Pollution source reverse tracing method based on digital twinborn and intelligent identification

Abstract

The invention relates to the technical field of pollution traceability, and particularly discloses a pollution source reverse traceability method based on digital twinning and intelligent identification, and the method comprises the steps: determining the working information in a workshop according to the filing information, selecting a static detection point based on the working information, and synchronously determining the type of the static detection point; obtaining and counting working data containing time labels based on the static detection points, and constructing and displaying a digital twinborn model; segmenting the constructed digital twin model to obtain sub-models, and synchronously determining model parameters of the sub-models; and selecting a dynamic detection point according to the risk level of the sub-model, acquiring an air parameter containing a time label based on the dynamic detection point, and identifying the air parameter containing the time label by using AI to determine a pollution source. According to the method, the air data are analyzed, the sub-regions where problems possibly exist are determined, pollutants of the sub-models are recognized by means of the AI, pollution sources are positioned, the intuition degree is extremely high, and workers can take emergency measures conveniently.

Description

Technical Field

[0001] The present invention relates to the technical field of pollution source tracing, and specifically to a pollution source reverse tracing method based on digital twins and intelligent identification. Background Art

[0002] In the process of industrial activities, many industrial raw materials themselves are pollutants or the processing process produces some pollutants. Some of these pollutants are solid pollutants, some are liquid pollutants, and some are gaseous pollutants. Since gaseous pollutants have strong diffusion capabilities, their impact range is very large, especially in indoor environments. Gaseous pollutants can cause very serious consequences. Therefore, real-time monitoring of gaseous pollutants is necessary.

[0003] Most existing monitoring solutions are simple alarm-type monitoring solutions, which select some monitoring points to monitor gas concentrations in real time. When the concentration is high, an alarm signal is generated. This method is too simple. Workers can only know which location has too high concentrations, and they need to manually trace the source and take corresponding measures. This makes the efficiency of the pollution emergency process very low. Therefore, how to improve the efficiency of the pollution emergency process is the technical problem that the technical solution of the present invention wants to solve. Summary of the Invention

[0004] The purpose of the present invention is to provide a pollution source reverse tracing method based on digital twins and intelligent identification to solve the problems raised in the above background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions: A pollution source reverse tracing method based on digital twins and intelligent identification, the method comprising: Determine the work information in the workshop according to the record information, select static detection points based on the work information, and simultaneously determine the type of the static detection points; the work information includes the work scope, work scene and personnel information; the type includes wind type, temperature type and visual type; Based on static detection points, work data with time tags is acquired and counted, and a digital twin model is constructed and displayed; The constructed digital twin model is segmented to obtain sub-models, and the model parameters of the sub-models are simultaneously determined; the model parameters include risk level and risk type; the risk type adopts pollutant type; Dynamic detection points are selected according to the risk level of the sub-model, and air parameters with time tags are obtained based on the dynamic detection points. AI is used to identify the air parameters with time tags and determine the pollution source.

[0006] As a further solution of the present invention, the steps of determining the work information in the workshop according to the filing information, selecting the static detection point based on the work information, and simultaneously determining the type of the static detection point include: Establish a connection channel with the filing database to read the work scope, work scene and personnel information in the workshop during the current work cycle; Obtaining the volume of the external space of the work scene, and determining a detection density based on the volume; the detection density is inversely proportional to the volume; and the detection density is expressed as a percentage; Input the staff information into the preset evaluation model, output the staff capability table, and adjust the inspection density based on the staff capability table; Based on the adjusted detection density, static detection points are inserted and selected within the working range, and the types of the static detection points are simultaneously determined.

[0007] As a further solution of the present invention, the step of inserting and selecting static detection points within the working range based on the adjusted detection density and simultaneously determining the type of the static detection points includes: Obtain the contour of the working range, and evenly set static detection points of the wind type on the contour of the working range based on the detection density; Select two mutually perpendicular directions in the working range, obtain the maximum length of the working range in the two mutually perpendicular directions, and determine two mutually perpendicular dividing lines; Based on the detection density, reference points are evenly set on two mutually perpendicular cutting lines, parallel lines are constructed based on the reference points, and the intersection of the parallel lines is selected as the sampling point; The sampling point is used as a static detection point of temperature type; Based on the sampling points, static detection points of the visual type are selected; and a union of collection ranges of the static detection points of the visual type includes all sampling points.

[0008] As a further solution of the present invention: the steps of acquiring and counting working data with time tags based on static detection points, and constructing and displaying a digital twin model include: Receive the temperature uploaded by static detection points of temperature type and build a temperature matrix based on the sampling points; Receive wind parameters uploaded by static detection points of wind type, and use the wind parameters to correct the temperature matrix; Obtain the working image uploaded by the static inspection point of the visual type, perform 3D modeling based on the map within the working range and the working image, and obtain a 3D map; The corrected temperature matrix is ​​inserted into the three-dimensional graph based on the preset color value table to obtain and display the digital twin model.

[0009] As a further solution of the present invention, the step of inserting the corrected temperature matrix into the three-dimensional graph based on the preset color value table to obtain and display the digital twin model includes: Receive the correspondence between temperature and color value uploaded by the staff and build a color value table; Read the corrected temperature matrix, diffuse the temperature matrix according to the preset diffusion rule, and obtain a three-dimensional array; Based on the color value table, the data in the three-dimensional array is converted into color values, inserted into the three-dimensional scene, and the digital twin model is obtained and displayed.

[0010] As a further solution of the present invention, the step of reading the corrected temperature matrix and diffusing the temperature matrix according to a preset diffusion rule to obtain a three-dimensional array includes: For any row or column position in the temperature matrix, query the nearest known temperature in the four directions of up, down, left, and right; The vertical vector is determined based on the known temperatures in the up and down directions, and the horizontal vector is determined based on the known temperatures in the left and right directions. The modulus of the vector is proportional to the temperature difference, and the direction of the vector is from high temperature to low temperature. Calculating a resultant vector of the vertical vector and the horizontal vector, and querying the nearest known temperature in the opposite direction of the resultant vector based on the resultant vector; The temperature at that row or column position is calculated based on the nearest known temperature in the opposite direction of the resultant vector.

[0011] As a further solution of the present invention: the steps of dividing the constructed digital twin model to obtain sub-models and simultaneously determining the model parameters of the sub-models include: Select the digital twin model regularly and traverse each point in the selected digital twin model; When the color value difference between adjacent points is less than the preset difference threshold, the two points are connected; Count the interconnected points to get the sub-model, and determine the risk level by the color value of each point in the sub-area; For any sub-model, AI is applied to identify the subject in the sub-model, locate the pollutants, and query the type of pollutants as the risk type; The process of determining the risk level is as follows: Calculating a risk value and determining a risk level based on the step value reached by the risk value; each value in the step value corresponds to a risk level; The calculation process of risk value is: Where, is the preset correction factor, is the color value of the w-th point in the sub-area, For sub-regions.

[0012] As a further solution of the present invention, the steps of selecting dynamic detection points based on the risk level of the sub-model, obtaining air parameters with time tags based on the dynamic detection points, and applying AI to identify the air parameters with time tags to determine the pollution source include: Select the sub-model whose risk level reaches the preset level threshold; Determine the number of dynamic detection points based on the risk level; randomly inserting the number of dynamic detection points into the sub-model, executing the operation in a loop, and exiting the loop and outputting the insertion result when the distance between any two dynamic detection points is not less than a preset distance threshold; Based on dynamic detection points, air parameters with time tags are obtained, and AI is used to identify the air parameters with time tags to determine the source of pollution.

[0013] As a further solution of the present invention, the steps of obtaining air parameters with time tags based on dynamic detection points, identifying the air parameters with time tags using AI, and determining the pollution source include: Count the dynamic detection points of all sub-models and use the counted dynamic detection points as waypoints to generate a path; Sending the route to at least one UAV, and obtaining air parameters at each dynamic detection point based on an air detector installed on the UAV; the air parameters contain a time tag, and the air parameters are air type and concentration; The pollution source is determined based on the air parameters at each dynamic detection point.

[0014] As a further solution of the present invention: the step of determining the pollution source based on the air parameters at each dynamic detection point includes: For each dynamic detection point, query the air parameters at the most recent moment based on the time tag; Determine a concentration threshold according to the air type in the air parameter, and when the concentration in the air parameter reaches a preset concentration threshold, determine the radiation radius according to the concentration; A spherical area is created based on the radiation radius with the dynamic detection point as the center, and the sub-models that intersect with the spherical area are marked as contaminated areas; The number of times each sub-model is marked is counted. When the number of marks reaches the preset threshold, AI is used to identify the subject in the sub-model and locate the pollutants.

[0015] Compared with the existing technology, the beneficial effects of the present invention are: the present invention installs static monitoring equipment according to the registered work information, obtains work data and builds a model. After the modeling is completed, the model is divided according to the work data to obtain several sub-models, the sub-models are identified, the dynamic detection points are determined, the air data at the dynamic detection points are analyzed, and then the sub-models with possible problems are determined. With the help of AI, the pollutants in the sub-models are identified and the pollution sources are located. The method is highly intuitive and convenient for staff to take emergency measures. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present invention.

[0017] Figure 1 This is a flowchart of the pollution source reverse tracing method based on digital twins and intelligent identification.

[0018] Figure 2 This is the first sub-process flowchart of the pollution source reverse tracing method based on digital twins and intelligent identification.

[0019] Figure 3 This is the second sub-process flowchart of the pollution source reverse tracing method based on digital twins and intelligent identification.

[0020] Figure 4 This is the third sub-process flowchart of the pollution source reverse tracing method based on digital twins and intelligent identification.

[0021] Figure 5 This is the fourth sub-process flowchart of the pollution source reverse tracing method based on digital twins and intelligent identification. DETAILED DESCRIPTION

[0022] In order to make the technical problems, technical solutions and beneficial effects to be solved by the present invention more clearly understood, the present invention is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0023] Figure 1 This is a flowchart of a pollution source reverse tracing method based on digital twins and intelligent identification. In an embodiment of the present invention, a pollution source reverse tracing method based on AI includes: Step S100: Determine the work information in the workshop according to the record information, select static detection points based on the work information, and simultaneously determine the type of the static detection points; the work information includes the work scope, work scene and personnel information; the type includes wind type, temperature type and visual type; The present invention is applied to a production workshop, where existing workshop work activities are first filed and then carried out. For example, at the meeting every morning, today's work projects are reported, and after approval by the safety officer, the staff will carry out the relevant projects. The technical solution of the present invention is applied to this work system, so the work information is known data for the technical solution of the present invention. The work information reported daily will be pre-stored in the filing database. For the execution subject of this method, the work information is directly read from the filing database, analyzed, and some static detection points are selected. Detection equipment is installed at the static detection points to obtain information. The meaning of a static detection point is not a completely stationary detection point. It is a detection point with a very low frequency of change, which is generally related to the project cycle. There are many types of static detection points, including wind type, temperature type and visual type, and different detection equipment is installed respectively to obtain different types of data.

[0024] Step S200: Acquire and count working data with time tags based on static detection points, and build and display a digital twin model; By obtaining the working data at each moment based on the detection equipment installed at the static detection points and counting the working data at each static detection point at each moment, a digital twin model can be constructed. The digital twin model is essentially a three-dimensional model that can be displayed directly.

[0025] Step S300: Segment the constructed digital twin model to obtain sub-models, and simultaneously determine the model parameters of the sub-models; the model parameters include risk level and risk type; the risk type is the pollutant type; By identifying and dividing the constructed digital twin model, multiple sub-areas can be obtained, called sub-models. The risk level and risk type of each sub-model can be determined based on the working data already obtained in the sub-model, which are called model parameters.

[0026] Step S400: Select dynamic detection points based on the risk level of the sub-model, obtain air parameters with time tags based on the dynamic detection points, apply AI to identify the air parameters with time tags, and determine the pollution source; Identify the risk level of the sub-model and determine some dynamic detection points. These dynamic detection points are detected by unmanned equipment. The positions of dynamic detection points can change frequently because unmanned equipment has a high degree of freedom, thereby obtaining more valuable data in industrial scenarios. It is worth mentioning that the data acquisition frequency of dynamic detection points will be lower, because within an hour, a dynamic detection point may be detected every five minutes, while the equipment at the static detection point is fixed and its data acquisition frequency is very high to ensure the comprehensiveness of the data.

[0027] By acquiring time-tagged air parameters based on dynamic detection points and applying AI to identify these time-tagged air parameters, the pollution source can be determined. The AI's primary goal is to identify pollutants. The air parameters are first used to determine the area to be identified, and then the AI ​​detects the cause of pollution within the area, ultimately locating the pollutants.

[0028] Figure 2 This is a flowchart of the first sub-process of the pollution source reverse tracing method based on digital twins and intelligent identification. The steps of determining the work information in the workshop based on the filing information, selecting static detection points based on the work information, and simultaneously determining the type of the static detection points include: Step S101: Establish a connection channel with the filing database to read the work scope, work scene and personnel information in the workshop during the current work cycle; Step S102: obtaining the volume of the external space of the working scene, and determining the detection density according to the volume; the detection density is inversely proportional to the volume; the detection density is expressed as a percentage; Step S103: Inputting staff information into a preset evaluation model, outputting a staff capability table, and adjusting the inspection density based on the staff capability table; Step S104: inserting and selecting static detection points within the working range based on the adjusted detection density, and simultaneously determining the types of the static detection points.

[0029] In an example of the technical solution of the present invention, the process of determining static detection points is explained, a connection channel with a filing database is established, the working scope, working scene and personnel information in the workshop during the current working cycle are read, and the volume of the external space of the working scene is obtained. The larger the volume, the higher the safety. The detection density is determined by the volume. The detection density is inversely proportional to the volume, and an inverse proportional function with a constant can be used. Among them, regarding the description of the external space, the external space is the space where the staff is located, such as a small room in a workshop or some confined space. The larger the volume, the safer it is.

[0030] Furthermore, the staff information is input into a preset evaluation model, and a staff capability table is output. The staff capability table includes staff label items and capability score items, which are used to characterize the staff and their capabilities. The detection density is adjusted by the staff capability table. The adjustment method is to calculate the average capability score, determine a correction coefficient based on the average capability score, and multiply it by the detection density. The higher the average capability score, the smaller the correction coefficient, and the smaller the detection density. The correction coefficient is in the range of 0 to 1. Furthermore, the step of inserting and selecting static detection points within the working range based on the adjusted detection density and simultaneously determining the type of the static detection points includes: Obtain the contour of the working range, and evenly set static detection points of the wind type on the contour of the working range based on the detection density; Select two mutually perpendicular directions in the working range, obtain the maximum length of the working range in the two mutually perpendicular directions, and determine two mutually perpendicular dividing lines; Based on the detection density, reference points are evenly set on two mutually perpendicular cutting lines, parallel lines are constructed based on the reference points, and the intersection of the parallel lines is selected as the sampling point; The sampling point is used as a static detection point of temperature type; Based on the sampling points, static detection points of the visual type are selected; and a union of collection ranges of the static detection points of the visual type includes all sampling points.

[0031] There are three types of static detection points involved in the present invention, namely wind type, temperature type and visual type. Wind detection occurs at the boundary of the working range. Therefore, the wind parameter detectors can be evenly arranged at the same interval on the boundary of the working range according to the detection density. The interval is the boundary length multiplied by the density.

[0032] In addition, two rows of mutually perpendicular parallel lines are constructed in the working range, and the intersection of the parallel lines is used as the sampling point, and a temperature parameter detector is installed at the sampling point; at the same time, some cameras are installed and the camera parameters are determined so that all sampling points can be monitored.

[0033] The process of constructing parallel lines is to first determine two mutually perpendicular directions, query the maximum length within the working range in the two mutually perpendicular directions, determine the baseline, then evenly select points on the baseline based on the detection density, and finally construct multiple rows of parallel lines based on the points; this process is equivalent to constructing a grid within the working range.

[0034] Figure 3 This is a flowchart of the second sub-process of the pollution source reverse tracing method based on digital twins and intelligent identification. The steps of acquiring and counting working data with time tags based on static detection points and constructing and displaying a digital twin model include: Step S201: receiving the temperature uploaded by a static detection point of temperature type, and constructing a temperature matrix based on the sampling points; Step S202: receiving wind parameters uploaded by a static detection point of wind type, and correcting the temperature matrix according to the wind parameters; Step S203: Obtain a working image uploaded by a static inspection point of the visual type, perform 3D modeling based on the map within the working range and the working image, and obtain a 3D map; Step S204: inserting the corrected temperature matrix into the three-dimensional graph based on the preset color value table to obtain and display the digital twin model.

[0035] Receive the temperature uploaded by the temperature parameter detector, perform statistics according to the sampling points, and obtain the temperature matrix; receive the wind parameters uploaded by the wind parameter detector, determine the wind direction and wind speed by the wind parameters, determine the temperature conduction direction between two adjacent sampling points based on the wind direction, and determine the temperature change by the wind speed, and then fine-tune the two adjacent values ​​in the temperature matrix. Among them, the wind speed and change are determined independently by the staff. For example, 5% of the temperature of the previous sampling point in a certain wind direction is extracted and superimposed on the temperature of the next sampling point.

[0036] On this basis, the operation image is obtained, and a three-dimensional map of the current moment can be constructed from the basic layout information of the operation range and the operation image. By introducing the temperature matrix into the three-dimensional map, a three-dimensional model containing temperature information can be obtained, called a digital twin model. Finally, the digital twin model is displayed with the help of a display device.

[0037] Furthermore, the step of inserting the corrected temperature matrix into the three-dimensional graph based on the preset color value table to obtain and display the digital twin model includes: Receive the correspondence between temperature and color value uploaded by the staff and build a color value table; Read the corrected temperature matrix, diffuse the temperature matrix according to the preset diffusion rule, and obtain a three-dimensional array; Based on the color value table, the data in the three-dimensional array is converted into color values, inserted into the three-dimensional scene, and the digital twin model is obtained and displayed.

[0038] Receive the correspondence between temperature and color value uploaded by the staff, that is, what temperature corresponds to what color value; read the corrected temperature matrix. There is data only in the row and column positions corresponding to the temperature detection points in the temperature matrix. Therefore, for the row and column positions of unknown data, a model needs to be made, which is the diffusion process mentioned above. After diffusion, the obtained matrix is ​​called a three-dimensional matrix. Combined with the correspondence uploaded by the staff, the three-dimensional temperature matrix can be inserted into the three-dimensional model.

[0039] Specifically, the step of reading the corrected temperature matrix and diffusing the temperature matrix according to a preset diffusion rule to obtain a three-dimensional array includes: For any row or column position in the temperature matrix, query the nearest known temperature in the four directions of up, down, left, and right; The vertical vector is determined based on the known temperatures in the up and down directions, and the horizontal vector is determined based on the known temperatures in the left and right directions. The modulus of the vector is proportional to the temperature difference, and the direction of the vector is from high temperature to low temperature. Calculating a resultant vector of the vertical vector and the horizontal vector, and querying the nearest known temperature in the opposite direction of the resultant vector based on the resultant vector; The temperature at that row or column position is calculated based on the nearest known temperature in the opposite direction of the resultant vector.

[0040] In an example of the technical solution of the present invention, a simple diffusion process is provided. For any row and column position in the temperature matrix, a vertical vector is determined based on the known temperatures in the up and down directions, and a horizontal vector is determined based on the known temperatures in the left and right directions. The vertical vector and the horizontal vector are added to obtain a resultant vector. The resultant vector represents the direction of temperature from high to low and the temperature difference. In order to simplify the calculation process, the modulus of the resultant vector can be ignored. A nearest temperature is directly queried in the opposite direction of the resultant vector, and a nearest temperature is queried in the direction of the resultant vector. The two temperatures are regarded as a linear change relationship to determine the temperature at the row and column position; for example, the row and column positions of the two temperatures are connected to obtain a line segment, and the temperature difference is determined based on the ratio of the row and column positions to be calculated on the line segment. The temperature difference is then added to the temperature at one end.

[0041] Figure 4 This is a flowchart of the third sub-process of the pollution source reverse tracing method based on digital twins and intelligent identification. The steps of dividing the constructed digital twin model to obtain sub-models and simultaneously determining the model parameters of the sub-models include: Step S301: regularly selecting a digital twin model and traversing each point in the selected digital twin model; Step S302: When the color value difference between adjacent points is less than a preset difference threshold, the two points are connected; Step S303: Count the interconnected points to obtain a sub-model, and determine the risk level based on the color value of each point in the sub-area; Step S304: For any sub-model, apply AI to identify the subject in the sub-model, locate the pollutants, and query the type of pollutants as the risk type; The digital twin models at different times are different. By selecting the digital twin model at a certain time, obtaining the color value of each point, calculating the color value difference between adjacent points, and comparing it with the preset tolerance, similar points can be classified into one category to obtain different sub-areas. The risk level can be determined based on the color value of each point in the sub-area.

[0042] Furthermore, by applying AI to identify the subjects in the sub-model, pollutants can be located and their types can be queried as risk types. In fact, there is a better way to do this, which is to query the working image corresponding to the sub-model, apply AI to identify the working image, locate pollutants, and check the type of pollutants as risk types.

[0043] Specifically, the process of determining the risk level is as follows: Calculating a risk value and determining a risk level based on the step value reached by the risk value; each value in the step value corresponds to a risk level; The calculation process of risk value is: Where, is the preset correction factor, is the color value of the w-th point in the sub-area, For sub-regions.

[0044] In fact, the above content is essentially a contour recognition process in a three-dimensional scene, which can be migrated and applied based on the existing two-dimensional contour recognition technology.

[0045] Figure 5 This is a flowchart of the fourth sub-process of the AI-based pollution source reverse tracing method. The steps of selecting dynamic detection points based on the risk level of the sub-model, obtaining air parameters with time tags based on the dynamic detection points, and applying AI to identify the air parameters with time tags to determine the pollution source include: Step S401: selecting a sub-model whose risk level reaches a preset level threshold; Step S402: Determine the number of dynamic detection points according to the risk level; Step S403: randomly inserting the number of dynamic detection points into the sub-model, executing the process in a loop, and exiting the loop when the distance between any two dynamic detection points is not less than a preset distance threshold, outputting the insertion result; Step S404: Acquire air parameters with time tags based on dynamic detection points, apply AI to identify the air parameters with time tags, and determine the pollution source.

[0046] In an example of the technical solution of the present invention, the working process of the dynamic detection point is explained, and a sub-model whose risk level reaches a preset level threshold is selected. If the risk level is not enough, it does not need to be considered. The number of dynamic detection points is determined according to the risk level, and the said number of dynamic detection points are randomly inserted into the sub-model. The loop is executed to obtain multiple insertion methods. When there is a method in which the distance between any two dynamic detection points is not less than the preset distance threshold (large enough), the loop is exited and the insertion result is output; based on the dynamic detection points, air parameters with time tags are obtained, and AI is used to identify the air parameters with time tags to determine the pollution source.

[0047] Furthermore, the steps of obtaining air parameters with time tags based on dynamic detection points, applying AI to identify the air parameters with time tags, and determining the pollution source include: Count the dynamic detection points of all sub-models and use the counted dynamic detection points as waypoints to generate a path; Sending the route to at least one UAV, and obtaining air parameters at each dynamic detection point based on an air detector installed on the UAV; the air parameters contain a time tag, and the air parameters are air type and concentration; The pollution source is determined based on the air parameters at each dynamic detection point.

[0048] When the dynamic detection points in each sub-model with a sufficiently large risk level are determined, the counted dynamic detection points are used as waypoints to generate a path. The path generation process can use the existing path generation algorithm. The path is sent to at least one drone, and the air parameters at each dynamic detection point are obtained based on the air detector installed on the drone. The pollution source is determined based on the air parameters at each dynamic detection point.

[0049] It should be noted that, assuming that a drone takes half an hour to work along a path, the data acquisition frequency for each dynamic detection point is more than half an hour (including the reset time). If you want to increase the data acquisition frequency, you don't need to increase the drone's movement speed, but add another drone with the same working conditions to double the data acquisition frequency. The more drones there are, the greater the data acquisition frequency.

[0050] Specifically, the step of determining the pollution source based on the air parameters at each dynamic detection point includes: For each dynamic detection point, query the air parameters at the most recent moment based on the time tag; Determine a concentration threshold according to the air type in the air parameter, and when the concentration in the air parameter reaches a preset concentration threshold, determine the radiation radius according to the concentration; A spherical area is created based on the radiation radius with the dynamic detection point as the center, and the sub-models that intersect with the spherical area are marked as contaminated areas; The number of times each sub-model is marked is counted. When the number of marks reaches the preset threshold, AI is used to identify the subject in the sub-model and locate the pollutants.

[0051] In an example of the technical solution of the present invention, the process of determining the pollution source is explained. For each dynamic detection point, the air parameters at the most recent moment are queried according to the time tag (because the data acquisition frequency is very low, the air parameters at the most recent moment need to be queried as the air parameters at the current moment), and the concentration threshold is determined according to the air type in the air parameters. When the concentration in the air parameters reaches the preset concentration threshold, the radiation radius is determined according to the concentration. Based on the radiation radius, a spherical area is created with the dynamic detection point as the center of the circle, and the sub-model that intersects with the spherical area is marked as a polluted area. When the number of markings of a sub-model is large enough, AI is used to identify the subject in the sub-model and locate the pollutants.

[0052] It should be noted that the sub-model itself is obtained from the working image. In fact, you can also query the working image corresponding to the sub-model, and use AI to identify the working image to obtain more accurate recognition results.

[0053] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A pollution source reverse tracing method based on digital twins and intelligent identification, characterized in that: The method comprises: Obtaining work information in the workshop based on the filing information, selecting static inspection points based on the work information, and simultaneously determining the types of static inspection points; the work information includes work scope, work scene, and personnel information; the types of static inspection points include wind type, temperature type, and visual type; Acquire and count working data with time tags according to the static detection points, and construct and display a digital twin model; The constructed digital twin model is segmented to obtain sub-models, and model parameters of the sub-models are simultaneously determined; the model parameters include risk level and risk type; the risk type is the pollutant type; Dynamic detection points are selected according to the risk level of the sub-model, air parameters containing time tags are obtained based on the dynamic detection points, and the air parameters containing time tags are identified through the pollution source adaptive identification model to determine the pollution source.

2. The method according to claim 1, characterized in that The method for setting the static detection point is: Establish a connection channel with the filing database to read the work scope, work scene and personnel information in the workshop during the current work cycle; Obtaining the volume of the external space of the work scene, and determining a detection density based on the volume; the detection density is inversely proportional to the volume; and the detection density is expressed as a percentage; Input the staff information into the preset evaluation model, output the staff capability table, and adjust the inspection density based on the staff capability table; Based on the adjusted detection density, static detection points are inserted and selected within the working range, and the types of the static detection points are simultaneously determined.

3. The method according to claim 2, characterized in that The method for selecting the static detection points is: Obtain the contour of the working range, and evenly set static detection points of the wind type on the contour of the working range based on the detection density; Select two mutually perpendicular directions in the working range, obtain the maximum length of the working range in the two mutually perpendicular directions, and determine two mutually perpendicular dividing lines; Based on the detection density, reference points are evenly set on two mutually perpendicular cutting lines, parallel lines are constructed based on the reference points, and the intersection of the parallel lines is selected as the sampling point; The sampling point is used as a static detection point of temperature type; Based on the sampling point selection type, the static detection point is of visual type; The union of the collection ranges of static detection points of the vision type includes all sampling points.

4. The method according to claim 1, wherein The steps of building the digital twin model include: Receive the temperature uploaded by static detection points of temperature type and build a temperature matrix based on the sampling points; Receive wind parameters uploaded by static detection points of wind type, and use the wind parameters to correct the temperature matrix; Obtain the working image uploaded by the static inspection point of the visual type, perform 3D modeling based on the map within the working range and the working image, and obtain a 3D map; The corrected temperature matrix is ​​inserted into the three-dimensional graph based on the preset color value table to obtain and display the digital twin model.

5. The method according to claim 4, characterized in that The display steps of the digital twin model include: Receive the correspondence between temperature and color value uploaded by the staff and build a color value table; Read the corrected temperature matrix, diffuse the temperature matrix according to the preset diffusion rule, and obtain a three-dimensional array; Based on the color value table, the data in the three-dimensional array is converted into color values, inserted into the three-dimensional scene, and the digital twin model is obtained and displayed.

6. The method according to claim 5, characterized in that The diffusion step of the temperature matrix includes: For any row or column position in the temperature matrix, query the nearest known temperature in the four directions of up, down, left, and right; The vertical vector is determined based on the known temperatures in the up and down directions, and the horizontal vector is determined based on the known temperatures in the left and right directions. The modulus of the vector is proportional to the temperature difference, and the direction of the vector is from high temperature to low temperature. Calculating a resultant vector of the vertical vector and the horizontal vector, and querying the nearest known temperature in the opposite direction of the resultant vector based on the resultant vector; The temperature at that row or column position is calculated based on the nearest known temperature in the opposite direction of the resultant vector.

7. The method according to claim 1, characterized in that The steps of dividing the digital twin model into sub-models include: Select the digital twin model regularly and traverse each point in the selected digital twin model; When the color value difference between adjacent points is less than the preset difference threshold, the two points are connected; Count the interconnected points to obtain a sub-model, and determine the risk level by the color value of each point in the sub-area; For any sub-model, identify the subject in the sub-model, locate the pollutants, and use the pollutant type obtained from the query as the risk type; The process of determining the risk level is as follows: Calculating a risk value and determining a risk level based on the step value reached by the risk value; each value in the step value corresponds to a risk level; The calculation process of the risk value is as follows: ; Where L is the risk value, is the preset correction factor, is the color value of the w-th point in the sub-area, For sub-regions.

8. The method according to claim 1, characterized in that The steps of determining the pollution source include: Select the sub-model whose risk level reaches the preset level threshold; Determine the number of dynamic detection points based on risk level; randomly inserting the number of dynamic detection points into the sub-model, executing the operation in a loop, and exiting the loop and outputting the insertion result when the distance between any two dynamic detection points is not less than a preset distance threshold; Based on the dynamic detection points, the air parameters with time tags are obtained, and the pollution source adaptive identification model is applied to identify the air parameters with time tags to determine the pollution source.

9. The method according to claim 8, characterized in that The pollution source adaptive identification model includes: Count the dynamic detection points of all sub-models and use the counted dynamic detection points as waypoints to generate a path; Sending the route to at least one UAV, and obtaining air parameters at each dynamic detection point based on an air detector installed on the UAV; the air parameters contain a time tag, and the air parameters are air type and concentration; The pollution source is determined based on the air parameters at each dynamic detection point.

10. The method according to claim 9, characterized in that The steps of determining the pollution source based on the air parameters at each dynamic detection point include: For each dynamic detection point, query the air parameters at the most recent moment based on the time tag; Determine a concentration threshold according to the air type in the air parameter, and when the concentration in the air parameter reaches a preset concentration threshold, determine the radiation radius according to the concentration; A spherical area is created based on the radiation radius with the dynamic detection point as the center, and the sub-models that intersect with the spherical area are marked as contaminated areas; The number of times each sub-model is marked is counted. When the number of times the mark reaches a preset threshold, the pollution source adaptive identification model is applied to identify the subject in the sub-model to locate the pollution source.

Images