Visual gas leakage detection system and method based on laser spectrum

Abstract

The invention discloses a visual gas leakage detection system and method based on a laser spectrum, and relates to the technical field of gas detection visualization. According to the visual gas leakage detection system and method based on the laser spectrum, the method comprises the following steps that 1, gas monitoring data are collected and preprocessed; 2, extracting a characteristic spectrum from the gas monitoring data, obtaining a gas spectrum reduction value, and separating an overlapped absorption spectrum; 3, estimating the diffusion trajectory of the gas, and calculating a gas diffusion coefficient estimation value and a diffusion rate estimation value; outputting a gas concentration estimated value; 4, gas leakage source points are calculated, gas mass tracking is carried out, and different gas leakage sources are separated; and 5, generating a scatter diagram to display gas concentrations at different positions, and marking the position of a leakage source. The problem that a complex multi-gas environment cannot be recognized due to the fact that homogeneity and heterogeneity possibly exist in the same environment due to leakage of different types of gases in the prior art is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of gas detection visualization, and in particular to a laser spectroscopy-based visualized gas leak detection system and method. Background Art

[0002] In industrial production, energy transportation, chemical processes, and many other scenarios, the leakage of flammable and toxic gases remains a significant safety hazard. In recent years, with the development of laser spectroscopy technology, the use of laser absorption spectroscopy for gas detection has become a significant trend. This technology offers the advantages of non-contact, high sensitivity, and high selectivity, making it suitable for trace gas detection in complex environments.

[0003] However, traditional gas leak detection methods mainly rely on fixed-point sensor arrays or portable detectors. These methods have problems such as limited detection range, insufficient positioning accuracy, and slow response time. They are difficult to meet practical needs such as simultaneous detection of multiple gases, rapid location of leak sources, and visualization of concentration distribution. Traditional laser spectroscopy technology usually only focuses on gas concentration estimation and lacks the ability to model and visualize gas diffusion dynamics, making it even more difficult to achieve accurate identification of multiple gases and tracking of leak sources.

[0004] Therefore, in order to solve the above problems, there is an urgent need for a visual gas leak detection system and method based on laser spectroscopy. Summary of the Invention

[0005] Technical problems solved In response to the shortcomings of the existing technology, the present invention provides a visual gas leak detection system and method based on laser spectroscopy, which solves the problem that different types of gas leaks may have homogeneity and heterogeneity in the same environment, resulting in the inability to identify complex multi-gas environments.

[0006] Technical Solution To achieve the above objectives, the present invention is implemented through the following technical solutions: a laser spectroscopy-based visualized gas leak detection system and method, comprising the following steps: step one, collecting gas monitoring data and preprocessing the gas monitoring data; step two, extracting characteristic spectra from the gas monitoring data, obtaining gas spectrum restoration values ​​based on the gas monitoring data, and separating overlapping absorption spectra of multiple gases; step three, using a diffusion model to estimate the diffusion trajectory of the gas under the influence of wind speed and other environmental conditions, and calculating the estimated gas diffusion coefficient and diffusion rate; fusing the gas spectral features with the diffusion features, outputting the gas concentration estimate, and further distinguishing the gas type; step four, reversely inferring the gas leakage source based on the gas concentration gradient, and tracking gas clusters to separate different gas leakage sources; step five, generating a scatter plot to display the gas concentration at different locations, and marking the location of the leakage source with annotations.

[0007] Furthermore, the specific process of collecting gas monitoring data and preprocessing the gas monitoring data is as follows: the gas monitoring data includes spectral characteristic data and diffusion characteristic data, the gas absorption spectrum and absorption intensity are obtained through the sensor array to obtain the spectral characteristic data, the time, gas position, wind speed, temperature and humidity and gas concentration are synchronously collected to obtain the diffusion characteristic data of the gas, Gaussian filtering is used to smooth the laser spectrum curve, the noise of the gas monitoring data is removed, the adaptive baseline fitting method is used for baseline correction, the gas monitoring data is normalized using the linear normalization method, and the standardization process is used to make each gas monitoring data feature have the same scale.

[0008] Furthermore, the characteristic spectrum of the gas monitoring data is extracted, and the specific process of obtaining the gas spectrum restoration value based on the gas monitoring data is as follows: based on the derivative spectroscopy method, the significant absorption peaks in the gas absorption spectrum are extracted to obtain the absorption spectrum of the gas, and the gas spectrum restoration value is obtained according to the gas monitoring data; the specific analysis process of the gas spectrum restoration value is as follows: according to the spectral characteristic data, the absorption intensity of each gas at different wavelengths is obtained to obtain the gas absorption spectrum, according to the diffusion characteristic data, the gas concentration at the spatial position (x, y) and time t is obtained, the absorption spectrum of each gas is multiplied by its corresponding gas concentration, and the sum is calculated to obtain the gas spectrum restoration value.

[0009] Furthermore, the specific process of separating the overlapping absorption spectra of multiple gases is as follows: the main features of the gas spectrum restoration value are extracted by principal component analysis, and the gas spectrum restoration value after dimensionality reduction by principal component analysis is used as input, and the overlapping absorption component matrix in the gas spectrum restoration value is decomposed into non-negative and independent basic absorption components through non-negative matrix decomposition and independent component analysis, and the individual absorption characteristics of each gas and the concentration distribution of each gas at different observation points are decoded.

[0010] Furthermore, the diffusion model is used to estimate the diffusion trajectory of the gas under the influence of wind speed and other environmental conditions, and the specific process of calculating the estimated value of the gas diffusion coefficient and the estimated value of the diffusion rate is as follows: based on the gas monitoring data, a Gaussian diffusion model is established to estimate the diffusion trajectory under the influence of wind speed, temperature and humidity, and the gas diffusion path and air mass boundary in a complex environment are simulated; based on the gas monitoring data, the horizontal coordinate position, vertical coordinate position, time information and gas concentration in the monitoring area of ​​the gas leakage location are obtained, and the sum of the second-order derivative of the gas concentration in the horizontal direction and the second-order derivative in the vertical direction in the two-dimensional space is calculated to obtain the Laplace operator, and the absolute value of the Laplace operator is taken and the inverse is calculated; the first-order partial derivative of the gas concentration with respect to time is calculated to obtain the gas concentration change rate and the absolute value is taken; the above two factors are multiplied to obtain the estimated value of the gas diffusion coefficient; based on the water content of the gas concentration in the two-dimensional space, the estimated value of the gas diffusion coefficient is obtained. The rate of change in the horizontal and vertical directions is used to obtain the gas concentration gradient, and the ratio of the estimated gas diffusion coefficient to the gas concentration is multiplied by the gas concentration gradient to obtain the estimated gas diffusion rate. The gas diffusion rate is compared with the diffusion threshold in real time. When the estimated gas diffusion rate is greater than the diffusion threshold, it is considered that there is a gas leak in the area, and the laser scanner is controlled to encrypt the scanning frequency and wavelength range in the area, focusing on high-risk areas. Based on the current diffusion rate, it is predicted that the area may be affected at the next moment, and a reminder for further inspection is sent to relevant personnel. When the estimated gas diffusion rate is less than or equal to the diffusion threshold, it is considered that there is no gas leak in the area, and the sensor data is automatically checked for errors, data loss and errors. If the data is abnormal, the sensor is calibrated and reset, and the data in the area is cached for a time window. It is considered safe only when it is continuously below the threshold within the time window.

[0011] Furthermore, the gas spectral features and diffusion features are fused to output the gas concentration estimate. The specific process for further distinguishing the gas types is as follows: the spectral feature data and diffusion feature data are processed separately using a dual-branch network algorithm, the spectral feature data and diffusion feature data are fused through the self-attention mechanism to obtain a new fused feature representation, the spectral feature data and diffusion feature data are input to establish a neural network model, the fully connected layer weight matrix is ​​obtained through training, and the bias term is set when the neural network is initialized, the fused feature representation is multiplied by the fully connected layer weight matrix and the bias term is added; the gas diffusion coefficient estimate is multiplied by the diffusion behavior weight and a constant of one is added; finally, the above two factors are multiplied to obtain the gas concentration estimate; the gas concentration estimate and the concentration threshold, the gas diffusion rate and the diffusion threshold are compared in real time. When the gas concentration estimate is greater than the gas concentration threshold and the gas diffusion rate estimate is greater than the diffusion rate threshold, the high concentration and high diffusion of the gas is determined to be an obvious leak. The system automatically increases the sampling frequency, pushes the leak alarm to the safety system, and determines the gas type according to the concentration distribution and diffusion feature data of each gas in combination with the gas absorption spectrum, and enters the leak source positioning. and separation from multiple gas clouds; when the estimated gas concentration is greater than the gas concentration threshold and the estimated gas diffusion rate is less than or equal to the diffusion rate threshold, the high-concentration, low-diffusion gas is judged to be a near-source, high-concentration, concentrated leak, and a small-scale laser rescan is activated to continuously track the key areas of leaking gas accumulation, send equipment failure and loose interface inspection reminders to relevant personnel, and enter the leakage source location and multiple gas cloud separation step; when the estimated gas concentration is less than or equal to the gas concentration threshold and the estimated gas diffusion rate is greater than the diffusion rate threshold, the low-concentration, high-diffusion gas is judged to be an early, rarefied leak, and the early warning mechanism is automatically activated. The phenomenon is recorded as a warning event, and the leakage area is marked as a "suspected leak zone". The monitoring of this area is intensified in the next cycle, and the leakage source location and multiple gas cloud separation step is entered; when the estimated gas concentration is less than or equal to the gas concentration threshold and the estimated gas diffusion rate is less than or equal to the diffusion rate threshold, the low-concentration, low-diffusion gas is judged to be no obvious leak, the laser scanning frequency and spectral sampling resolution are reduced, the area is automatically adjusted to a low-priority area, and high-resolution sampling is automatically restored once every fixed period for verification.

[0012] Furthermore, the specific process of reversely inferring the gas leakage source based on the gas concentration gradient is as follows: reversely inferring the gas leakage source based on the gas concentration gradient, calculating the first-order partial derivative of the gas concentration in the horizontal direction and the first-order partial derivative in the vertical direction in two-dimensional space, multiplying the first-order partial derivative in the horizontal direction with the unit vector in the horizontal direction, multiplying the first-order partial derivative in the vertical direction with the unit vector in the vertical direction, and summing the two items to obtain the gradient vector in the two-dimensional space, and integrating the gradient vector to obtain the gas leakage source.

[0013] Furthermore, gas cluster tracking is carried out to separate different gas leakage sources: the calculated gas leakage source point is used to automatically mark the possible leakage area, and the Kalman filter diffusion model is used to track the diffusion path of the gas cloud in real time to predict the future distribution of the gas cloud. The monitoring points and monitoring frequencies are dynamically adjusted according to the changes in the gas source point and diffusion path; the leakage source is confirmed by tracking the diffusion path of the gas cloud to verify whether the source point is consistent with the actual leakage point; the density clustering algorithm is used to cluster the gas concentration clouds in the space, distinguish different gas clouds, separate the leakage sources of different gases, and display the leakage sources of different gases respectively. According to the gas type and concentration, emergency response measures are automatically formulated, including ventilation, isolation and evacuation, and sent to relevant personnel.

[0014] Furthermore, the specific process of generating a scatter plot to display the gas concentration at different locations and marking the location of the leakage source with annotations is as follows: drawing a scatter plot of the gas concentration, visualizing the gas concentration at each coordinate point, and reflecting the high and low concentrations through color, marking the location of the leakage source, and using different symbols and colors to highlight the coordinates of the leakage source.

[0015] Furthermore, it includes a gas monitoring data preprocessing module, an absorption feature extraction module, a spectrum and diffusion information fusion and identification module, a leakage source positioning and multi-gas cloud separation module and a visualization module: wherein the gas monitoring data preprocessing module is used to collect gas monitoring data and preprocess the gas monitoring data; the absorption feature extraction module is used to extract the characteristic spectrum of the gas monitoring data, obtain the gas spectrum restoration value according to the gas monitoring data, and separate the overlapping absorption spectra of multiple gases; the spectrum and diffusion information fusion and identification module is used to estimate the diffusion trajectory of the gas under the influence of wind speed and other environmental conditions by using the diffusion model, and calculate the gas diffusion coefficient estimate and the diffusion rate estimate; the gas spectrum feature and the diffusion feature are fused to output the gas concentration estimate to further distinguish the gas type; the leakage source positioning and multi-gas cloud separation module is used to reversely infer the gas leakage source point based on the gas concentration gradient, and track the gas clusters to separate different gas leakage sources; the visualization module is used to generate a scatter plot to display the gas concentration at different locations, and mark the location of the leakage source with annotations.

[0016] Beneficial effects The present invention has the following beneficial effects: (1) The present invention can efficiently and accurately determine the location of the leakage source by combining the gas concentration gradient reverse calculation method with the gas diffusion model; the reverse calculation method based on the concentration gradient can reduce the time delay of the traditional leakage source positioning, improve the real-time performance, and improve the accuracy of the source point positioning, especially in a complex gas diffusion environment.

[0017] (2) The present invention processes the monitoring data of multiple gases simultaneously and separates the overlapping absorption spectra through advanced signal processing methods such as principal component analysis and non-negative matrix decomposition, thereby realizing the identification and differentiation of multiple gas types; the leakage sources of different gases can be accurately separated, and personalized emergency response can be carried out according to the gas type and concentration.

[0018] (3) The present invention integrates laser spectral feature data with diffusion feature data, utilizes a dual-branch network algorithm and a self-attention mechanism, and can comprehensively consider the spectral features and diffusion features of the gas, thereby improving the accuracy of the gas concentration estimation value and being able to distinguish different types of gas leaks in real time.

[0019] (4) The present invention adaptively adjusts the frequency and scanning range of gas monitoring through real-time gas concentration estimation and diffusion rate estimation. When the gas diffusion rate exceeds a predetermined threshold, the system automatically increases the scanning frequency and focuses on high-risk areas, thereby improving the monitoring sensitivity. When the diffusion rate is lower than the threshold, the system reduces the sampling frequency to reduce resource waste.

[0020] Of course, any product implementing the present invention does not necessarily need to achieve all of the advantages described above at the same time. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 This is a flow chart of a visual gas leak detection method based on laser spectroscopy; Figure 2 This is a structural diagram of a visual gas leak detection system based on laser spectroscopy; Figure 3 A laser spectroscopy-based visualized gas leak detection system and method provides gas concentration distribution and leak source location maps; DETAILED DESCRIPTION

[0022] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0023] See also Figure 1-Figure 3, an embodiment of the present invention provides a technical solution: a laser spectroscopy-based visualized gas leak detection system and method, comprising the following steps: step one, collecting gas monitoring data and preprocessing the gas monitoring data; step two, extracting characteristic spectra from the gas monitoring data, obtaining gas spectrum restoration values ​​based on the gas monitoring data, and separating overlapping absorption spectra of multiple gases; step three, estimating the diffusion trajectory of the gas under the influence of wind speed and other environmental conditions using a diffusion model, and calculating an estimated value of the gas diffusion coefficient and an estimated value of the diffusion rate; fusing the gas spectral features with the diffusion features, outputting a gas concentration estimate, and further distinguishing the gas type; step four, reversely inferring the gas leakage source based on the gas concentration gradient, tracking gas clusters, and separating different gas leakage sources; step five, generating a scatter plot to display the gas concentration at different locations, and marking the location of the leakage source with annotations.

[0024] Specifically, the specific process of collecting gas monitoring data and preprocessing the gas monitoring data is as follows: gas monitoring data includes spectral characteristic data and diffusion characteristic data. The spectral characteristic data of the gas is obtained through the sensor array, and the time, gas position, wind speed, temperature and humidity and gas concentration are synchronously collected to obtain the diffusion characteristic data of the gas. In this way, the system can accurately obtain gas concentration and diffusion characteristics under different environmental conditions, ensuring the comprehensiveness and reliability of the data; use Gaussian filtering to smooth the laser spectrum curve, remove the noise of the gas monitoring data, use the adaptive baseline fitting method to perform baseline correction, use the linear normalization method to normalize the gas monitoring data, and through standardization processing, make each gas monitoring data feature have the same scale, thereby improving the comparability between different features and laying the foundation for subsequent feature fusion and classification recognition.

[0025] In this implementation, by eliminating problems such as noise, baseline drift, and scale inconsistency, the quality of spectral data is optimized, the stability and comparability of the data are improved, and accurate and reliable input data is provided for the gas leak detection system, ensuring the efficiency and accuracy of subsequent analysis and model output.

[0026] Specifically, the characteristic spectrum of the gas monitoring data is extracted, and the specific process of obtaining the gas spectrum restoration value based on the gas monitoring data is as follows: extracting significant absorption peaks based on the derivative spectroscopy method to obtain the absorption spectrum of the gas, effectively improving the resolution of the gas spectrum characteristics, so that the absorption characteristics of different gases can be more clearly distinguished; obtaining the gas spectrum restoration value based on the gas monitoring data; the specific analysis process of the gas spectrum restoration value is as follows: obtaining the absorption intensity of each gas at different wavelengths based on the spectral characteristic data to obtain the gas absorption spectrum, obtaining the gas concentration at the spatial position (x, y) and time t based on the diffusion characteristic data, multiplying the absorption spectrum of each gas by its corresponding gas concentration and adding them up to obtain the gas spectrum restoration value.

[0027] The gas spectrum reduction value is obtained by the following method: = Where, is the horizontal position coordinate in the gas leakage monitoring area; Indicates the longitudinal position coordinates in the gas leakage monitoring area; is the time variable; Indicates wavelength; is the gas spectrum reduction value, which represents the spectral intensity at different wavelengths λ at the spatial position (x, y) and time t; represents the absorption spectrum of the i-th gas; represents the gas concentration of the i-th gas at the spatial position (x, y) and time t; K represents the total number of gas types.

[0028] In this implementation, derivative spectroscopy is used to extract significant absorption peaks and calculate spectral reduction values. This process significantly improves the resolution of gas spectral characteristics, ensuring that the absorption characteristics of different gases can be clearly distinguished. Furthermore, by combining spectral characteristics with diffusion characteristics, spatial and temporal variations are taken into account, resulting in more accurate gas concentration estimates. This method optimizes dynamic gas monitoring, improves the accuracy of leak detection, and provides more reliable support for source location and emergency response. Overall, it enhances the real-time performance, accuracy, and data processing capabilities of the gas monitoring system.

[0029] Specifically, the specific process of separating the overlapping absorption spectra of multiple gases is as follows: the main features of the gas spectrum restoration value are extracted through principal component analysis, which effectively reduces the data dimension, reduces the interference of noise, and improves the efficiency and accuracy of subsequent processing. The gas spectrum restoration value after dimensionality reduction by principal component analysis is used as input, and the overlapping absorption component matrix in the gas spectrum restoration value is decomposed into non-negative and independent basic absorption components through non-negative matrix decomposition and independent component analysis. The individual absorption characteristics of each gas and the concentration distribution of each gas at different observation points are decoded, avoiding interference between different gases and improving the accurate identification and separation of multiple gas sources.

[0030] In this implementation, principal component analysis (PCA) reduces the dimensionality of gas spectral data, thereby reducing the complexity and computational effort of data processing and improving computational efficiency. Furthermore, non-negative matrix factorization (NMF) and independent component analysis (ICA) effectively separate the overlapping absorption spectra of multiple gases, accurately extracting the individual absorption characteristics and concentration distribution of each gas. This provides a reliable basis for the precise identification and source location of multiple gases. This method significantly improves the ability to detect multiple gas leak sources in complex environments, enhancing the robustness and accuracy of the system.

[0031] Specifically, the specific process of using the diffusion model to estimate the diffusion trajectory of gas under the influence of wind speed and other environmental conditions, and calculating the estimated value of the gas diffusion coefficient and the estimated value of the diffusion rate is as follows: based on the gas monitoring data, a Gaussian diffusion model is established to estimate the diffusion trajectory under the influence of wind speed, temperature and humidity, and the gas diffusion path and air mass boundary in a complex environment are simulated; based on the gas monitoring data, the horizontal coordinate position, vertical coordinate position, time information and gas concentration in the monitoring area of ​​the gas leakage location are obtained, and the sum of the second-order derivative of the gas concentration in the horizontal direction and the second-order derivative in the vertical direction in the two-dimensional space is calculated to obtain the Laplace operator, the Laplace operator is taken as the absolute value and then the inverse is calculated, the first-order partial derivative of the gas concentration with respect to time is calculated to obtain the gas concentration change rate and take the absolute value, and the above two factors are multiplied to obtain the estimated value of the gas diffusion coefficient; based on the gas concentration in the two-dimensional space, the horizontal and vertical directions are obtained. The rate of change in the horizontal and vertical directions is used to obtain the gas concentration gradient, and the ratio of the estimated gas diffusion coefficient to the gas concentration is multiplied by the gas concentration gradient to obtain the estimated gas diffusion rate; the gas diffusion rate is compared with the diffusion threshold in real time. When the estimated gas diffusion rate is greater than the diffusion threshold, it is considered that there is a gas leak in the area, and the laser scanner is controlled to encrypt the scanning frequency and wavelength range in the area, focusing on high-risk areas. Based on the current diffusion rate, it is predicted that the area may be affected at the next moment, and a reminder for further inspection is sent to relevant personnel; when the estimated gas diffusion rate is less than or equal to the diffusion threshold, it is considered that there is no gas leak in the area, and the sensor data is automatically checked for errors, data loss and errors. If the data is abnormal, the sensor is calibrated and reset, and the data in the area is cached for a time window. It is considered safe only when it is continuously below the threshold within the time window.

[0032] The estimated gas diffusion coefficient is obtained by the following method: = Where, is the horizontal position coordinate in the gas leakage monitoring area; Indicates the longitudinal position coordinates in the gas leakage monitoring area; is the time variable; represents the estimated value of gas diffusion coefficient; is the Laplace operator of concentration; Indicates the local time rate of change of gas concentration.

[0033] The estimated gas diffusion rate is obtained by the following method: = Where, is the horizontal position coordinate in the gas leakage monitoring area; Indicates the longitudinal position coordinates in the gas leakage monitoring area; is the time variable; represents the estimated value of gas diffusion rate; represents the estimated value of gas diffusion coefficient; represents the gas concentration at the spatial position (x, y) and time t; The gas concentration gradient represents the rate of change of concentration with spatial position.

[0034] In this implementation, by establishing a Gaussian diffusion model and incorporating environmental factors such as wind speed, temperature, and humidity, it is possible to accurately estimate the diffusion trajectory and air mass boundaries of gases in complex environments, thereby improving the accuracy and real-time performance of gas leak detection. The calculated gas diffusion coefficient and diffusion rate estimates provide a scientific basis for determining leak areas, particularly in high-risk areas, enabling enhanced detection accuracy through intensified scanning. Furthermore, a real-time comparison mechanism based on diffusion rates can rapidly identify potential leak areas and automatically adjust monitoring strategies based on different risk levels, ensuring efficient system response and stability. Verifying the reliability of sensor data and automatically calibrating the system enhances the system's adaptability and fault tolerance.

[0035] Specifically, the gas spectral features and diffusion features are fused to output the gas concentration estimate. The specific process for further distinguishing the gas type is as follows: the spectral feature data and diffusion feature data are processed separately using a dual-branch network algorithm, the spectral feature data and diffusion feature data are fused through the self-attention mechanism to obtain a new fused feature representation, the spectral feature data and diffusion feature data are input to establish a neural network model, the fully connected layer weight matrix is ​​obtained through training, and the bias term is set when the neural network is initialized, the fused feature representation is multiplied by the fully connected layer weight matrix and the bias term is added, the gas diffusion coefficient estimate is multiplied by the diffusion behavior weight and a constant of one is added, and finally the above two factors are multiplied to obtain the gas concentration estimate; the gas concentration estimate and the concentration threshold, the gas diffusion rate and the diffusion threshold are compared in real time. When the gas concentration estimate is greater than the gas concentration threshold and the gas diffusion rate estimate is greater than the diffusion rate threshold, the high concentration and high diffusion of the gas is determined to be an obvious leak. The system automatically increases the sampling frequency, pushes the leak alarm to the safety system, and determines the gas type according to the concentration distribution and diffusion feature data of each gas in combination with the gas absorption spectrum, and enters the leak source location and Multiple gas cloud separation step; when the gas concentration estimate is greater than the gas concentration threshold and the gas diffusion rate estimate is less than or equal to the diffusion rate threshold, the gas high concentration and low diffusion is judged to be a near-source high-concentration concentrated leak, a small-scale laser rescan is activated, the key areas of leaking gas accumulation are continuously tracked, equipment failure and interface loosening inspection reminders are sent to relevant personnel, and the leak source location and multiple gas cloud separation step is entered; when the gas concentration estimate is less than or equal to the gas concentration threshold and the gas diffusion rate estimate is greater than the diffusion rate threshold, the gas low concentration and high diffusion is judged to be an early rarefied leak, the early warning mechanism is automatically activated, the phenomenon is recorded as a warning event, the leakage area is marked as a "suspected leak zone", the monitoring of the area is intensified in the next cycle, and the leak source location and multiple gas cloud separation step is entered; when the gas concentration estimate is less than or equal to the gas concentration threshold and the gas diffusion rate estimate is less than or equal to the diffusion rate threshold, the gas low concentration and low diffusion is judged to be no obvious leak, the laser scanning frequency and spectral sampling resolution are reduced, the area is automatically adjusted to a low-concern priority area, and high-resolution sampling is automatically restored once every fixed period for verification.

[0036] The estimated gas concentration is obtained by the following method: (x,y,t)=( )⋅(1+ (x,y,t)) Where, is the horizontal position coordinate in the gas leakage monitoring area; Indicates the longitudinal position coordinates in the gas leakage monitoring area; is the time variable; (x, y, t) represents the estimated gas concentration; represents the weight matrix of the fully connected layer obtained by training the gas monitoring data using the neural network model; Represents the bias term obtained by initializing the neural network model; Indicates the spectrum restoration value, i.e., the spectrum characteristic data; Represents diffusion characteristic data; represents the fused feature representation obtained by fusion through the self-attention mechanism; (x, y, t) represents the estimated gas diffusion coefficient; represents the diffusion behavior weight factor obtained by cross-validation based on gas diffusion characteristic data, The range is between 0 and 1.

[0037] In this implementation, by integrating gas spectral and diffusion characteristics, it is possible to accurately estimate gas concentrations and accurately distinguish gas types based on their diffusion behaviors. This process enhances the intelligence of gas leak detection, enabling the identification of the nature and scale of gas leaks based on real-time data, such as high concentrations with high diffusion or low concentrations with high diffusion. This allows for dynamic adjustment of monitoring strategies, optimizing sampling frequency and area scanning. During gas leak detection, the system automatically identifies potential leak sources and, based on gas type and concentration distribution, rapidly responds to varying risk levels. Precise gas source location and multi-gas cloud separation ensure system efficiency and accuracy. Furthermore, the system's adaptive capabilities and early warning mechanisms enhance fault detection and prevention capabilities, ensuring safe operation.

[0038] Specifically, the specific process of reversely inferring the gas leakage source based on the gas concentration gradient is as follows: reversely inferring the gas leakage source based on the gas concentration gradient, accurately identifying the source position of the gas leakage, calculating the first-order partial derivative of the gas concentration in the horizontal direction and the first-order partial derivative in the vertical direction in two-dimensional space, multiplying the first-order partial derivative in the horizontal direction with the unit vector in the horizontal direction, multiplying the first-order partial derivative in the vertical direction with the unit vector in the vertical direction, and summing the two items to obtain the gradient vector in the two-dimensional space, and integrating the gradient vector to obtain the gas leakage source.

[0039] The gas leakage source point is obtained by the following method: ( , )= + ) Where, is the horizontal position coordinate in the gas leakage monitoring area; Indicates the longitudinal position coordinates in the gas leakage monitoring area; represents the time variable; ( , ) indicates the source of gas leakage; Represents the gas concentration at the spatial position (x, y) and time t; Represents the unit vector in the horizontal direction; A unit vector representing the vertical direction.

[0040] Several gas concentration distribution points are randomly set. According to the gas concentration at different positions, the gradients of the gas concentration in the horizontal and vertical directions are calculated, and the gas leakage source point is calculated, as shown in Table 1, Gas Concentration Distribution and Leakage Source Location Data Table of Gas Leak Detection System.

[0041] like Figure 3 As shown in Table 1 and Table 2, a gas concentration distribution and leakage source location diagram of a visual gas leakage detection system and method based on laser spectroscopy is provided in this application example. Figure 3 It can be seen that the location of the leakage source can be inferred based on the different gas concentration distributions at different spatial locations.

[0042] In this implementation, by reversely inferring the source of a gas leak, the leak's origin can be precisely located and the gas diffusion path accurately traced in two-dimensional space. This method combines gas concentration gradients and spatial position to effectively infer the leak's origin through integral calculations, providing high-precision support for leak source location. This precise location method improves the leak detection system's response speed and decision-making accuracy, particularly in complex environments. It can accurately identify and locate the source of a gas leak, thereby enhancing the system's intelligence and automation.

[0043] Specifically, gas cluster tracking is carried out to separate different gas leakage sources: the calculated gas leakage source point is used to automatically mark the possible leakage area, providing a basis for further positioning and analysis, and the Kalman filter diffusion model is used to track the diffusion path of the gas cloud in real time to predict the future distribution of the gas cloud. The changes in the gas source point and diffusion path dynamically adjust the monitoring point and monitoring frequency to ensure that other problems in the diffusion process are discovered in time; the leakage source is confirmed by tracking the diffusion path of the gas cloud to verify whether the source point is consistent with the actual leakage point; the density clustering algorithm is used to cluster the gas concentration clouds in space, distinguish different gas clouds, and separate the leakage sources of different gases, ensuring that each gas source point can be correctly tracked, and the leakage sources of different gases are displayed separately. According to the gas type and concentration, emergency response measures are automatically formulated, including ventilation, isolation and evacuation, and sent to relevant personnel.

[0044] In this implementation, the process of tracking and separating different gas leak sources by gas clusters can effectively improve the accuracy of leak source identification and response efficiency. The Kalman filter diffusion model helps track the diffusion path of the gas cloud in real time and adjusts the monitoring points and frequency based on dynamic changes to ensure that potential problems in the diffusion process are promptly discovered. The density clustering algorithm is used to accurately distinguish different gas clouds, ensuring that each gas leak source can be accurately tracked and displayed. This method enhances the ability to locate and monitor leak sources in multi-gas environments, supports personalized emergency response measures for each gas type and concentration, and improves the efficiency of safety precautions and accident handling.

[0045] Specifically, the specific process of generating a scatter plot to display the gas concentration at different locations and marking the location of the leakage source with annotations is as follows: draw a scatter plot of the gas concentration, visualize the gas concentration at each coordinate point, and reflect the high and low concentrations through color to help users quickly identify the distribution of gas concentrations, mark the location of the leakage source, and use different symbols and colors to highlight the coordinates of the leakage source, making the location of the leakage source more intuitive and easy to understand.

[0046] In this implementation, by generating a scatter plot of gas concentration and annotating the leak source, we can visually demonstrate the distribution and changing trends of gas concentrations, helping users quickly identify gas leak areas. This visualization method improves data interpretability and enhances the real-time and responsiveness of the monitoring system, helping decision-makers to promptly locate the leak source and implement effective emergency measures.

[0047] Reference Figure 2 As shown, the second aspect of the present invention provides a laser spectroscopy-based visualized gas leak detection system, which is applied to the above-mentioned laser spectroscopy-based visualized gas leak detection method, and includes a gas monitoring data preprocessing module, an absorption feature extraction module, a spectrum and diffusion information fusion and identification module, a leakage source positioning and multi-gas cloud separation module, and a visualization module: wherein the gas monitoring data preprocessing module is used to collect and preprocess gas monitoring data; the absorption feature extraction module is used to extract the characteristic spectrum of the gas monitoring data, obtain the gas spectrum restoration value based on the gas monitoring data, and separate the overlapping absorption spectra of multiple gases; the spectrum and diffusion information fusion and identification module is used to estimate the diffusion trajectory of the gas under the influence of wind speed and other environmental conditions using a diffusion model, and calculate the estimated value of the gas diffusion coefficient and the estimated value of the diffusion rate; the gas spectrum feature is fused with the diffusion feature to output the gas concentration estimate to further distinguish the gas type; the leakage source positioning and multi-gas cloud separation module is used to reversely infer the gas leakage source point based on the gas concentration gradient, and to track gas clusters to separate different gas leakage sources; the visualization module is used to generate a scatter plot to display the gas concentration at different locations and mark the location of the leakage source with annotations.

[0048] In this implementation, the collaborative work of multiple modules enables comprehensive and accurate monitoring, analysis, and location of gas leaks. In the gas monitoring data preprocessing module, data cleaning and feature extraction provide high-quality data input for subsequent analysis. The absorption feature extraction module extracts key absorption features and separates overlapping absorption spectra, ensuring accurate identification of the presence of different gases. The spectrum and diffusion information fusion identification module combines environmental impact and diffusion models to accurately estimate gas concentration and diffusion rate, improving the accuracy of gas concentration and leak source determination. The leak source location and multi-gas cloud separation module accurately locates leak sources and effectively separates different gas clouds through gradient back-calculation and gas cluster tracking. The visualization module enhances the visibility and readability of monitoring data through scatter plots and annotation tags, helping decision makers quickly respond to emergencies. The combination of these modules improves the comprehensiveness, real-time nature, and accuracy of gas leak detection and response.

[0049] It should be noted that, in this document, relational terms such as first and second, etc., are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the terms "comprises," "comprising," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not explicitly listed, or elements inherent to such process, method, article, or apparatus.

[0050] The preferred embodiments of the present invention disclosed above are intended only to help illustrate the present invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the present invention to the specific embodiments described. Obviously, many modifications and variations are possible based on the content of this specification. These embodiments are selected and described in detail in this specification to better explain the principles and practical applications of the present invention, thereby enabling those skilled in the art to better understand and utilize the present invention. The present invention is limited only by the claims and their full scope and equivalents.

Claims

1. A visual gas leak detection method based on laser spectroscopy, characterized in that: The following steps are involved: Step 1: Collect gas monitoring data and pre-process the gas monitoring data; Step 2: Extract characteristic spectra from the gas monitoring data, obtain gas spectrum restoration values ​​based on the gas monitoring data, and separate overlapping absorption spectra of multiple gases; Step 3: Use the diffusion model to estimate the diffusion trajectory of the gas under the influence of wind speed and other environmental conditions, and calculate the estimated value of the gas diffusion coefficient and diffusion rate. Fusion of the gas spectral characteristics with the diffusion characteristics to output the estimated gas concentration value, and further distinguish the gas type. Step 4: Reversely infer the gas leakage source based on the gas concentration gradient, and track the gas clusters to separate different gas leakage sources; Step 5: Generate a scatter plot to show the gas concentration at different locations and mark the location of the leak source with annotations.

2. The method for visual gas leak detection based on laser spectroscopy according to claim 1, characterized in that: The specific process of collecting gas monitoring data and preprocessing the gas monitoring data is as follows: Gas monitoring data includes spectral characteristic data and diffusion characteristic data. The spectral characteristic data is obtained by acquiring the gas absorption spectrum and absorption intensity through the sensor array. The diffusion characteristic data of the gas is obtained by synchronously collecting time, gas spatial position, wind speed, temperature and humidity, and gas concentration. Gaussian filtering is used to smooth the laser spectrum curve and remove noise from the gas monitoring data. The adaptive baseline fitting method is used for baseline correction. The gas monitoring data is normalized using the linear normalization method. Through standardization, each gas monitoring data feature has the same scale.

3. The method for visual gas leak detection based on laser spectroscopy according to claim 1, characterized in that: The specific process of extracting the characteristic spectrum of the gas monitoring data and obtaining the gas spectrum restoration value according to the gas monitoring data is as follows: Based on the derivative spectroscopy method, the significant absorption peaks in the gas absorption spectrum are extracted to obtain the gas absorption spectrum, and the gas spectrum reduction value is obtained according to the gas monitoring data; The specific analysis process of the gas spectrum restoration value is as follows: the absorption intensity of each gas at different wavelengths is obtained according to the spectral characteristic data to obtain the gas absorption spectrum, the gas concentration at the spatial position (x, y) and time t is obtained according to the diffusion characteristic data, and the absorption spectrum of each gas is multiplied by its corresponding gas concentration and the sum is calculated to obtain the gas spectrum restoration value.

4. The method for visual gas leak detection based on laser spectroscopy according to claim 1, characterized in that: The specific process of separating the overlapping absorption spectra of multiple gases is as follows: The main features of the gas spectrum restoration value are extracted by principal component analysis, and the gas spectrum restoration value after principal component analysis dimensionality reduction is used as input. The overlapping absorption component matrix in the gas spectrum restoration value is decomposed into non-negative and independent basic absorption components through non-negative matrix decomposition and independent component analysis, decoding the individual absorption characteristics of each gas and the concentration distribution of each gas at different observation points.

5. The method for visual gas leak detection based on laser spectroscopy according to claim 1, characterized in that: The specific process of using the diffusion model to estimate the diffusion trajectory of the gas under the influence of wind speed and other environmental conditions, and calculating the estimated value of the gas diffusion coefficient and the estimated value of the diffusion rate is as follows: A Gaussian diffusion model is established based on gas monitoring data to estimate diffusion trajectories under the influence of wind speed, temperature and humidity, and to simulate gas diffusion paths and air mass boundaries in complex environments. The horizontal and vertical positions, time information, and gas concentration of the gas leak location in the monitoring area are obtained based on the gas monitoring data. The sum of the second-order derivative of the gas concentration in the horizontal direction and the second-order derivative in the vertical direction in two-dimensional space is calculated to obtain the Laplace operator. The absolute value of the Laplace operator is taken and the inverse is calculated. The first-order partial derivative of the gas concentration with respect to time is calculated to obtain the gas concentration change rate and the absolute value is taken. The above two factors are multiplied to obtain the estimated value of the gas diffusion coefficient. The gas concentration gradient is obtained according to the rate of change of the gas concentration in the horizontal and vertical directions in the two-dimensional space, and the estimated gas diffusion rate is obtained by multiplying the ratio of the estimated gas diffusion coefficient to the gas concentration by the gas concentration gradient; The system compares the gas diffusion rate with the diffusion threshold in real time. When the estimated gas diffusion rate exceeds the diffusion threshold, it is considered that there is a gas leak in the area. The laser scanner is controlled to increase the scanning frequency and wavelength range in the area, focusing on high-risk areas. Based on the current diffusion rate, the system predicts the areas that may be affected in the next moment and sends a reminder to relevant personnel for further inspection. When the estimated gas diffusion rate is less than or equal to the diffusion threshold, it is considered that there is no gas leakage in the area. The sensor data is automatically checked for errors, data loss and errors. If the data is abnormal, the sensor is calibrated and reset. The data of the area is cached for a time window. It is considered safe only when the value is continuously below the threshold within the time window.

6. The method for visual gas leak detection based on laser spectroscopy according to claim 1, characterized in that: The specific process of fusing the gas spectral characteristics with the diffusion characteristics, outputting the gas concentration estimation value, and further distinguishing the gas types is as follows: A dual-branch network algorithm is used to process spectral feature data and diffusion feature data separately. The spectral feature data and diffusion feature data are fused through the self-attention mechanism to obtain a new fused feature representation. The spectral feature data and diffusion feature data are input to establish a neural network model. The fully connected layer weight matrix is ​​trained and a bias term is set during neural network initialization. The fused feature representation is multiplied by the fully connected layer weight matrix and the bias term is added. The gas diffusion coefficient estimate is multiplied by the diffusion behavior weight and a constant of one is added. Finally, the gas concentration estimate is multiplied by the above two factors. The system compares the estimated gas concentration with the concentration threshold, and the gas diffusion rate with the diffusion threshold in real time. When the estimated gas concentration is greater than the gas concentration threshold and the estimated gas diffusion rate is greater than the diffusion rate threshold, it determines that the high concentration and high diffusion of gas is an obvious leak. The system automatically increases the sampling frequency, pushes the leak alarm to the safety system, and determines the gas type based on the concentration distribution and diffusion characteristic data of each gas in combination with the gas absorption spectrum. It then enters the leak source location and multi-gas cloud separation steps; When the estimated gas concentration is greater than the gas concentration threshold and the estimated gas diffusion rate is less than or equal to the diffusion rate threshold, the high-concentration, low-diffusion gas is judged to be a near-source, high-concentration, concentrated leak. A small-scale laser rescan is activated to continuously track the key areas where the leaking gas is concentrated. Equipment failure and loose interface inspection reminders are sent to relevant personnel, and the leak source location and multi-gas cloud separation steps are initiated. When the estimated gas concentration is less than or equal to the gas concentration threshold and the estimated gas diffusion rate is greater than the diffusion rate threshold, the low-concentration, high-diffusion gas is judged to be an early-stage rarefaction leak. The early warning mechanism is automatically activated, the phenomenon is recorded as a warning event, and the leakage area is marked as a "suspected leakage zone." In the next cycle, monitoring of this area will be intensified, and the leak source location and multi-gas cloud separation steps will be entered. When the estimated gas concentration is less than or equal to the gas concentration threshold and the estimated gas diffusion rate is less than or equal to the diffusion rate threshold, the low gas concentration and low diffusion are judged to be without obvious leakage, the laser scanning frequency and spectral sampling resolution are reduced, the area is automatically adjusted to a low-concern priority area, and high-resolution sampling is automatically restored every fixed period for verification.

7. The method for visual gas leak detection based on laser spectroscopy according to claim 1, characterized in that: The specific process of reversely calculating the gas leakage source based on the gas concentration gradient is as follows: The gas leakage source point is inferred based on the gas concentration gradient. The first-order partial derivative of the gas concentration in the horizontal direction and the first-order partial derivative in the vertical direction in two-dimensional space are calculated. The first-order partial derivative in the horizontal direction is multiplied by the unit vector in the horizontal direction, and the first-order partial derivative in the vertical direction is multiplied by the unit vector in the vertical direction. The two terms are summed to obtain the gradient vector in two-dimensional space. The gradient vector is integrated to obtain the gas leakage source point.

8. The method for visual gas leak detection based on laser spectroscopy according to claim 1, characterized in that: The above also conducts gas cluster tracking and separates different gas leakage sources: The calculated gas leakage source is used to automatically mark the possible leakage area, and the Kalman filter diffusion model is used to track the diffusion path of the gas cloud in real time to predict the future distribution of the gas cloud. The monitoring points and monitoring frequency are dynamically adjusted according to the changes in the gas source and diffusion path. Confirm the leak source by tracing the diffusion path of the gas cloud and verify whether the source point is consistent with the actual leak point; Use density clustering algorithm to cluster gas concentration clouds in space, distinguish different gas clouds, separate the leakage sources of different gases, and display the leakage sources of different gases respectively. According to the gas type and concentration, emergency response measures including ventilation, isolation and evacuation are automatically formulated and sent to relevant personnel.

9. The method for visual gas leak detection based on laser spectroscopy according to claim 1, characterized in that: The specific process of generating a scatter plot showing the gas concentration at different locations and marking the location of the leak source with annotations is as follows: Draw a scatter plot of gas concentration to visualize the gas concentration at each coordinate point and use color to reflect the high and low concentrations. Mark the location of the leak source and use different symbols and colors to highlight the coordinates of the leak source.

10. The laser spectroscopy-based visual gas leak detection system according to claim 1, characterized in that: It includes gas monitoring data preprocessing module, absorption feature extraction module, spectrum and diffusion information fusion identification module, leakage source positioning and multi-gas cloud separation module and visualization module: The gas monitoring data preprocessing module is used to collect gas monitoring data and preprocess the gas monitoring data; The absorption feature extraction module is used to extract the characteristic spectrum of the gas monitoring data, obtain the gas spectrum restoration value based on the gas monitoring data, and separate the overlapping absorption spectra of multiple gases; The spectrum and diffusion information fusion identification module is used to estimate the diffusion trajectory of the gas under the influence of wind speed and other environmental conditions using a diffusion model, and calculate the estimated value of the gas diffusion coefficient and diffusion rate; it fuses the gas spectral characteristics with the diffusion characteristics, outputs the gas concentration estimate, and further distinguishes the gas type; The leakage source positioning and multi-gas cloud separation module is used to reversely infer the gas leakage source point based on the gas concentration gradient, and to track the gas clusters to separate different gas leakage sources; The visualization module is used to generate a scatter plot to display the gas concentration at different locations and mark the location of the leakage source with annotations.

Images