A method for constructing a gas leakage infrared imaging dataset based on computational fluid dynamics

Through the construction method of gas leakage infrared imaging data set based on computational fluid mechanics, three-dimensional gas concentration data is generated, which solves the problem that two-dimensional data sets cannot reflect spatial scenes in the existing technology, and achieves a higher reliability data set, which is suitable for Complex gas leakage scenarios and reduce production costs.

CN118586308BActive Publication Date: 2024-11-29UNIV OF ELECTRONICS SCI & TECH OF CHINA
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202410639497.5
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-05-22
Publication Date
2024-11-29
Estimated Expiration
2044-05-22

AI Technical Summary

Technical Problem

The existing gas leakage data sets produced using computational fluid mechanics methods only use two-dimensional concentration data, which cannot reflect the complete concentration information of the spatial scene, and it is difficult to adapt to complex gas leakage scenes. The cost of building a field scene on its own is high and poses safety risks.

Method used

The gas leakage infrared imaging data set construction method based on computational fluid mechanics is adopted to create a more reliable and realistic grayscale image data by establishing a three-dimensional physical model, simulating the gas leakage process, and using optical gas imaging technology to generate three-dimensional gas concentration data.

Benefits of technology

It significantly reduces the cost of data set production, improves the reliability and authenticity of the data set, can be used in traditional leakage classification and positioning tasks, and can use the gas concentration path length to perform training tasks related to leakage concentration.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118586308B_ABST
    Figure CN118586308B_ABST
Patent Text Reader

Abstract

The present invention discloses a method for constructing a gas leakage infrared imaging data set based on computational fluid dynamics, which belongs to the technical field of gas leakage detection and prevention and control, and specifically comprises the following steps: establishing a three-dimensional physical model of a gas leakage field scene, simulating it based on the principle of computational fluid dynamics after grid division, obtaining the mole fraction of the leaked gas of each grid under the time step, and forming the three-dimensional gas concentration data corresponding to each frame; using optical gas imaging based on a pinhole camera model, the three-dimensional gas concentration data is imaged and processed into an initial image, and the gas concentration path length corresponding to each pixel point in each frame of the initial image is calculated, and after normalization to the maximum and minimum values, each frame of grayscale image is synthesized, thereby constructing a gas leakage infrared imaging data set. The present invention is not only used for leakage classification tasks and positioning tasks, but also for training tasks related to leakage concentration. The synthesized grayscale image is more reliable and authentic, which solves the problem of the small number of public data sets for gas leakage infrared imaging and significantly reduces the cost of data set production.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of gas leakage detection and prevention, and in particular relates to a method for constructing a gas leakage infrared imaging data set based on computational fluid dynamics. Background Art

[0002] Gas leak detection technology is a key technology used to monitor and identify potential hazards in industrial environments. With the continuous development of my country's chemical industry, production safety issues are becoming increasingly urgent. Dangerous gas leaks, such as oil and natural gas, may not only cause major safety accidents such as explosions and poisoning, but may also have long-term negative impacts on the natural environment, causing massive economic losses and casualties. Therefore, fast and effective gas leak detection technology is essential to ensure industrial production safety and reduce environmental pollution.

[0003] Gas leak detection technology based on deep learning has been widely used in recent years. It specifically includes the use of gas infrared imaging and optical flow algorithms to locate leak sources and measure gas diffusion rates through video sequences, as well as measuring time-series concentration data through spectral and concentration detectors to construct data sets for training neural networks for gas concentration prediction and accident classification.

[0004] On the one hand, the cost of installing a large number of sensor networks is difficult to estimate, and it is more suitable for fixed working scenes within a certain range, and it is difficult to deal with sudden gas leaks on long pipelines. On the other hand, although the method of using infrared imaging to monitor gas leaks is more flexible and can quickly locate leaks and classify accident levels, there is a lack of relevant public data sets for algorithm training. Building a large number of field scenes to shoot video data sets for training still requires high costs and has certain safety risks.

[0005] Therefore, using computational fluid dynamics to build a simulated gas leakage scenario can reduce training costs and avoid experimental risks. However, the current datasets produced using computational fluid dynamics methods only use two-dimensional concentration data of the cross section, which cannot reflect the complete concentration information of the spatial scene and is difficult to adapt to complex gas leakage scenarios. The main application is limited to leakage classification and location tasks. Summary of the invention

[0006] In response to the problems existing in the existing data sets produced using computational fluid dynamics methods, the present invention provides a method for constructing a gas leakage infrared imaging data set based on computational fluid dynamics. It uses optical gas imaging to synthesize images that reflect three-dimensional gas concentration data, which has higher reliability and authenticity, and obtains gas concentration data of pixel points in each frame of the image, which can be used for training tasks related to leakage concentration and significantly reduces the cost of data set production.

[0007] The technical solution adopted by the present invention is as follows:

[0008] A method for constructing a gas leakage infrared imaging data set based on computational fluid dynamics comprises the following steps:

[0009] S1. In the field scene of gas leakage, collect the geometric structure dimensions and leakage aperture of the field pipeline, and then establish a three-dimensional physical model of the field scene of gas leakage;

[0010] S2, meshing the three-dimensional physical model, and determining the inlet surface, outlet surface and wall surface of the three-dimensional physical model;

[0011] S3. Based on the principle of computational fluid dynamics, the three-dimensional physical model after meshing is simulated, and the mole fraction of the leaked gas of each mesh under the time step is obtained by simulation by setting the inlet velocity of at least one group of inlet faces, so as to form the three-dimensional gas concentration data corresponding to each frame;

[0012] During the simulation, the components of the leaked gas and ambient gas determined in the actual gas leak scenario are used to construct the component transport equation, and the turbulence model is selected based on the geometric structure of the on-site pipeline and the flow rate of the leaked gas;

[0013] S4, using optical gas imaging based on a pinhole camera model to image the three-dimensional gas concentration data into an initial image, and calculating the gas concentration path length (concentration path-length) corresponding to each pixel point in each frame of the initial image;

[0014] S5. Perform maximum-minimum normalization on the gas concentration path lengths corresponding to all pixels in the initial image, and synthesize each frame of grayscale image according to the normalization result; construct a gas leakage infrared imaging dataset based on the obtained grayscale image and the gas concentration path lengths corresponding to each pixel therein.

[0015] Furthermore, the size of the grid divided in S2 is 1 / 100 to 1 / 1000 of the geometric structure size of the on-site pipeline.

[0016] Furthermore, the grids divided in S2 are polyhedrons such as tetrahedrons or hexahedrons.

[0017] Furthermore, during the simulation of S3, the Courant number is required not to exceed 2.

[0018] Furthermore, the turbulence model in S3 is a Standard k-ε model, a RNG (Re-normalization group) k-ε model, a Realizable k-ε model, a Standard k-ω model or a SST (Shear Stress Transport) k-ω model.

[0019] Furthermore, the mole fraction of the leaked gas in each grid in S3 is equal to the ratio of the amount of substance of the leaked gas to the amount of substance of all gases in the grid.

[0020] Furthermore, in the imaging process of S4, the gas concentration path length corresponding to each pixel point is obtained by performing path integration on the leaked gas mole fractions of all grids included in each pixel point.

[0021] Furthermore, before the imaging process of S4, the three-dimensional gas concentration data is interpolated by nearest neighbor to expand the data nodes and improve the photo quality.

[0022] Compared with the prior art, the present invention has the following beneficial effects:

[0023] 1. The present invention proposes a method for constructing a gas leakage infrared imaging dataset based on computational fluid dynamics, which solves the problem of the small number of public gas leakage infrared imaging datasets. Compared with building a large number of field scenes to shoot video datasets by oneself, the present invention can significantly reduce the dataset production cost;

[0024] 2. The present invention uses optical gas imaging based on a pinhole camera model to perform imaging processing on three-dimensional gas concentration data, and then synthesizes grayscale images. Compared with a data set produced using only two-dimensional cross-sectional concentration data, the grayscale images synthesized by the present invention have higher reliability and authenticity;

[0025] 3. The gas leakage infrared imaging dataset constructed by the present invention is not only used for traditional leakage classification tasks and positioning tasks, but also can be used for training tasks related to leakage concentration by utilizing the gas concentration path length of each pixel point. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a flow chart of the method for constructing a gas leakage infrared imaging data set based on computational fluid dynamics proposed in Example 1;

[0027] Figure 2 It is a schematic diagram of the structure of the three-dimensional physical model in Example 1;

[0028] Figure 3 The three-dimensional physical model after the Fluent is imported in Example 1;

[0029] Figure 4 Schematic diagram of the pinhole camera model used in Example 1;

[0030] Figure 5 A grayscale picture frame synthesized in Example 1;

[0031] Figure 6 For Example 1 Figure 5 2D cross-sectional concentration grayscale image of the same frame. DETAILED DESCRIPTION

[0032] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with embodiments and drawings. The exemplary embodiments of the present invention and their description are only used to explain the present invention and are not intended to limit the present invention.

[0033] Example 1

[0034] This embodiment proposes a method for constructing a gas leakage infrared imaging dataset based on computational fluid dynamics. The process is as follows: Figure 1 As shown, the following steps are included:

[0035] S1. In the actual gas leakage scene, collect the geometric structure dimensions and leakage aperture of the on-site pipeline, and then establish a three-dimensional physical model of the actual gas leakage scene in the XYZ rectangular coordinate system in the geometric modeling software (such as Discovery), that is, Figure 2 The dimensions of the gas box shown in the X, Y, and Z directions are 0.8 m, 0.5 m, and 0.7 m, respectively;

[0036] S2. Import the 3D physical model into meshing software (such as Ansys Meshing), perform tetrahedral meshing on the 3D physical model, set the maximum mesh size to 0.005 m, and determine the inlet surface, outlet surface and wall surface of the 3D physical model;

[0037] S3, import the meshed 3D physical model into computational fluid dynamics simulation software (such as Fluent), and obtain the following Figure 3 The three-dimensional physical model shown in the figure undergoes a pre-processing process, specifically:

[0038] Select transient calculation, turn on gravity, and set the acceleration in the z direction to -9.81m / s 2 ; Select the SST k-ω model as the turbulence model; construct the component transport equation, set the components of the leaked gas and the ambient gas to methane and air respectively, select incompressible gas for the density term, set the methane inlet velocity at the inlet surface to 2.5m / s, the initial methane component ratio to 1, the temperature to 300K, and the initial methane component ratio at the outlet surface to 0;

[0039] After the standard initialization pre-processing process, the PISO (Pressure-Implicit with Splitting of Operators) algorithm is used to solve the calculation with a time step of 0.005s. The ASCII format is selected during the solution process, and the grid center option is used to save the simulation data. By setting the inlet velocity of at least one group of inlet surfaces, the leakage gas mole fraction of each grid under the time step is simulated to form the three-dimensional gas concentration data corresponding to each frame.

[0040] S4. Import the three-dimensional gas concentration data into mathematical software (such as MATLAB). Considering that the three-dimensional gas concentration data generated by simulation is uneven, the data points at the edge of the three-dimensional physical model are denser, and the data points inside are sparser. The image generated by directly using the original three-dimensional gas concentration data without interpolation will have a large number of noise points and poor quality. Therefore, it is necessary to establish an interpolation grid in three-dimensional space and use the nearest neighbor interpolation method to interpolate the concentration data to improve the image quality, and then obtain the interpolated three-dimensional gas concentration data; wherein the established interpolation grid is consistent with the spatial position of the three-dimensional physical model, and the grid division rule is: 0.004m as the step length on x∈[-0.8,0], 0.01m as the step length on y∈[0,0.5], and 0.005m as the step length on z∈[0,0.7];

[0041] S5. Using Figure 4 The optical gas imaging of the pinhole camera model shown in the figure processes the three-dimensional gas concentration data into an initial image and calculates the gas concentration path length corresponding to each pixel point in each frame of the initial image;

[0042] Among them, the lens position of the pinhole camera model is (-0.4, -50, 0.35), and the pixel size is set to 0.005m×0.005m. Since the pixel is very small, the pixel can be regarded as a small square grid on the imaging plane. The four sides of the square grid and the point where the lens is located form four planes respectively. These four planes surround a calculation area. The path integration of the leaked gas mole fraction of all grids in the calculation area (that is, the data in the feasible domain) can be performed to obtain the gas concentration path length corresponding to the pixel point, in units of ppm×m. Each frame of the initial image includes 224×224 pixels. Parfor parallel computing is used in mathematical software to speed up the calculation.

[0043] S6. Normalize the gas concentration path lengths corresponding to all pixels in the initial image to the maximum value and the minimum value. The calculation formula is:

[0044]

[0045] Among them, C is the gas concentration path length corresponding to the current pixel point, C min is the minimum value of the gas concentration path length among all pixels, C max is the maximum value of the gas concentration path length among all pixels, C * is the normalized result;

[0046] And synthesize each frame of grayscale image based on all normalization results, such as Figure 5As shown; based on the obtained grayscale image and the gas concentration path length corresponding to each pixel point therein, a gas leakage infrared imaging video dataset is constructed.

[0047] Figure 6 For Figure 5 The 2D cross-sectional concentration grayscale image of the same frame specifically captures the 2D concentration data at y=0.25m. Figure 6 It only reflects part of the gas leakage situation in the actual gas leakage scene, while this embodiment uses optical gas imaging based on the pinhole camera model to perform imaging processing on the three-dimensional gas concentration data to synthesize a grayscale image, which can cover the complete concentration information of the spatial area and has higher reliability and authenticity.

[0048] At the same time, the video data set constructed in this embodiment also includes a specific gas concentration path length, which can be used for training tasks related to leakage concentration.

[0049] It should be noted that these are only specific implementation modes of the present invention, but the protection scope of the present invention is not limited thereto, and any modifications, equivalent substitutions and improvements made by any technician familiar with the technical field within the technical scope disclosed by the present invention and within the spirit and principles of the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for constructing a gas leakage infrared imaging dataset based on computational fluid dynamics, characterized in that: The following steps are involved: S1. In the field scene of gas leakage, collect the geometric structure dimensions and leakage aperture of the field pipeline, and then establish a three-dimensional physical model of the field scene of gas leakage; S2, meshing the three-dimensional physical model, and determining the inlet surface, outlet surface and wall surface of the three-dimensional physical model; S3. Based on the principle of computational fluid dynamics, the three-dimensional physical model after meshing is simulated, and the mole fraction of the leaked gas of each mesh under the time step is obtained by simulation by setting the inlet velocity of at least one group of inlet faces, so as to form the three-dimensional gas concentration data corresponding to each frame; During the simulation, the components of the leaked gas and ambient gas determined in the actual gas leak scenario are used to construct the component transport equation, and the turbulence model is selected based on the geometric structure of the on-site pipeline and the flow rate of the leaked gas; S4, using optical gas imaging based on a pinhole camera model to process the three-dimensional gas concentration data into an initial image, and calculating the gas concentration path length corresponding to each pixel point in each frame of the initial image; S5. Perform maximum-minimum normalization on the gas concentration path lengths corresponding to all pixels in the initial image, and synthesize each frame of grayscale image according to the normalization result; construct a gas leakage infrared imaging dataset based on the obtained grayscale image and the gas concentration path lengths corresponding to each pixel therein.

2. The method for constructing a gas leakage infrared imaging data set based on computational fluid dynamics according to claim 1, characterized in that: The grid size divided in S2 is 1 / 100 to 1 / 1000 of the geometric structure size of the on-site pipeline.

3. The method for constructing a gas leakage infrared imaging data set based on computational fluid dynamics according to claim 1, characterized in that: During the S3 simulation process, the Courant number is required to be no more than 2.

4. The method for constructing a gas leakage infrared imaging data set based on computational fluid dynamics according to claim 1, characterized in that: The turbulence model described in S3 is the Standard k-ε model, the RNG k-ε model, the Realizable k-ε model, the Standard k-ω model or the SST k-ω model.

5. The method for constructing a gas leakage infrared imaging data set based on computational fluid dynamics according to claim 1, characterized in that: During the imaging process of S4, the gas concentration path length corresponding to each pixel point is obtained by performing path integration on the leaked gas mole fractions of all grids contained in each pixel point.

6. The method for constructing a gas leakage infrared imaging data set based on computational fluid dynamics according to claim 1, characterized in that: Prior to the imaging process in S4, nearest neighbor interpolation is performed on the three-dimensional gas concentration data.

Citation Information

Patent Citations

  • Numerical analog method for impact of gas pipeline leakage on internal flow field

    CN107590336A

  • Actual gas physical property simulation method and system based on computational fluid mechanics

    CN113723030A