alarm, the sound and light linkage alarm component will simultaneously emit a clear sound and light signal to indicate that the device is working. The wireless network communication component reports heartbeat data periodically and alarm data in real time, allowing platform monitoring personnel to view and process the data in real time.

Images