A cloud service-based intelligent fire-fighting big data analysis method

Abstract

The present application relates to a kind of wisdom fire-fighting big data analysis methods based on cloud service, comprising the following steps: obtaining the time series data of multiple-source gas concentration that fire scene is collected by gas sensor array, obtain smooth gas concentration time series curve;According to gas concentration time series curve, the concentration change of different gas is analyzed;Establish the coupling relationship matrix of environmental parameter and chemical reaction chain;Through coupling relationship matrix, the diffusion speed and mixing degree of different gas under different environmental conditions are calculated;Identify the typical gas concentration change mode caused by different types of dangerous source, generate preliminary estimation result of dangerous source type and location;Analysis of the type and location of dangerous source in current fire scene, generate real-time positioning information of dangerous source;Real-time adjust coupling coefficient, and generate dynamic adjustment dangerous source identification result.The present application realizes the accurate, real-time and dynamic identification of dangerous source in fire scene by fusing multiple-source data, chemical reaction chain model and machine learning algorithm.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of data analysis, in particular to a wisdom fire-fighting big data analysis method based on cloud service. BACKGROUND

[0002] In the component feature research of gas sensor array in wisdom fire-fighting, there is a key technical problem. In the fire scene, the concentrations of multiple gases will present complex change law with time, especially when the concentration of a certain type of gas appears sharp fluctuation, how to accurately identify the dangerous source becomes a big challenge.

[0003] The sharp fluctuation of gas concentration may be caused by various reasons, such as the change of burning material, the spread of fire, the occurrence of chemical reaction, etc., it is difficult to accurately determine the type and location of dangerous source simply by relying on the change trend of gas concentration, at this time, how to combine the time series change law of gas concentration with the chemical reaction chain becomes the key to the problem; different burning materials will produce different gas components in the combustion process, and there may be complex chemical reactions between these gas components, if the characteristics of chemical reaction chain can be deeply analyzed and matched with the change law of gas concentration, it is possible to realize the accurate identification of dangerous source; however, the analysis of chemical reaction chain needs to consider various factors, such as temperature, pressure, catalyst, etc., which further increases the complexity of the problem, how to real-time process and analyze sensor data in complex environment and timely generate reliable fire-fighting tactics report is also a big technical challenge; this needs to comprehensively consider the timeliness, accuracy and interpretability of data, and combine with the experience and knowledge of fire-fighting experts, so as to provide strong support for fire-fighting decision. SUMMARY

[0004] In order to solve the problems existing in the prior art, the present application aims to provide a wisdom fire-fighting big data analysis method based on cloud service. The wisdom fire-fighting big data analysis method based on cloud service realizes the accurate, real-time and dynamic identification of dangerous source in fire scene by fusing multi-source data, chemical reaction chain model and machine learning algorithm.

[0005] The wisdom fire-fighting big data analysis method based on cloud service provided by the present application comprises the following steps:

[0006] S1, acquiring multi-source gas concentration time series data collected by gas sensor array in fire scene, and respectively pre-processing the concentration change trend of different gases to remove noise interference, and obtaining smooth gas concentration time series curve;

[0007] S2, according to the gas concentration time series curve, analyzing the concentration change of different gases, combining with the pre-established chemical reaction chain model, if there is a chemical reaction of target gas, generating the mapping relationship of chemical reaction chain related to target gas;

[0008] S3, obtaining environmental parameters of the fire scene, establishing a coupling relationship matrix of the environmental parameters and the chemical reaction chain in combination with a mapping relationship of the chemical reaction chain, and determining an influence of the environmental parameters on gas concentration fluctuation;

[0009] S4, calculating diffusion speeds and mixing degrees of different gases under different environmental conditions through the coupling relationship matrix, generating a predicted value of target gas concentration fluctuation, and analyzing a deviation degree of the predicted value of the target gas concentration fluctuation from an actual measured value;

[0010] S5, if the target gas is related to the position and type of the hazard source, extracting a relationship between gas concentration fluctuation and the hazard source from historical fire data, identifying typical gas concentration change patterns caused by different types of hazard sources, training a classification model, and generating a preliminary estimation result of the type and position of the hazard source;

[0011] S6, analyzing the type and position of the hazard source of the current fire scene according to the preliminary estimation result in combination with real-time collected gas concentration time series data, and generating real-time positioning information of the hazard source;

[0012] S7, updating the predicted value of the gas concentration fluctuation through the real-time positioning information, adjusting a coupling coefficient of the environmental parameters and the chemical reaction chain in real time in combination with a change trend of the environmental factors, and generating a dynamically adjusted hazard source identification result.

[0013] Preferably, the step S1 specifically comprises:

[0014] obtaining various types of gas concentration time series data from a gas sensor array, and performing normalization processing on the concentration time series data according to sensitivity coefficients of various types of gas sensors to obtain first gas concentration data;

[0015] performing smoothing processing on the first gas concentration data by using a sliding mean filter, and performing wavelet transform denoising on data points exceeding a preset noise interference amplitude threshold according to the preset noise interference amplitude threshold to obtain second gas concentration data;

[0016] performing fitting on discrete data points in a time series segment by using a cubic spline interpolation function to obtain third gas concentration data from the second gas concentration data;

[0017] calculating a first concentration change gradient value according to a concentration threshold value set according to physical characteristics of various types of gases, and performing smoothing processing on the gradient value by using a Gaussian kernel function to obtain a smoothed concentration time series curve of each type of gas from the third gas concentration data.

[0018] Preferably, the step S2 specifically comprises:

[0019] Acquire target gas concentration data according to the gas concentration time series curve, calculate similarity values ​​between target gas molecules and other gas molecules using molecular structure feature vectors stored in a chemical reaction chain database, and obtain candidate chemical reaction chain data;

[0020] For the candidate chemical reaction chain data, a pre-established gas phase reaction rate constant database is used to calculate the reaction rate constant value of each elementary reaction step in the reaction chain within a temperature range to obtain reaction chain kinetic data;

[0021] Calculating the Pearson correlation coefficient between the reactant concentration change value and the measured gas concentration time series curve within the temperature range based on the reaction chain kinetics data, and grouping the reaction chains whose correlation coefficient is greater than a coefficient threshold using a hierarchical clustering algorithm to obtain reaction chain correlation data;

[0022] For the reaction chain associated data, the standard formation enthalpy change and standard entropy change values ​​of the reactants and products are obtained from the thermodynamic database, the chemical reaction equilibrium constant of the reaction chain under temperature and pressure conditions is calculated, and a reaction chain mapping model is established through the support vector regression method.

[0023] Preferably, the step S3 specifically includes:

[0024] The effects of environmental parameters on gas concentration fluctuations include gas reaction rate, gas diffusion rate, and gas mixing degree;

[0025] A sensor network is used to obtain temperature data, humidity data, and pressure data at the fire scene, and a radial basis function is used to calculate a temperature distribution field, a humidity distribution field, and a pressure distribution field based on the temperature data, the humidity data, and the pressure data.

[0026] Extracting the activation energy value and the pre-exponential factor value of the reactants from the temperature distribution field, the humidity distribution field, and the pressure distribution field through a chemical reaction chain mapping relationship, and obtaining a reaction rate influence coefficient based on the activation energy value and the pre-exponential factor value in combination with the Arrhenius equation;

[0027] According to the reaction rate influence coefficient, the gas diffusion flux value is calculated using Fick's law, and a deep neural network model is established based on the gas diffusion flux value to obtain gas diffusion velocity distribution data;

[0028] The Shannon entropy value is calculated for the gas diffusion velocity distribution data through the gas concentration time series curve, and a multivariate autoregressive model is established based on the Shannon entropy value to obtain the environmental parameter influence coefficient matrix.

[0029] Preferably, the step S4 specifically includes:

[0030] According to the coupling relationship matrix, the numerical values of the temperature field, the pressure field and the humidity field on the spatial grid points are obtained, and the concentration gradient data of the target gas in the three-dimensional space are calculated through the numerical values;

[0031] The mixing entropy value of the target gas and the environmental gas is calculated by using the concentration gradient data, and the gas concentration prediction data are obtained by establishing a Bayesian network containing three nodes of temperature, pressure and humidity through the mixing entropy value;

[0032] The gas concentration prediction data are divided into a plurality of time windows, the deviation feature vector of the predicted value and the measured value is calculated by using the time window, and the deviation feature data are obtained by grouping the deviation feature vector through a hierarchical clustering algorithm;

[0033] According to the deviation feature data, it is judged whether the standard deviation exceeds a preset upper limit of the standard deviation threshold value, if the preset upper limit of the standard deviation threshold value is exceeded, the frequency spectrum feature of the deviation feature data is extracted by using the Fourier transform, and the target gas concentration change mode is judged through the frequency spectrum feature.

[0034] Preferably, the step S5 specifically comprises:

[0035] The time series data of the gas concentration are obtained according to the three-dimensional spatial coordinates of the hazard source, the low-frequency approximation coefficients and the high-frequency detail coefficients of the time series data are extracted on a plurality of decomposition scales by using the wavelet transform, and the first feature data are obtained;

[0036] For the first feature data, a mapping relationship between the first feature data and the hazard source type is established by using a random forest classifier, the classifier parameters are optimized through cross-validation, and the second classification data are obtained;

[0037] The Euclidean distance matrix between the gas concentration fluctuation features of different types of hazard sources is calculated according to the second classification data, the hazard sources are grouped by using a spectral clustering algorithm, and the third template data are obtained by extracting a typical fluctuation template of each group.

[0038] If the Pearson correlation coefficient between the measured gas concentration curve and the third template data is greater than a preset Pearson correlation coefficient judgment threshold value, the posterior probability of the hazard source position is calculated by using a naive Bayes method, and the spatial position area of the hazard source and the type of hazardous material are determined.

[0039] Preferably, the step S6 specifically comprises:

[0040] The concentration values between the gas concentration sampling points are spatially interpolated by using a bicubic interpolation algorithm to obtain continuous concentration distribution field data;

[0041] The continuous concentration distribution field data are filtered by using a Gaussian kernel function to obtain first filtered field data;

[0042] According to the first filtering field data division, a concentric circle search area centered on the initial position of the danger source is divided, a particle swarm algorithm is used to iteratively search the concentration gradient in the search area, and second gradient field data is obtained;

[0043] The second gradient field data is calculated to obtain the statistical moment of the gas concentration in the search area, a hidden Markov chain is used to estimate the state transition of the danger source position coordinates and type identification, and the spatial coordinates of the danger source are determined.

[0044] Preferably, the step S7 specifically comprises:

[0045] The real-time adjustment of the coupling coefficient of the environmental parameter and the chemical reaction chain includes adjusting the reaction rate constant;

[0046] The generated dynamic adjustment of the danger source identification result includes the danger source type, position and danger level information;

[0047] According to the measured values of the environmental parameters collected by the temperature sensor, the pressure sensor and the humidity sensor, an environmental change trend vector is constructed, and the future environmental parameter change prediction value is obtained through the environmental change trend vector;

[0048] According to the future environmental parameter change prediction value, the reaction rate constant is calculated by using the temperature correction term and the pressure correction term, and the gas concentration prediction value is obtained by using the reaction rate constant and the recursive least square method to establish a Kalman filter;

[0049] For the gas concentration prediction value, the gradient vector and the Hessian matrix of the concentration field are calculated, and the support vector regression with the kernel function as the radial basis is used to obtain the danger source position coordinates;

[0050] For the danger source position coordinates, the posterior distribution of the danger source type is updated by using Bayesian probability estimation, and the danger level is obtained according to the hazard degree index vector constructed by the gas concentration exceeding multiple, the spread speed and the diffusion range.

[0051] The intelligent fire-fighting big data analysis method based on cloud service has the advantages that:

[0052] The cloud service-based intelligent fire-fighting big data analysis method can remove noise interference and obtain a smooth gas concentration time sequence curve, thereby improving the accuracy of data; in combination with a chemical reaction chain model and environmental parameters, the diffusion speed and mixing degree of different gases under different environmental conditions can be calculated, and a prediction value of target gas concentration fluctuation is generated, which helps firefighters to understand the gas concentration change situation of the fire scene in advance and make more accurate judgments; by analyzing the relationship between gas concentration fluctuation and the hazard source, a typical gas concentration change mode caused by different types of hazard sources can be identified, and a classification model is trained to generate a preliminary estimation result of the hazard source type and location, which greatly shortens the identification time of the hazard source; according to the preliminary estimation result and the real-time collected gas concentration time sequence data, the hazard source type and location of the current fire scene can be further analyzed, and real-time positioning information of the hazard source is generated, which provides strong support for rapid positioning of the fire scene and elimination of the hazard source; by integrating advanced technologies such as the Internet of Things, big data, cloud computing and the like, comprehensive perception and intelligent analysis of the fire scene are realized, which helps to promote the intelligent development of fire-fighting work and improve the efficiency and accuracy of fire-fighting work. The present application realizes accurate, real-time and dynamic identification of the hazard source of the fire scene by integrating multi-source data, a chemical reaction chain model and a machine learning algorithm. BRIEF DESCRIPTION OF DRAWINGS

[0053] Figure 1 is a flow chart of the cloud service-based intelligent fire-fighting big data analysis method. DETAILED DESCRIPTION

[0054] As Figure 1 shown, the cloud service-based intelligent fire-fighting big data analysis method includes the following steps:

[0055] S1, acquiring multi-source gas concentration time sequence data collected by a gas sensor array at a fire scene, and respectively preprocessing the concentration change trend of different gases to remove noise interference and obtain a smooth gas concentration time sequence curve;

[0056] S2, analyzing the concentration change of different gases according to the gas concentration time sequence curve, and combining a pre-established chemical reaction chain model, if there is a chemical reaction of the target gas, a mapping relationship of the chemical reaction chain related to the target gas is generated;

[0057] S3, acquiring environmental parameters of the fire scene, combining the mapping relationship of the chemical reaction chain, establishing a coupling relationship matrix of the environmental parameters and the chemical reaction chain, and determining the influence of the environmental parameters on the gas concentration fluctuation;

[0058] S4. Calculate the diffusion speed and mixing degree of different gases under different environmental conditions through the coupling matrix, generate the predicted value of the target gas concentration fluctuation, and analyze the deviation degree of the predicted value of the target gas concentration fluctuation and the actual measured value;

[0059] S5. If the target gas is related to the position and type of the hazard source, extract the relationship between the gas concentration fluctuation and the hazard source from historical fire data, identify the typical gas concentration change mode caused by different types of hazard sources, train a classification model, and generate a preliminary estimation result of the type and position of the hazard source;

[0060] S6. According to the preliminary estimation result, combine the real-time collected gas concentration time series data, analyze the type and position of the hazard source in the current fire scene, and generate real-time positioning information of the hazard source;

[0061] S7. Through the real-time positioning information, update the predicted value of the gas concentration fluctuation, combine the change trend of the environmental factors, real-time adjust the coupling coefficient of the environmental parameters and the chemical reaction chain, and generate a dynamically adjusted hazard source identification result.

[0062] Further, in the embodiment, step S1 specifically includes:

[0063] Obtain various types of gas concentration time series data from the gas sensor array, normalize the concentration time series data according to the sensitivity coefficients of various types of gas sensors to obtain first gas concentration data;

[0064] Smooth the first gas concentration data using a sliding mean filter, and perform wavelet transform denoising on data points exceeding the preset noise interference amplitude threshold according to the preset noise interference amplitude threshold to obtain second gas concentration data;

[0065] Fit the discrete data points in the time series segment using a cubic spline interpolation function to obtain third gas concentration data;

[0066] According to the concentration threshold set according to the physical properties of various types of gases, calculate the first-order concentration change gradient value, and smooth the gradient value using a Gaussian kernel function to obtain the smooth concentration time series curve of each type of gas;

[0067] Specifically, carbon monoxide, carbon dioxide, ammonia and nitrogen oxide concentration data are collected from the gas sensor array, and the gas concentration values of a plurality of time series points are obtained at each sensor position according to the sampling frequency of ten times per second. Then, the collected concentration data is normalized according to the sensitivity coefficients of various types of gas sensors to obtain first gas concentration data;

[0068] For the first gas concentration data, a sliding mean filter with a window length of five data points is used for preliminary smoothing, and when the concentration change rate between adjacent data points exceeds the preset noise disturbance amplitude threshold of each type of gas, the fourth-order wavelet transform is used to denoise the abnormal data points to obtain the second gas concentration data;

[0069] For the second gas concentration data, each time sequence segment is divided according to every ten seconds according to the data acquisition period, and a cubic spline interpolation function is used to fit the discrete data points in each time sequence segment. The cubic spline interpolation function uses natural boundary conditions to obtain the third gas concentration data;

[0070] For the third gas concentration data, the concentration threshold is set according to the physical properties of each type of gas for segmentation, and the first-order concentration change gradient value is calculated in each time sequence step;

[0071] For the first-order concentration change gradient value, a Gaussian kernel function with a standard deviation of zero point five is used to perform final smoothing on the gradient curve to obtain the smoothed concentration time curve of each type of gas;

[0072] The example is as follows:

[0073] In fire scene gas monitoring, the concentration change trend of carbon monoxide, carbon dioxide, ammonia, nitrogen oxides and other harmful gases has an important influence on the safety protection of firefighters;

[0074] The gas sensor array is arranged at different positions in the fire scene, and 10 data points are collected per second in a high temperature environment. Due to the interference of temperature and humidity, the original gas concentration data collected by the sensor often fluctuates and has noise;

[0075] For carbon monoxide gas sensors, the sensitivity coefficient is 2.5, and when the measured carbon monoxide concentration is 500ppm, the normalized value is 0.5. Other gases are also normalized according to their respective sensitivity coefficients;

[0076] In the sliding mean filter processing, the normalized gas concentration data is slid with a window length of 5 data points, and the average value of the data in the window is calculated to smooth the data fluctuations. For example, the original values of carbon monoxide concentration in the continuous 5 data points are 0.48, 0.52, 0.49, 0.51 and 0.50, and the value obtained after sliding average is 0.50;

[0077] When the concentration change amplitude of a certain data point exceeds 0.1, the fourth-order wavelet transform is used to denoise the abnormal point, thereby eliminating the data anomalies caused by temperature fluctuations;

[0078] In the data segmentation processing, every 10 seconds of data is taken as a time sequence segment, and a cubic spline interpolation function is used to fit 100 discrete data points in the time segment;

[0079] The interpolation function adopts a natural boundary condition at the boundary of the sequence segment, i.e., the second derivative is zero, which ensures the smooth transition of the curve at the boundary;

[0080] For example, in the time sequence segment from 0 to 10 seconds, the carbon monoxide concentration rises from 0.2 to 0.8, and the continuous concentration change curve in this time segment is obtained by cubic spline interpolation;

[0081] In the gas concentration segmentation threshold setting, the thresholds are set according to the physical properties of carbon monoxide, carbon dioxide and other gases, and the concentration threshold of carbon monoxide is set to 0.6 and the concentration threshold of carbon dioxide is set to 0.8. When calculating the first order gradient, the concentration change rate between adjacent two data points reflects the speed of gas diffusion;

[0082] A Gaussian kernel function with a standard deviation of 0.5 is used to smooth the gradient curve to obtain a more smooth concentration change trend curve, which eliminates the slight fluctuations introduced in the data acquisition and processing process. After normalization, denoising, interpolation and smoothing processing of the original gas concentration data, the time sequence curve which can accurately reflect the change law of various gas concentrations in the fire scene is obtained, providing reliable data support for the safety evaluation of the fire scene.

[0083] Further, in the embodiment, step S2 specifically comprises:

[0084] According to the gas concentration time sequence curve, target gas concentration data is obtained, the similarity value between the target gas molecules and other gas molecules is calculated through the molecular structure feature vector stored in the chemical reaction chain database, and candidate chemical reaction chain data is obtained;

[0085] For the candidate chemical reaction chain data, the reaction rate constant values of each elementary reaction step in the reaction chain in the temperature interval are calculated by using the pre-established gas phase reaction rate constant database, and reaction chain kinetics data are obtained;

[0086] According to the reaction chain kinetics data, the Pearson correlation coefficient of the reactant concentration change value in the temperature interval and the measured gas concentration time sequence curve is calculated, the hierarchical clustering algorithm is used to group the reaction chains with a correlation coefficient greater than a coefficient threshold, and reaction chain association data are obtained;

[0087] For the reaction chain association data, the standard enthalpy of formation and standard entropy change values of the reactants and products are obtained from the thermodynamic database, the chemical reaction equilibrium constant of the reaction chain under the temperature and pressure conditions is calculated, and a reaction chain mapping model is established by using the support vector regression method;

[0088] Specifically, the concentration data of the target gas including carbon monoxide, carbon dioxide, ammonia and nitrogen oxides is extracted from the gas concentration time curve, the similarity between the target gas molecules and other gas molecules is calculated by the molecular structure feature vector stored in the chemical reaction chain database, a plurality of candidate chemical reaction paths in which the target gas acts as a reactant or product are formed, and the reaction chains are sorted according to the conversion rate of the reactants to obtain first reaction chain structure data;

[0089] The first reaction chain structure data is filtered according to the activation energy required for bond breaking of the reactant molecules in each reaction chain, the reaction rate constant of each elementary reaction step in each reaction chain within the temperature range of 298 to 1500 Kelvin is calculated by using a pre-established gas phase reaction rate constant database, and the reaction rate value is calculated by combining the reactant concentration to generate second reaction chain kinetic data;

[0090] The second reaction chain kinetic data is correlated based on the change value of the reactant concentration within the temperature range of 298 to 1500 Kelvin and the measured gas concentration time curve, the correlation coefficient is calculated by using the Pearson correlation coefficient, and the reaction chains with a correlation coefficient greater than 0.8 are grouped by using a hierarchical clustering algorithm to obtain third reaction chain correlation data;

[0091] The third reaction chain correlation data is used to obtain the standard enthalpy of formation and standard entropy of the reactants and products from a thermodynamic database, and the chemical reaction equilibrium constant of each reaction chain under the current temperature and pressure conditions is calculated;

[0092] The product concentration under the theoretical equilibrium state is calculated according to the reaction equilibrium constant, the root mean square error is used to evaluate the degree of agreement between the theoretical concentration and the measured concentration data, and the support vector regression method with a kernel function as a radial basis function is used to establish the mapping relationship between the reaction chain and the target gas;

[0093] In a fire scene, chemical reactions between multiple gases form a complex reaction network, and it is of great significance to identify key reaction chains by analyzing the gas concentration change trend for fire development trend judgment, and the chemical reaction chain database stores gas molecular structure feature vectors;

[0094] Examples are as follows:

[0095] Taking a carbon monoxide molecule as an example, its feature vector includes information such as the carbon-oxygen single bond length of 124 picometers and the molecular polarity moment of 0.112 debye, and by calculating the cosine similarity of the feature vectors of the target gas and other gas molecules, the molecules with a similarity greater than 0.9 are likely to have a chemical reaction;

[0096] In the screening of reaction chains, the activation energy required for the breaking of molecular bonds is an important reference index. For example, the activation energy for the breaking of the carbon-oxygen single bond in a carbon monoxide molecule is 1076 kilojoules per mole, and the activation energy for the breaking of the carbon-oxygen double bond in a carbon dioxide molecule is 532 kilojoules per mole.

[0097] The gas-phase reaction rate constant database records the reaction rate constants at different temperatures. For example, the reaction rate constant of carbon monoxide with oxygen at 298 kelvins is 2.3 x 10 -4 cubic meters per mole per second, and as the temperature increases to 1500 kelvins, the reaction rate constant increases to 3.8 x 10 -2 cubic meters per mole per second.

[0098] Based on the measured time-series curves of gas concentrations, the Pearson correlation coefficient between the change trend of reactant concentrations and the theoretically predicted values can be calculated. For the reaction of carbon monoxide oxidizing to carbon dioxide, when the correlation coefficient is greater than 0.8, it indicates that the reaction indeed occurs in the fire scene.

[0099] The hierarchical clustering algorithm classifies reaction chains with close correlation coefficients into the same class. For example, the oxidation reactions of carbon monoxide and ammonia both belong to exothermic reactions and can be classified into the same class.

[0100] The thermodynamic database provides the thermodynamic parameters of reactants and products. For example, the standard enthalpy of formation of carbon monoxide is -110.5 kilojoules per mole, and the standard entropy is 197.6 joules per mole kelvin. The standard enthalpy of formation of carbon dioxide is -393.5 kilojoules per mole, and the standard entropy is 213.7 joules per mole kelvin.

[0101] According to these parameters, the equilibrium constant of the reaction at different temperatures can be calculated, and then the product concentration at the theoretical equilibrium state can be obtained.

[0102] When the root mean square error between the theoretical concentration and the measured concentration is less than 10%, the support vector regression with a radial basis kernel function is used to establish the mapping relationship of the reaction chain, the scaling parameter of the kernel function is set to 0.1, and the penalty coefficient is set to 100, to achieve the balance between accuracy and generalization of reaction chain prediction.

[0103] Further, in this embodiment, step S3 specifically includes:

[0104] The influence of environmental parameters on gas concentration fluctuations includes gas reaction rate, gas diffusion speed, and gas mixing degree.

[0105] Temperature data, humidity data, and pressure data of the fire scene are obtained by using a sensor network. According to the temperature data, humidity data, and pressure data, the temperature distribution field, humidity distribution field, and pressure distribution field are calculated by using a radial basis function.

[0106] The activation energy value and the pre-exponential factor value of the reactants are extracted from the temperature distribution field, the humidity distribution field and the pressure distribution field through the chemical reaction chain mapping relationship, and the reaction rate influence coefficient is obtained according to the activation energy value and the pre-exponential factor value combined with the Arrhenius equation;

[0107] According to the reaction rate influence coefficient, the gas diffusion flux value is calculated by using Fick's law, and the gas diffusion velocity distribution data is obtained by establishing a deep neural network model through the gas diffusion flux value;

[0108] According to the gas diffusion velocity distribution data, the Shannon entropy value is calculated through the gas concentration time series curve, and the environmental parameter influence coefficient matrix is obtained by establishing a multiple autoregressive model according to the Shannon entropy value;

[0109] Specifically, the multi-point time series data of temperature, humidity and pressure are collected from the fire scene sensor network, the three-dimensional interpolation algorithm based on radial basis function is used to perform spatial interpolation on the sampling point data, the temperature distribution field, the humidity distribution field and the pressure distribution field are constructed, and the first environmental field data is obtained by calculating the three-dimensional gradient vector of each distribution field;

[0110] Based on the chemical reaction chain mapping relationship, the activation energy and the pre-exponential factor of the reactants are extracted from the first environmental field data, the influence coefficient of temperature on the reaction rate constant is calculated by combining the Arrhenius equation, the correction factor of pressure on the collision frequency of gas molecules is calculated by using the state equation, and the second reaction kinetics data is generated;

[0111] According to the second reaction kinetics data, the diffusion flux of gas under the action of temperature gradient, pressure gradient and humidity gradient is calculated by using Fick's first law and second law, a deep neural network containing three convolutional layers and two fully connected layers is established to predict the distribution of gas diffusion velocity in space, and the third diffusion field data is obtained;

[0112] According to the third diffusion field data, the Shannon entropy is calculated as the gas mixing degree index according to the gas concentration time series curve, and a multiple autoregressive model with temperature, pressure and humidity as independent variables is established to quantify the influence coefficient of environmental parameters on the gas mixing degree;

[0113] The coupling relationship matrix of environmental parameters and chemical reaction chain is constructed by combining the second reaction kinetics data, the third diffusion field data and the gas mixing degree index, the main influencing factors are extracted by using the singular value decomposition method, and the influence strength of environmental parameters on gas concentration fluctuation is judged;

[0114] The example is as follows:

[0115] There is a complex coupling relationship between the environmental parameters and the chemical reaction chain at the fire scene. The sensor network is arranged at different positions in the fire scene to collect environmental parameters. Taking the temperature sensor as an example, 27 measuring points are uniformly arranged in a space of 3m x 3m x 3m, and each measuring point collects 1 temperature value per second, with a measurement range of 298 to 1500 Kelvin;

[0116] When performing three-dimensional interpolation using a radial basis function, a Gaussian kernel function is selected as the basis function, and the influence radius is set to 1 meter. The continuous temperature distribution field of the entire space is obtained through interpolation;

[0117] The relationship between the reaction rate of the gas and the temperature follows the Arrhenius equation. Taking the oxidation reaction of carbon monoxide as an example, the activation energy is 125 kilojoules per mole, and the pre-exponential factor is 4.6 x 10 11 cubic meters per mole per second;

[0118] When the temperature rises from 298 Kelvin to 1500 Kelvin, the reaction rate constant increases by about 5 orders of magnitude. The effect of pressure on the reaction rate mainly reflects in the molecular collision frequency. According to the state equation, the collision frequency increases by about 0.3 times for every 100 kilopascal increase in pressure;

[0119] The diffusion process of the gas is described by Fick's diffusion law. For the diffusion of carbon monoxide in air, the diffusion coefficient is 2 x 10 -5 square meters per second at 298 Kelvin and 101 kilopascal. The diffusion coefficient increases by about 15% for every 100 Kelvin increase in temperature, and by about 12% for every 10 kilopascal decrease in pressure;

[0120] The deep neural network contains 3 convolutional layers, each with 64, 128, and 256 convolutional kernels respectively, with a kernel size of 3 x 3 x 3, for extracting spatial features;

[0121] The number of neurons in the 2 fully connected layers is 512 and 256 respectively, for predicting the diffusion velocity distribution;

[0122] The degree of gas mixing is represented by Shannon entropy. Taking the mixing of two gases as an example, when the volume ratio is 1:1, the entropy value is maximum at 0.693, and the entropy value decreases as the ratio deviates from this value. The multivariate autoregressive model takes the temperature, pressure, and humidity time series with a lag order of 3 as input to predict the change in mixing entropy;

[0123] The dimension of the coupling relationship matrix is the product of the number of parameters and the number of influence indicators. The main influence mode can be extracted through singular value decomposition;

[0124] When the cumulative contribution rate of singular values reaches 90%, the corresponding singular vectors reflect the main way in which environmental parameters affect the fluctuation of gas concentration. Temperature plays a role by affecting the reaction rate, and its singular value is maximum at 0.85;

[0125] The pressure plays a role by affecting the collision frequency and diffusion velocity, with a singular value of 0.62, and the influence of humidity is relatively small, with a singular value of 0.31.

[0126] Further, in this embodiment, step S4 specifically comprises:

[0127] According to the coupling relationship matrix, the numerical values of the temperature field, the pressure field and the humidity field on the spatial grid points are obtained, and the concentration gradient data of the target gas in the three-dimensional space is obtained through numerical calculation;

[0128] The concentration gradient data is used to calculate the mixing entropy value of the target gas and the environment gas, and the Bayesian network containing three nodes of temperature, pressure and humidity is established to obtain the gas concentration prediction data;

[0129] The gas concentration prediction data is divided into multiple time windows, and the deviation feature vector of the prediction value and the measured value is calculated using the time window, and the deviation feature vector is grouped by the hierarchical clustering algorithm to obtain the deviation feature data;

[0130] According to the deviation feature data, it is judged whether the standard deviation exceeds the preset upper limit of the standard deviation threshold, and if it exceeds the preset upper limit of the standard deviation threshold, the frequency spectrum features of the deviation feature data are extracted using the Fourier transform, and the target gas concentration change mode is judged through the frequency spectrum features;

[0131] Specifically, the numerical values of the temperature field, the pressure field and the humidity field on the spatial grid points are read from the environmental parameter and the chemical reaction chain coupling matrix, the concentration gradient of the target gas in the three-dimensional space is calculated based on the Fick diffusion law, and the mean free path and the collision frequency of the gas molecules are calculated through the gas dynamics equation to obtain the first spatiotemporal diffusion data;

[0132] The first spatiotemporal diffusion data is used to calculate the mixing uniformity of the target gas and the environment gas using Shannon entropy and Gibbs mixing entropy, a Bayesian conditional probability network containing three nodes of temperature, pressure and humidity is established to predict the gas concentration change trend, and the second time series prediction data is obtained;

[0133] The second time series prediction data is divided into multiple time windows to calculate the mean square error, the standard deviation and the maximum deviation of the predicted concentration value and the measured concentration value in each window, and the deviation feature vectors of all time windows are grouped using the hierarchical clustering algorithm to obtain the third deviation feature data;

[0134] The third deviation feature data is determined according to the preset standard deviation threshold and the maximum deviation threshold, and when the standard deviation exceeds the preset upper limit of the threshold or the maximum deviation exceeds the preset lower limit of the threshold, the frequency spectrum features of the deviation sequence are extracted using the Fourier transform;

[0135] The main frequency component and energy distribution of the deviation sequence are calculated based on the spectral characteristics, whether the target gas concentration fluctuation presents a periodic change mode, a monotone increasing mode or a shock decay mode is judged, and a deviation mode judgment result is obtained;

[0136] The example is as follows:

[0137] In the prediction of environmental parameters and gas concentration in a fire scene, the coupling relationship matrix of environmental parameters and chemical reaction chain records the temperature, pressure and humidity values at the spatial grid points. Taking a 3m x 3m x 3m space as an example, it is divided into 27 grid points, and the temperature range at each grid point is recorded as 298 to 1500 Kelvin, the pressure range is 90 to 110 kilopascals, and the relative humidity range is 20% to 90%;

[0138] When calculating the concentration gradient of the target gas based on the Fick diffusion law, a grid spacing of 0.5 meters is selected, and the diffusion coefficient of carbon monoxide under standard conditions is calculated to be 2 x 10 -5 square meters per second. The average free path of molecules is calculated to be 70 nanometers by the gas dynamics equation, and the collision frequency is 7 x 10 9 times per second;

[0139] The calculation of gas mixing uniformity uses Shannon entropy and Gibbs mixing entropy. For a binary gas mixture, when the volume ratio of the two gases is 1:1, the Shannon entropy reaches a maximum value of 0.693, and gradually decreases as the ratio deviates from 1:1;

[0140] The conditional probability between the three nodes of temperature, pressure and humidity in the Bayesian network is obtained by training historical data. When the temperature increases by 100 Kelvin, the probability of increase in carbon monoxide concentration is 0.8, and when the pressure decreases by 10 kilopascals, the probability of increase in concentration is 0.6;

[0141] In the time window division, 10 seconds are selected as the window length, and the adjacent windows overlap by 5 seconds. The statistical characteristics of the predicted value and the measured value are calculated for each window, the mean square error threshold is set to 0.01, the standard deviation threshold is set to 0.1, and the maximum deviation threshold is set to 0.2;

[0142] Hierarchical clustering uses Euclidean distance as a similarity measure to cluster the deviation feature vectors into 3 categories, corresponding to stable, fluctuating and dramatic changes respectively;

[0143] When extracting the spectral characteristics of the deviation sequence by Fourier transform, the sampling frequency is 1 Hz and the transform length is 128 points;

[0144] The determination standard of the main frequency component is that the energy ratio exceeds 50%. When a main frequency component of 0.1 Hz appears, it indicates that the deviation presents a 10-second periodic fluctuation;

[0145] When the spectrum presents a continuous rising trend, it indicates that the deviation presents a monotone increasing mode; when the energy of high frequency components gradually attenuates, it indicates that the deviation presents a damped oscillation mode;

[0146] In practical applications, taking carbon monoxide concentration prediction as an example, when the environmental temperature rises from 500 Kelvin to 800 Kelvin, the deviation sequence of the predicted value and the measured value shows a main frequency component of 0.05 Hz after Fourier transform, and the energy ratio reaches 65%, indicating that the concentration prediction deviation presents a 20-second period fluctuation mode;

[0147] At the same time, the standard deviation of the deviation sequence is 0.15, which exceeds the preset threshold value 0.1, and the maximum deviation is 0.25, which exceeds the preset threshold value 0.2, and the comprehensive judgment indicates that the prediction result has significant periodic deviation.

[0148] Further, in the embodiment, step S5 specifically comprises:

[0149] According to the three-dimensional space coordinates of the hazard source, time series data of gas concentration is obtained, and wavelet transform is used to extract low-frequency approximation coefficients and high-frequency detail coefficients of the time series data at several decomposition scales to obtain first feature data;

[0150] For the first feature data, a random forest classifier is used to establish a mapping relationship between the first feature data and the hazard source type, and the classifier parameters are optimized through cross-validation to obtain second classification data;

[0151] According to the second classification data, the Euclidean distance matrix between the gas concentration fluctuation characteristics of different types of hazard sources is calculated, and a spectral clustering algorithm is used to group the hazard sources, and a typical fluctuation template is extracted for each group to obtain third template data;

[0152] If the Pearson correlation coefficient between the measured gas concentration curve and the third template data is greater than a preset Pearson correlation coefficient judgment threshold, then the posterior probability of the hazard source position is calculated using the Naive Bayes method, and the spatial position area of the hazard source and the type of hazardous material are determined;

[0153] Specifically, the three-dimensional space coordinates of the hazard source, the type of the hazard source and the corresponding time series data of the gas concentration are read from the historical fire database, the statistical characteristics of the gas concentration fluctuation curve are extracted according to a fixed sampling period, and the low-frequency approximation coefficients and high-frequency detail coefficients of the fluctuation curve are extracted at three decomposition scales using wavelet transform to obtain first-level feature data;

[0154] For the first-level feature data, the relative distance vector and the azimuth angle vector are calculated according to the spatial coordinates of the hazard source and the coordinates of the gas sampling points, a sample matrix containing fluctuation characteristics, distance characteristics and azimuth characteristics is constructed, a random forest classifier is used to establish a corresponding relationship between the feature samples and the type of the hazard source, and the classifier parameters are optimized through five-fold cross-validation to obtain second classification mapping data.

[0155] For the second classification mapping data, the Euclidean distance matrix between the gas concentration fluctuation characteristics caused by different types of dangerous sources is calculated, the spectral clustering algorithm is used to group the dangerous sources, and the third template feature data is obtained by extracting the typical fluctuation template of each group;

[0156] For the third template feature data, the Pearson correlation coefficient is used to calculate the similarity between the measured concentration curve of the target gas and the typical fluctuation template of the dangerous source. When the correlation coefficient is greater than the preset judgment threshold, the fourth matching correlation data is generated;

[0157] According to the fourth matching correlation data, the posterior probability of the dangerous source position falling in different spatial grid regions is calculated using the Naive Bayes method, and the spatial position region and the type of dangerous substance of the dangerous source are determined by combining the conditional probability distribution of the type of dangerous source;

[0158] The example is as follows:

[0159] In the identification of dangerous sources in fire scenes, the historical database records the gas concentration change characteristics caused by different types of dangerous sources;

[0160] Taking a methanol storage tank as an example, its spatial coordinates are represented by a three-dimensional rectangular coordinate system, and the origin is set at the entrance of the fire scene. The coordinates of the storage tank are (10 meters, 5 meters, 2 meters), and the gas concentration sampling points are arranged around the storage tank at an interval of 1 meter;

[0161] The sampling period of the gas concentration time series data is 1 second, and the continuous concentration change curve is obtained within 120 seconds of observation time;

[0162] When wavelet transforming the concentration fluctuation curve, db4 wavelet basis function is selected, and low-frequency approximation coefficients and high-frequency detail coefficients are extracted at three decomposition scales;

[0163] The first layer of decomposition reflects the change trend of 120 seconds, the second layer of decomposition reflects the fluctuation characteristics of 60 seconds, and the third layer of decomposition reflects the local details of 30 seconds;

[0164] For the fire caused by leakage of the methanol storage tank, the low-frequency coefficients show the overall rising trend of the concentration, and the high-frequency coefficients show obvious periodic fluctuations;

[0165] In the construction of feature samples, the distance feature includes the Euclidean distance from the sampling point to the dangerous source, and the distance distribution of the methanol storage tank and the surrounding 8 sampling points is in the range of 1 to 3 meters;

[0166] The azimuth feature includes the horizontal azimuth angle and the pitch angle, forming a 16-dimensional position feature vector;

[0167] The random forest classifier contains 100 decision trees with a maximum depth of 8 layers, and the optimal parameter combination is determined by five-fold cross-validation.

[0168] The clustering results of the hazard source types show that flammable liquid hazard sources, such as methanol and ethanol storage tanks, exhibit similar concentration fluctuation characteristics, and their typical templates exhibit a rapid rise and then tend to be stable;

[0169] The concentration fluctuations of gaseous flammable substances, such as natural gas pipelines, exhibit periodic oscillation characteristics, and the concentration changes of solid flammable substances, such as plastic products, are relatively slow. The correlation coefficient threshold of the typical template is set to 0.8, and if it exceeds this threshold, it is considered to be a match;

[0170] In Bayesian probability reasoning, the fire space is divided into 27 grid regions, each with a size of 2m x 2m x 2m;

[0171] When the target gas concentration is detected to exhibit a rapid rise characteristic, and the correlation coefficient with the flammable liquid typical template is 0.85, the hazard source position is most likely to fall within the 9 grid regions within a range of 2 meters from the sampling point, with the center grid having the highest posterior probability of 0.45, and the probability of the hazard source type being determined as a flammable liquid is 0.82, achieving probability estimation of the hazard source position and type.

[0172] Further, in the embodiment, step S6 specifically includes:

[0173] A bicubic interpolation algorithm is used to perform spatial interpolation on the concentration values between the gas concentration sampling points to obtain continuous concentration distribution field data;

[0174] The continuous concentration distribution field data is filtered using a Gaussian kernel function to obtain first filtered field data;

[0175] The first filtered field data is used to divide a concentric circular search region centered on the initial position of the hazard source, and a particle swarm algorithm is used to iteratively search the concentration gradient in the search region to obtain second gradient field data;

[0176] The gas concentration statistical moments in the search region are calculated from the second gradient field data, a hidden Markov chain is used to estimate the state transition of the hazard source position coordinates and type identifier, and the spatial coordinates of the hazard source are determined;

[0177] Specifically, the gas concentration data is collected from the sensor network, a concentration distribution matrix is constructed based on the spatial arrangement of the sampling point coordinates, a bicubic interpolation algorithm with a cubic polynomial weight parameter is used to perform spatial interpolation on the concentration values between the sampling points, and a continuous concentration distribution field with a spatial resolution of one-fourth of the sampling point spacing is generated;

[0178] The concentration field data is filtered using a Gaussian kernel function with a standard deviation of one-half of the sampling point spacing to obtain first filtered field data;

[0179] For the first filtered field data, in combination with the initial position of the hazard source generated by the classifier, three concentric circular search regions with radii being integer times of the sampling point spacing are divided with the initial position as the center, and the particle swarm algorithm with a population size of fifty is used to iteratively search the concentration gradient in each search region to obtain second gradient field data;

[0180] For the second gradient field data, the statistical moments of the gas concentration in each search region are calculated based on a sliding time window, and the hidden Markov chain with an observation probability matrix size of the number of hazard source types is used to estimate the state transition of the hazard source position coordinates and type identification to obtain third state estimation data;

[0181] For the third state estimation data, the normalized cross-correlation coefficient of the current gas concentration field and the preset hazard source feature template is calculated, and the spatial coordinates of the hazard source are determined according to the peak position and numerical value of the cross-correlation coefficient matrix;

[0182] Based on the spatial coordinates and type identification of the hazard source, in combination with the real-time change trend of the gas concentration field, the hazard source positioning information is updated according to a fixed data sampling period to generate real-time tracking data of the hazard source position and type;

[0183] The example is as follows:

[0184] In the hazard source positioning in a fire scene, the sensor network is uniformly arranged in space at an interval of 2 meters to form an 8x8x4 three-dimensional sampling grid;

[0185] For the discrete concentration data obtained by the sampling points, a bicubic interpolation algorithm is used to generate a continuous concentration distribution field with a resolution of 0.5 meters;

[0186] The interpolation algorithm uses a cubic polynomial weight function to calculate the concentration value of the interpolation point based on the data of the adjacent 16 sampling points, ensuring the continuity and smoothness of the spatial concentration distribution;

[0187] The Gaussian filter uses a kernel function with a standard deviation of 1 meter to effectively eliminate sensor noise interference;

[0188] After preliminary positioning of the hazard source, three concentric circular search regions are divided with the initial position as the center, with radii of 2 meters, 4 meters and 6 meters respectively;

[0189] The population size of the particle swarm algorithm is 50, and each particle has position and velocity information, tracking the direction of the maximum concentration gradient in the search region;

[0190] Taking the methanol tank leakage as an example, when the search region radius is 4 meters, after 20 iterations, the particle swarm finds the local maximum value of the concentration gradient at a distance of 3.2 meters from the initial position, indicating the possible position of the hazard source;

[0191] In the state estimation process, a 30-second sliding time window is used to calculate the statistical characteristics;

[0192] For flammable liquid hazard sources, the skewness of their concentration distribution is usually greater than 0.8, showing a clear right-biased characteristic;

[0193] The kurtosis is greater than 3.5, indicating that the distribution curve is steeper than the normal distribution;

[0194] The observation probability matrix of the hidden Markov chain has a dimension of 5x5, corresponding to the 5 main types of hazard sources: flammable liquid, flammable gas, flammable solid, explosive, and toxic substance;

[0195] The hazard source feature template is a 5x5x5 three-dimensional array that describes the typical concentration distribution patterns of different types of hazard sources in the surrounding space;

[0196] When calculating the normalized cross-correlation coefficient between the current concentration field and the feature template, the sliding step is set to 0.5 meters, and the cross-correlation peak is searched within the entire search space;

[0197] When the cross-correlation coefficient exceeds 0.85, it is considered that the matching position of the feature template has been found. For the methanol tank leakage scenario, within 120 seconds after the fire, the tracking error of the hazard source position remains within 1 meter, and the accuracy of type identification reaches 90%;

[0198] The data update period is set to 1 second, matching the sampling frequency of the sensor. In each update period, based on the latest concentration field data and state estimation results, the position coordinates and type identification of the hazard source are dynamically adjusted;

[0199] When the position change is less than 0.5 meters and the type determination remains consistent for 5 consecutive periods, it is considered that the positioning result is stable. Through this real-time tracking mechanism, the hazard source positioning information can timely reflect the dynamic changes of the fire scene situation.

[0200] Further, in the embodiment, step S7 specifically includes:

[0201] The real-time adjustment of the coupling coefficient between the environmental parameters and the chemical reaction chain includes adjusting the reaction rate constant;

[0202] Generating a dynamically adjusted hazard source identification result includes hazard source type, position, and hazard level information;

[0203] According to the measured values of the environmental parameters collected by the temperature sensor, pressure sensor, and humidity sensor, an environmental change trend vector is constructed, and the future environmental parameter change prediction value is obtained through the environmental change trend vector;

[0204] According to the future environmental parameter change prediction value, the reaction rate constant is calculated by using the temperature correction term and the pressure correction term, and the gas concentration prediction value is obtained by establishing the Kalman filter through the reaction rate constant and the recursive least square method;

[0205] For the gas concentration prediction value, the gradient vector and the Hessian matrix of the concentration field are calculated, and the support vector regression with the kernel function as the radial basis is used to obtain the dangerous source position coordinates;

[0206] For the dangerous source position coordinates, the Bayesian probability estimation is used to update the posterior distribution of the dangerous source type, and the hazard degree index vector is constructed according to the gas concentration exceeding multiple, the spread speed and the diffusion range to obtain the danger level;

[0207] Specifically, according to the real-time positioning information of the dangerous source and the measured value of the environmental parameter, the change amount of temperature, pressure and humidity per unit time is used to construct the environmental change trend vector, the environmental parameter change value at the future time is predicted through the third-order autoregressive equation of the sliding time window, the temperature correction term and the pressure correction term of the reaction rate constant are calculated based on the activation energy and the pre-exponential factor, and the first rate parameter data is obtained;

[0208] For the first rate parameter data, the reaction rate of the gas is calculated by using the corrected reaction rate constant, the linear coupling coefficient of the gas concentration and the environmental parameter is updated by using the recursive least square method, the Kalman filter of the measurement equation and the state equation is established to predict the fluctuation trend of the gas concentration, and the second concentration prediction data is obtained;

[0209] For the second concentration prediction data, the first-order spatial derivative and the second-order spatial derivative of the gas concentration field around the dangerous source are calculated, the gradient vector and the Hessian matrix of the concentration field are constructed, the support vector regression with the kernel function as the radial basis is used to update the dangerous source position coordinates, and the third coordinate data is obtained;

[0210] For the third coordinate data, the Bayesian probability estimation is used to update the posterior distribution of the dangerous source type, and the type identification is determined combined with the characteristics of the dangerous source type, and the fourth type data is obtained;

[0211] Based on the three dimensions of the gas concentration exceeding multiple, the spread speed and the diffusion range, a hazard degree index vector is constructed, and the danger level is quantitatively graded according to the Euclidean norm of the index vector, and a dynamic recognition result containing the dangerous source type identification, the spatial position coordinates and the danger level is generated;

[0212] The example is as follows:

[0213] In the process of dynamic identification of dangerous sources in fire scenes, the rapid change of environmental parameters has a significant impact on the distribution of gas concentration;

[0214] The environmental change trend vector adopts a 10-second sliding time window to record the temperature, pressure, and humidity change rates, and when the temperature rise rate exceeds 20 kelvins per minute, a third-order autoregressive equation is used to predict the temperature change trend in the next 30 seconds;

[0215] Taking the carbon monoxide oxidation reaction as an example, its activation energy is 125 kilojoules per mole, and the pre-exponential factor is 4.6*10^11, and when the temperature rises from 500 kelvins to 800 kelvins, the reaction rate constant increases by about 1000 times;

[0216] The coupling coefficient of gas concentration and environmental parameters is updated by recursive least squares, and the initial covariance matrix is set to the unit matrix with a forgetting factor of 0.95;

[0217] The state vector of the Kalman filter includes gas concentration and concentration change rate, and the measurement equation is based on the measured values of the sensor, and the state equation describes the evolution law of the concentration over time;

[0218] For carbon monoxide, the state noise covariance is 0.01, the measurement noise covariance is 0.02, and the filter updates the state estimation every second;

[0219] In the dangerous source position update, the spatial derivative of the gas concentration field reflects the direction and intensity of the concentration change;

[0220] Taking a 3m*3m*3m spatial region as an example, the concentration gradient vector is calculated on a 0.5m grid, and the x, y, and z components reflect the spatial distribution characteristics of the concentration change;

[0221] The eigenvalues of the Hessian matrix characterize the curvature characteristics of the concentration field, and positive eigenvalues indicate convex distribution of the concentration, and negative eigenvalues indicate concave distribution;

[0222] The support vector regression adopts a radial basis kernel function, the kernel parameter is 0.1, and the penalty coefficient is 100. The Bayesian probability estimation of the dangerous source type is based on the prior distribution established based on historical data, and the real-time observation data is used to update the posterior probability;

[0223] Taking the flammable liquid type dangerous source as an example, when a rapidly rising concentration curve is observed and the spatial distribution shows obvious point source characteristics, the posterior probability of this type is significantly improved;

[0224] In the hazard degree index vector, the concentration exceeding multiple is calculated based on the occupational exposure limit, the spread speed is determined by the expansion rate of the high concentration area per unit time, and the diffusion range is calculated based on the volume of the space where the concentration exceeds the threshold;

[0225] When the Euclidean norm of the index vector is less than 1.0, it is determined to be a mild hazard, the norm is between 1.0 and 2.0 for a moderate hazard, and greater than 2.0 for a severe hazard;

[0226] The dynamic identification result is updated once per second, contains the spatial coordinates of the danger source, the type determination result and the danger level, and realizes real-time tracking of the danger source characteristics and dynamic assessment of the harm degree.

[0227] In the description of the present application, it should be understood that the orientation words such as "front, back, up, down, left, right", "transverse, vertical, horizontal" and "top, bottom" and the like indicate the orientation or positional relationship shown in the drawings, which are only for the convenience of describing the present application and simplifying the description, and do not indicate and imply that the devices or elements referred to must have a particular orientation or be constructed and operated in a particular orientation, therefore cannot be understood as a limitation on the protection scope of the present application.

[0228] For those skilled in the art, other various corresponding changes and deformations can be made according to the above described technical solutions and concepts, and all of these changes and deformations should belong to the protection scope of the claims of the present application.

Claims

1.A cloud service-based intelligent fire-fighting big data analysis method, characterized in that, The method comprises the following steps: S1, acquiring multi-source gas concentration time series data collected by a gas sensor array at a fire scene, and respectively preprocessing concentration change trends of different gases to remove noise interference and obtain smooth gas concentration time series curves; S2, analyzing concentration changes of different gases according to the gas concentration time series curves, and combining a pre-established chemical reaction chain model to generate a mapping relationship of a chemical reaction chain related to a target gas if there is a chemical reaction of the target gas; S3, acquiring environmental parameters of the fire scene, combining the mapping relationship of the chemical reaction chain, establishing a coupling relationship matrix of the environmental parameters and the chemical reaction chain, and determining the influence of the environmental parameters on gas concentration fluctuations; S4, calculating diffusion speeds and mixing degrees of different gases under different environmental conditions through the coupling relationship matrix, generating a prediction value of target gas concentration fluctuations, and analyzing the deviation degree of the prediction value of the target gas concentration fluctuations from an actual measured value; S5, if the target gas is related to the location and type of a hazard source, extracting the relationship between gas concentration fluctuations and the hazard source from historical fire data, identifying typical gas concentration change patterns caused by different types of hazard sources, training a classification model, and generating preliminary estimation results of the type and location of the hazard source; S6, according to the preliminary estimation results, combining real-time collected gas concentration time series data, analyzing the type and location of the hazard source at the current fire scene, and generating real-time positioning information of the hazard source; S7, updating the prediction value of the gas concentration fluctuations through the real-time positioning information, combining the change trend of the environmental factors, real-time adjusting the coupling coefficients of the environmental parameters and the chemical reaction chain, and generating a dynamically adjusted hazard source identification result; The step S3 specifically comprises: The influence of the environmental parameters on the gas concentration fluctuations includes gas reaction rate, gas diffusion speed and gas mixing degree; Temperature data, humidity data and pressure data of the fire scene are acquired by using a sensor network, and temperature distribution field, humidity distribution field and pressure distribution field are calculated by radial basis function according to the temperature data, the humidity data and the pressure data; Reactant activation energy values and pre-exponential factor values are extracted from the temperature distribution field, the humidity distribution field and the pressure distribution field through a chemical reaction chain mapping relationship, and reaction rate influence coefficients are obtained according to the activation energy values and the pre-exponential factor values combined with Arrhenius equation; Gas diffusion flux values are calculated by using Fick's law according to the reaction rate influence coefficients, and gas diffusion speed distribution data are obtained by establishing a deep neural network model through the gas diffusion flux values; Shannon entropy values are calculated from the gas diffusion speed distribution data through gas concentration time series curves, and environmental parameter influence coefficient matrices are obtained by establishing a multivariate autoregressive model according to the Shannon entropy values. 2.The cloud service-based intelligent fire-fighting big data analysis method according to claim 1, characterized in that, The step S1 specifically comprises: Gas concentration time series data of various types of gases are acquired from a gas sensor array, and first gas concentration data are obtained by normalizing the concentration time series data according to sensitivity coefficients of various types of gas sensors; The first gas concentration data is smoothed by using a sliding mean filter, and wavelet transform denoising is performed on data points exceeding a preset noise interference amplitude threshold to obtain second gas concentration data; The second gas concentration data is fitted by using a cubic spline interpolation function to obtain third gas concentration data. The third gas concentration data is smoothed by using a Gaussian kernel function to obtain a smoothed concentration time curve of each type of gas. 3.The cloud service-based intelligent fire-fighting big data analysis method according to claim 1, characterized in that, The step S2 specifically comprises: Target gas concentration data is obtained from the gas concentration time curve, and a similarity value between a target gas molecule and other gas molecules is calculated by using a molecular structure feature vector stored in a chemical reaction chain database to obtain candidate chemical reaction chain data; The reaction rate constant values of each elementary reaction step in the reaction chain within a temperature interval are calculated by using a pre-established gas phase reaction rate constant database to obtain reaction chain kinetics data; The Pearson correlation coefficient between the reactant concentration change value within the temperature interval and the measured gas concentration time curve is calculated based on the reaction chain kinetics data, the reaction chains with a correlation coefficient greater than a coefficient threshold are grouped by using a hierarchical clustering algorithm to obtain reaction chain association data; The standard enthalpy of formation and standard entropy change values of reactants and products are obtained from a thermodynamic database, the chemical reaction equilibrium constant of the reaction chain under temperature and pressure conditions is calculated, and a reaction chain mapping model is established by using a support vector regression method. 4.The cloud service-based intelligent fire-fighting big data analysis method according to claim 1, wherein, The step S4 specifically comprises: The temperature field, pressure field and humidity field values on the spatial grid points are obtained based on the coupling relationship matrix, and the concentration gradient data of the target gas in the three-dimensional space are calculated based on the values; The mixing entropy value of the target gas and the environment gas is calculated based on the concentration gradient data, and the Bayesian network including the temperature, pressure and humidity nodes is established based on the mixing entropy value to obtain gas concentration prediction data; The gas concentration prediction data is divided into a plurality of time windows, the deviation feature vector of the prediction value and the measured value is calculated based on the time windows, and the deviation feature data is obtained by grouping the deviation feature vector based on the hierarchical clustering algorithm; It is judged whether the standard deviation exceeds a preset upper limit of the standard deviation threshold based on the deviation feature data, and if the preset upper limit of the standard deviation threshold is exceeded, the frequency spectrum features of the deviation feature data are extracted by using Fourier transform, and the target gas concentration change mode is judged based on the frequency spectrum features. 5.The cloud service-based intelligent fire-fighting big data analysis method according to claim 1, wherein, The step S5 specifically comprises: Gas concentration time series data is obtained based on the three-dimensional spatial coordinates of the hazard source, and the low-frequency approximation coefficients and high-frequency detail coefficients of the time series data are extracted by using wavelet transform at a plurality of decomposition scales to obtain first feature data; The mapping relationship between the first feature data and the hazard source type is established by using a random forest classifier based on the first feature data, the classifier parameters are optimized by cross-validation, and second classification data is obtained. According to the second classification data, a Euclidean distance matrix between gas concentration fluctuation characteristics of different types of hazard sources is calculated, a spectral clustering algorithm is used to group the hazard sources, a typical fluctuation template of each group is extracted, and third template data is obtained; If a Pearson correlation coefficient between the actual gas concentration curve and the third template data is greater than a preset Pearson correlation coefficient determination threshold, a posterior probability of the hazard source position is calculated using a Naive Bayes method, and the spatial position area of the hazard source and the type of hazardous material are determined. 6.The cloud service-based intelligent fire-fighting big data analysis method according to claim 1, wherein, The step S6 specifically includes: A bicubic interpolation algorithm is used to perform spatial interpolation on the concentration values between the gas concentration sampling points, and continuous concentration distribution field data is obtained; The continuous concentration distribution field data is filtered using a Gaussian kernel function, and first filtered field data is obtained; According to the first filtered field data, a concentric circle search area centered on the initial position of the hazard source is divided, and a particle swarm algorithm is used to iteratively search the concentration gradient in the search area, and second gradient field data is obtained; The second gradient field data is used to calculate the statistical moment of the gas concentration in the search area, a hidden Markov chain is used to estimate the state transition of the hazard source position coordinates and type identifier, and the spatial coordinates of the hazard source are determined. 7.The cloud service-based intelligent fire-fighting big data analysis method according to claim 1, characterized in that, The step S7 specifically includes: Real-time adjustment of the coupling coefficient of the environmental parameters and the chemical reaction chain includes adjusting the reaction rate constant; Generating a dynamically adjusted hazard source identification result includes hazard source type, position, and hazard level information; An environment change trend vector is constructed according to the actual measured values of the environmental parameters collected by the temperature sensor, the pressure sensor, and the humidity sensor, and future environmental parameter change prediction values are obtained through the environment change trend vector; According to the future environmental parameter change prediction values, the reaction rate constant is calculated using a temperature correction term and a pressure correction term, and a Kalman filter is established through the reaction rate constant and the recursive least squares method to obtain a gas concentration prediction value; For the gas concentration prediction value, the gradient vector and the Hessian matrix of the concentration field are calculated, and the support vector regression with a kernel function as a radial basis is used to obtain the hazard source position coordinates; For the hazard source position coordinates, the Bayesian probability estimation is used to update the posterior distribution of the hazard source type, and a hazard degree index vector is constructed according to the gas concentration exceeding multiple, the spread speed, and the diffusion range to obtain the hazard level.

Images