Hazardous chemical substance device leakage concentration field dynamic prediction method based on infrared thermal imaging correction

Abstract

The invention relates to a hazardous chemical substance device leakage concentration field dynamic prediction method based on infrared thermal imaging correction. The method comprises the following steps: (1) acquiring real-time parameters from a site; (2) adding a time variable on the basis of the Gaussian calculation model, and carrying out integration on instantaneous leakage in a time infinite range to obtain a critical condition of a concentration field prediction algorithm; (3) carrying out integration on the original model within finite time, and solving an integration result through the obtained critical condition to obtain a new dynamic prediction model of which the leakage influence range changes along with time; (4) calculating to obtain an isoconcentration curve of the new dynamic prediction model; (5) drawing an actual isoconcentration curve by using an edge detection technology in combination with the infrared image and the real-time monitoring data; (6) positioning the same coordinate position point on the theoretical prediction model, correcting the theoretical concentration curve by using the actual concentration curve, performing overall coordinate transformation on the concentration field, and correcting the whole concentration field; and (7) after a certain time, shooting the infrared thermal image again, and correcting the concentration field again. And the accuracy and effectiveness of emergency disposal are improved.

Description

Technical Field

[0001] The present invention belongs to the field of leakage analysis and prediction of hazardous chemicals equipment, and relates to a method for dynamically predicting leakage concentration fields of hazardous chemicals equipment based on infrared thermal imaging correction. Background Art

[0002] The traditional continuous leakage model can only predict the maximum impact range when the accident develops to the most dangerous situation and the concentration field reaches a steady state through static data, and changes the prediction to the prediction result under ideal conditions with no obstruction and constant wind speed and direction. It does not consider the impact of environmental factors such as obstruction of on-site equipment and changes in wind speed and direction on the concentration field due to actual leakage, and cannot reflect the real-time situation and changing process of the accident. On the one hand, it misses the opportunity to deal with the small leakage in the early stage of the accident, and on the other hand, it is easy to cause waste of emergency resources, which does not meet the industry needs of precise prevention and control.

[0003] With the advancement of technology, through infrared thermal imagers and image recognition edge detection technology, it is possible to capture and identify the actual concentration field contour curve in real time. Although the absolute concentration information of image recognition is inaccurate, the relative image gradient recognition result is highly reliable. Combined with the accurate concentration values ​​obtained by online monitoring sensors, the results of the theoretical prediction model can be optimized, and the prediction result errors caused by the ideal state assumptions of the theoretical prediction model can be corrected, so that the corrected concentration values ​​of the entire leakage impact range can be obtained in real time and efficiently.

[0004] Xi'an University of Science and Technology's "Dynamic Evaluation Method for Leakage Range of H2S Transport Pipelines" (Patent No.: ZL202310244925.X) requires a large number of concentration sensors to be evenly arranged on the pipeline, and continuous monitoring requires high power supply endurance. As the length of the pipeline increases, the sensors are densely distributed, and the power consumption continues to increase. At the same time, this method does not clearly define the technical solution for monitoring power supply endurance. First, the engineering transformation of the additional dense installation of a large number of concentration sensors adds safety risks to the pipeline; second, the application cost is high, and the burden on enterprises increases. Third, once an accident occurs, there is a possibility of sensor failure or power supply interruption, making the method impossible to implement.

[0005] Guangxi University's "A calculation model based on the continuous real-time leakage of normal pressure vertical storage tank body" (patent number: ZL201810724745.0) and China Energy Rong'an (Beijing) Technology Co., Ltd., Jilin Electric Power Co., Ltd.'s "A high-pressure heater leakage prediction and analysis method and system" (ZL202210344120.8) can only determine leakage and calculate leakage amount by liquid level changes for liquid media, but cannot solve the problem of gas medium leakage. At the same time, the above methods only perform real-time prediction of leakage warning and leakage amount, and cannot predict the impact range of leakage in real time. Summary of the invention

[0006] Purpose of the invention: In view of the above-mentioned problems existing in the prior art, in order to present the impact range of leakage of hazardous chemicals equipment in real time, seize the opportunity to deal with small leakage in the early stage of the accident, correct the prediction result error caused by the ideal state assumption, carry out targeted emergency rescue, save emergency resources, and achieve precise prevention and control, the present invention proposes a dynamic prediction method for leakage concentration field of hazardous chemicals equipment based on infrared thermal imaging correction.

[0007] Technical solution: In order to achieve the above-mentioned invention objectives, the present invention adopts the following technical solution: a dynamic prediction method for leakage concentration field of hazardous chemical equipment based on infrared thermal imaging correction, the first step is to obtain the real-time pressure, temperature, initial mass, volume, medium density and other parameters of the hazardous chemical equipment from the on-site DCS system; the second step is to add a time variable on the basis of the Gaussian calculation model for the specific working conditions of the hazardous chemical equipment leakage, subdivide the continuous leakage into instantaneous leakage within a small time period, and integrate the instantaneous leakage in the infinite time range to deduce the critical conditions of the concentration field prediction algorithm at this time; the third step is to integrate the original model within a finite time, and resolve the integration result through the critical conditions in step 2 to obtain a new dynamic prediction model in which the leakage influence range changes with time; the fourth step is to calculate the isoconcentration curve of the new dynamic prediction model, that is, the concentration field formula; the fifth step is to correct the concentration field in combination with infrared images and real-time monitoring data. Use an unmanned aerial vehicle equipped with an online optical gas thermal imager to take thermal imaging pictures of the leakage range, locate the boundary points of the concentration field, and draw approximate iso-concentration contour curves based on the principle of drawing lines along the direction of the smallest image gradient in combination with image recognition edge detection technology. Use the on-site DCS system or online monitoring sensor to obtain the actual point concentration value of the contour curve; the sixth step is to locate the same coordinate position point on the theoretical prediction model, use the actual concentration curve to correct the theoretical prediction concentration curve, and perform overall coordinate transformation on the concentration field, repeat the above steps, and correct the entire concentration field. The seventh step is to retake the infrared thermal imaging picture after a certain period of time and correct the concentration field again.

[0008] The specific model construction process includes the following steps:

[0009] Step S1. Real-time acquisition and initialization of relevant parameters. Obtain the real-time pressure, temperature, and hazardous chemical medium parameters (medium type, density, hazardous concentration range) of the hazardous chemical device from the on-site DCS system and online monitoring device, and equipment parameters: total mass, equipment volume, internal height, etc.).

[0010] Step S2. Obtain the analytical solution of the concentration field prediction algorithm under the condition of infinite leakage time. For the specific working condition of leakage of hazardous chemicals, a time variable is added on the basis of the Gaussian calculation model, and the continuous leakage is subdivided into instantaneous leakage in a small time period, and the instantaneous leakage is integrated in the infinite time range. Without considering the three turbulent diffusion coefficients σx, y, and z x, σ y , σ z In the case of the difference between continuous leakage and instantaneous leakage (see Table 1 and Table 2 for details), when t→∞, substituting the leakage diffusion coefficient σ, we can get: It can be seen that when x>1, there is The solution is:

[0011]

[0012] Step S3. Construct a dynamic concentration field prediction model at any time t. Integrate the original model in a finite time, and resolve the integral result through the steady-state critical condition in step 2 to obtain a new dynamic prediction model of the leakage impact range changing with time. x , σ y , σ z In the case of the difference between continuous leakage and instantaneous leakage, the relationship between concentration and coordinate position, time function and special position law can be obtained by integration and simplification:

[0013]

[0014] Step S4. Based on the new prediction model, a prediction algorithm for the leakage impact range is obtained. The isoconcentration curve of the new dynamic prediction model is calculated (take σ = 0.18x 0.92 ), thereby predicting the scope of leakage impact;

[0015]

[0016] Step S5. Use a drone equipped with an online optical gas thermal imager to inspect and shoot near the equipment, obtain infrared thermal imaging images of the actual concentration field on site parallel to the ground (xoy plane), use image edge detection technology to identify the boundary of the concentration field, and draw an approximate iso-concentration contour curve based on the principle of drawing lines along the direction of the smallest image gradient; then use the on-site DCS system or online monitoring sensor to obtain the actual point concentration value of the contour curve. Then, locate the same position point on the theoretical prediction model, use the actual concentration field to correct the theoretical prediction concentration field, and perform an overall coordinate transformation on the concentration field to correct the entire concentration field. The following steps are included:

[0017] (1) Use a drone equipped with an online optical gas thermal imager to inspect and shoot equipment accessories to obtain the optical gas thermal imager's gas sensitivity (ppm*m), infrared resolution (pixel), detector pixel spacing d (μm), focal length fmm, clear imaging range mm, etc.;

[0018] (2) Fly around the alarm device, parallel to the ground (xoy plane), and take thermal imaging pictures of the leaking gas.

[0019] (3) Using image edge detection technology, the absolute value of the identified concentration is inaccurate, but the relative concentration gradient is accurate. The image RGB color change gradient, that is, the amplitude of the image RGB color change, is calculated, and the change rate of the image x-axis and y-axis is calculated. Specifically:

[0020] 1) Assume that I(x,y) represents the image pixel value (such as RGB color), the first-order derivative can be used to obtain the amplitude of change. Here, the derivative should be infinitely close to 0 in order to obtain an infinitesimal partial derivative, but the image cannot take an infinitesimal value. The pixels in the image are discrete, and the minimum distance is 1 pixel. Therefore, the gradient in the x and y directions is:

[0021]

[0022] Gradient size:

[0023]

[0024] Gradient direction:

[0025]

[0026] 2) Considering adjacent positions, the gradient directions include vertical, horizontal, and diagonal directions, with a total of 8 possible values, as shown in Figure 6:

[0027] I(x m ,y m )={I(x n +1,y n )、I(x n ,y n +1)、I(x n -1,y n )、I(x n ,y n -1), I(x n

[0028] +1,y n +1)、I(x n +1,y n -1), I(x n -1,y n -1), I(x n -1,y n

[0029] +1)}

[0030] 3) In order to connect the discrete points into an approximate isoconcentration contour curve, the principle is to draw lines along the direction of the minimum image gradient. The image gradient calculation formula is as follows:

[0031]

[0032] 4) The maximum difference in image gradient between all points on the obtained concentration profile curve should be less than the set acceptable error range G RGB , whose value is obtained based on the statistical results of a series of image recognition and experimental results:

[0033] G(x,y) max ≤D RGB

[0034] 5) Repeat the above steps to obtain a complete approximate isoconcentration contour curve, and combine the on-site DCS system and online monitoring sensors to obtain the actual concentration C at any point on the contour curve. RGBi Its actual coordinates (x RGBi ,y RGBi );

[0035] 6) According to the corresponding coordinates of the approximate isoconcentration contour curve, obtain and replace the corresponding coordinate concentration value in the theoretical prediction model;

[0036] 7) Perform overall coordinate transformation on the theoretical model concentration field and correct the concentration field. Connect the leakage source point O(0,0) and any point C on the approximate isoconcentration contour curve. RGBi (x RGBi ,y RGBi ), we get the equation of the line or its extension:

[0037] direction:

[0038] Size: y = x tanσ i

[0039] Get the actual concentration C RGBi (x RGBi ,y RGBi ) The corresponding coordinates (x Gi ,y Gi ), we can get any theoretical prediction model point y on the straight line i The coordinate transformation position y′ relationship is:

[0040]

[0041] Repeat the above steps to perform overall coordinate transformation on the theoretical prediction model.

[0042] (4) At time t+Δt, the infrared thermal imaging image is retaken and compared with the calculated concentration at the same time, the actual concentration curve is corrected and presented on the platform map, and the corrected dynamic isoconcentration curve is displayed in real time.

[0043] Beneficial effects: The advantages and positive effects of the present invention are:

[0044] (1) Follow up the development trend of the accident in real time throughout the entire process. It takes a period of time for the leakage to develop into a large-scale leakage. Only if no prevention and control measures are implemented during this process can the most dangerous impact range predicted by the traditional continuous leakage model through static data when the accident develops into a large-scale leakage and the concentration field reaches a steady state be achieved. The present invention can follow up the development trend of the accident throughout the entire process, dynamically feedback the leakage situation from the initial moment of the leakage, capture the change process of the accident impact range, which is conducive to the targeted implementation of disposal measures, avoid the waste of emergency resources, and significantly improve the accuracy and effectiveness of emergency disposal.

[0045] (2) Image gradient calculation is the edge information of the image. Due to the differences in shooting angle, light intensity, illumination angle, environmental background, and superposition of shooting targets, the absolute information error of image recognition is large, but the relative image gradient is accurate and reliable. Using a drone equipped with an online optical gas thermal imager to patrol and shoot near the equipment, the image edge detection technology can identify the boundary of the concentration field, and draw an approximate isoconcentration contour curve based on the principle of drawing lines along the direction with the smallest image gradient, which can obtain a relatively accurate approximate isoconcentration contour curve.

[0046] (3) The actual point concentration value on the approximate isoconcentration contour curve can be obtained with the help of the on-site DCS system or online monitoring sensor, which can solve the problem of large absolute concentration error caused by image recognition and ensure the concentration detection accuracy on the basis of obtaining the actual concentration field curve.

[0047] (4) Although the method of combining image recognition with precise concentration measurement can obtain a single isoconcentration curve more accurately, the leakage concentration field changes dynamically over time. To ensure the accuracy of the measurement, the entire concentration field correction needs to be completed at the same time. To achieve this in actual work, it is necessary to densely arrange infrared thermal imagers and online monitoring sensors, which is not feasible. This method is based on the above-mentioned approximate isoconcentration contour curve, locates the same coordinate position points of the curve on the theoretical prediction model, uses the actual concentration field to correct the theoretically predicted concentration field, and performs an overall coordinate transformation on the concentration field to correct the entire concentration field. It can obtain the corrected concentration value of the entire leakage impact range in real time and efficiently.

[0048] (5) Wide scope of application. The present invention is not limited to specific media and specific equipment, and can be applied to any hazardous chemical that leaks into the environment in a gaseous medium. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is the graph of the function N = erf(n);

[0050] Figure 2 It is the graph of the function y = erfc(p);

[0051] Figure 3It is a workflow diagram of the digital twin prediction model of the leakage concentration field of the hazardous chemicals device of the present invention;

[0052] Figure 4 is a schematic diagram of concentration field correction;

[0053] Figure 5 is the image gradient map. DETAILED DESCRIPTION

[0054] The present invention will be further described below with reference to the accompanying drawings and through specific embodiments.

[0055] Example 1

[0056] A dynamic prediction method for leakage concentration field of hazardous chemical installations based on infrared thermal imaging correction is constructed as follows:

[0057] 1. Obtain monitoring parameters in real time

[0058] Acquire and initialize relevant parameters in real time. Obtain real-time pressure, temperature and hazardous chemical medium parameters (medium type, density, hazardous concentration range) of hazardous chemical equipment from the on-site DCS system and online monitoring devices, as well as equipment parameters: total mass, equipment volume, internal height, etc.

[0059] 2. Dynamic concentration field construction

[0060] Since the continuous leakage Gaussian model can only obtain the final steady-state condition, the concentration field under the condition of constant concentration, that is:

[0061]

[0062] In order to reflect the change process from the initial leakage concentration of 0 to the stable concentration field, the present invention regards the continuous leakage as an instantaneous point source that is continuously and evenly released over a period of time, and uses the instantaneous leakage model integral to solve the real-time concentration field of the continuous leakage model:

[0063] Transient leakage model:

[0064]

[0065] Continuous Leakage Model:

[0066]

[0067] Diffusion coefficient calculation formula σ=0.18x 0.92 :

[0068] Among them, the gas leakage flow (unit: kg / s) is calculated according to the Bernoulli equation:

[0069]

[0070] 2.1 Limiting case mixed concentration field

[0071] At a constant leak rate If continuous leakage occurs, the concentration at (x, y, z) at time t is the time integral of the instantaneous leakage solution. First, consider the limiting case solution from time 0 to infinity:

[0072]

[0073] Solve the above integral equation:

[0074] Initial conditions, at t = 0:

[0075] <c>(x,y,z)=0

[0076] Boundary conditions, (x,y,z)=∞ far away:

[0077] <c>(x,y,z)=0

[0078] The solution is:

[0079]

[0080] because:

[0081] sign(k1)=1

[0082] That is, when t→∞, the above formula can be transformed into:

[0083]

[0084] Simplify the equation and let:

[0085]

[0086] N = erf(n)

[0087] The above function graph can be seen in Figure 1 , we can know that:

[0088] When n = 2, N = 0.9953 ≈ 1

[0089] Without considering the three turbulent diffusion coefficients σ of x, y, and z x , σ y , σ z In the case of the difference between continuous leakage and instantaneous leakage, when t→∞, When , we substitute the leakage diffusion coefficient σ into the equation:

[0090] σ=0.18x 0.92

[0091] x=0.8963

[0092] It can be seen that when x>1, we have The above solution can be further simplified as:

[0093]

[0094] That is, by integrating the instantaneous leakage solution over time, we can obtain the leakage concentration field under the condition of infinite time (t→∞). When the downwind distance x>1, it can be simplified to the solution of stable continuous point source release.

[0095] 2.2 Dynamic concentration field at any time t

[0096] (1) At any time during the leakage development process, t = t a Concentration range, without considering the three turbulent diffusion coefficients σ of x, y, and z x , σ y , σ z In case of difference between continuous leakage and instantaneous leakage, take σ x =σ y =σ z =0.18x 0.92 We can get:

[0097]

[0098] (2) Simplify the above formula:

[0099] because

[0100]

[0101] is the downwind x-axis position, wind speed u, and leakage diffusion coefficient σ with respect to time t x Functions of multiple variables can simplify the above solution:

[0102] From 2.1 above, we can see that when x>1, Substituting into the above formula, we get:

[0103]

[0104] Further simplifying, let:

[0105]

[0106] but:

[0107] P = 1 - erf(p) = erf c(p)

[0108] The function diagram is shown in Figure 2 , we can know that:

[0109] 1) When When P≈0, we can get:

[0110] <c>(x,y,z,t a )=0

[0111] at this time:

[0112]

[0113] To satisfy the above formula, we need:

[0114]

[0115] Here, take:

[0116]

[0117] but:

[0118]

[0119] Then the p error is less than 0.5%.

[0120]

[0121] In actual use, the boundary value is taken according to the concentration value corresponding to the isoconcentration curve drawn according to actual needs.

[0122] 2) When When P≈1, we can get:

[0123]

[0124] at this time:

[0125]

[0126] or

[0127]

[0128] 3) When p≤-2, P≈2, we can get:

[0129] <c>(x,y,z,t a )= <c>(x,y,z) l

[0130] at this time:

[0131]

[0132] or

[0133]

[0134] (4) In summary, the concentration formula at any time is:

[0135]

[0136] 1) When x≤x erfc＝2 ,Right now hour:

[0137] <c>(x,y,z,t a )= <c>(x,y,z) l

[0138] 2) When x = x erfc＝1 ,Right now hour:

[0139]

[0140] 3) When x→∞, σ x →∞, the p error is less than 0.5%, that is hour:

[0141] <c>(x,y,z,t a )=0

[0142]

[0143] 2.3 Dynamic concentration field isoconcentration curves

[0144] For a given concentration value C * ,make:

[0145]

[0146] but:

[0147]

[0148] Remove the negative sign and simplify to get:

[0149]

[0150] Taking the logarithm of both sides of the equation, simplifying it, we get:

[0151]

[0152] The final isoconcentration curve is:

[0153]

[0154] 3. Concentration field correction

[0155] (1) Use a drone equipped with an online optical gas thermal imager to inspect and shoot equipment accessories to obtain the optical gas thermal imager's gas sensitivity (ppm*m), infrared resolution (pixel), detector pixel spacing d (μm), focal length fmm, clear imaging range mm, etc.;

[0156] (2) Fly around the alarm device, parallel to the ground (xoy plane), and take thermal imaging pictures of the leaking gas.

[0157] (3) Due to different environments, different shooting angles (overhead, sideways, and upward), and different lighting conditions (front light, backlight), the photos taken have different effects. The absolute value of the identified concentration is inaccurate, but the relative concentration gradient is accurate. Therefore, the image edge detection technology is used to calculate the image RGB color change gradient, that is, the amplitude of the image RGB color change, and calculate the change rate of the image x-axis and y-axis. Specifically:

[0158] 1) Assume that I(x,y) represents the pixel value of the image (such as RGB color). The first-order derivative can be used to obtain the amplitude of change. The derivative value solution here should be infinitely close to 0 in order to obtain an infinitesimal partial derivative. However, the image cannot obtain an infinitesimal value. The pixels in the image are discrete, and the minimum distance is 1 pixel. Therefore, the gradient in the x direction is:

[0159] G x =I(x+1,y)-I(x,y)

[0160] Gradient in y direction:

[0161] G y =I(x,y+1)-I(x,y)

[0162] Gradient size:

[0163]

[0164] Gradient direction:

[0165]

[0166] 2) Considering adjacent positions, the gradient directions are vertical, horizontal, and diagonal, with a total of 8 possible values, such as Figure 5 As shown:

[0167] I(x m ,y m )={I(x n +1,y n )、I(x n ,y n +1)、I(x n -1,y n )、I(x n ,y n -1), I(x n

[0168] +1,y n +1)、I(x n +1,y n -1), I(x n -1,y n -1), I(x n -1,y n

[0169] +1)}

[0170] 3) In order to connect the discrete points into an approximate isoconcentration contour curve, the principle is to draw lines along the direction of the minimum image gradient. The image gradient calculation formula is as follows:

[0171]

[0172] 4) The maximum difference in image gradient between all points on the obtained concentration profile curve should be less than the set acceptable error range G RGB , whose value is based on the statistical acquisition of a series of image recognition and experimental results:

[0173] G(x,y) max ≤G RGB

[0174] 5) Repeat the above steps to obtain a complete approximate isoconcentration contour curve, and combine the on-site DCS system and online monitoring sensors to obtain the actual concentration C at any point on the contour curve. RGBi Its actual coordinates (x RGBi ,y RGBi );

[0175] 6) According to the corresponding coordinates of the approximate isoconcentration contour curve, obtain and replace the corresponding coordinate concentration value in the theoretical prediction model;

[0176] 7) Perform overall coordinate transformation on the overall concentration field of the theoretical model and correct the concentration field. Connect the leakage source point O(0,0) and any point C on the approximate isoconcentration contour curve. RGBi (x RGBi ,y RGBi ), we get the equation of the line or its extension:

[0177] direction:

[0178]

[0179] size:

[0180] y=x·tanα i

[0181] Get the actual concentration C RGBi (x RGBi ,y RGBi ) The corresponding coordinates (x Gi ,y Gi ), we can get any theoretical prediction model point y on the straight line i The coordinate transformation position y′ relationship is:

[0182]

[0183] Similarly, we can get: Repeat the above steps to complete the overall coordinate change of the theoretical prediction model.

[0184] (4) At time t+Δt, retake the infrared thermal imaging image and compare it with the calculated concentration at the same time. Correct the actual concentration curve and present it on the platform map. Display the corrected dynamic isoconcentration curve in real time. Figure 4 shown.

[0185] The above description is only an embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent substitution, improvement, etc. made within the spirit and principle of the present invention should be included in the scope of the claims of the present invention.< / c> < / c> < / c> < / c> < / c> < / c> < / c> < / c>

Claims

1. A method for dynamic prediction of leakage concentration field of hazardous chemical equipment based on infrared thermal imaging correction, characterized in that: For the leakage accident scenario of the pressure-bearing device involving hazardous chemicals, the relevant parameters are selected and initialized, and the instantaneous leakage solution is integrated over time to obtain the functional relationship between the downwind position and the point concentration when the time is infinite (t→∞); based on the above solution, the function solution of the downwind x-axis position and the multivariate variable at any time t in the development of the leakage is obtained, and the functional relationship between the downwind position and the point concentration at any time t is obtained; then, a drone equipped with an online optical gas thermal imager is used to patrol and shoot near the equipment to obtain infrared thermal imaging images of the actual concentration field on site, locate the boundary points of the concentration field, and combine image recognition edge detection technology to locate the boundary points along the direction with the smallest image gradient. Draw an approximate isoconcentration contour curve based on the principle of line drawing, and use the on-site DCS system or online monitoring sensor to obtain the actual point concentration value of the contour curve; then, locate the same position point on the theoretical prediction model, use the actual concentration curve to correct the theoretical prediction concentration curve, and perform an overall coordinate transformation on the concentration field to correct the entire concentration field; repeat the above steps to finally obtain a dynamic prediction model that fits the actual leakage impact range. It can be corrected in real time according to the on-site measured data, realize the whole process of following up the development of the accident, dynamically feedback the leakage situation from the initial moment of the leakage, capture the change process of the accident impact range, and significantly improve the accuracy and effectiveness of emergency response.

2. According to claim 1, a method for dynamic prediction of leakage concentration field of hazardous chemical equipment based on infrared thermal imaging correction is characterized in that: The following steps are involved: Step S1. Obtain and initialize relevant parameters in real time: Step S2. Obtain the analytical solution of the critical condition of the concentration field when the device leaks at infinite time: For the specific working condition of the device leaking hazardous chemicals, add the time variable on the basis of the Gaussian calculation model, subdivide the continuous leakage into instantaneous leakage in a small time period, and integrate the instantaneous leakage in the infinite time range. Without considering the three turbulent diffusion coefficients σx, y, and z, x , σ y , σ z In the case of the difference between continuous leakage and instantaneous leakage (taking σ = 0.18x 0.92 ), when t→∞, we can get: When x>1, we have The above solution can be further simplified as: Parameter Description: <c>(x,y,z) l : The mass concentration of any point (x, y, z) in the continuous leakage diffusion space in the air, mg / m 3 < / c> <c>(x,y,z,t) s : Mass concentration of the medium in the air at any point (x, y, z) in the instantaneous leakage diffusion space at time t, mg / m 3 < / c> Leakage flow, kg / s Q m : Leakage quality, for the same leakage scenario, u: average wind speed in the environment during leakage, m / s σ x , σ y , σ z : are the diffusion parameters of the horizontal axis, horizontal axis and vertical upward respectively; x, y, z: downwind distance, crosswind distance, vertical wind distance, respectively, m; t a : Any corresponding time selected for the diffusion consequence prediction, s C * : mass concentration specified on the isoconcentration curve, mg / m 3 I(x,y): image pixel value at coordinate (x,y) (e.g. RGB value) I(x m ,y m ): coordinate (x m ,y m ) image pixel value at the position (e.g. RGB value) G x : x-direction image gradient, vector G y : y-direction image gradient, vector G(x,y) min : Image gradient of xoy plane, vector θ: gradient direction angle, ° C RGBi (x RGBi ,y RGBi ): The position on the approximate isoconcentration contour curve is (x RGBi ,y RGBi ) point concentration, mg / m 3 α i :Connect the leakage source point O(0,0) and any point on the approximate isoconcentration contour curve C RGBi (x RGBi ,y RGBi ) The angle between the straight line or its extension and the x-axis direction, ° T0: Initial temperature of gas in the device, K T: Gas temperature in the device at time t, K t: the time corresponding to the parameter extraction, s t0: initial time of leakage, s n: amount of gas, mol R g : Ideal gas constant, J / (mol·K) A: Leakage orifice area, m 2 C D : Leakage coefficient, dimensionless P: The pressure inside the tank at time t, Pa P w : atmospheric pressure outside the leak hole, Pa M: gas molar mass, kg / mol K: gas heat capacity ratio, the ratio of specific heat at constant pressure to specific heat at constant volume Step S3. Construct a dynamic concentration field prediction model at any time t: Integrate the original model within a finite time, and resolve the integral result through the steady-state critical condition in step 2 to obtain a new dynamic prediction model of the leakage impact range changing with time; specific steps: (1) Without considering the three turbulent diffusion coefficients σ of x, y, and z x , σ y , σ z In the case of the difference between continuous leakage and instantaneous leakage (taking σ = 0.18x 0.92 ), integrating and simplifying gives: (2) Bring in critical conditions: According to x>1, Bring in: (3) Further simplify, let: but: P = 1 - erf(p) = erf c(p) Solving the original model piecewise, we can get: 1) When or When P≈2, at this time: 2) When or hour, P≈1, we can get: 3) When x→∞, σ x →∞, P≈0, we can get: At this time, if the p error is less than 0.5%, then hour: <c>(x,y,z,t a )=0< / c> In summary, the relationship between the scale and coordinate position, time function and special position law at any time is: Step S4. Based on the new prediction model, the leakage impact range prediction algorithm is obtained: the isoconcentration curve of the new dynamic prediction model is calculated (take σ = 0.18x 0.92 ), thereby predicting the scope of leakage impact; Step S5. Use a drone equipped with an online optical gas thermal imager to patrol and shoot near the equipment, obtain infrared thermal imaging pictures of the actual concentration field on site parallel to the ground (xoy plane), use image edge detection technology to identify the boundary of the concentration field, and draw an approximate iso-concentration contour curve based on the principle of drawing lines along the direction of the minimum image gradient; then use the on-site DCS system or online monitoring sensor to obtain the actual point concentration value of the contour curve; then, locate the same position point on the theoretical prediction model, use the actual concentration field to correct the theoretical prediction concentration field, and perform an overall coordinate transformation on the concentration field to correct the entire concentration field.

3. According to claim 2, a method for dynamic prediction of leakage concentration field of hazardous chemical equipment based on infrared thermal imaging correction is characterized in that: In step S1, the relevant parameters include but are not limited to the following parameters obtained from the on-site DCS system and the online monitoring device: (1) Real-time pressure and temperature of hazardous chemicals equipment; (2) Hazardous chemical medium parameters: medium type, density, hazardous concentration range; (3) Equipment parameters: total mass, equipment volume, and internal height.

4. According to claim 2, a method for dynamically predicting leakage concentration field of hazardous chemical equipment based on infrared thermal imaging correction is characterized in that: In step S5, the concentration field correction specifically includes the following steps: (1) Use a drone equipped with an online optical gas thermal imager to inspect and photograph equipment accessories to obtain the optical gas thermal imager's gas sensitivity (ppm*m), infrared resolution (pixel), detector pixel spacing d (μm), focal length f (mm), and clear imaging range (mm); (2) Fly around the alarm device, parallel to the ground (xoy plane), and take thermal imaging pictures of the leaking gas; (3) Using image edge detection technology, calculate the image RGB color change gradient, that is, the amplitude of the image RGB color change, and find its change rate on the image x-axis and y-axis; (4) At time t+Δt, the infrared thermal imaging image is retaken and compared with the calculated concentration at the same time, the actual concentration curve is corrected and presented on the platform map, and the corrected dynamic isoconcentration curve is displayed in real time.

5. According to claim 4, a method for dynamically predicting leakage concentration field of hazardous chemicals equipment based on infrared thermal imaging correction is characterized in that: In step (3), the step of calculating the image RGB color gradient specifically includes the following steps: 1) Assume that I(x,y) represents the pixel value of the image. The first-order derivative can be used to obtain the amplitude of change. The derivative here should be infinitely close to 0 in order to obtain an infinitesimal partial derivative. However, the image cannot take an infinitesimal value. The pixels in the image are discrete, and the minimum distance is 1 pixel. Therefore, the gradient in the x and y directions is: Gradient size: Gradient direction: 2) Considering adjacent positions, the gradient directions are vertical, horizontal, and diagonal, with a total of 8 possible values: I(x m ,y m )={I(x n +1,y n )、I(x n ,y n +1)、I(x n -1,y n )、I(x n ,y n -1)、I(x n +1,y n +1)、I(x n +1,y n -1)、I(x n -1,y n -1)、I(x n -1,y n +1)} 3) In order to connect the discrete points into an approximate isoconcentration contour curve, the principle is to draw lines along the direction of the minimum image gradient. The image gradient calculation formula is as follows: 4) The maximum difference in image gradient between all points on the obtained concentration profile curve should be less than the set acceptable error range G RGB , whose value is obtained based on the statistical results of a series of image recognition and experimental results: G(x,y) max ≤G RGB 5) Repeat the above steps to obtain a complete approximate isoconcentration contour curve, and combine the on-site DCS system and online monitoring sensors to obtain the actual concentration C at any point on the contour curve. RGBi Its actual coordinates (x RGBi ,y RGBi ); 6) According to the corresponding coordinates of the approximate isoconcentration contour curve, obtain and replace the corresponding coordinate concentration value in the theoretical prediction model; 7) Perform overall coordinate transformation on the theoretical model concentration field and correct the concentration field. Connect the leakage source point O(0,0) and any point C on the approximate isoconcentration contour curve. RGBi (x RGBi ,y RGBi ), we get the equation of the line or its extension: direction: size: y=x·tanα i Get the actual concentration C RGBi (x RGBi ,y RGBi ) The corresponding coordinates (x Gi ,y Gi ), we can get any theoretical prediction model point y on the straight line i The coordinate transformation position y′ relationship is: Repeat the above steps to perform overall coordinate transformation on the theoretical prediction model.

Images