A dynamic prediction method for leakage concentration field of hazardous chemical installations based on infrared thermal imaging correction

Abstract

The present invention relates to a method for dynamic prediction of leakage concentration field of hazardous chemical equipment based on infrared thermal imaging correction, which comprises the following steps: (1) obtaining real-time parameters from the scene; (2) adding a time variable to the Gaussian calculation model, integrating the instantaneous leakage in the infinite time range, and obtaining the critical condition of the concentration field prediction algorithm; (3) integrating the original model in a finite time, and resolving the integral result by the obtained critical condition, thereby obtaining a new dynamic prediction model in which the leakage influence range changes with time; (4) calculating the isoconcentration curve of the new dynamic prediction model; (5) combining infrared images and real-time monitoring data, and drawing the actual isoconcentration curve by edge detection technology; (6) locating the same coordinate position point on the theoretical prediction model, correcting the theoretical concentration curve with the actual concentration curve, performing overall coordinate transformation on the concentration field, and correcting the entire concentration field; (7) retaking the infrared thermal image after a certain period of time, and correcting the concentration field again. The method improves the accuracy and effectiveness of emergency response.

Description

Technical Field

[0001] The present invention belongs to the field of leakage analysis and prediction of hazardous chemical equipment, and relates to a method for dynamically predicting the leakage concentration field of hazardous chemical 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. The prediction is changed to the prediction result under ideal conditions of no obstruction and constant wind speed and direction. It does not take into account the impact of environmental factors such as obstruction of on-site equipment, changes in wind speed and direction on the concentration field during actual leakage. It 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 stage in the early stage of the accident. On the other hand, it easily causes waste of emergency resources and 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 results are 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 in H2S Transport Pipelines" (Patent No.: ZL202310244925.X) requires a large number of concentration sensors to be evenly distributed along the pipeline. Continuous monitoring also places high demands on power supply endurance. As the pipeline length increases, the sensors become densely distributed, and power consumption continues to increase. Furthermore, this method does not clearly define a technical solution for monitoring power supply endurance. First, the engineering modification of densely installing a large number of additional concentration sensors increases pipeline safety risks. Second, the application cost is high, increasing the burden on enterprises. Third, in the event of an accident, there is the possibility of sensor failure or power supply interruption, making the method unfeasible.

[0005] Guangxi University's "A calculation model based on the continuous real-time leakage of atmospheric vertical storage tank bodies" (patent number: ZL201810724745.0) and China Energy Rong'an (Beijing) Technology Co., Ltd. and 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 volume based on 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 warnings and leakage volume, and cannot predict the scope of leakage impact 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 chemical 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 the leakage concentration field of hazardous chemical equipment based on infrared thermal imaging correction.

[0007] Technical solution: In order to achieve the above-mentioned purpose, the present invention adopts the following technical solution:

[0008] A dynamic prediction method for the leakage concentration field of hazardous chemical installations based on infrared thermal imaging correction is provided. The first step is to obtain the real-time pressure, temperature, initial mass, volume, medium density and other parameters of the hazardous chemical installation 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 installation leakage, subdivide the continuous leakage into instantaneous leakage within a small time period, and integrate the instantaneous leakage in the infinite time range to derive the critical condition 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 condition in step 2 to obtain a new dynamic prediction model of the leakage influence range changing 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 by combining the infrared image and real-time monitoring data. Using an unmanned aerial vehicle (UAV) equipped with an online optical gas thermal imager, thermal images of the leak area are captured. The boundary points of the concentration field are located. Using image recognition and edge detection technology, an approximate isoconcentration contour curve is drawn along the direction of minimum image gradient. The actual concentration values at these points on the contour curve are obtained using an on-site DCS system or online monitoring sensors. In the sixth step, points with identical coordinates are located on the theoretical prediction model. The theoretical prediction curve is then corrected using the actual concentration curve. The overall coordinate transformation of the concentration field is then performed, and the above steps are repeated to correct the entire concentration field. In the seventh step, after a certain period of time, the infrared thermal image is retaken and the concentration field is corrected again.

[0009] The specific model construction process includes the following steps:

[0010] Step S1. Real-time acquisition and initialization of relevant parameters. 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. Equipment parameters (gross mass, equipment volume, internal height, etc.)

[0011] 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 hazardous chemical device leakage, 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 σ of x, y, and z, the concentration field prediction algorithm is used to predict the leakage field. 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→∞, the leakage diffusion coefficient σ is substituted into the equation: It can be seen that when x>1, The solution is:

[0012]

[0013] 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. x , σ y , σ z In the case of the difference between continuous leakage and instantaneous leakage, the relationship between concentration, coordinate position, time function and special position law can be obtained by integration and simplification:

[0014]

[0015] 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;

[0016]

[0017] Step S5. Use a drone equipped with an online optical gas thermal imager to conduct inspections near the equipment, obtain infrared thermal images of the actual concentration field on site parallel to the ground (xoy plane), use image edge detection technology to identify the concentration field boundary, and draw an approximate isoconcentration contour curve based on the principle of drawing lines along the direction of 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. The following steps are included:

[0018] (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, and clear imaging range mm;

[0019] (2) Fly around the alarm device, parallel to the ground (xoy plane), and take thermal imaging pictures of the leaking gas.

[0020] (3) Using image edge detection technology, the absolute value of the identified concentration is not accurate, 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:

[0021] 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. 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 and y directions is:

[0022]

[0023] Gradient size:

[0024]

[0025] Gradient direction:

[0026]

[0027] 2) Considering adjacent positions, the gradient direction includes vertical direction, horizontal direction, and diagonal direction, with a total of 8 possible values, such as Figure 5 As shown:

[0028] 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

[0029] -1), I(x n +1,y n +1)、I(x n +1,y n -1), I(x n -1,y n

[0030] -1), I(x n -1,y n +1)}

[0031] 3) To connect discrete points into an approximate isoconcentration contour curve, the principle is to draw a line along the direction of the minimum image gradient. The image gradient calculation formula is as follows:

[0032]

[0033] 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:

[0034] G(x, y) max ≤D RGB

[0035] 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 );

[0036] 6) Obtain and replace the corresponding coordinate concentration value in the theoretical prediction model according to the corresponding coordinate of the approximate isoconcentration contour curve;

[0037] 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:

[0038] direction:

[0039]

[0040] size:

[0041] y=x·tanα i

[0042] 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:

[0043]

[0044] Repeat the above steps to perform overall coordinate transformation on the theoretical prediction model.

[0045] (4) At time t+Δt, retake the infrared thermal imaging picture and compare it with the calculated concentration at the same time. Then, correct the actual concentration curve and present it on the platform map, displaying the corrected dynamic isoconcentration curve in real time.

[0046] Beneficial effects: The advantages and positive effects of the present invention are:

[0047] (1) Follow up the development trend of the accident in real time throughout the entire process. It takes a period of time for a leak to develop into a large-scale leak. 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 be achieved when the accident develops into a large-scale leak and the concentration field reaches a steady state. 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 leak, and capture the changes in the scope of the accident impact, which is conducive to the targeted implementation of disposal measures, avoids the waste of emergency resources, and significantly improves the accuracy and effectiveness of emergency response.

[0048] (2) Image gradient calculation is based on the edge information of the image. Due to 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 highly reliable. Using a drone equipped with an online optical gas thermal imager to inspect and shoot near the equipment, image edge detection technology can identify the concentration field boundary. Approximate isoconcentration contour curves are drawn based on the principle of drawing lines along the direction of minimum image gradient, which can obtain relatively accurate approximate isoconcentration contour curves.

[0049] (3) By using the on-site DCS system or online monitoring sensor to obtain the actual point concentration value on the approximate isoconcentration contour curve, the problem of large absolute concentration error caused by image recognition can be solved, and the concentration detection accuracy is guaranteed on the basis of obtaining the actual concentration field curve.

[0050] (4) Although the method of combining image recognition with precise concentration measurement can obtain a single isoconcentration curve with relative accuracy, the leakage concentration field changes dynamically over time. To ensure the accuracy of the measurement, the entire concentration field needs to be corrected at the same time. To achieve this in actual work, infrared thermal imagers and online monitoring sensors need to be densely arranged, which is not feasible. This method is based on the above-mentioned approximate isoconcentration contour curve. The same coordinate position points of the curve are located on the theoretical prediction model. The theoretical predicted concentration field is corrected with the actual concentration field, and the overall coordinate transformation of the concentration field is performed to correct the entire concentration field. This method can obtain the corrected concentration value of the entire leakage impact range in real time and efficiently.

[0051] (5) Wide range of applications. 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 state. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is the graph of the function N = erf(n);

[0053] Figure 2 is the graph of the function y = erf c(p);

[0054] Figure 3 This is a workflow diagram of the digital twin prediction model for the leakage concentration field of hazardous chemical devices according to the present invention;

[0055] Figure 4 is a schematic diagram of concentration field correction;

[0056] Figure 5 is the image gradient map. DETAILED DESCRIPTION

[0057] The present invention will be further described below with reference to the accompanying drawings and through specific embodiments.

[0058] Example 1

[0059] A dynamic prediction method for leakage concentration field of hazardous chemical installations based on infrared thermal imaging correction is constructed as follows:

[0060] 1. Obtain monitoring parameters in real time

[0061] 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 such as total mass, equipment volume, and internal height.

[0062] 2. Dynamic concentration field construction

[0063] 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:

[0064]

[0065] 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 releases uniformly over a period of time, and uses the instantaneous leakage model integral to solve the real-time concentration field of the continuous leakage model:

[0066] Transient leakage model:

[0067]

[0068] Continuous leakage model:

[0069]

[0070] Diffusion coefficient calculation formula σ=0.18x 0.92 :

[0071] The gas leakage flow rate (unit: kg / s) is calculated according to the Bernoulli equation:

[0072]

[0073] 2.1 Limiting case mixed concentration field

[0074] At a constant leak rate If a continuous leak occurs, the concentration at (x, y, z) at time t is the time integral of the instantaneous leak solution. First, consider the limiting case solution from time 0 to infinity:

[0075]

[0076] Solve the above integral equation:

[0077] Initial conditions, at t = 0:

[0078] <c>(x,y,z)=0 boundary condition, (x,y,z)=∞ far away:

[0079] <c>(x,y,z)=0, and we get:

[0080]

[0081] because:

[0082] sign(k1)=1

[0083] That is, when t→∞, the above formula can be transformed into:

[0084]

[0085] Simplify the equation and let:

[0086]

[0087] N=erf(n)

[0088] The above function graph is shown in Figure 1 , we can know that:

[0089] When n=2, N=0.9953≈1

[0090] Without considering the three turbulent diffusion coefficients σ of x, y, and z x , σ y , σ z In the case of difference between continuous leakage and instantaneous leakage, when t→∞, When , we can substitute the leakage diffusion coefficient σ into the equation:

[0091] σ=0.18x 0.92

[0092] x=0.8963

[0093] It can be seen that when x>1, there is The above solution can be further simplified as:

[0094]

[0095] 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.

[0096] 2.2 Dynamic concentration field at any time t

[0097] (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:

[0098]

[0099] (2) Simplify the above formula:

[0100] because

[0101]

[0102] is the downwind x-axis position, wind speed u, and leakage diffusion coefficient σ with respect to time t x Functions of multiple variables such as simplifies the above solution:

[0103] From 2.1 above, we can see that when x>1, Substituting into the above formula we get:

[0104]

[0105] Simplify further and let:

[0106]

[0107] but:

[0108] P=1-erf(p)=erf c(p) Figure 2 , we can know that:

[0109] 1) When When P≈0, we can get:

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

[0111]

[0112] To satisfy the above formula, we need:

[0113]

[0114] Here, take:

[0115]

[0116] but:

[0117]

[0118] Then the p error is less than 0.5%.

[0119]

[0120] In actual use, the boundary value is taken according to the concentration value corresponding to the isoconcentration curve drawn according to actual needs.

[0121] 2) When When P≈1, we can get:

[0122]

[0123] at this time:

[0124]

[0125] or

[0126]

[0127] 3) When p≤-2, P≈2, we can get:

[0128] <c>(x,y,z,t a )= <c>(x,y,z) l

[0129] at this time:

[0130]

[0131] or

[0132]

[0133] (4) In summary, the concentration formula at any time is:

[0134]

[0135] 1) When x≤x erfc＝2 ,Right now hour:

[0136] <c>(x,y,z,t a )= <c>(x,y,z) l

[0137] 2) When x = x erfc＝1 ,Right now hour:

[0138]

[0139] 3) When x→∞, σ x →∞, the error of p is less than 0.5%, that is hour:

[0140] <c>(x,y,z,t a )=0

[0141]

[0142] 2.3 Dynamic concentration field isoconcentration curve for a given concentration value C * ,make:

[0143]

[0144] but:

[0145]

[0146] Remove the negative sign and simplify to get:

[0147]

[0148] Taking the logarithm of both sides of the equation and simplifying it, we get:

[0149]

[0150] The final isoconcentration curve is:

[0151]

[0152] 3. Concentration field correction

[0153] (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, and clear imaging range mm;

[0154] (2) Fly around the alarm device, parallel to the ground (xoy plane), and take thermal imaging pictures of the leaking gas.

[0155] (3) Due to the different effects of photos taken in different environments, different shooting angles (overhead, side, and upward), and different lighting conditions (front light, back light), the absolute value of the recognized 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 its change rate on the image x-axis and y-axis. Specifically:

[0156] 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. The derivative value 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 x-direction gradient is:

[0157] G x =I(x+1,y)-I(x,y)

[0158] Gradient in the y direction:

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

[0160] Gradient size:

[0161]

[0162] Gradient direction:

[0163]

[0164] 2) Considering adjacent positions, the gradient direction includes vertical direction, horizontal direction, and diagonal direction, with a total of 8 possible values, such as Figure 5 As shown:

[0165] 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

[0166] -1), I(x n +1,y n +1)、I(x n +1,y n -1), I(x n -1,y n

[0167] -1), I(x n -1,y n +1)}

[0168] 3) To connect discrete points into an approximate isoconcentration contour curve, the principle is to draw a line along the direction of the minimum image gradient. The image gradient calculation formula is as follows:

[0169]

[0170] 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:

[0171] G(x, y) max ≤G RGB

[0172] 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 );

[0173] 6) Obtain and replace the corresponding coordinate concentration value in the theoretical prediction model according to the corresponding coordinate of the approximate isoconcentration contour curve;

[0174] 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:

[0175] direction:

[0176]

[0177] size:

[0178] y=x·tanα i

[0179] 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:

[0180]

[0181] Similarly, we can get:

[0182]

[0183] Repeat the above steps to complete the theoretical prediction model and perform overall coordinate changes.

[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. The corrected dynamic isoconcentration curve is displayed in real time. Figure 4 shown.

[0185] The foregoing is merely an embodiment of the present invention and is not intended to limit the present invention. It will be apparent to those skilled in the art that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention are intended to be included within the scope of the claims of the present invention.< / c> < / c> < / c> < / c> < / c> < / c> < / c> < / c>

Claims

1. A dynamic prediction method for leakage concentration field of hazardous chemical equipment based on infrared thermal imaging correction, characterized in that: For the leakage accident scenario of the pressure-containing device involving hazardous chemicals, relevant parameters are selected and initialized, and the instantaneous leakage is integrated over time. The functional relationship between the downwind position and the point concentration can be obtained when the time is infinite and t→∞. Based on the above solution, the functional solution of the downwind x-axis position and the multivariate variable at any time t during 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 draw a line along the direction of the minimum image gradient. In principle, an approximate isoconcentration contour curve is drawn, and the concentration value of the actual point on the contour curve is obtained with the help of the on-site DCS system or online monitoring sensor; then, the same position point is located on the theoretical prediction model, and the theoretical prediction concentration curve is corrected with the actual concentration curve, and the concentration field is transformed as a whole to correct the entire concentration field; the above steps of correcting the entire concentration field are repeated, and the final dynamic prediction model that fits the actual leakage impact range is obtained. It can be corrected in real time according to the on-site measured data, and the development trend of the accident can be followed up throughout the entire process. The leakage situation is dynamically fed back from the initial moment of the leakage, and the change process of the accident impact range is captured, which significantly improves the accuracy and effectiveness of emergency response.

2. The method for dynamic prediction of leakage concentration field of hazardous chemical equipment based on infrared thermal imaging correction according to claim 1 is characterized in that: The following steps are involved: Step S1. Real-time acquisition and initialization of relevant parameters: Step S2. Obtain the analytical solution for the critical condition of the concentration field when the device leaks at infinite time: For the accident scenario involving the leakage of hazardous chemicals, a time variable is added to the Gaussian calculation model, and the continuous leakage is subdivided into instantaneous leakage in a small time period. The instantaneous leakage is integrated within the infinite time range. Without considering the three turbulent diffusion coefficients σ of x, y, and z, the critical condition of the concentration field is obtained. x , σ y , σ z In case of difference between continuous leakage and instantaneous leakage, take 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 : Mass concentration in the air at any point in the continuous leakage diffusion space, with coordinates x, y, and z, mg / m 3 < / c> <c>(x,y,z,t) s : The mass concentration of the medium in the air at time t at any point in the instantaneous leakage diffusion space, with coordinates x, y, and z, is mg / m 3 < / c> Leakage flow, kg / s Q m : Leakage quality, for the same leakage scenario, u: average wind speed in the environment at the time of leakage, m / s σ x , σ y , σ z : are the diffusion parameters of the horizontal vertical axis, horizontal horizontal axis and vertical upward respectively; x, y, z: downwind distance, crosswind distance, and perpendicular wind distance, respectively, in m; t a : Any corresponding moment selected for 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) I(x m ,y m ): coordinate (x m ,y m ) image pixel value at the position 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 C on the approximate isoconcentration contour curve RGBi (x RGBi ,y RGBi ) The angle between the straight line or its extension and the x-axis, °T0: initial temperature of the gas in the device, K T: Gas temperature in the device at time t, K t: the time when the parameters are extracted, 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: within a limited time Integrate the original model 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 case of difference between continuous leakage and instantaneous leakage, take σ = 0.18x 0.92 , integrating and simplifying, we can get: (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 at this time: 2) When or hour, We can get: 3) When x→∞, σ x →∞, 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 any time scale and coordinate position, time function and special position rules are as follows: 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, and 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 conduct inspections and take photos near the equipment, obtain infrared thermal imaging images of the actual concentration field on site parallel to the ground, use image edge detection technology to identify the boundary of the concentration field, and draw an approximate isoconcentration contour curve based on the principle of drawing lines along the direction of 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. The method for dynamic prediction of leakage concentration field of hazardous chemical equipment based on infrared thermal imaging correction according to claim 2 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 online monitoring device: (1) Real-time pressure and temperature of hazardous chemical equipment; (2) Hazardous chemical medium parameters: medium type, density, hazardous concentration range; (3) Equipment parameters: total mass, equipment volume, and internal height.

4. The method for dynamic prediction of leakage concentration field of hazardous chemical equipment based on infrared thermal imaging correction according to claim 2 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 shoot equipment accessories, and obtain the optical gas thermal imager's gas sensitivity (in ppm*m); infrared resolution (in pixels); detector pixel spacing d (in μm); focal length f (in mm); and clear imaging range (in mm). (2) Fly around the alarm device, parallel to the ground, 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 calculate its change rate on the x-axis and y-axis of the image; (4) At time t+Δt, retake the infrared thermal imaging picture and compare it with the calculated concentration at the same time. Then, correct the actual concentration curve and present it on the platform map, displaying the corrected dynamic isoconcentration curve in real time.

5. The method for dynamic prediction of leakage concentration field of hazardous chemical equipment based on infrared thermal imaging correction according to claim 4 is characterized in that: In step (3), the calculation of the image RGB color gradient specifically includes the following steps: 1) Assume that I(x,y) represents the image pixel value. 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. 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 and y directions is: Gradient size: Gradient direction: 2) Considering adjacent positions, the gradient direction can be vertical, horizontal, or 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) To connect discrete points into an approximate isoconcentration contour curve, the principle is to draw a line 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) Obtain and replace the corresponding coordinate concentration value in the theoretical prediction model according to the corresponding coordinate of the approximate isoconcentration contour curve; 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