A multi-mode adaptive flame detection system and method

Abstract

The application provides a multi-mode adaptive flame detection system and method, and relates to the technical field of flame detection. The system comprises a sensing module for collecting ultraviolet light signals and infrared light signals in a flame signal, which are denoted as flame detection signals; a signal processing module for performing flame identification according to the flame detection signals and determining the type of the flame; a mode selection module for selecting a flame detection algorithm matched with the type of the flame according to the type of the flame, and detecting the flame; and a cooling module for cooling the sensing module. The application can identify the type of fuel in real time according to the characteristics of the flame signal, and can adapt to different fuels such as gas, light oil, coal powder or biomass particles without manual replacement of hardware, so that reliable monitoring of multi-fuel combustion can be realized by a single detection system.

Description

Technical Field

[0001] This application relates to the field of flame detection technology, and in particular to a multi-mode adaptive flame detection system and method. Background Technology

[0002] In the field of industrial combustion control and safety monitoring, flame detection technology is one of the key technologies to ensure the safe and stable operation of combustion equipment. It is widely used in various industrial scenarios with combustion chambers, such as gas-fired or oil-fired boilers in the power industry, petrochemical and chemical heating furnaces, ceramic and glass melting furnaces, and district heating boiler rooms. With increasing demands for energy conservation, emission reduction, and flexible multi-fuel switching, combustion systems are gradually developing towards multi-fuel adaptability and intelligence, placing higher requirements on the adaptability and reliability of flame detection systems.

[0003] Currently, mainstream flame detection technologies are primarily based on single-principle sensors, including ultraviolet (UV) flame sensors and infrared (IR) flame sensors. UV sensors respond rapidly to short-wave radiation generated by the combustion of gaseous or light fuels, while IR sensors are more sensitive to thermal radiation generated by the combustion of solid or heavy fuels. Existing systems typically select one type of sensor based on fuel type during installation and commissioning, and match it with the corresponding algorithm. However, this "one fuel, one sensor" approach has significant drawbacks: when the fuel type changes, the system must be shut down to replace the sensor and readjust the parameters, a cumbersome process with a risk of misoperation; the system cannot automatically recognize fuel changes and lacks adaptive capability. Summary of the Invention

[0004] To address the aforementioned shortcomings, this application provides a multi-mode adaptive flame detection system and method.

[0005] Firstly, this application provides a multi-mode adaptive flame detection system, which adopts the following technical solution: A multi-mode adaptive flame detection system, the system comprising: The sensing module includes an ultraviolet flame sensor and an infrared flame sensor, which are used to collect ultraviolet light signals and infrared light signals in the flame signal, respectively, record them as flame detection signals, and send them to the signal processing module and the mode selection module. The signal processing module, connected to the sensing module, is used to receive the flame detection signal sent by the sensing module, and to identify the flame and determine the type of flame based on the flame detection signal. The mode selection module is connected to the signal processing module and the sensing module respectively. It is used to select a flame detection algorithm that matches the flame type to detect the flame. The cooling module includes an ultraviolet flame sensor cooling unit and an infrared flame sensor cooling unit, which are used to cool the ultraviolet flame sensor and the infrared flame sensor, respectively.

[0006] Optionally, the sensing module further includes an optical path unit, which includes a single lens and a beam splitter. The single lens is used to collect and converge the light emitted by the flame; the beam splitter is used to separate the light converged by the single lens into different wavelengths.

[0007] Optionally, the signal processing module includes a signal reliability evaluation unit and a signal conditioning unit; After receiving the flame detection signal sent by the sensing module, and before identifying the flame based on the flame detection signal and determining the type of flame, the signal reliability assessment unit is executed. The signal reliability assessment unit is used to extract the signal-to-noise ratio, baseline drift, main frequency of fluctuation, and standard deviation of amplitude fluctuation of the flame detection signal, obtain the environmental parameters of the current flame detection signal, calculate the reliability of the current flame detection signal based on the signal reliability assessment formula, and determine whether the calculated reliability is greater than the preset reliability threshold. If it is, no processing is performed; if not, the signal conditioning unit is executed to adjust the flame detection signal collected by the sensing module. The signal conditioning unit includes a first conditioning subunit, a second conditioning subunit, and a third conditioning subunit; The first adjustment subunit is used to execute a preset signal stability adjustment strategy; The second adjustment subunit is used to execute a preset sensor adjustment strategy; The third adjustment subunit is used to execute preset optical path and environmental adjustment strategies.

[0008] Optionally, the signal reliability evaluation formula is: ; in, The weighting coefficients represent the signal-to-noise ratio coupling term. This indicates the coupling relationship between the signal-to-noise ratio and the standard deviation of the amplitude fluctuation of the flame detection signal. , Indicates the standard deviation of amplitude fluctuation. This represents the normalized signal-to-noise ratio. , Indicates the signal-to-noise ratio. Indicates the maximum reference signal-to-noise ratio. Indicates the dynamic weight of the signal-to-noise ratio. Indicates the penalty factor. The weighting coefficients of the baseline drift coupling term are represented. This indicates the coupling relationship between the baseline drift of the flame detection signal and the dominant frequency of the fluctuation. , , Indicates the baseline drift amount. Indicates the dominant frequency of the fluctuation. Indicates the Nyquist frequency. Indicates the effective baseline drift. Indicates the frequency influence coefficient. Represents the sensitivity coefficient. This represents the weighting coefficient of the environmental coupling term. This indicates the coupling relationship between environmental parameters and flame detection signals. , This represents the difference between the current ambient temperature and the reference temperature. This represents the difference between the current standard deviation of amplitude fluctuation and the mean of the historical standard deviation of amplitude fluctuation. This represents the difference between the current ambient humidity and the reference humidity. Indicates the ambient background light radiation intensity. Indicates the amplitude of environmental electromagnetic interference. This represents the coupling coefficient between temperature and the standard deviation of amplitude fluctuation. This represents the coupling coefficient between humidity and baseline drift. This represents the coupling coefficient between light radiation intensity and electromagnetic interference.

[0009] Optionally, the signal processing module further includes a credibility determination unit, which includes a first determination subunit and a second determination subunit; The first judgment unit is used to predict the predicted value of the amplitude fluctuation standard deviation of the current flame detection signal, denoted as the first associated predicted value; obtain the amplitude fluctuation standard deviation of the current flame detection signal, denoted as the first actual associated value; calculate the relative deviation between the first associated predicted value and the first actual associated value, denoted as the first relative deviation; and determine whether the first relative deviation is greater than a preset first deviation threshold. If so, it is determined that the amplitude fluctuation standard deviation and the signal-to-noise ratio are abnormally correlated; otherwise, it is determined that the amplitude fluctuation standard deviation and the signal-to-noise ratio are normally correlated. The second judgment unit is used to predict the standard deviation of the amplitude fluctuation of the flame detection signal at the next moment based on the standard deviation of the amplitude fluctuation of the flame detection signal and the baseline drift, and record it as the second associated predicted value; obtain the standard deviation of the amplitude fluctuation of the flame detection signal at the next moment and record it as the second actual associated value; calculate the relative deviation between the second associated predicted value and the second actual associated value and record it as the second relative deviation; and determine whether the second relative deviation is greater than the preset second deviation threshold. If it is, the correlation between the standard deviation of amplitude fluctuation and the baseline drift is determined to be abnormal; if not, the correlation between the standard deviation of amplitude fluctuation and the baseline drift is determined to be normal.

[0010] Optionally, when the confidence level of the flame detection signal is greater than the preset confidence level threshold, and the correlation between the standard deviation of amplitude fluctuation and the signal-to-noise ratio and the correlation between the standard deviation of amplitude fluctuation and the baseline drift are both normal, no processing is performed. When the reliability of the flame detection signal is not greater than the preset reliability threshold, and the correlation between the amplitude fluctuation standard deviation and the signal-to-noise ratio and the correlation between the amplitude fluctuation standard deviation and the baseline drift are both normal, the first adjustment subunit is executed. When the amplitude fluctuation standard deviation is abnormally correlated with the signal-to-noise ratio but the amplitude fluctuation standard deviation is normally correlated with the baseline drift, the second adjustment subunit is executed; When the amplitude fluctuation standard deviation is normally correlated with the signal-to-noise ratio but abnormally correlated with the baseline drift, the third adjustment subunit is executed; When the correlation between the standard deviation of amplitude fluctuation and the signal-to-noise ratio and the correlation between the standard deviation of amplitude fluctuation and the baseline drift are both abnormal, the second adjustment subunit, the first adjustment subunit, and the third adjustment subunit are executed sequentially.

[0011] Optionally, the signal processing module further includes a parameter query unit, which stores historical flame detection signals and corresponding signal adjustment strategies. The parameter query unit is used to compare the current flame detection signal with the historical flame detection signal after the signal reliability assessment unit is executed and before the signal adjustment unit is executed. If the current flame detection signal is consistent with the historical flame detection signal, the signal adjustment is performed according to the signal adjustment strategy of the historical flame detection signal; otherwise, the signal adjustment unit is executed. The signal conditioning strategies include signal stability adjustment strategies, sensor adjustment strategies, and optical path and environment adjustment strategies.

[0012] Optionally, the signal processing module further includes a flame authenticity determination unit; The flame authenticity determination unit is used to construct a flame state vector based on the flame detection signal of the sensing module after executing the parameter query unit or the signal conditioning unit. The flame feature vector includes ultraviolet signal amplitude, infrared signal amplitude, ultraviolet-infrared ratio, wave main frequency, scintillation intensity, signal-to-noise ratio, baseline drift, amplitude fluctuation standard deviation, and environmental parameters. Based on a preset fuel feature database, an ideal flame feature vector is constructed, and the correlation between the flame state vector and the ideal flame feature vector is calculated as an authenticity parameter. The authenticity parameter is then compared with a preset authenticity judgment table. If the authenticity parameter is less than the first threshold, the flame detected by the flame detection signal is determined to be a real flame. If the authenticity parameter is not less than the first threshold and the authenticity parameter is less than the second threshold, then the flame of the flame detection signal is determined to be an unknown flame. If the authenticity parameter is not less than the second threshold, then the flame of the flame detection signal is determined to be subject to non-flame interference. The first threshold is less than the second threshold.

[0013] Optionally, the signal processing unit includes a flame recognition unit; The flame recognition unit is used to construct a flame recognition model based on determining that the flame in the flame detection signal is a real flame. The flame state vector is input into the flame recognition model, and the flame category probability distribution is output. The flame category probabilities are sorted in descending order, with the flame category probability ranked first being recorded as the first probability and the flame category probability ranked second being recorded as the second probability. The difference between the first probability and the second probability is calculated and recorded as the category confidence. If the category confidence is greater than a preset confidence threshold and the first probability is greater than a preset probability threshold, then the flame category with the first probability is taken as the flame category of the flame detection signal; otherwise, the flame category of the flame detection signal is determined to be uncertain.

[0014] Secondly, this application provides a multi-mode adaptive flame detection method, which is applicable to the system described in any one of the first aspects above, and the method adopts the following technical solution: The ultraviolet and infrared light signals in the flame signal are collected and recorded as flame detection signals; Flame identification is performed based on flame detection signals to determine the type of flame. Based on the type of flame, a flame detection algorithm that matches the flame type is selected to detect the flame.

[0015] In summary, this application includes at least one of the following beneficial technical effects: 1. This application integrates ultraviolet and infrared flame sensors to achieve automatic selection and switching of flame detection modes. The system can identify fuel type in real time based on flame signal characteristics, adapting to different fuels such as natural gas, light oil, pulverized coal, or biomass pellets without manual hardware replacement, enabling reliable monitoring of multi-fuel combustion by a single detection system. Compared to existing technologies that require manual sensor replacement, this application significantly improves the flexibility and operating efficiency of combustion equipment while reducing downtime and maintenance costs.

[0016] 2. This application extracts the signal-to-noise ratio, baseline drift, amplitude fluctuation, and fluctuation frequency of the flame detection signal, and calculates the signal reliability in conjunction with environmental parameters. This system can determine and classify abnormal signals, further triggering multi-subunit adjustment strategies (including signal stability adjustment, sensor adjustment, and optical path environment adjustment). This effectively reduces the impact of environmental interference, sensor contamination, and signal fluctuations on flame recognition, thereby ensuring the high reliability of the flame recognition input signal. The system also achieves rapid and adaptive parameter adjustment through historical flame signal comparison and strategy invocation, maintaining stable detection performance without manual intervention. This not only reduces the workload of operators but also lowers the safety risks caused by human error, improving the long-term reliability and safety of the combustion furnace.

[0017] 3. By predicting the correlation between amplitude fluctuation and signal-to-noise ratio, and amplitude fluctuation and baseline drift, and by judging deviations, this application enables the system to proactively identify potential abnormal states of flame signals and achieve predictive adjustment. This technology can not only detect sudden signal anomalies but also identify the impact of sensor drift or environmental interference on the detection results in advance, thereby improving system safety and stability.

[0018] 4. In the signal processing, this application constructs a flame state vector, an ideal flame state equation, and an environmental compensation neural network model to determine the authenticity of the flame detection signal, effectively distinguishing between real flames, unknown flames, and non-flame interference, significantly reducing false alarm and false negative rates. Furthermore, in the flame recognition unit, the flame state vector, ideal flame detection signal, and environmental parameters are input into the flame recognition model, outputting a flame category probability distribution. Judgment is then made using confidence levels and probability thresholds, ensuring the reliability and interpretability of the flame category determination. This system not only responds to environmental changes in real time and adaptively but also considers historical experience and physical constraints, achieving high-precision, multi-mode, and low-false-alarm flame detection and classification, providing reliable technical support for applications such as industrial combustion control and fire safety monitoring. Attached Figure Description

[0019] Figure 1 This is a schematic block diagram of Embodiment 1 of this application; Figure 2 This is a schematic block diagram of the signal processing module of Embodiment 1 of this application; Figure 3 This is a schematic block diagram of the signal conditioning unit in Embodiment 1 of this application. Detailed Implementation

[0020] The following combination Figures 1 to 3 This application will be described in further detail.

[0021] Example 1: This example discloses a multi-mode adaptive flame detection system, such as... Figure 1As shown, the system includes: a sensing module for collecting ultraviolet and infrared light signals from flame signals, denoted as flame detection signals; a signal processing module for flame identification based on the flame detection signals to determine the flame type; a mode selection module for selecting a flame detection algorithm matching the flame type to detect the flame; and a cooling module for cooling the sensing module. The specific steps of this embodiment are as follows: The sensing module includes an ultraviolet (UV) flame sensor and an infrared (IR) flame sensor, which are used to collect UV and IR light signals from the flame signal, respectively, recording them as flame detection signals, and sending them to the signal processing module and the mode selection module. Specifically, the sensing module also includes an optical path unit, which includes a single lens and a beam splitter. The single lens collects and converges the light emitted by the flame; the beam splitter separates the light converged by the single lens into different wavelengths, transmitting the light to the UV and IR flame sensors respectively. This common optical path design ensures that the two flame sensors observe the exact same flame area, eliminating recognition errors caused by different viewing angles and improving the accuracy of flame detection.

[0022] The signal processing module is connected to the sensing module and is used to receive the flame detection signal sent by the sensing module, and to identify the flame and determine its type based on the flame detection signal. Specifically, for example... Figure 2 As shown, the signal processing module includes a signal credibility assessment unit, a credibility determination unit, a parameter query unit, a signal conditioning unit, a flame authenticity discrimination unit, and a flame identification unit.

[0023] The signal reliability assessment unit extracts the signal-to-noise ratio, baseline drift, dominant frequency of fluctuation, and standard deviation of amplitude fluctuation of the flame detection signal. It obtains the environmental parameters of the current flame detection signal, calculates the reliability of the current flame detection signal based on the signal reliability assessment formula, and determines whether the calculated reliability exceeds a preset reliability threshold. If it does, no action is taken; otherwise, the signal conditioning unit is executed to adjust the flame detection signal acquired by the sensor module. Environmental parameters include ambient temperature, ambient humidity, ambient background light radiation intensity, and ambient electromagnetic interference amplitude.

[0024] The formula for evaluating the reliability of the signal is: ; in, The weighting coefficients represent the signal-to-noise ratio coupling term. This indicates the coupling relationship between the signal-to-noise ratio and the standard deviation of the amplitude fluctuation of the flame detection signal. , Indicates the standard deviation of amplitude fluctuation. This represents the normalized signal-to-noise ratio. , Indicates the signal-to-noise ratio. This indicates the maximum reference signal-to-noise ratio, the optimal signal-to-noise ratio value for the system design, such as 50dB. Indicates the dynamic weight of the signal-to-noise ratio. , This represents the attenuation coefficient, which is set to 2.0 in this embodiment. This represents the standard deviation of the reference amplitude fluctuation, which can be set according to the fuel type in this embodiment. This represents the penalty factor, which is 0.5 in this embodiment. The weighting coefficients of the baseline drift coupling term are represented. This indicates the coupling relationship between the baseline drift of the flame detection signal and the dominant frequency of the fluctuation. , , Indicates the baseline drift amount. Indicates the dominant frequency of the fluctuation. This represents the Nyquist frequency, which in this embodiment is taken as half the sampling frequency, i.e., 500 Hz. Indicates the effective baseline drift. This represents the frequency influence coefficient, which is set to 2.0 in this embodiment. This represents the sensitivity coefficient, which is set to 5.0 in this embodiment. This represents the weighting coefficient of the environmental coupling term. This indicates the coupling relationship between environmental parameters and flame detection signals. , This represents the difference between the current ambient temperature and the reference temperature. This represents the difference between the current standard deviation of amplitude fluctuation and the mean of the historical standard deviation of amplitude fluctuation. This represents the difference between the current ambient humidity and the reference humidity. Indicates the ambient background light radiation intensity. Indicates the amplitude of environmental electromagnetic interference. This represents the coupling coefficient between temperature and the standard deviation of amplitude fluctuation, which is set to 0.1 in this embodiment. This represents the coupling coefficient between humidity and baseline drift, which is set to 0.2 in this embodiment. This represents the coupling coefficient between light radiation intensity and electromagnetic interference, which is set to 0.05 in this embodiment. , and The value is adjusted according to the fuel type. In this embodiment, the fuel type of the current flame detection signal can be roughly determined based on the ratio of ultraviolet intensity to infrared intensity in the flame detection signal, so as to evaluate the signal reliability of the flame detection signal.

[0025] The signal conditioning unit includes a first conditioning subunit, a second conditioning subunit, and a third conditioning subunit, such as... Figure 3 As shown, the first adjustment subunit is used to execute a preset signal stability adjustment strategy, which includes amplitude fluctuation suppression (i.e., adjusting the gain of the ultraviolet flame sensor and the infrared flame sensor), frequency filtering optimization, and signal integration adjustment.

[0026] Specifically, amplitude fluctuation suppression: Obtain the standard deviation of the amplitude fluctuation of the current flame detection signal, determine whether the standard deviation of the amplitude fluctuation of the current flame detection signal is greater than the preset fluctuation threshold. If it is, reduce the sensor gain; if not, increase the sensor gain. Then re-execute the signal reliability evaluation unit. If the reliability of the flame detection signal is still not greater than the preset reliability threshold, perform frequency filtering optimization; otherwise, do not process it.

[0027] Frequency filtering optimization: Obtain the main fluctuation frequency of the current flame detection signal, and determine whether the main fluctuation frequency of the current flame detection signal is greater than the preset frequency threshold. If yes, enable low-pass digital filtering; if no, enable high-pass digital filtering. Then, re-execute the signal reliability evaluation unit. If the reliability of the flame detection signal is still not greater than the preset reliability threshold, perform signal integration adjustment; otherwise, do not process it.

[0028] Signal integration adjustment: Obtain the signal-to-noise ratio (SNR) of the current flame detection signal, determine whether the SNR of the current flame detection signal is greater than the preset SNR threshold. If yes, extend the signal integration time; otherwise, shorten the signal integration time and then re-execute the signal reliability assessment unit. If the reliability of the flame detection signal is still not greater than the preset reliability threshold, generate an alarm prompt; otherwise, do not take any action.

[0029] The second adjustment subunit is used to execute a preset sensor adjustment strategy, which includes sensor susceptibility checking, sensor gain adjustment, and electromagnetic interference troubleshooting.

[0030] Specifically, sensor responsiveness check: Calculate the sensor responsiveness attenuation rate. If the attenuation rate is greater than 30%, mark the sensor as aged and activate the backup sensor; otherwise, do not take any action and re-execute the reliability judgment unit. If the amplitude fluctuation standard deviation is abnormally correlated with the signal-to-noise ratio, perform sensor gain adjustment; otherwise, do not take any action.

[0031] Sensor gain adjustment: Query the historical best gain value, adjust the current gain to within ±10% of the historical best value, and re-execute the confidence judgment unit. If the amplitude fluctuation standard deviation is abnormally correlated with the signal-to-noise ratio, perform electromagnetic interference investigation; otherwise, do not take any action.

[0032] Electromagnetic interference troubleshooting: Perform spectrum analysis on the flame detection signal. If the harmonic components are greater than 20mV, activate the notch filter and re-execute the reliability determination unit. If the amplitude fluctuation standard deviation is abnormally correlated with the signal-to-noise ratio, generate an alarm prompt; otherwise, do not take any action.

[0033] The third adjustment subunit is used to execute a preset optical path and environment adjustment strategy, which includes optical path pollution remediation and environmental compensation adjustment.

[0034] Specifically, optical path contamination remediation involves: determining whether condensation exists in the current environment; if so, lens heating is performed, and the reliability assessment unit is re-executed; if the amplitude fluctuation standard deviation and baseline drift are abnormally correlated, the optical path contamination status is checked; otherwise, no action is taken. Optical path contamination status includes calculating the lens contamination index; if the lens contamination index exceeds a preset contamination threshold, pulsed airflow cleaning is triggered, and the reliability assessment unit is re-executed; if the amplitude fluctuation standard deviation and baseline drift are abnormally correlated, environmental compensation adjustments are performed; otherwise, no action is taken.

[0035] The formula for the lens contamination index is: ; in, Indicates the lens contamination index. Indicates the initial flame signal amplitude. This indicates the current flame signal amplitude.

[0036] Environmental compensation and adjustment: Obtain current and initial environmental parameters, including temperature, humidity, background light intensity, and noise signal; based on the current and initial environmental parameters, calculate the environmental interference deviation value and determine whether the environmental interference deviation value exceeds a preset environmental threshold. If the environmental interference deviation value is less than the preset environmental threshold, no action will be taken; If the environmental interference deviation value is greater than the preset environmental threshold, switch to narrowband optical filter mode, shorten the exposure time, or reduce noise, and re-execute the confidence determination unit. If the amplitude fluctuation standard deviation is abnormally correlated with the baseline drift, it is recommended to shut down for protection.

[0037] The formula for the environmental interference deviation value is: ; in, , , , Indicates the weighting coefficients. Indicates the current ambient temperature. Indicates the initial ambient temperature. Indicates the current ambient temperature. Indicates the initial ambient temperature. Indicates the current background light intensity. Indicates the initial background light intensity. Indicates the current noise signal. This represents the initial noise signal, in this embodiment, , , , Those skilled in the art can design according to actual needs. , , , The value of .

[0038] Further, following the signal reliability assessment unit, a reliability determination unit is executed, which includes a first determination subunit and a second determination subunit. The first determination unit is used to predict the predicted value of the amplitude fluctuation standard deviation of the current flame detection signal, denoted as the first associated predicted value; obtain the amplitude fluctuation standard deviation of the current flame detection signal, denoted as the first actual associated value; calculate the relative deviation between the first associated predicted value and the first actual associated value, denoted as the first relative deviation; and determine whether the first relative deviation is greater than a preset first deviation threshold. If so, the correlation between the amplitude fluctuation standard deviation and the signal-to-noise ratio is determined to be abnormal; otherwise, the correlation between the amplitude fluctuation standard deviation and the signal-to-noise ratio is determined to be normal.

[0039] Specifically, based on historical flame monitoring data, a regression prediction model for amplitude fluctuation standard deviation and signal-to-noise ratio is constructed. The measured value of the amplitude fluctuation standard deviation of the current flame detection signal is obtained and substituted into the regression prediction model of amplitude fluctuation standard deviation and signal-to-noise ratio to calculate the predicted value of the amplitude fluctuation standard deviation of the flame detection signal, which is denoted as the first correlation prediction value. The actual correlation value of the amplitude fluctuation standard deviation and signal-to-noise ratio of the current flame detection signal is obtained and denoted as the first actual correlation value. Based on the formula... The relative deviation between the first associated predicted value and the first actual associated value is calculated and denoted as the first relative deviation. It is then determined whether the first relative deviation is greater than the preset first deviation threshold. If so, it is determined that the correlation between the amplitude fluctuation standard deviation and the signal-to-noise ratio is abnormal, indicating that there is a situation such as sensor performance degradation, electromagnetic interference, or sudden change in fuel type. If not, the prediction is consistent with the actual value, and the correlation between the amplitude fluctuation standard deviation and the signal-to-noise ratio is normal.

[0040] in, , and The regression coefficients are updated online using the least squares method, with an update period of 50 time windows. This represents the first associated predicted value. This represents the first actual associated value.

[0041] The second judgment unit is used to predict the standard deviation of the amplitude fluctuation of the flame detection signal at the next moment based on the standard deviation of the amplitude fluctuation of the flame detection signal and the baseline drift, and record it as the second associated predicted value; obtain the standard deviation of the amplitude fluctuation of the flame detection signal at the next moment and record it as the second actual associated value; calculate the relative deviation between the second associated predicted value and the second actual associated value and record it as the second relative deviation; and determine whether the second relative deviation is greater than the preset second deviation threshold. If it is, the correlation between the standard deviation of amplitude fluctuation and the baseline drift is determined to be abnormal; if not, the correlation between the standard deviation of amplitude fluctuation and the baseline drift is determined to be normal.

[0042] Specifically, a first-order inertial prediction model is established. Obtain the standard deviation of the amplitude fluctuation of the current flame detection signal. and baseline drift Substitute the values ​​into the first-order inertial prediction model to calculate the predicted standard deviation of the amplitude fluctuation of the flame detection signal at the next moment. Obtain the standard deviation of amplitude fluctuation at the next moment. Based on the formula The relative deviation between the second predicted value and the second actual value is calculated and denoted as the second relative deviation. It is then determined whether the second relative deviation is greater than the preset second deviation threshold. If it is, the correlation between the amplitude fluctuation standard deviation and the baseline drift is determined to be abnormal, indicating that there is a sudden external interference or optical path contamination that accelerates the deterioration. If not, the correlation between the amplitude fluctuation standard deviation and the baseline drift is determined to be normal.

[0043] in, This represents the standard deviation of the amplitude fluctuation of the current flame detection signal. , express, This indicates the baseline drift of the current flame detection signal. This represents the standard deviation of the amplitude fluctuation of the flame detection signal at the next moment. This represents the coupling coefficient between the standard deviation of amplitude fluctuation and the baseline drift, with a value of 0.5.

[0044] No action is taken when the confidence level of the flame detection signal is greater than the preset confidence level threshold, and the correlation between the standard deviation of amplitude fluctuation and the signal-to-noise ratio and the correlation between the standard deviation of amplitude fluctuation and the baseline drift are both normal.

[0045] When the reliability of the flame detection signal is not greater than the preset reliability threshold, and the correlation between the amplitude fluctuation standard deviation and the signal-to-noise ratio and the correlation between the amplitude fluctuation standard deviation and the baseline drift are both normal, the first adjustment subunit is executed.

[0046] When the amplitude fluctuation standard deviation is abnormally correlated with the signal-to-noise ratio but the amplitude fluctuation standard deviation is normally correlated with the baseline drift, the second adjustment subunit is executed.

[0047] When the amplitude fluctuation standard deviation is normally correlated with the signal-to-noise ratio but abnormally correlated with the baseline drift, the third adjustment subunit is executed.

[0048] When the correlation between the standard deviation of amplitude fluctuation and the signal-to-noise ratio and the correlation between the standard deviation of amplitude fluctuation and the baseline drift are both abnormal, the second adjustment subunit, the first adjustment subunit, and the third adjustment subunit are executed sequentially.

[0049] Furthermore, the parameter query unit stores historical flame detection signals and corresponding signal conditioning strategies. This parameter query unit is used to compare the current flame detection signal with historical flame detection signals after the signal reliability assessment unit and before the signal conditioning unit. If the current flame detection signal matches the historical flame detection signal, the signal conditioning strategy of that historical flame detection signal is applied. If the current flame detection signal does not match the historical flame detection signal, the signal conditioning unit is executed. It should be noted that when determining whether the current flame detection signal matches the historical flame detection signal, the signal-to-noise ratio, baseline drift, dominant frequency of the fluctuation, standard deviation of amplitude fluctuation, and corresponding environmental parameters of the flame detection signal need to be compared one by one.

[0050] The flame authenticity determination unit is used after the execution parameter query unit or signal conditioning unit to construct a flame state vector, an ideal flame state equation and an environmental compensation neural network model based on the flame detection signal of the sensing module. Based on the environmental compensation neural network model, the ideal flame state equation is parameter corrected. Based on the corrected parameters, the corrected ideal flame state equation is obtained. The flame detection signal is input into the corrected ideal flame state equation, and the ideal flame detection signal is output. The difference between the flame state vector and the ideal flame detection signal is calculated as the authenticity parameter. The authenticity parameter is compared with the preset authenticity judgment table. If the authenticity parameter is less than the first threshold, the flame detected by the flame detection signal is determined to be a real flame. If the authenticity parameter is not less than the first threshold and the authenticity parameter is less than the second threshold, then the flame of the flame detection signal is determined to be an unknown flame. If the authenticity parameter is not less than the second threshold, then the flame in the flame detection signal is determined to be subject to non-flame interference.

[0051] Specifically, a flame state vector is constructed based on flame detection signals and environmental parameters. This flame state vector includes ultraviolet light signals collected by an ultraviolet flame sensor, infrared signals collected by an infrared flame sensor, and environmental parameters. The flame feature vector includes ultraviolet signal amplitude, infrared signal amplitude, ultraviolet-infrared ratio, dominant frequency of fluctuation, scintillation intensity, signal-to-noise ratio, baseline drift, amplitude fluctuation standard deviation, and environmental parameters. Based on an experimentally calibrated fuel feature database, an ideal flame feature vector is constructed. This database records typical radiation characteristic parameters of various fuels under standard combustion conditions through experimental calibration. The Euclidean or Mahalanobis distance between the flame state vector and the ideal flame feature vector is calculated as a authenticity parameter, and a comparison is made between this authenticity parameter and a preset authenticity judgment table. If the authenticity parameter is less than the first threshold, the flame detected by the flame detection signal is determined to be a real flame. If the authenticity parameter is not less than the first threshold and the authenticity parameter is less than the second threshold, then the flame of the flame detection signal is determined to be an unknown flame. If the authenticity parameter is not less than the second threshold, then the flame in the flame detection signal is determined to be subject to non-flame interference.

[0052] The preset authenticity judgment table includes a first threshold, a second threshold, and a flame authenticity result, wherein the first threshold is less than the second threshold.

[0053] In this application, by constructing a flame state vector, an ideal flame state equation, and an environmental compensation neural network model, the authenticity of flame detection signals can be determined, effectively distinguishing between real flames, unknown flames, and non-flame interference, and significantly reducing the false alarm rate and false negative rate.

[0054] The flame recognition unit is used to construct a flame recognition model based on a multi-layer feedforward neural network, based on the determination that the flame in the flame detection signal is a real flame. Historical experimental data of each flame and corresponding environmental parameters are acquired as training data. The training data is input into the flame recognition model for iterative training. During iterative training, the model loss is calculated according to the loss function, the model gradient is updated based on the model loss, and the model parameters are iteratively optimized using a gradient descent algorithm to obtain a trained flame recognition model. This trained flame recognition model is recorded as the new flame recognition model. The flame state vector is input into the new flame recognition model, and the flame category probability distribution is output. The flame category probabilities are sorted in descending order, with the highest probability being recorded as the first probability and the second probability as the second probability. The difference between the first and second probabilities is calculated and recorded as the category confidence. If the category confidence is greater than a preset confidence threshold and the first probability is greater than a preset probability threshold, then the flame category with the first probability is taken as the flame category of the flame detection signal. Otherwise, the flame category of the flame detection signal is considered uncertain, and the flame category is confirmed through manual verification or additional data collection.

[0055] The mode selection module is connected to both the signal processing module and the sensing module. It is used to select a flame detection algorithm that matches the flame type to detect the flame. Specifically, based on the flame type identified by the flame recognition unit, a flame detection algorithm matching the flame type is selected for flame detection. Specifically, when the flame detection signal is determined to be a flame generated by liquid or gaseous fuel, an ultraviolet light signal detection algorithm (such as high-frequency pulse counting) is used to detect the flame; when the flame detection signal is determined to be a flame generated by solid fuel, an infrared light signal detection algorithm (such as low-frequency intensity monitoring) is used to detect the flame.

[0056] In other embodiments, those skilled in the art can collect multi-dimensional features of flame signals (such as the mean, variance, peak value, and dominant frequency in the frequency domain of ultraviolet and infrared light signals) and input these features into a pre-trained classification model (such as a support vector machine, neural network, or other machine learning model) to classify flame types, thereby achieving higher accuracy and stronger anti-interference capabilities in flame identification. This method can not only be used to distinguish different fuel types, but can also be further adapted to flame changes in complex combustion environments, achieving intelligent and adaptive flame detection.

[0057] In this application, the system includes a cooling module, which is an integrated metal probe housing. Internally, it is physically separated into two chambers by a heat insulation layer: an ultraviolet (UV) photodiode cooling unit and an infrared (IR) photodiode cooling unit, used to cool the UV and IR photodiodes respectively. Each chamber of the UV and IR photodiode cooling units has its own dedicated optical window at its front end. The optical window in the UV photodiode cooling unit is made of synthetic quartz glass; the optical window in the IR photodiode cooling unit is made of calcium fluoride crystal to ensure high transmittance in their respective wavelength bands. The entire cooling module housing integrates a compressed air channel, achieving both air cooling and positive pressure purging. The continuous airflow cools the sensing module and prevents the lens from being contaminated by smoke and dust, ensuring the flame sensor's long-term stable operation in the harsh high-temperature, high-dust furnace environment.

[0058] Example 2: This example discloses a multi-mode adaptive flame detection method, which is applicable to the system described in Example 1. The method includes: The ultraviolet and infrared light signals in the flame signal are collected and recorded as flame detection signals; Flame identification is performed based on flame detection signals to determine the type of flame. Based on the type of flame, a flame detection algorithm that matches the flame type is selected to detect the flame.

[0059] This application integrates ultraviolet and infrared flame sensors to achieve automatic selection and switching of flame detection modes. It can identify fuel type in real time based on flame signal characteristics, adapting to different fuels such as natural gas, light oil, pulverized coal, or biomass pellets without manual hardware replacement, enabling reliable monitoring of multi-fuel combustion with a single detection system. Compared to existing technologies that require manual sensor replacement, this application significantly improves the flexibility and operating efficiency of combustion equipment while reducing downtime and maintenance costs.

[0060] The above are all preferred embodiments of this application and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A multi-mode adaptive flame detection system, characterized in that, The system includes: The sensing module includes an ultraviolet flame sensor and an infrared flame sensor, which are used to collect ultraviolet light signals and infrared light signals in the flame signal, respectively, record them as flame detection signals, and send them to the signal processing module and the mode selection module. The signal processing module, connected to the sensing module, is used to receive the flame detection signal sent by the sensing module, and to identify the flame and determine the type of flame based on the flame detection signal. The mode selection module is connected to the signal processing module and the sensing module respectively. It is used to select a flame detection algorithm that matches the flame type to detect the flame. The cooling module includes an ultraviolet flame sensor cooling unit and an infrared flame sensor cooling unit, which are used to cool the ultraviolet flame sensor and the infrared flame sensor, respectively.

2. The multi-mode adaptive flame detection system according to claim 1, characterized in that, The sensing module further includes an optical path unit, which includes a single lens and a beam splitter. The single lens is used to collect and converge the light emitted by the flame; the beam splitter is used to separate the light converged by the single lens into different wavelengths.

3. The multi-mode adaptive flame detection system according to claim 1, characterized in that, The signal processing module includes a signal reliability evaluation unit and a signal conditioning unit; After receiving the flame detection signal sent by the sensing module, and before identifying the flame based on the flame detection signal and determining the type of flame, the signal reliability assessment unit is executed. The signal reliability assessment unit is used to extract the signal-to-noise ratio, baseline drift, main frequency of fluctuation, and standard deviation of amplitude fluctuation of the flame detection signal, obtain the environmental parameters of the current flame detection signal, calculate the reliability of the current flame detection signal based on the signal reliability assessment formula, and determine whether the calculated reliability is greater than the preset reliability threshold. If it is, no processing is performed; if not, the signal conditioning unit is executed to adjust the flame detection signal collected by the sensing module. The signal conditioning unit includes a first conditioning subunit, a second conditioning subunit, and a third conditioning subunit; The first adjustment subunit is used to execute a preset signal stability adjustment strategy; The second adjustment subunit is used to execute a preset sensor adjustment strategy; The third adjustment subunit is used to execute preset optical path and environmental adjustment strategies.

4. The multi-mode adaptive flame detection system according to claim 3, characterized in that, The formula for evaluating the reliability of the signal is: ； in, The weighting coefficients represent the signal-to-noise ratio coupling term. This indicates the coupling relationship between the signal-to-noise ratio and the standard deviation of the amplitude fluctuation of the flame detection signal. , Indicates the standard deviation of amplitude fluctuation. This represents the normalized signal-to-noise ratio. , Indicates the signal-to-noise ratio. Indicates the maximum reference signal-to-noise ratio. Indicates the dynamic weight of the signal-to-noise ratio. Indicates the penalty factor. The weighting coefficients of the baseline drift coupling term are represented. This indicates the coupling relationship between the baseline drift of the flame detection signal and the dominant frequency of the fluctuation. , , Indicates the baseline drift amount. Indicates the dominant frequency of the fluctuation. Indicates the Nyquist frequency. Indicates the effective baseline drift. Indicates the frequency influence coefficient. Represents the sensitivity coefficient. This represents the weighting coefficient of the environmental coupling term. This indicates the coupling relationship between environmental parameters and flame detection signals. , This represents the difference between the current ambient temperature and the reference temperature. This represents the difference between the current standard deviation of amplitude fluctuation and the mean of the historical standard deviation of amplitude fluctuation. This represents the difference between the current ambient humidity and the reference humidity. Indicates the ambient background light radiation intensity. Indicates the amplitude of environmental electromagnetic interference. This represents the coupling coefficient between temperature and the standard deviation of amplitude fluctuation. This represents the coupling coefficient between humidity and baseline drift. This represents the coupling coefficient between light radiation intensity and electromagnetic interference.

5. The multi-mode adaptive flame detection system according to claim 3, characterized in that, The signal processing module further includes a credibility determination unit, which includes a first judgment subunit and a second judgment subunit. The first judgment unit is used to predict the predicted value of the amplitude fluctuation standard deviation of the current flame detection signal, denoted as the first associated predicted value; obtain the amplitude fluctuation standard deviation of the current flame detection signal, denoted as the first actual associated value; calculate the relative deviation between the first associated predicted value and the first actual associated value, denoted as the first relative deviation; and determine whether the first relative deviation is greater than a preset first deviation threshold. If so, it is determined that the amplitude fluctuation standard deviation and the signal-to-noise ratio are abnormally correlated; otherwise, it is determined that the amplitude fluctuation standard deviation and the signal-to-noise ratio are normally correlated. The second judgment unit is used to predict the standard deviation of the amplitude fluctuation of the flame detection signal at the next moment based on the standard deviation of the amplitude fluctuation of the flame detection signal and the baseline drift, and record it as the second associated predicted value; obtain the standard deviation of the amplitude fluctuation of the flame detection signal at the next moment and record it as the second actual associated value; calculate the relative deviation between the second associated predicted value and the second actual associated value and record it as the second relative deviation; and determine whether the second relative deviation is greater than the preset second deviation threshold. If it is, the correlation between the standard deviation of amplitude fluctuation and the baseline drift is determined to be abnormal; if not, the correlation between the standard deviation of amplitude fluctuation and the baseline drift is determined to be normal.

6. The multi-mode adaptive flame detection system according to claim 5, characterized in that, When the confidence level of the flame detection signal is greater than the preset confidence level threshold, and the correlation between the standard deviation of amplitude fluctuation and the signal-to-noise ratio and the correlation between the standard deviation of amplitude fluctuation and the baseline drift are both normal, no processing is performed. When the reliability of the flame detection signal is not greater than the preset reliability threshold, and the correlation between the amplitude fluctuation standard deviation and the signal-to-noise ratio and the correlation between the amplitude fluctuation standard deviation and the baseline drift are both normal, the first adjustment subunit is executed. When the amplitude fluctuation standard deviation is abnormally correlated with the signal-to-noise ratio but the amplitude fluctuation standard deviation is normally correlated with the baseline drift, the second adjustment subunit is executed; When the amplitude fluctuation standard deviation is normally correlated with the signal-to-noise ratio but abnormally correlated with the baseline drift, the third adjustment subunit is executed; When the correlation between the standard deviation of amplitude fluctuation and the signal-to-noise ratio and the correlation between the standard deviation of amplitude fluctuation and the baseline drift are both abnormal, the second adjustment subunit, the first adjustment subunit, and the third adjustment subunit are executed sequentially.

7. The multi-mode adaptive flame detection system according to claim 3, characterized in that, The signal processing module further includes a parameter query unit, which stores historical flame detection signals and corresponding signal adjustment strategies. The parameter query unit is used to compare the current flame detection signal with the historical flame detection signal after the signal reliability assessment unit is executed and before the signal adjustment unit is executed. If the current flame detection signal is consistent with the historical flame detection signal, the signal adjustment is performed according to the signal adjustment strategy of the historical flame detection signal. Conversely, the signal conditioning unit is activated; The signal conditioning strategies include signal stability adjustment strategies, sensor adjustment strategies, and optical path and environment adjustment strategies.

8. The multi-mode adaptive flame detection system according to claim 3 or 7, characterized in that, The signal processing module also includes a flame authenticity determination unit; The flame authenticity determination unit is used to construct a flame state vector based on the flame detection signal of the sensing module after executing the parameter query unit or the signal conditioning unit. The flame feature vector includes ultraviolet signal amplitude, infrared signal amplitude, ultraviolet-infrared ratio, wave main frequency, scintillation intensity, signal-to-noise ratio, baseline drift, amplitude fluctuation standard deviation, and environmental parameters. Based on a preset fuel feature database, an ideal flame feature vector is constructed, and the correlation between the flame state vector and the ideal flame feature vector is calculated as an authenticity parameter. The authenticity parameter is then compared with a preset authenticity judgment table. If the authenticity parameter is less than the first threshold, the flame detected by the flame detection signal is determined to be a real flame. If the authenticity parameter is not less than the first threshold and the authenticity parameter is less than the second threshold, then the flame of the flame detection signal is determined to be an unknown flame. If the authenticity parameter is not less than the second threshold, then the flame of the flame detection signal is determined to be subject to non-flame interference. The first threshold is less than the second threshold.

9. The multi-mode adaptive flame detection system according to claim 8, characterized in that, The signal processing unit includes a flame recognition unit; The flame recognition unit is used to construct a flame recognition model based on determining that the flame in the flame detection signal is a real flame. The flame state vector is input into the flame recognition model, and the flame category probability distribution is output. The flame category probabilities are sorted in descending order. The flame category probability ranked first is recorded as the first probability, and the flame category probability ranked second is recorded as the second probability. The difference between the first probability and the second probability is calculated and recorded as the category confidence. If the category confidence is greater than a preset confidence threshold and the first probability is greater than a preset probability threshold, then the flame category with the first probability is taken as the flame category of the flame detection signal. Conversely, if the flame detection signal does not show a flame type, then the flame type is determined to be uncertain.

10. A multi-mode adaptive flame detection method, characterized in that, The system is applicable to the system as described in any one of claims 1-9, and the method includes: The ultraviolet and infrared light signals in the flame signal are collected and recorded as flame detection signals; Flame identification is performed based on flame detection signals to determine the type of flame. Based on the type of flame, a flame detection algorithm that matches the flame type is selected to detect the flame.

Images