A fire smoke detection method with combustible material recognition capability

Abstract

The application discloses a fire smoke detection method with combustible material identification capability, and relates to the technical field of fire smoke detection. Characteristic information of combustible material particles is obtained based on SEM / AFM observation and chemical element component analysis, and then three scattering matrix elements are selected according to the characteristic information. A combustible material identification model is set by taking the related parameters of the selected three scattering matrix elements as the input of the combustible material identification model, and a combustible material particle identification threshold is set. Suspicious particulate matters are marked by screening according to the characteristic information of on-site particulate matters, and then a target function of the suspicious particulate matters is output according to the combustible material identification model. The target function is compared with the combustible material particle identification threshold to determine whether there is suspicious combustible material particles, and then a fire smoke alarm is triggered. The combustible material particles are accurately identified through scattering matrix and particulate matter characteristic analysis, so that the accuracy and reliability of fire warning are improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of fire smoke detection, in particular to a fire smoke detection method with combustible material recognition capability. BACKGROUND

[0002] Fire smoke detectors are a crucial part of fire alarm systems. With the acceleration of industrialization and the increasing demand for safety, fire prevention has become an important issue in modern society, especially in high-density population areas, industrial areas, data centers, and other places. The risk and loss of fire are enormous, so there is an increasing demand for intelligent smoke detectors with combustible material recognition capability.

[0003] Existing fire smoke detection technologies, such as smoke detectors based on photoelectric, ion, or temperature changes, are often disturbed by various factors (such as air humidity, environmental pollution, etc.) in practical applications, leading to false positives or false negatives. Traditional smoke detectors only work based on smoke concentration and temperature changes, and cannot deeply understand the properties of smoke particles, making it difficult to distinguish between combustible and non-combustible smoke. In addition, traditional smoke detectors have poor particle recognition ability when the smoke concentration is low or in the early stages of fire, and cannot timely detect potential fire risks. Moreover, they only issue an alarm when the smoke concentration or temperature change reaches a certain threshold, which may miss the early stage of fire or misjudge the smoke caused by other environmental factors.

[0004] Therefore, there is an urgent need for a fire smoke detection method with combustible material recognition capability to address the above problems. SUMMARY

[0005] To address the shortcomings of the prior art, the present application provides a fire smoke detection method with combustible material recognition capability, which solves the problem of traditional fire smoke detection methods that cannot accurately distinguish between combustible and non-combustible particles.

[0006] To achieve the above object, the present application is implemented by the following technical solutions: A fire smoke detection method with combustible material identification capability, comprising the following steps: obtaining characteristic information of combustible material particles based on SEM / AFM observation and chemical element composition analysis; simulating the relationship between the characteristic information of the combustible material particles and scattering matrix elements, identifying the influence degree of the characteristic information change of the combustible material particles on different scattering matrix elements; selecting three scattering matrix elements based on the relationship simulation results, and obtaining the related parameters of the selected three scattering matrix elements; setting a combustible material discrimination model by taking the related parameters of the selected three scattering matrix elements as the input of the combustible material discrimination model; optimizing the output results of the combustible material discrimination model, and then setting the minimum value of the output results of the combustible material discrimination model, i.e. the combustible material particle discrimination threshold; real-time monitoring of on-site particulate matter, obtaining the characteristic information of the on-site particulate matter, and then screening and marking the suspicious particulate matter according to the characteristic information of the on-site particulate matter; obtaining the related parameters of the three scattering matrix elements of the suspicious particulate matter, and then outputting the objective function of the suspicious particulate matter according to the combustible material discrimination model; comparing the objective function with the combustible material particle discrimination threshold to determine whether there is suspicious combustible material particles, and then triggering a fire smoke alarm.

[0007] Further, the specific analysis of the relationship simulation between the characteristic information of the combustible material particles and the scattering matrix elements, and the identification of the influence degree of the characteristic information change of the combustible material particles on different scattering matrix elements is as follows: obtaining the characteristic information of the combustible material particles, and obtaining the unique characteristic index of the combustible material particles based on the characteristic information; simulating the relationship between the unique characteristic index of the combustible material particles and the scattering matrix elements, and constructing the relationship between the scattering matrix elements and the unique characteristic index: S meofsp =λ*A uciffsp , wherein S meofsp represents the scattering matrix element, A uciffsp represents the unique characteristic index, and λ represents the relationship coefficient between the scattering matrix element and the unique characteristic index; substituting the unique characteristic index of different combustible material particles and the scattering matrix elements of each different combustible material particle into the relationship between the scattering matrix elements and the unique characteristic index, obtaining the relationship coefficient between the scattering matrix elements of each different combustible material particle and the unique characteristic index of different combustible material particles, and then taking the average of the relationship coefficient between the scattering matrix elements of each different combustible material particle and the unique characteristic index of different combustible material particles as the relationship coefficient between each scattering matrix element and the unique characteristic index, which is used to represent the influence degree of the characteristic information change of the combustible material particles on different scattering matrix elements.

[0008] Further, the specific analysis of the obtaining the characteristic information of the combustible particles based on the characteristic information to obtain the unique characteristic index of the combustible particles is that: the characteristic information of the combustible particles specifically includes surface roughness, size and main chemical elements and their relative concentrations; the surface roughness, size and main chemical elements and their relative concentrations of the combustible particles are standardized, and then combined according to the surface roughness, size and main chemical elements and their relative concentrations to obtain the unique characteristic index of the combustible particles.

[0009] Further, the specific analysis of the selecting three scattering matrix elements based on the relationship simulation results and obtaining the related parameters of the selected three scattering matrix elements is that: the relationship coefficients of each scattering matrix element and the unique characteristic index are compared, and then each scattering matrix element is sorted according to the relationship coefficient size, and the top three scattering matrix elements are selected as the three scattering matrix elements; the related parameters of the selected three scattering matrix elements specifically include the incident wavelength of the scattering matrix element, the scattering angle of the scattering matrix element and the polarization direction.

[0010] Further, the specific analysis of the setting the combustible particle discrimination model by taking the related parameters of the selected three scattering matrix elements as the input of the combustible particle discrimination model is that: the unique physical index of each scattering matrix element is obtained according to the related parameters of the three scattering matrix elements, and then the unique physical index of the three scattering matrix elements of different combustible particles is taken as the input to obtain the combustible particle discrimination model, and the specific combustible particle discrimination model expression is: In the formula, P(y=flammable) represents the probability that the particulate matter is a combustible particle, U1, U2 and U3 respectively represent the unique physical index of the three scattering matrix elements.

[0011] Further, the output result of the combustible material discrimination model is optimized, and the minimum value of the output result of the combustible material discrimination model is set, that is, the specific analysis of the combustible particle discrimination threshold is: obtaining the output results of different combustible particle samples according to the combustible material discrimination model, obtaining non-combustible particle samples, obtaining the output results of different non-combustible particle samples according to the combustible material discrimination model, and then combining the output results of different combustible particle samples and the output results of different non-combustible particle samples into output probability samples; randomly generating data with a value range of 0 to 1 as a traversal threshold, comparing the output probability samples according to different traversal thresholds respectively, and then outputting the comparison results, combining the comparison results into a confusion matrix, and the specific confusion matrix includes the number of samples that are actually combustible particles and are determined to be combustible particles according to the traversal threshold, the number of samples that are actually combustible particles and are determined to be non-combustible particles according to the traversal threshold, the number of samples that are actually non-combustible particles and are determined to be combustible particles according to the traversal threshold, and the number of samples that are actually non-combustible particles and are determined to be non-combustible particles according to the traversal threshold; obtaining the precision and recall rate corresponding to different traversal thresholds according to different confusion matrices obtained by comparison, obtaining the harmonic mean of the precision and recall rate, and comparing the harmonic mean of the precision and recall rate corresponding to different traversal thresholds, and marking the traversal threshold corresponding to the maximum harmonic mean of the precision and recall rate as the minimum value of the output result of the combustible material discrimination model, that is, the combustible particle discrimination threshold.

[0012] Further, the on-site particles are monitored in real time, the characteristic information of the on-site particles is obtained, and then the on-site particles are screened according to the characteristic information of the on-site particles, and the specific analysis of the suspicious particles is: obtaining the characteristic information of the on-site particles based on the real-time monitoring of the on-site particles, and then obtaining the unique characteristic index of the on-site particles according to the characteristic information of the on-site particles; obtaining the range of the unique characteristic index of the combustible particles based on the characteristic information of the combustible particles, and then comparing the unique characteristic index of the on-site particles with the range of the unique characteristic index of the combustible particles, screening the on-site particles within the range of the unique characteristic index of the combustible particles, and marking them as suspicious particles.

[0013] Further, the target function is compared with the combustible particle discrimination threshold to determine whether there is a suspicious combustible particle, and then the fire smoke alarm is triggered. The specific analysis is: the related parameters of the three scattering matrix elements of the suspicious particles are brought into the combustible material discrimination model, and the target function of the suspicious particles is output; the target function of the suspicious particles is compared with the combustible particle discrimination threshold, and if there is a suspicious particle whose target function is greater than or equal to the combustible particle discrimination threshold, the suspicious particle is determined to be a suspicious combustible particle, and the fire smoke alarm is triggered.

[0014] The present application has the following beneficial effects:

[0015] The fire smoke detection method with combustible material recognition capability can more accurately identify and distinguish different types of combustible materials by obtaining characteristic information of combustible material particles based on SEM / AFM observation and chemical composition analysis. These particle characteristic information can effectively help determine whether the particulate matter is combustible and further speculate whether it can cause a fire. Compared with traditional fire detection methods based on smoke concentration or temperature change, the method is more accurate and sensitive. By simulating the relationship between scattering matrix elements and combustible material particle characteristic information, the influence of different particles on scattering characteristics can be understood in depth. The analysis method based on scattering matrix elements provides a more comprehensive detection mechanism that can distinguish different types of particles and identify the presence of combustible materials. Real-time monitoring of on-site particulate matter and screening based on the characteristic information of the particulate matter can quickly mark suspicious particulate matter, meaning that the system can respond at the initial stage of a fire or before a potential fire, greatly reducing the response time when a fire occurs and reducing the risk of fire spread. Based on the selected scattering matrix element parameter, a combustible material discrimination model can be set to achieve accurate discrimination threshold setting by optimizing the model output result, further improving the reliability and accuracy of combustible material identification. By comparing the objective function and threshold value, the system can accurately distinguish suspicious particulate matter, thereby effectively reducing false positives and false negatives and minimizing losses after a fire occurs. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 The present application is a fire smoke detection method with combustible material recognition capability. DETAILED DESCRIPTION

[0017] The present application is a fire smoke detection method with combustible material recognition capability.

[0018] The general idea of the present application is to analyze the characteristics of combustible material particles by combining SEM / AFM technology, and to establish a combustible material discrimination model through scattering matrix simulation. During real-time monitoring, the characteristic information of the particulate matter is compared with the discrimination model to determine whether there are suspicious combustible material particles, thereby triggering a fire smoke alarm.

[0019] Please refer to Figure 1The embodiment of the present application provides a technical scheme: a fire smoke detection method with combustible material identification capability, comprising the following steps: obtaining characteristic information of combustible material particles based on SEM / AFM observation and chemical element composition analysis; simulating the relationship between the characteristic information of the combustible material particles and the scattering matrix elements, identifying the influence degree of the characteristic information change of the combustible material particles on the change of different scattering matrix elements; selecting three scattering matrix elements based on the relationship simulation result, and obtaining the related parameters of the selected three scattering matrix elements; setting a combustible material discrimination model by taking the related parameters of the selected three scattering matrix elements as the input of the combustible material discrimination model; optimizing the output result of the combustible material discrimination model, and then setting the minimum value of the output result of the combustible material discrimination model, that is, the combustible material particle discrimination threshold; real-time monitoring of on-site particulate matter, obtaining the characteristic information of the on-site particulate matter, and then screening according to the characteristic information of the on-site particulate matter to mark the suspicious particulate matter; obtaining the related parameters of the three scattering matrix elements of the suspicious particulate matter, and then outputting the objective function of the suspicious particulate matter according to the combustible material discrimination model; comparing the objective function with the combustible material particle discrimination threshold to determine whether there is suspicious combustible material particle, and then triggering the fire smoke alarm.

[0020] Specifically, the relationship between the characteristic information of the combustible material particles and the scattering matrix elements is simulated, and the specific analysis of identifying the influence degree of the characteristic information change of the combustible material particles on the change of different scattering matrix elements is as follows: obtaining the characteristic information of the combustible material particles, and obtaining the unique characteristic index of the combustible material particles based on the characteristic information; simulating the relationship between the unique characteristic index of the combustible material particles and the scattering matrix elements, and constructing the relationship between the scattering matrix elements and the unique characteristic index: S meofsp =λ*A uciffsp , wherein S meofsp represents the scattering matrix element, A uciffsp represents the unique characteristic index, and λ represents the relationship coefficient between the scattering matrix element and the unique characteristic index; the unique characteristic index of different combustible material particles and the scattering matrix elements of each different combustible material particle are substituted into the relationship between the scattering matrix elements and the unique characteristic index, the relationship coefficient between the scattering matrix elements of each different combustible material particle and the unique characteristic index of different combustible material particles is obtained, and then the average of the relationship coefficient between the scattering matrix elements of each different combustible material particle and the unique characteristic index of different combustible material particles is taken as the relationship coefficient between each scattering matrix element and the unique characteristic index, which is used to represent the influence degree of the characteristic information change of the combustible material particles on the change of different scattering matrix elements.

[0021] In this implementation scheme, the characteristic information of combustible particles is obtained. The specific analysis for obtaining the unique characteristic index of combustible particles based on this characteristic information is as follows: The characteristic information of combustible particles specifically includes surface roughness, size, and major chemical elements and their relative concentrations. The surface roughness, size, and major chemical elements and their relative concentrations of the combustible particles are standardized, and then combined based on these factors to obtain the unique characteristic index of the combustible particles. An example of the expression for obtaining the unique characteristic index is as follows:

[0022] A uciffsp =Sr*ε1+Dim*ε2+Mce*ε3+Rc*ε4, where Sr represents

[0023] Surface roughness, Dim represents size, Mce represents major chemical elements, Rc represents the relative concentration of major chemical elements, and ε1, ε2, ε3, and ε4 represent the weight values ​​of surface roughness, size, major chemical elements, and their relative concentrations, respectively. The specific weight values ​​are obtained by: directly specifying the weight of each feature based on expert experience or prior knowledge; or using statistical methods, such as Principal Component Analysis (PCA), to automatically determine the importance of each feature based on the variance distribution in the data. PCA can help identify which features have the greatest impact on the final feature index, thus providing a basis for weight allocation. Alternatively, regression analysis or other machine learning algorithms (such as random forests and support vector machines) can be used to train the model, learn the weights of each feature, and automatically adjust the weights by fitting them to actual data, so that the final feature index has the strongest predictive ability for the target (such as flammability). Furthermore, hierarchical analysis of the relative importance of features can be performed, using expert scores and judgment matrices to calculate the weights.

[0024] Based on the simulation results, three scattering matrix elements were selected, and the specific analysis of the relevant parameters of the selected three scattering matrix elements was as follows: the relationship coefficients between each scattering matrix element and the unique characteristic index were compared, and then the scattering matrix elements were sorted according to the relationship coefficients. The top three scattering matrix elements were selected as the three scattering matrix elements. The relevant parameters of the selected three scattering matrix elements specifically include the incident wavelength, scattering angle and polarization direction of the scattering matrix elements.

[0025] The surface roughness represents the irregularity of the particle surface, which affects the scattering direction and intensity of the scattered light. A rougher surface will result in stronger scattering, and the surface morphology can be obtained by techniques such as scanning electron microscopy (SEM) and atomic force microscopy (AFM), so as to calculate the roughness; the size represents the physical size (such as diameter, thickness) of the particle, which has an important influence on its scattering characteristics, and the size of the particle can be measured by particle size analyzers (such as laser particle size analyzers) or microscopic techniques (such as SEM) when the particle size is large; the main chemical elements and their relative concentrations determine the absorption and scattering characteristics of the particle to different wavelengths of light, and the concentration change of the elements will affect the scattering intensity and the distribution of the scattering matrix, and the chemical composition and concentration of the particle can be determined by X-ray fluorescence spectroscopy (XRF) and electron probe analysis techniques.

[0026] The specific surface roughness quantitative index is the average roughness, and the size is quantified by using the mean or median of the particle size as a representative; the main chemical elements are quantified by numbering different chemical elements, and then the numbered values of different chemical elements are used as the quantitative indicators of the chemical elements; the concentration of the chemical elements is obtained by element analysis to obtain the mass fraction or molar fraction of each element, and for each main element, the mass fraction is selected as the quantitative indicator.

[0027] The scattering matrix describes the reflection, transmission and scattering characteristics of the particle to different incident light waves, and by analyzing the scattering matrix elements under different incident wavelengths, scattering angles and polarization directions, the scattering characteristics of different substances can be effectively distinguished, thereby improving the recognition accuracy. The incident wavelength affects the way the light interacts with the particle, and different wavelengths of light have different sensitivities to the particle, which can be set by the wavelength range of the light source or obtained by using a tunable laser; the scattering angle determines the scattering direction of the light after interacting with the particle, and is closely related to the shape, size and surface characteristics of the particle, and the scattering intensity of the light at different angles is recorded by a detector to obtain the scattering angle distribution; the polarization direction is the polarization state of the light, which has a significant influence on the scattering of the particle, and different polarization directions correspond to different scattering modes, and the polarization direction of the scattered light is measured by a polarized light source and a polarization detector.

[0028] The scattering matrix is a mathematical tool for describing the interaction between different waveguides in a linear system, which is applied in quantum mechanics, electromagnetism, particle physics and other fields, and is a matrix describing the relationship between input and output waves, which is usually used for analysis of multi-port networks. The scattering matrix element is understood as the relationship between the amplitude and phase of the input wave and the output wave of the system. Assuming that there is a scattering system involving multiple ports (such as the incident and scattering processes of particles and waves), for a scattering system containing (N) ports, the scattering matrix is an N × N matrix, denoted as (S), where each element S ijThe scattering matrix elements represent how the signals of input ports j affect the signals of output ports i, i.e., the scattering matrix elements describe the interaction between different ports, and the scattering matrix elements are distinguished by different incident wavelengths, scattering angles, and polarization directions.

[0029] By analyzing how different characteristic information of combustible particles affects the scattering matrix elements, more accurate discriminators can be provided for combustible particle classification. Modeling the relationship between the unique characteristic index of the particles and the scattering matrix elements can help distinguish different types of combustible particles, especially in complex environments, and can help improve the reliability and accuracy of identification. By considering multiple characteristic information (such as surface roughness, size, and chemical elements), information can be extracted from multiple dimensions, which can help improve the sensitivity of the system to changes in different combustible particles, and the robustness of the model will also be improved because it can better adapt to different particle materials and environmental conditions. By standardizing and combining different characteristic information, a comprehensive unique characteristic index can be obtained, which avoids over-reliance on individual characteristics, reduces data dimensionality, and enhances the representativeness of the characteristics. By sorting the relationship coefficients between the scattering matrix elements and the unique characteristic index, the most discriminative scattering matrix elements can be selected, reducing the impact of irrelevant or redundant information on model performance, and optimizing the efficiency and accuracy of the measurement system.

[0030] Specifically, by selecting the relevant parameters of the three scattering matrix elements as the input of the combustible discrimination model, the specific analysis of the combustible discrimination model is as follows: three unique physical indexes of the three scattering matrix elements are obtained according to the relevant parameters of the three scattering matrix elements, and the specific expression of the unique physical index is as follows: x = IW * κ1 + SA * κ2 + PD * κ3, where U x represents the unique physical index, xThe scattering matrix element number is represented by S, the incident wavelength is represented by IW, the scattering angle is represented by SA, the polarization direction is represented by PD, and κ1, κ2, and κ3 represent the weight values of the incident wavelength, the scattering angle, and the polarization direction, respectively. The specific weight values are obtained by manually setting the weights according to existing experimental data or the understanding of the mechanism of different scattering matrix elements in the literature. For example, if a certain related parameter is more important to the scattering matrix element, it can be given a greater weight. Machine learning algorithms such as support vector machines and neural networks can also be used to model training data, and the optimal weight of each related parameter can be automatically learned through an optimization algorithm. The weights are obtained by minimizing the prediction error such as classification error or regression error. The contributions of different related parameters to the discrimination results in multiple samples can also be analyzed, and statistical methods such as principal component analysis and correlation analysis can be used to evaluate the contribution of each physical index, thereby obtaining the corresponding weights. Based on the unique physical indicators of the three scattering matrix elements of different combustible particles as input, a combustible discrimination model is obtained, and the specific combustible discrimination model expression is: In the formula, P(y=flammable) represents the probability that the particulate matter is a combustible particle, and U1, U2, and U3 represent the unique physical indicators of the three scattering matrix elements, respectively.

[0031] In this embodiment, the elements of the scattering matrix can reflect the electromagnetic wave scattering characteristics of the material. By combining information of different angles, wavelengths, and polarization directions, the scattering characteristics of the particulate matter can be more comprehensively described, thereby improving the discrimination ability of combustibles. Each combustible particle exhibits unique scattering characteristics under different scattering conditions, and the unique physical indicators extracted from multiple scattering matrix elements can establish more distinctive models for different types of combustibles, thereby improving the accuracy of discrimination. The adjustment of different scattering angles, wavelengths, and polarization directions is flexible and adaptable to different experimental conditions and environmental changes, and has strong adaptability. By combining multiple scattering matrix elements, not only the characteristics of single data are utilized, but also multiple angle data are fully integrated, thereby improving the robustness and precision of combustible detection.

[0032] Specifically, the output result of the combustible matter discrimination model is optimized, and the minimum value of the output result of the combustible matter discrimination model is set, that is, the specific analysis of the combustible matter particle discrimination threshold is as follows: the output results of different combustible matter particles are obtained according to the combustible matter discrimination model, and non-combustible matter particle samples are obtained, the output results of different non-combustible matter particle samples are obtained according to the combustible matter discrimination model, and then the output results of different combustible matter particles and the output results of different non-combustible matter particle samples are combined into output probability samples; data with a value range of 0 to 1 is randomly generated as a traversal threshold, and the output probability samples are compared according to different traversal thresholds, and then the comparison results are output, and the comparison results are combined into a confusion matrix, and the specific confusion matrix includes the number of samples that are actually combustible matter particles and are determined to be combustible matter particles according to the traversal threshold, the number of samples that are actually combustible matter particles and are determined to be non-combustible matter particles according to the traversal threshold, the number of samples that are actually non-combustible matter particles and are determined to be combustible matter particles according to the traversal threshold, and the number of samples that are actually non-combustible matter particles and are determined to be non-combustible matter particles according to the traversal threshold; different confusion matrices are obtained according to different traversal thresholds, and the precision and recall rates corresponding to different traversal thresholds are obtained, and the harmonic mean of the precision and recall rates is obtained; the harmonic mean of the precision and recall rates corresponding to different traversal thresholds is compared, and the traversal threshold corresponding to the maximum harmonic mean of the precision and recall rates is marked as the minimum value of the output result of the combustible matter discrimination model, that is, the combustible matter particle discrimination threshold.

[0033] In the present embodiment, the non-combustible matter particle sample output result represents the prediction result of the model for the non-combustible matter particle sample, and a probability value is also output, which represents the possibility that the particle belongs to non-combustible matter. Specifically, the model is input with the non-combustible matter particle sample after training, and the probability value output by the model is obtained.

[0034] The traversal threshold represents a plurality of candidate thresholds for evaluating the classification result of the model, and the range is between 0 and 1. Each traversal threshold represents that the samples with a probability value greater than the threshold are determined to be combustible matter particles, and the samples with a probability value less than the threshold are determined to be non-combustible matter particles. A plurality of thresholds are randomly generated with a step size between 0 and 1, or the thresholds are defined according to the actual application requirements.

[0035] The confusion matrix specifically represents a matrix for evaluating the classification performance of the model, and records the comparison between the true label and the predicted label, including: true positives (TP), the number of samples that are actually combustible matter and are predicted to be combustible matter; false negatives (FN), the number of samples that are actually combustible matter but are predicted to be non-combustible matter; false positives (FP), the number of samples that are actually non-combustible matter but are predicted to be combustible matter; and true negatives (TN), the number of samples that are actually non-combustible matter and are predicted to be non-combustible matter. The prediction results of all samples according to different thresholds are compared to generate a confusion matrix.

[0036] Precision specifically measures the accuracy of the model in predicting flammable materials, calculated by the true positives and false positives in the confusion matrix, with the formula: P recision represents the precision.

[0037] Recall measures the model's ability to identify flammable materials, calculated by the true positives and false negatives in the confusion matrix, with the formula: R ecall represents the recall.

[0038] F1-score is used to evaluate the classification effect of the model, obtained by calculating the harmonic mean of precision and recall, with the formula: F1 represents the harmonic mean.

[0039] By traversing different thresholds and combining the confusion matrix, the classification effect of flammable and non-flammable materials can be evaluated, and the classification threshold of the model can be accurately adjusted to optimize the discrimination performance of the model. By finding the threshold that maximizes the harmonic mean of precision and recall (F1-score), the model can find the best balance between precision and recall; in practical applications, misjudgments of the model (such as judging non-flammable materials as flammable or judging flammable materials as non-flammable) may have serious consequences, and by optimizing the discrimination threshold of flammable particles, false positives and false negatives can be effectively reduced, thereby improving the robustness of the model; by traversing different thresholds, the sensitivity of the model can be adjusted according to different application requirements, for example, if the tolerance for misjudging non-flammable materials as flammable is low in some scenarios, a higher threshold can be selected to reduce the risk of misjudgment; through a large number of samples and different threshold combinations, the model can better adapt to different test data and have stronger generalization ability.

[0040] Specifically, real-time monitoring of on-site particulate matter is performed to obtain characteristic information of the on-site particulate matter, and then the characteristic information of the on-site particulate matter is used for screening to mark the specific analysis of the suspicious particulate matter: real-time monitoring of on-site particulate matter is performed to obtain characteristic information of the on-site particulate matter, and then the characteristic information of the on-site particulate matter is used to obtain a unique characteristic index of the on-site particulate matter; a range of unique characteristic indexes of flammable particles is obtained based on the characteristic information of the flammable particles, specifically, a light scattering characteristic database of flammable particles is established based on known unique characteristic indexes of flammable particles, and then the range of unique characteristic indexes of flammable particles is obtained based on the light scattering characteristic database, which can also be set by relevant professionals, and then the unique characteristic index of the on-site particulate matter is compared with the range of unique characteristic indexes of flammable particles to screen out on-site particulate matter with a unique characteristic index within the range of unique characteristic indexes of flammable particles and mark it as suspicious particulate matter.

[0041] In this embodiment, when monitoring particulate matter in real time, common devices include: optical sensors: particulate matter sensors using laser scattering principles can measure the concentration of particulate matter of different sizes in the air in real time; aerosol particle counters: used to detect the number and size distribution of suspended particulate matter in the air, which can accurately analyze the particle size, concentration, etc. of particulate matter; infrared sensors: use infrared spectral characteristics to monitor the characteristic wavelength band of particulate matter, which can detect the absorption characteristics of particulate matter and infer the composition of particulate matter.

[0042] This embodiment can adopt a three-emission-two-reception optical path design, and the specific device configuration includes a transmission end and a reception end for detecting the light scattering characteristics of particulate matter. Through multi-angle optical path design, more accurate measurement of particulate matter concentration and characteristic information can be achieved. The three-emission-two-reception optical path design is an application of optical particulate matter monitoring technology, which includes three different wavelength light sources that can respectively excite different light scattering characteristics of particulate matter. Different wavelengths of light can provide different optical characteristics of particulate matter, helping to distinguish the types and sizes of particulate matter. The light receiving end is used to receive the scattering and reflection of light by particulate matter. By setting two receiving ends, scattered light signals can be collected from different angles and directions, which can enhance the reliability of the signal and improve the accuracy of extracting particulate matter characteristic information.

[0043] Real-time monitoring of particulate matter characteristic information can quickly identify possible combustible particulate matter, which is of great significance for fire prevention and environmental pollution monitoring. By screening and marking suspicious particulate matter, potential fire hazards can be discovered in time, reducing the probability of accidents and improving on-site safety. Comparing the unique characteristic indicators of particulate matter can more accurately screen samples that meet the combustible particulate matter, thereby reducing false positives and false negatives. In cooperation with the automatic alarm system, the alarm is triggered based on real-time data or further manual confirmation, reducing the pressure of manual inspection.

[0044] Specifically, the target function is compared with the combustible particulate matter discrimination threshold to determine whether there is suspicious combustible particulate matter, and then the fire smoke alarm is triggered. The specific analysis is as follows: the related parameters of the three scattering matrix elements of the suspicious particulate matter are brought into the combustible discrimination model, and the target function of the suspicious particulate matter is output. The target function of the suspicious particulate matter is compared with the combustible particulate matter discrimination threshold. If the target function of the suspicious particulate matter is greater than or equal to the combustible particulate matter discrimination threshold, the suspicious particulate matter is determined to be suspicious combustible particulate matter, and the fire smoke alarm is triggered.

[0045] In the embodiment, by comparing the scattering matrix elements of the suspicious particles with the predetermined combustible particle identification threshold, it can be accurately identified which particles are suspicious combustible particles, avoiding the risk of false triggering of fire alarm and improving the accuracy of the alarm; the traditional fire smoke detector may frequently misreport due to the presence of more non-fire smoke particles (such as dust, insect debris, etc.) in the environment, and the method greatly reduces the probability of misreporting through specific scattering matrix analysis, not only considering the existence of particles, but also considering the properties and scattering characteristics of particles; the scattering matrix reflects the scattering characteristics of particles under light irradiation, which are directly related to the size, shape and composition of particles, and by analyzing these three parameters, the system can distinguish different types of particles (such as smoke, dust, etc.) from the physical layer, thereby more accurately determining which particles have combustible characteristics; by calculating the objective function of the suspicious particles, different particle types can be effectively matched with specific fire smoke characteristics, thereby identifying suspicious particles; without manual observation or manual intervention, the system automatically determines and responds, and as long as the objective function of the suspicious particles is greater than or equal to the combustible particle identification threshold, the system will automatically trigger the alarm. This intelligent automatic identification system not only improves the response speed and efficiency, but also reduces the risk of human negligence, which helps to take effective preventive measures at the early stage of fire and reduce fire losses.

[0046] In summary, the present application has at least the following effects:

[0047] Through SEM / AFM observation and chemical element composition analysis, the characteristic information of combustible particles can be accurately obtained, thereby realizing accurate identification of combustible particles; by simulating the relationship between the characteristic information of combustible particles and the scattering matrix elements, the influence degree of the change of the characteristic information on different scattering matrix elements can be identified, providing a scientific basis for subsequent model setting; based on the selected scattering matrix elements and related parameters selected based on the relationship simulation results, a combustible identification model is constructed, which can realize effective identification of on-site particles and improve the accuracy of fire detection; real-time monitoring of on-site particles and screening according to the characteristic information can mark suspicious particles, thereby realizing timely warning of fire smoke; by obtaining the scattering matrix element related parameters of suspicious particles and outputting the objective function according to the combustible identification model, it can be compared with the combustible particle identification threshold to quickly determine whether there are suspicious combustible particles and trigger the fire smoke alarm, thereby gaining valuable time for fire emergency treatment; the automatic identification of combustible particles through the combustible identification model reduces manual intervention and improves the intelligent level of fire prevention and control. Real-time monitoring and data analysis of on-site particles can realize the prediction and evaluation of fire risk and provide more scientific and effective decision support for fire prevention and control.

[0048] Those skilled in the art will appreciate that embodiments of the present application can be devised for a method. Accordingly, the present application can be embodied in the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) embodying computer readable program code.

[0049] The present application is described in reference to the flow diagrams according to the methods of embodiments of the present application. It will be understood that each block of the flow diagram, and combinations of blocks in the flow diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processing machine, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flow diagram block or blocks. Figure 1 The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flow diagram block or blocks.

[0050] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the function specified in the flow diagram block or blocks. Figure 1 The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flow diagram block or blocks.

[0051] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flow diagram block or blocks. Figure 1 The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flow diagram block or blocks.

[0052] While preferred embodiments of the application have been described, it should be apparent that a person of ordinary skill in the art can make modifications and variations to the described embodiments without departing from the spirit and scope of the application. It is, therefore, intended that the application be interpreted as including all such modifications and variations as fall within the scope of the claims and equivalents thereof.

[0053] Obviously, numerous modifications and variations of the present application are possible in light of the above teachings. It is therefore to be understood that within the scope of the claims and their equivalents, the application can be practiced otherwise than as specifically described.

Claims

1. A fire smoke detection method with a combustible material recognition capability, characterized by, The method comprises the following steps: Obtaining characteristic information of combustible particles based on SEM / AFM observation and chemical element composition analysis; Simulating the relationship between the characteristic information of combustible particles and scattering matrix elements, and identifying the influence degree of the change of the characteristic information of combustible particles on different scattering matrix elements; Selecting three scattering matrix elements based on the simulation results, and obtaining the related parameters of the selected three scattering matrix elements; Setting a combustible discrimination model by taking the related parameters of the selected three scattering matrix elements as inputs of the combustible discrimination model; Optimizing the output results of the combustible discrimination model, and then setting the minimum value of the output results of the combustible discrimination model as a combustible particle discrimination threshold; Real-time monitoring of on-site particulate matter, obtaining the characteristic information of on-site particulate matter, and then screening and marking suspicious particulate matter according to the characteristic information of on-site particulate matter; Obtaining the related parameters of the three scattering matrix elements of the suspicious particulate matter, and then outputting a target function of the suspicious particulate matter according to the combustible discrimination model; Comparing the target function with the combustible particle discrimination threshold to determine whether there is suspicious combustible particle, and then triggering a fire smoke alarm.

2. The fire smoke detection method with combustible material recognition capability according to claim 1, characterized in that, The specific analysis of simulating the relationship between the characteristic information of combustible particles and scattering matrix elements, and identifying the influence degree of the change of the characteristic information of combustible particles on different scattering matrix elements comprises: Obtaining the characteristic information of combustible particles, and obtaining a unique characteristic index of combustible particles based on the characteristic information; A relationship between a unique characteristic index of the combustible particles and a scattering matrix element of the combustible particles is simulated, and a relationship between the scattering matrix element and the unique characteristic index is constructed: S meofsp = λ * A uciffsp , wherein S meofsp represents the scattering matrix element, A uciffsp represents the unique characteristic index, and λ represents a relationship coefficient between the scattering matrix element and the unique characteristic index. Substituting the unique characteristic index of different combustible particles and the scattering matrix elements of each different combustible particle into a relationship between the scattering matrix elements and the unique characteristic index to obtain the relationship coefficients of the scattering matrix elements of each different combustible particle and the unique characteristic index of different combustible particles, and then marking the average of the relationship coefficients of the scattering matrix elements of each different combustible particle and the unique characteristic index of different combustible particles as the relationship coefficients of each scattering matrix element and the unique characteristic index, which are used to represent the influence degree of the change of the characteristic information of combustible particles on different scattering matrix elements.

3. The fire smoke detection method with combustible material recognition capability according to claim 2, characterized in that, The specific analysis of obtaining the characteristic information of combustible particles, and obtaining a unique characteristic index of combustible particles based on the characteristic information comprises: The characteristic information of the combustible particles specifically includes surface roughness, size, and main chemical elements and their relative concentrations; Standardizing the surface roughness, size, and main chemical elements and their relative concentrations of combustible particles, and then combining the surface roughness, size, and main chemical elements and their relative concentrations to obtain the unique characteristic index of combustible particles.

4. The fire smoke detection method with combustible material recognition capability according to claim 2, wherein, The specific analysis of selecting three scattering matrix elements based on the simulation results, and obtaining the related parameters of the selected three scattering matrix elements comprises: Comparing the relationship coefficients of each scattering matrix element and the unique characteristic index, and then sorting the scattering matrix elements according to the relationship coefficients to select the top three scattering matrix elements as the three scattering matrix elements. The correlation parameters of the selected three scattering matrix elements specifically include an incident wavelength of the scattering matrix element, a scattering angle of the scattering matrix element, and a polarization direction.

5. The fire smoke detection method with combustible material recognition capability according to claim 1, wherein, The specific analysis of setting the combustible substance discrimination model by taking the correlation parameters of the selected three scattering matrix elements as inputs of the combustible substance discrimination model is as follows: According to the correlation parameters of the three scattering matrix elements, unique physical indexes of the three scattering matrix elements are obtained, and then according to the unique physical indexes of the three scattering matrix elements of different combustible particles as input, a combustible discrimination model is obtained, and a specific combustible discrimination model expression is as follows: In the formula, P(y=flammable) represents the probability of the particulate matter being a combustible particle, U1, U2 and U3 represent the unique physical indexes of the three scattering matrix elements respectively.

6. The fire smoke detection method with combustible material recognition capability according to claim 1, wherein, The specific analysis of optimizing the output result of the combustible substance discrimination model and then setting a minimum value of the output result of the combustible substance discrimination model, i.e., a combustible substance particle discrimination threshold, is as follows: According to the output result of the combustible substance discrimination model, output results of different combustible substance particles are obtained, and non-combustible substance particle samples are obtained. According to the output result of the combustible substance discrimination model, output results of different non-combustible substance particle samples are obtained, and then the output results of different combustible substance particles and the output results of different non-combustible substance particle samples are combined into output probability samples. Data randomly generated in a range of 0 to 1 are taken as traversal thresholds, and the output probability samples are compared according to different traversal thresholds, and then comparison results are output. The comparison results are combined into a confusion matrix. The specific confusion matrix includes the number of samples that are actually combustible substance particles and are determined as combustible substance particles according to the traversal thresholds, the number of samples that are actually combustible substance particles and are determined as non-combustible substance particles according to the traversal thresholds, the number of samples that are actually non-combustible substance particles and are determined as combustible substance particles according to the traversal thresholds, and the number of samples that are actually non-combustible substance particles and are determined as non-combustible substance particles according to the traversal thresholds. According to different confusion matrices obtained by comparison according to different traversal thresholds, the precision and recall rates corresponding to different traversal thresholds are obtained, and a harmonic mean of the precision and recall rates is obtained. The harmonic means of the precision and recall rates corresponding to different traversal thresholds are compared, and the traversal threshold corresponding to the maximum harmonic mean of the precision and recall rates is marked as the minimum value of the output result of the combustible substance discrimination model, i.e., the combustible substance particle discrimination threshold.

7. The fire smoke detection method with combustible material recognition capability according to claim 2, wherein, The specific analysis of monitoring the on-site particles in real time, obtaining characteristic information of the on-site particles, and then screening and marking suspicious particles according to the characteristic information of the on-site particles is as follows: The characteristic information of the on-site particles is obtained based on the real-time monitoring of the on-site particles, and then the unique characteristic index of the on-site particles is obtained according to the characteristic information of the on-site particles. The unique characteristic index of the combustible substance particles is obtained according to the characteristic information of the combustible substance particles, and then the unique characteristic index of the on-site particles is compared with the range of the unique characteristic index of the combustible substance particles. The on-site particles whose unique characteristic index is within the range of the unique characteristic index of the combustible substance particles are screened out and marked as suspicious particles.

8. The fire smoke detection method with combustible material recognition capability according to claim 1, wherein, The specific analysis of comparing the objective function with the combustible substance particle discrimination threshold to determine whether there is a suspicious combustible substance particle, and then triggering a fire smoke alarm is as follows: The correlation parameters of the three scattering matrix elements of the suspicious particles are input into the combustible substance discrimination model, and the objective function of the suspicious particles is output. The target function of the suspicious particle is compared with the combustible particle discrimination threshold, and if the target function of the suspicious particle is greater than or equal to the combustible particle discrimination threshold, the suspicious particle is determined to be a suspicious combustible particle, triggering a fire smoke alarm.

Images