Temperature gradient measurement method based on striae and neural network and air supply and exhaust regulation system

Abstract

The application relates to a temperature gradient measuring method based on a schlieren and a neural network and an air supply and exhaust regulation system, and relates to the technical field of air flow temperature field detection. The regulation method comprises the following steps: collecting background schlieren images and air flow diffusion schlieren images of a heat source monitoring area through a first background schlieren temperature measuring platform and a second schlieren diffusion imaging platform; extracting light deflection displacement features of the background schlieren images, inputting the trained BP neural network model, and obtaining grid temperature field and temperature gradient distribution features; extracting position coordinates of a transient air flow diffusion front end of the air flow diffusion schlieren images and mapping the position coordinates to the grid temperature field to determine the temperature gradient and the corresponding temperature feature value mean value of the grid; comparing the mean value with a preset early warning threshold value, generating air supply and exhaust regulation and alarm instructions in combination with the number of on-site operators, controlling the linkage air supply and exhaust of the same layer ventilation gate, the bottom air supply unit and the top exhaust unit, and stopping the operation when the mean value is lower than a stop threshold value.

Description

Technical Field

[0001] This invention relates to the field of airflow temperature field detection technology, and more specifically, to a method for measuring temperature gradient based on schlieren and neural networks, and an air supply and exhaust control system. Background Technology

[0002] In numerous applications of high-temperature heat sources in industrial production processes (such as molten metal tapping, casting, or high-temperature slag pouring in metallurgy and casting processes; furnace feeding and crucible casting in glass production; and slag pouring in fly ash melting furnaces or plasma gasification furnaces for solid waste and hazardous waste treatment), and in indoor high-temperature scientific experiments (such as aero-engine exhaust plume injection and transient combustion chamber combustion gas ignition), the monitoring of the diffusion characteristics of rapidly heating airflows around indoor high-temperature work areas and the control of air supply and exhaust are directly related to the occupational health and safety of workers. Rapidly and accurately acquiring the temperature field characteristics and temperature gradient distribution of the transient diffusion front of flowing high-temperature airflow within a large field of view, and adaptively controlling the air supply and exhaust devices accordingly, is a key requirement for improving high-temperature working environments and achieving low-energy operation. Existing temperature measurement technologies are insufficient to accurately capture the high-temperature characteristic values ​​of the transient airflow diffusion front, and cannot provide effective data support for the precise linkage of air supply and exhaust devices.

[0003] Currently, common methods for detecting airflow temperature fields in industry include thermocouple contact temperature measurement methods with single or multiple locations, as well as non-contact temperature measurement methods such as cooled infrared measurement, far-infrared measurement, acoustic measurement, and background schlieren method. Thermocouple contact temperature measurement requires direct contact with the gas flow field being measured, and can only obtain temperature amplitude within a short distance range. Furthermore, it suffers from long temperature rise and fall exchange times, low testing frequency, and insufficient real-time performance. Non-contact temperature measurement methods, such as the cooled infrared measurement method using a cooled high-sensitivity detector and assisted spectral filtering, and the far-infrared measurement method based on the characteristic radiation (absorption spectrum) characteristics of gas molecules in the far-infrared band, are all based on Planck's blackbody radiation law, Stephan-Boltzmann law, and the theory of gas molecule vibrational spectroscopy. The specific infrared or far-infrared bands selected for gas molecules are easily affected by the flow field of the gas being measured. The algorithms for calculating the temperature field by inverting the spectral radiation signal are complex, resulting in insufficient real-time performance. Furthermore, both methods require specific narrowband filtering processing, taking into account the molecular composition of the gas flow field and its characteristic radiation or absorption spectrum characteristics in specific infrared or far-infrared bands. Acoustic measurement methods face challenges in decoupling the acoustic transmission velocity field from the temperature field being measured, difficulties in acquiring acoustic wave transmission delay signals, and complex inverse solutions from acoustic imaging. Background schlieren measurement methods, which use pixel window cross-correlation algorithms or optical flow algorithms to solve the temperature field, are time-consuming and struggle to accurately capture the temperature gradient distribution characteristics and nonlinear diffusion characteristics at the gas diffusion front.

[0004] The aforementioned existing technologies all have significant limitations: contact temperature measurement methods cannot obtain large field-of-view temperature field characteristic values ​​and cannot reflect the nonlinear diffusion law of airflow. Non-contact temperature measurement methods, due to insufficient algorithm efficiency, are difficult to meet the real-time requirements under transient conditions, and none of them can quantitatively express the temperature field gradient distribution characteristics. At the same time, the optical path systems of these technologies are often quite complex, requiring a large number of precision optical and acoustic measurement instruments, resulting in high engineering implementation costs and making them difficult to adapt to the adaptive supply and exhaust ventilation control requirements based on high-temperature early warning at the airflow diffusion front end in industrial scenarios. Summary of the Invention

[0005] This invention provides a method for measuring temperature gradients based on schlieren and neural networks, and a supply and exhaust ventilation control system, to improve at least one of the above-mentioned technical problems.

[0006] In a first aspect, the present invention provides a method for measuring temperature gradients based on schlieren and neural networks, which includes steps S1 to S6.

[0007] S1. Background schlieren images and airflow diffusion schlieren images are acquired by using the first background schlieren temperature measurement platform and the second schlieren diffusion imaging platform arranged in the heat source monitoring area.

[0008] S2. Extract the light deflection displacement features from the background schlieren image, input them into the trained BP neural network model, and calculate the gridded temperature field information and temperature gradient distribution information of the heat source monitoring area. The neural network model is trained using the light deflection displacement features and the temperature field training dataset, and the temperature field and temperature gradient are calculated using the light deflection displacement features and the neural network model.

[0009] S3. Extract the position coordinate information of the transient airflow diffusion front end from the airflow diffusion schlieren image, and map the position coordinate information to the gridded temperature field information to determine the grid set where the transient airflow diffusion front end is located.

[0010] S4. Determine the mean value of the temperature characteristic value at the front end of the transient airflow diffusion based on the temperature information and temperature gradient information in the grid set.

[0011] S5. Compare the average temperature characteristic value of the transient airflow diffusion front end with the preset temperature warning threshold, and generate supply and exhaust ventilation control instructions and alarm instructions in combination with the number of workers in the heat source monitoring area.

[0012] S6. Control the normally closed ventilation damper, the bottom air supply unit and the top exhaust unit to operate in conjunction according to the air supply and exhaust control command, so that the high-temperature indoor airflow in the heat source monitoring area is discharged through the vertical air supply flue, and the air supply and exhaust operation is stopped when the average temperature characteristic value at the front end of the transient airflow diffusion is lower than the preset stop threshold.

[0013] Secondly, the present invention provides a supply and exhaust air control system based on schlieren and neural network for measuring temperature gradients, used to execute the method for measuring temperature gradients based on schlieren and neural network described in any paragraph of the first aspect.

[0014] The control system includes:

[0015] The first background schlieren temperature measurement platform is used to acquire background schlieren images before and after airflow temperature field disturbance.

[0016] The second schlieren diffusion imaging platform is used to acquire airflow diffusion schlieren images.

[0017] The computing platform is used to generate gridded temperature field information and temperature gradient distribution information based on the background schlieren image, extract the position coordinate information of the transient airflow diffusion front end based on the airflow diffusion schlieren image, and determine the mean value of the temperature feature value of the transient airflow diffusion front end.

[0018] Personnel detection device, used to detect the number of workers in the heat source monitoring area.

[0019] The control platform is used to generate air supply and exhaust control commands and alarm commands based on the average temperature characteristic value of the transient airflow diffusion front end and the number of operators.

[0020] A supply and exhaust ventilation actuator is used to exhaust indoor airflow through a vertical supply and exhaust duct in response to the supply and exhaust ventilation control command.

[0021] By adopting the above technical solution, the present invention can achieve the following technical effects: This invention enables rapid and accurate measurement of transient airflow temperature gradient distribution characteristics in high-temperature indoor work areas of buildings, achieving a large field of view, quantitative analysis, and visualization. It utilizes background schlieren technology to build a light deflection displacement field measurement platform and combines a robust and highly accurate neural network model to construct a nonlinear mapping relationship between the light deflection displacement field and the airflow temperature field. Furthermore, it employs a partially dependent graph method to accurately calculate the temperature gradient distribution characteristic values. This effectively overcomes the limitations of traditional contact and non-contact temperature measurement methods, such as limited measurement field of view, slow temperature response, and difficulty in capturing the temperature gradient distribution characteristics of large-field transient temperature fields. The computational efficiency is significantly improved compared to classical optical and acoustic temperature measurement methods, fully meeting the engineering application needs for transient monitoring and real-time early warning of high-temperature airflow in industrial settings. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the structural layout from the first horizontal perspective of the control system.

[0023] Figure 2 This is a schematic diagram of the structural layout from the second-level perspective of the control system.

[0024] Figure 3This is the front view of the bottom-level air supply unit.

[0025] Figure 4 This is a side view of the bottom-level air supply unit.

[0026] Figure 5 This is the front view of the top-floor exhaust unit.

[0027] Figure 6 This is a side view of the top-floor exhaust unit.

[0028] Figure 7 This is a flowchart of the control method.

[0029] Figure 8 A schematic diagram to visualize the airflow temperature field distribution characteristics of the entire operating area.

[0030] Figure 9 A schematic diagram for visualizing the location of the grid at the front end of the airflow diffusion.

[0031] Figure 10 A schematic diagram for quantitatively calibrating the temperature gradient distribution in different grid regions along the horizontal direction.

[0032] Figure 11 A schematic diagram for quantitatively calibrating the temperature gradient distribution in different grid regions along the vertical direction.

[0033] Attached reference numerals: 101-Top-level exhaust unit, 102-Normal closed ventilation damper, 103-Buzzer alarm, 104-Background schlieren pattern splicing panel, 105-First black-and-white industrial camera, 106-LED light source, 107-Computer (including temperature field neural network calculation and high-temperature early warning monitoring platform), 108-Microcontroller control platform (connected to computer 107), 109-Heat source monitoring area (i.e., high-temperature working area), 401-Infrared camera, 110-Bottom-level air supply unit, 111-Vertical air supply flue, 112-Second black-and-white industrial camera, 113-Knife edge, 114-White LED point light source, 115-Concave spherical mirror. 201-First housing, 202-First fixed base, 203-First duct system, 204-First flexible joint, 205-First drive AC motor, 206-First belt conveyor system, 207-First centrifugal impeller, 208-First fan volute, 209-First impeller support frame. 301-Second housing, 302-Second fixed base, 303-Second duct system, 304-Second flexible joint, 305-Second drive AC motor, 306-Second belt conveyor system, 307-Second centrifugal impeller, 308-Second fan volute, 309-Second impeller support frame. Detailed Implementation

[0034] Example 1, please refer to Figures 1 to 6The first embodiment of the present invention provides a supply and exhaust air control system based on schlieren and neural network for measuring temperature gradients, used to execute the supply and exhaust air control method based on schlieren and neural network for measuring temperature gradients and the supply and exhaust air control method described in any paragraph of Embodiment 2.

[0035] The control system includes a first background schlieren temperature measurement platform, a second schlieren diffusion imaging platform, a computing platform, a personnel detection device, a control platform, and a ventilation and exhaust execution device.

[0036] The first background schlieren temperature measurement platform is used to acquire background schlieren images before and after airflow temperature field disturbance. The first background schlieren temperature measurement platform includes a background schlieren pattern splicing board 104, a first black and white industrial camera 105, and an LED light source 106. The background schlieren pattern splicing board 104 includes multiple random noise pattern boards. The first black and white industrial camera 105 is coaxially arranged with the center position of the background schlieren pattern splicing board 104. The LED light source 106 is a stable, controllable, and uniformly illuminated dot matrix distributed LED light source.

[0037] The second schlieren diffusion imaging platform is used to acquire airflow diffusion schlieren images. The second schlieren diffusion imaging platform includes a second monochrome industrial camera, a concave spherical mirror 115, a knife edge 113, and a white LED point light source 114. The optical axis of the second monochrome industrial camera is coaxially arranged with the principal optical axis of the concave spherical mirror 115. The knife edge 113 and the white LED point light source 114 are arranged in parallel, and the knife edge 113 is located at the radius of curvature of the coaxial optical path of the concave spherical mirror 115.

[0038] The blade 113 is a light-blocking element with a straight edge (similar to the edge of a blade), positioned at the radius of curvature of the concave spherical mirror 115. When there is airflow temperature disturbance, parallel light rays passing through the airflow are deflected. The position of the deflected light rays at the radius of curvature changes, causing the light rays that should have passed through entirely to be blocked to varying degrees by the blade 113. This difference in the amount of blocking ultimately translates into contrast in the camera image: areas with large temperature gradients show stronger light deflection and more noticeable changes in image brightness.

[0039] The computing platform generates gridded temperature field information and temperature gradient distribution information based on the background schlieren image, extracts the position coordinates of the transient airflow diffusion front end from the airflow diffusion schlieren image, and determines the average temperature feature value of the transient airflow diffusion front end. The computing platform includes a temperature field neural network calculation module and a temperature early warning monitoring module. The temperature field neural network calculation module establishes a mapping relationship between light deflection displacement characteristics and temperature field information, and the temperature early warning monitoring module generates early warning results based on the average temperature feature value of the transient airflow diffusion front end.

[0040] The personnel detection device is used to detect the number of workers within the heat source monitoring area. The personnel detection device is an infrared camera, which is used to acquire thermal images of the workers.

[0041] The control platform generates air supply and exhaust control commands and alarm commands based on the average temperature characteristic value of the transient airflow diffusion front end and the number of operators. The control platform includes a microcontroller control platform and a wireless communication module. The microcontroller control platform sends start / stop commands and gear control commands through the wireless communication module.

[0042] The air supply and exhaust actuator is used to discharge indoor airflow through a vertical air supply duct in response to the air supply and exhaust control command. The air supply and exhaust actuator includes a bottom-level air supply unit, a top-level exhaust unit, a vertical air supply duct, and normally closed ventilation dampers located on each floor of the building. The bottom-level air supply unit is connected to the bottom of the vertical air supply duct and is used to supply pressurized air into the vertical air supply duct. The top-level exhaust unit is connected to the top of the vertical air supply duct and is used to create a negative pressure flow within the vertical air supply duct. After the normally closed ventilation damper opens in response to the air supply and exhaust control command, it allows high-temperature indoor airflow from the heat source monitoring area to enter the vertical air supply duct and be discharged. Here, "high-temperature" in the high-temperature indoor airflow refers to a temperature greater than or equal to 35°C. In meteorology, high temperature refers to weather with a daily maximum temperature of 35°C or above. This embodiment classifies the air supply and exhaust control levels according to the high-temperature operation (GBZ / T·229.3-2025) grading standard based on the work intensity level of different work areas. Taking medium-intensity high-temperature operations as an example, when the temperature at the indoor airflow diffusion front reaches 35°C, 33°C, and 31°C, the high, medium, and low speed control levels of the drive air supply and exhaust motors are activated respectively.

[0043] The control system also includes a buzzer alarm, which is used to issue an alarm when the average temperature characteristic value at the front end of the transient airflow diffusion reaches a preset temperature warning threshold.

[0044] Specifically, air supply and exhaust control systems based on schlieren and neural networks to measure temperature gradients include... Figure 1 and Figure 2 As shown.

[0045] Figure 1 The schematic diagram of the first horizontal perspective of the control system shows the first working platform in a typical industrial building (taking a 2-story building as an example) for quickly and accurately measuring the temperature gradient distribution characteristics of transient airflow using background schlieren technology and neural network methods.

[0046] Figure 2The schematic diagram of the second-level perspective structure of the control system shows a second working platform that uses schlieren imaging technology to monitor the diffusion characteristics of transient airflow at high temperatures. Based on the coordinate information of the airflow diffusion front end and its temperature gradient distribution characteristics, the average temperature of the airflow diffusion front end is quantitatively calibrated. This value is then compared with the results of calculations by a computer-installed neural network and the high-temperature early warning threshold preset by the high-temperature early warning monitoring platform, and the air supply and exhaust devices are adaptively controlled in a linked manner.

[0047] Figure 1 The main functions of each component are as follows: The air inlet on the left side of the top floor exhaust unit 101 is installed on the top side of the vertical air supply duct 111, which plays the role of negative pressure top suction exhaust. It works at the same time as the pressure boosting and air supply of the bottom floor air supply unit 110, forming a negative pressure air suction effect above the vertical air supply duct, and forming a real-time flow guiding effect on the airflow diffusion in the indoor high-temperature working area and the airflow in the vertical air supply duct.

[0048] The drive motors of the top-floor exhaust fan unit 101 and the bottom-floor supply fan unit 110 are equipped with low, medium, and high speed control modules and 4G / 5G communication network signal receivers to receive start / stop and speed control commands sent by the microcontroller control platform. The normally closed ventilation dampers 102 on each floor contain 4G / 5G communication network signal receivers (model 23C44WR29111) to receive start / stop control commands from the microcontroller control platform.

[0049] Upon receiving the start signal, the gate device opens, and the high-temperature diffused airflow in the room is discharged through the gate to the vertical air supply flue 111. Under the negative pressure top suction, it is guided to the top of the flue on the top floor of the industrial building and discharged into the air through the vertical air supply flue and exhaust control device.

[0050] The buzzer alarm 103 is installed and fixed on a wall 2.5 meters above the ground. When the computer 107 measures and monitors in real time that the temperature characteristic value at the front end of the indoor high-temperature airflow diffusion exceeds the preset high-temperature warning value, it will sound an alarm to warn on-site workers to stop work and move away from the high-temperature work area, or to wear high-temperature work protection equipment to deal with the risk of high-temperature injury and potential occupational diseases. The computer 107 has built-in temperature field neural network calculation software and a high-temperature early warning monitoring platform.

[0051] The background schlieren pattern splicing panel 104 is set with a random noise pattern and consists of four square panels with a perimeter of 0.6 meters, spliced ​​together and fixed to the interior wall of the building. It is positioned approximately 1 meter above the heat source monitoring area 109, with its center coaxial with the optical path of the industrial camera. A first monochrome industrial camera 105, equipped with a camera lens and capable of taking pictures, is calibrated and mounted at a height coaxial with the center line of the background schlieren pattern splicing panel. Under stable, controllable, and uniform illumination provided by the dot matrix distributed LED light source 106, it continuously captures multiple real-time images of the background schlieren pattern changes according to a preset exposure time (50 milliseconds).

[0052] Figure 2 The second black-and-white industrial camera 112 and Figure 1 The first monochrome industrial camera 105 must possess high resolution, large aperture / depth-of-field performance, and a camera lens. Preferably, the parameters of the first monochrome industrial camera 105 and the second monochrome industrial camera 112 are: resolution 2592. 2048, pixel size 4.8μm 4.8μm, photosensitive area 12.44mm 9.83mm, 60fps frame rate, global shutter, C-Mount interface. Camera lens parameters: focal length 35mm, supports 1.1-inch target surface, optical resolution 150lp / mm, aperture range F2.8-F16, minimum focusing distance 2m. It employs the same camera configuration and height position calibration process. The concave spherical mirror 115 has a diameter of 406mm and a focal length of 1624mm. The blade edge 113 is positioned coaxially with the concave spherical mirror 115 at the same height as its centerline and is located at the radius of curvature of the concave spherical mirror 115. After calibration, the second monochrome industrial camera 112 is positioned coaxially with the concave spherical mirror 115 at the same height as its centerline. Under the condition of a high-intensity incident point light source provided by the white LED point light source 114, it continuously captures multiple real-time schlieren images formed by the light reflected from the concave spherical mirror 115 and transmitted through the blade edge 113, according to a preset exposure time (50ms).

[0053] Computer 107 is mounted on a wooden platform 20 centimeters off the ground. It receives high-resolution images of background mosaic patterns captured by a first black-and-white industrial camera 105 under conditions of airflow temperature field disturbance in the work area via a data transmission cable. The pixel resolution of the acquired high-resolution images is set to 2592. 2048. Following the steps of Example 2, based on the light deflection displacement field signal caused by the airflow temperature field disturbance measured by background schlieren measurement and the data predicted by neural network modeling, the temperature gradient distribution characteristics of different grid partitions in the high-temperature operation area are quickly calculated and visualized, maintaining real-time monitoring.

[0054] Computer 107 simultaneously receives high-definition images captured by the second black-and-white industrial camera 112, extracts visual images of airflow diffusion in different grid zones obtained based on schlieren imaging measurement, extracts the airflow temperature gradient feature values ​​and airflow diffusion position coordinate information in the same grid area, and obtains the average transient temperature feature values ​​at the front end of airflow diffusion in different directions.

[0055] The average value of the transient temperature gradient characteristic value at the front end of the airflow diffusion is compared with the high temperature warning threshold preset in the high temperature warning platform in computer 107. The comparison result is sent to the microcontroller control platform 108 to trigger the high temperature warning linkage response control signal.

[0056] The microcontroller control platform 108 carries a 4G / 5G communication device and uses a data transmission line to maintain a connection with the high temperature early warning monitoring platform built into the computer 107, and receives high temperature early warning start and stop control commands.

[0057] When the high temperature warning platform 107 monitors in real time that the average temperature characteristic of the front end of the high temperature airflow diffusion exceeds the preset high temperature warning threshold, it combines the image of the number of workers sent back by the infrared camera 401 and sends a circuit drive signal through the 4G / 5G communication device of the microcontroller control platform 108. After receiving the drive signal, the normally closed ventilation gate 102, the top exhaust unit 101 and the bottom air supply unit 110 start the high, medium and low speed control modules according to different warning response levels to accurately control the air supply and exhaust efficiency.

[0058] The bottom-level air supply unit 110 is installed on the outside of the bottom floor of the industrial building and is protected by a steel structure fence. It is used to supply pressurized air into the vertical air supply flue of the building and works together with the top-level exhaust unit to create a negative pressure air suction environment in the vertical air supply flue. The fan unit is equipped with a three-speed control module with low, medium and high speeds.

[0059] Infrared camera 401 is installed at the angle between the top surface of the working floor and the vertical facade of the building. It must ensure that the infrared wide-angle camera's coverage area can encompass the entire floor. In wide-angle mode, the camera can capture thermal images of personnel working inside the factory and transmit the real-time footage back to the computer via a 4G / 5G communication network. Preferably, infrared camera 401 is an infrared sensing wide-angle camera.

[0060] The ground floor ventilation unit is installed on the ground floor of the building. Figure 3 and Figure 4 They are respectively Figure 1 Front and side views of the middle and lower level air supply unit 110. The lower level air supply unit includes components such as the unit housing, fan system, first drive AC motor, connection structure, and mounting base.

[0061] The first housing 201 is the core load-bearing structure of the unit, mainly used to provide installation space for components such as centrifugal fans and motors, protect the internal structure, and also serve to insulate against heat and reduce noise and energy loss. It also guides and pressurizes the airflow entering the housing from the outside. The housing structure is equipped with an inspection door and an external air inlet. The first fixed base 202 secures the first housing 201 with fastening bolts.

[0062] The first drive AC motor 205 is connected to the 4G / 5G communication network signal receiving device. After receiving the warning level signal from the single-chip microcomputer control platform, it starts high, medium and low speed according to the corresponding warning level response, driving the centrifugal fan main shaft and impeller to rotate at high speed. With the help of the first belt conveyor system 206, the first centrifugal impeller 207 is driven to rotate around the shaft, which makes the airflow in the first fan volute 208 accelerate and generate pressurized airflow, which is guided and collected at the air outlet.

[0063] The first duct system 203 and the first flexible joint 204, which connect the air supply unit and the vertical air supply flue, output pressurized airflow to the relatively closed vertical ventilation flue, forming a top suction effect and negative pressure air suction environment with the top exhaust unit.

[0064] The first impeller support frame 209 is mainly used to fix the impeller shaft and the impeller itself, and to fasten them inside the first housing 201 to ensure reduced vibration when the first centrifugal impeller rotates. The first flexible joint 204, which connects the first duct system and the vertical air supply flue, mainly functions to isolate and reduce vibration and compensate for installation errors, preventing the vibration generated by the high-speed rotation of the fan system from being transmitted to the first duct system.

[0065] The top-floor exhaust unit is installed on the top floor of the building. Figure 5 and Figure 6 They are respectively Figure 1 Front and side views of the top-level exhaust fan unit 101. The top-level exhaust fan unit mainly includes the unit housing, fan system, second drive AC motor, connection structure, and mounting base components.

[0066] The structure of the top-level exhaust unit is similar to that of the bottom-level air supply unit, with components having the same function, installation and fastening methods, and corresponding positional relationships. It mainly includes a second housing 301, a second fixed base 302, a second drive AC motor 305, a second belt conveyor system 306, a second duct system 303, a second flexible joint 304, a second fan volute 308, and a second impeller support frame 309.

[0067] The second housing 301 is the core load-bearing structure of the top-level exhaust fan. The housing structure includes an inspection door and an external air outlet. The second fixed base 302 secures the second housing 301 with fastening bolts. The second drive AC motor 305 is connected to a 4G / 5G communication network signal receiving device. Upon receiving a warning level signal from the microcontroller control platform, it responds according to the corresponding warning level, starting at high, medium, or low speeds to drive the centrifugal fan main shaft and impeller to rotate at high speed. Through the second belt conveyor system 306, it drives the second centrifugal impeller 307 to rotate around its axis, causing the airflow within the second fan volute 308 to accelerate and generate pressurized airflow, which is then guided and collected at the air outlet.

[0068] The air is pressurized and supplied by the blower unit and transported to the exhaust control unit via the vertical air supply flue through the second duct system 303 and the second flexible joint 304. The exhaust control unit draws air to form a top suction effect and a negative pressure air suction environment in the vertical air supply flue. The second impeller support frame 309 is mainly used to fix the impeller shaft and the impeller itself, and is fastened inside the second housing 301.

[0069] The main difference between the operation modes of the exhaust control unit and the air supply control unit lies in the fact that when the second centrifugal impeller of the exhaust unit is rotating at high speed, a negative pressure environment is generated at the air inlet of the second fan casing. After being connected to the vertical air supply flue through the second air duct system, the negative pressure air extraction environment of the top-floor exhaust unit and the pressurized air supply environment of the bottom-floor air supply unit together form the top-floor negative pressure air suction effect along the vertical air supply flue of the building.

[0070] Therefore, when a normally closed ventilation damper on a certain floor is opened, the high-temperature airflow inside can be quickly diverted to the top floor of the building, carrying away the indoor temperature and collecting it for discharge until the indoor airflow temperature drops below the temperature warning threshold.

[0071] Example 2: This example utilizes a first working platform to measure the light deflection displacement characteristics under transient airflow temperature field disturbance conditions via background schlieren imaging, and a second working platform to monitor the nonlinear diffusion characteristics of airflow via schlieren imaging. Combined with neural network modeling and prediction methods, it can quickly, accurately, and with a large field of view measure the transient airflow temperature gradient distribution characteristics and their dynamic diffusion characteristics in high-temperature operating areas. Based on the comparison between these temperature gradient distribution characteristics and a preset high-temperature warning threshold, it can quickly and adaptively adjust the air supply units on the ground floor and the exhaust units on the top floor of the building. The measurement process for the average value of the transient airflow diffusion front-end temperature gradient distribution characteristics and the traditional technical route for measuring the temperature field are as follows: Figure 7 As shown.

[0072] Please see Figures 7 to 11 A method for measuring temperature gradients based on schlieren and neural networks and a method for controlling air supply and exhaust, comprising steps S1 to S6.

[0073] S1. Background schlieren images and airflow diffusion schlieren images are acquired by using the first background schlieren temperature measurement imaging platform and the second schlieren diffusion imaging platform arranged in the heat source monitoring area.

[0074] The first background schlieren temperature measurement platform includes a background schlieren pattern splicing board 104, a first black and white industrial camera 105, and an LED light source 106. The first black and white industrial camera 105 is coaxially arranged with the center position of the background schlieren pattern splicing board 104. The LED light source 106 is a stable, controllable, and uniformly illuminated dot matrix distributed LED light source.

[0075] The second schlieren diffusion imaging platform includes a second monochrome industrial camera 112, a concave spherical mirror 115, a knife edge 113, and a white LED point light source 114. The optical axis of the second monochrome industrial camera 112 is coaxially arranged with the principal optical axis of the concave spherical mirror 115. The knife edge 113 and the white LED point light source 114 are arranged in parallel. The knife edge 113 is located at the radius of curvature of the coaxial optical path of the concave spherical mirror 115.

[0076] The imaging direction of the first schlieren thermography imaging platform is perpendicular to the imaging direction of the second schlieren diffusion imaging platform.

[0077] In the absence of airflow and temperature field disturbance, the first black and white industrial camera is used to acquire static background schlieren images, and the coordinates of the reference noise points are determined based on multiple static background schlieren images.

[0078] Under the condition of airflow temperature field disturbance, the first black and white industrial camera is used to acquire background schlieren images, and the second black and white industrial camera is used to simultaneously acquire transient airflow dynamic diffusion schlieren images after reflection by the concave spherical mirror 115 and refraction by airflow temperature field disturbance.

[0079] Specifically, in order to quickly, accurately, and with a wide field of view measure the transient airflow temperature gradient distribution characteristics in high-temperature operating areas, in Figure 1 In the heat source monitoring area 109 shown in the room, a first working platform is built along the first horizontal viewpoint to measure the transient airflow temperature gradient distribution characteristics of the high-temperature airflow disturbance area based on background schlieren. The instantaneous heat source temperature of the heat source monitoring area 109 can reach 250℃ or above.

[0080] Under normal indoor temperature conditions, such as 20℃, the stability adjustment, center height coaxiality adjustment, focusing and optical parameter settings of the first black and white industrial camera 105 and the background schlieren pattern splicing board 104 are calibrated.

[0081] After calibration, five to ten static background schlieren images without airflow temperature field disturbance are captured using the first black and white industrial camera 105. Three images are randomly selected from these images, and the average pixel coordinates of the same noise point in the three images are calculated as the reference noise point coordinates for calculating the light deflection displacement field under the condition of no airflow temperature field disturbance.

[0082] During normal operation in the high-temperature work area, a first black-and-white industrial camera 105 captures 10 background schlieren images with airflow temperature field disturbances every 5 minutes, where the airflow temperature ranges from 20℃ to 250℃. Three images are then randomly selected. Based on the reference noise coordinates under the condition of no airflow temperature field disturbance, the light deflection displacement of the noise points in the corresponding background schlieren images under the condition of airflow temperature field disturbance is calculated. This data is then transmitted and stored in computer 107 via a data cable. This data is used to calculate the temperature gradient distribution characteristic values ​​of the work area and for neural network modeling training.

[0083] S2. Extract light deflection displacement features from the background schlieren image, and then input them into the trained neural network model to obtain the gridded temperature field information and temperature gradient distribution information of the heat source monitoring area. The neural network model is trained using the light deflection displacement features and the temperature field training dataset, and the temperature field is calculated based on the light deflection displacement features. Preferably, the step of extracting the light deflection displacement features includes S21 to S23.

[0084] S21. Determine the pixel deflection displacement based on the pixel coordinate difference of the same reference noise point before and after the airflow temperature field disturbance.

[0085] .

[0086] In the formula: For the first The pixel offset displacement vector of the reference noise point. The reference noise number. For the first The pixel offset displacement of each reference noise point in the horizontal direction of the image. For the first The pixel offset displacement of each reference noise point in the vertical direction of the image. For the first Each reference noise point at the sampling time The horizontal pixel coordinates of the image. For the first The horizontal reference pixel coordinates of the image under undisturbed airflow conditions. For the first Each reference noise point at the sampling time The vertical pixel coordinates of the image. For the first The vertical reference pixel coordinates of the reference noise point in the image under undisturbed airflow conditions. This refers to the image sampling time.

[0087] Specifically, the light deflection displacement field under airflow temperature field disturbance is solved based on classical optical transmission theory. Figure 1 In the process, the coordinate positions of all reference noise points in the left-side background schlieren pattern splicing board 104 are marked. Taking any single reference noise point as the research object, denoted as the [number]th [point]... A reference noise point was identified. MATLAB software and its embedded PIVLAB module were installed on computer 107. The cross-correlation algorithm built into the PIVLAB module was used to calculate the pixel deflection displacement field of the reference noise point within a 3×3 pixel window at the same reference noise point coordinate position.

[0088] S22. Convert the pixel deflection displacement into the actual light deflection displacement based on the pixel equivalent.

[0089] .

[0090] In the formula: For the first The actual light deflection displacement vector corresponding to each reference noise point. For the first The actual lateral light deflection displacement corresponding to each reference noise point. For the first The actual longitudinal light deflection displacement corresponding to each reference noise point. In pixel equivalents.

[0091] S23. Calculate the lateral light deflection angle and the longitudinal light deflection angle based on the actual light deflection displacement and the distance between the heat source monitoring area and the background schlieren pattern splicing board.

[0092] .

[0093] .

[0094] In the formula: For the first The lateral light deflection angle corresponding to each reference noise point. For the first The longitudinal light deflection angle corresponding to each reference noise point. This refers to the distance between the on-site heat source monitoring area and the background mosaic pattern splicing board.

[0095] Preferably, the step of inputting the light deflection displacement features into the trained neural network model to obtain the gridded temperature field information of the heat source monitoring area includes S24.

[0096] S24. The extracted light deflection displacement features are divided into grids, averaged, and normalized, and then input into the trained neural network model to output the temperature field feature values ​​of different grid regions. The average temperature feature value of each grid region is calculated based on the temperature field feature values ​​of different grid regions.

[0097] Specifically, using the MATLAB software installed on computer 107, the calculated arrays of background schlieren ray deflection displacements and airflow temperature field characteristic values ​​are used to visually reconstruct the airflow temperature field of the entire working area, such as... Figure 8 As shown. From Figure 8 The analysis region, characterized by large temperature variations (i.e., sensitivity to temperature gradient changes), was selected and divided into 25 rows and 25 columns of gridded areas based on its spatial distribution. In other words, the high-resolution background schlieren image of the sensitive area to airflow temperature field disturbances in the work area was divided into 25×25 gridded regions.

[0098] Using neural network calculation software in PC 107, the measurement error of the light deflection displacement field of the background schlieren pattern in different grid areas of 25×25 groups was corrected by averaging and normalization. The error correction algorithm includes: averaging the displacement field calculation results of the background schlieren pattern in the boundary areas of adjacent grids based on the displacement field amplitude of each background schlieren pattern splicing module; and normalizing the displacement field data in different grid areas to correct the measurement error of the light deflection displacement field caused by splicing plate installation error, pattern splicing error, and random noise.

[0099] Preferably, the step of obtaining temperature gradient distribution information includes S25 to S26.

[0100] S25. Select the target sample feature dataset from the light deflection displacement feature dataset after mean and normalization preprocessing, and combine the target sample feature dataset with the other sample feature datasets to form the complete input feature data of the neural network model. Then input the target sample feature dataset into the trained neural network model to obtain the predicted output of the airflow temperature field feature value dataset, and calculate the expected output.

[0101] Specifically, a partially dependent graph nonlinear analysis method is introduced. From a 25×25 gridded region of transient airflow temperature field sensitive to temperature gradient changes, the background schlieren ray deflection displacement field of any 7 rows and 7 columns of grid area is arbitrarily selected. The temperature field data corresponding to the row and column numbers of the deflection displacement field are replaced with the mean temperature characteristic values ​​of each vertical row and each horizontal column, thus establishing a sample dataset of 7 sets of background schlieren ray deflection displacement field-temperature field, such as... Figure 10 and Figure 11 In , , , , , and As shown, a pre-trained BP neural network is used to model a prediction model, and the expected output of the airflow temperature field feature value dataset is calculated.

[0102] The desired output is characterized by a partial dependency function, which satisfies the following: .

[0103] In the formula: This is a partially dependent function. This is a dataset of 7 randomly selected target sample features. This refers to the feature dataset of all samples other than the target sample feature dataset. For the remaining sample feature dataset The mathematical expectation. This is a trained neural network temperature field prediction model.

[0104] Based on the discontinuous and nonlinear distribution characteristics of the temperature field's characteristic values, a marginal expectation estimation algorithm for the target features is used to calculate the sample estimates of the partial dependency function. Specifically, based on the discontinuous and nonlinear distribution characteristics of the temperature field's characteristic values, the integral function employs a marginal expectation estimation algorithm for the target features.

[0105] .

[0106] In the formula: The sample estimates of the partially dependent functions are stored as a set of temperature feature values ​​with corresponding grid region identifiers. The number of samples in the remaining sample feature dataset. The sample number is the sample number in the feature dataset of the remaining samples. For the remaining sample feature dataset, the first Group sample characteristics.

[0107] S26. Based on the average temperature characteristics of adjacent grid regions, calculate the characteristic values ​​of the transverse temperature gradient, the longitudinal temperature gradient, and the combined temperature gradient. Specifically, based on the calculation results of the integral function, compare them with the temperature characteristic values ​​of adjacent grid regions to finally calculate the characteristic values ​​of the temperature gradient of the temperature field in different grid regions.

[0108] .

[0109] .

[0110] .

[0111] In the formula: For the first Line 1 The characteristic value of the temperature gradient in the horizontal direction of the column grid region. This represents the row number of the grid region. This represents the column number of the grid region. For the first Line 1 The characteristic value of the temperature gradient in the vertical direction of the column grid region. For the first Line 1 The comprehensive temperature gradient characteristic value of the grid region. For the first Line 1 The mean temperature characteristics of the grid region. For the first Line 1 The average temperature characteristics of adjacent grid regions. For the first Line 1 The average temperature characteristics of adjacent grid regions.

[0112] By quantitatively calibrating the relative height of color patches between adjacent temperature layers and the corresponding temperature gradient characteristic values, the quantified values ​​of the transient temperature gradient distribution characteristics of the nonlinear diffusion airflow in the horizontal and vertical directions within the calculated grid region are visualized using the relative height of color patches between adjacent temperature layers. Figure 10 and Figure 11 As shown.

[0113] This step uses a second-level perspective to establish... Coordinate plane. represents the vertical coordinate in space. These are the coordinates in the spatial depth direction.

[0114] Preferably, the temperature field is calculated based on the light deflection displacement characteristics, and a training dataset containing A1 to A4 is formed.

[0115] A1. Calculate the gas refractive index within the heat source monitoring area based on the stated transverse and longitudinal light refraction angles. Specifically, calculate the refractive index field at the high-temperature operating area location corresponding to the reference noise point location using the Poisson equation.

[0116] .

[0117] In the formula: Location within the heat source monitoring area The refractive index of the gas at that location. It is a two-dimensional Laplace operator in space. represents the horizontal coordinate in space. represents the vertical coordinate in space. The refractive index is the environmental refractive index. The effective thickness of the airflow temperature field along the optical path direction. For position The angle of refraction of lateral light rays at that location. For position The longitudinal light refracting angle at that location. This is for partial derivative operations.

[0118] A2. Calculate the gas density based on the gas refractive index. Specifically, calculate the high-temperature gas density field using the Glaston-Dale formula.

[0119] .

[0120] In the formula: For the first The airflow density at the location of the heat source monitoring area corresponding to each reference noise point. For the first The gas refractive index at the location of the heat source monitoring area corresponding to each reference noise point. This is the Glaston-Dale constant.

[0121] A3. Calculate the thermodynamic temperature based on the gas flow density. Specifically, calculate the characteristic value of the high-temperature gas flow temperature at the corresponding location in the working area using the ideal gas law.

[0122] .

[0123] .

[0124] .

[0125] In the formula: For the first The gas pressure at the location of the heat source monitoring area corresponding to each reference noise point. For the first The gas volume at the location of the heat source monitoring area corresponding to each reference noise point. For the first The amount of gaseous substance at the location of the heat source monitoring area corresponding to each reference noise point. is the molar gas constant. For the first The thermodynamic temperature at the location of the heat source monitoring area corresponding to each reference noise point. This is the molar mass of air.

[0126] Under approximately isobaric diffusion conditions of high-temperature airflow, the isobaric state relationship can also be used to approximate the temperature characteristic value of the high-temperature airflow at the corresponding location in the working area: .

[0127] In the formula: The ambient thermodynamic temperature is the temperature at room temperature. This refers to the ambient airflow density under normal indoor temperature conditions.

[0128] A4. Combine the light deflection displacement feature vectors and thermodynamic temperatures corresponding to multiple reference noise points to form a training dataset.

[0129] .

[0130] .

[0131] In the formula: This is the training dataset obtained from background schlieren measurement. For the first The light deflection displacement feature vector corresponding to each reference noise point. This represents the total number of reference noise points used in the calculation.

[0132] Specifically, based on the calculation principles and steps of S21 to S23 and A1 to A3, each is selected one by one. Figure 1 All single-point reference noise points in the left-side background schlieren pattern mosaic panel. Corresponding to the 2592×2048 pixel resolution of the high-definition image, the average data of each 3×3 pixel window is grouped into 864×683 sets of reference noise data.

[0133] Calculate the light deflection displacement corresponding to each reference noise point and its temperature field characteristic value in the high-temperature working area, and store them centrally in computer 107 to form a dataset of light deflection displacement fields of all reference noise points and corresponding transient airflow temperature fields of high-temperature working areas. Preferably, the step of training a neural network model to obtain gridded temperature field information includes B1 to B3.

[0134] B1. Divide the training dataset into sample datasets corresponding to gridded regions according to their spatial distribution.

[0135] B2. The calculation results of the light deflection displacement field of the background schlieren pattern in the boundary region of adjacent grids are averaged, and the light deflection displacement field data in different grid regions are normalized.

[0136] B3. Input the training dataset of the light deflection displacement field and the gridded airflow temperature field after mean and normalization processing into the double hidden layer BP neural network model for training, and obtain the neural network model used to characterize the nonlinear mapping relationship between the light deflection displacement field and the temperature field.

[0137] Specifically, based on the technical approach proposed in this invention—background schlieren, schlieren imaging, and neural network measurement of the average temperature characteristic value at the transient airflow diffusion front end—a dual-hidden-layer neural network optimization model is trained. This dual-hidden-layer neural network includes two hidden layers: the first hidden layer contains 80 neurons, and the second hidden layer contains 40 neurons.

[0138] During training, the robustness and generalization ability of the neural network model are continuously optimized, enabling the model to quickly calculate the transient airflow temperature field feature value dataset of the corresponding high-temperature working area within 30 to 45 seconds based on the reference noise light deflection displacement field data of all background schlieren patterns at different times.

[0139] The single-layer neural network output variable of the double-hidden-layer BP neural network model satisfies: .

[0140] .

[0141] In the formula: This represents the layer number of the neural network. This is the neuron number in the current layer. This refers to the sequence number of the neuron in the previous layer.

[0142] For the first Layer Weighted inputs to each neuron. For the first The number of neurons in a layer. For the first Layer The first neuron to the second Layer The connection weights of each neuron. For the first Layer The output value of each neuron. For the first Layer Threshold bias of each neuron. For the first Layer The output value of each neuron. For the first The node activation function of the layer.

[0143] For a given set of light deflection displacement feature vectors and temperature field feature value sample pairs, the output layer weight adjustment is expressed as follows: .

[0144] In the formula: Connection weights in the output layer The adjustment amount. This is the layer number of the output layer of the neural network. This is the learning rate. For the output layer Error term of each neuron. The first layer before the output layer The output value of each neuron.

[0145] The weights and threshold biases of each neuron in a hidden layer are adjusted according to the following formula: .

[0146] .

[0147] .

[0148] In the formula: For the first The updated connection weights of the layer. For the first Connection weights before layer update. For the first Adjustment amount of layer connection weights. For the first Layer The threshold bias after each neuron is updated. For the first Layer Threshold bias before each neuron is updated. For the first Layer The adjustment amount of the threshold bias of each neuron. For the first Layer Error term of each neuron.

[0149] The error function uses the squared error function: .

[0150] In the formula: This is the training error function for the neural network. This represents the number of neurons in the output layer. For the output layer A network of neurons predicts temperature values. For the output layer The target temperature value corresponding to each neuron.

[0151] The output layer error term is calculated according to the following formula: .

[0152] In the formula: For the output layer Error term of each neuron. It is the first derivative of the output layer activation function. For the output layer Weighted inputs to each neuron.

[0153] For any hidden layer, the adjustment of its connection weights is expressed as: .

[0154] In the formula: For the first Adjustment amount of layer connection weights. For the first Layer Error term of each neuron. For the first Layer The output value of each neuron.

[0155] The hidden layer error term is expressed according to the chain rule as follows: .

[0156] In the formula: This represents the partial derivative operation.

[0157] The first term in the above formula can be expressed as: .

[0158] In the formula: This refers to the sequence number of the neurons in the next layer. For the first Number of neurons in a layer. For the first Layer Error term of each neuron. For the first Layer The first neuron to the second Layer The connection weights of each neuron.

[0159] when At that time, the partial derivative of the output layer error function with respect to the output value is: .

[0160] In the formula: For the output layer The output value of each neuron.

[0161] When the output layer directly outputs the predicted temperature value and Indicates the same output quantity.

[0162] The second term in the above formula is the derivative of the excitation function: .

[0163] In the formula: For the first The first derivative of the layer activation function.

[0164] From this, we can obtain the hidden layer error term: .

[0165] Therefore, the adjustment amount of the neuron connection weights in a single hidden layer is: .

[0166] Based on the variables such as connection weights, threshold biases, and error terms derived above, a nonlinear neural network mapping model between the background schlieren pattern light deflection displacement field and the working area temperature field can be established.

[0167] By using the input-output link of the neural network model and taking the light deflection displacement field array of the background schlieren pattern as the input variable of the neural network, the output variable of the neural network, namely the temperature field characteristic value of different high-temperature grid regions, can be quickly calculated. Then, the temperature field characteristic value and its mean of different grid regions can be calculated.

[0168] By comparing with the reference temperature color scale, the transient airflow temperature field and its temperature gradient characteristics of the entire operating area were visually reconstructed, and the results are as follows: Figure 8 As shown.

[0169] S3. Extract the position coordinate information of the transient airflow diffusion front end from the airflow diffusion schlieren image, and map the position coordinate information to the gridded temperature field information to determine the grid set where the transient airflow diffusion front end is located.

[0170] To accurately visualize the location characteristics of the dynamic diffusion front of the transient airflow temperature field in a large field-of-view high-temperature operating area, Figure 2 Within the indoor heat source monitoring area 109 shown, a second working platform for schlieren large-field imaging measurement is constructed along a second horizontal viewing angle. This second horizontal viewing angle is perpendicular to the first horizontal viewing angle. Based on the working principle of the concave spherical mirror 115 imaging monitoring of the nonlinear diffusion characteristics of airflow and the position information of the diffusion front end, the field of view for visualization schlieren imaging measurement can be further expanded by adding a Z-shaped folded optical path system to the structural device.

[0171] Using the same principle as step S1, the second monochrome industrial camera 112 and the concave spherical mirror 115 are calibrated for stability adjustment, center height coaxiality adjustment, focusing, and optical parameter settings. During normal operation in the high-temperature working area, the second monochrome industrial camera 112 maintains a synchronized acquisition frequency with the first monochrome industrial camera 105, that is, capturing 10 transient airflow dynamic diffusion schlieren images every 5 minutes, after reflection by the concave spherical mirror 115 and refraction by the airflow temperature field disturbance. Subsequently, an optical flow algorithm is used to mesh the acquisition area according to 25 rows and 25 columns, and the schlieren images are then visually reconstructed, such as... Figure 9 As shown.

[0172] Preferably, S3 specifically includes steps S31 to S33.

[0173] S31. Perform optical flow analysis on the airflow diffusion schlieren images at adjacent sampling times to determine the pixel displacement of the pixels in the schlieren image.

[0174] The optical flow analysis employs a constant brightness constraint: .

[0175] In the formula: Sampling time In the location of the schlieren image The grayscale value at that location. represents the vertical coordinate in space. These are the coordinates in the spatial depth direction. This represents the vertical pixel displacement of a pixel in a schlieren image. This represents the pixel displacement of a pixel in a schlieren image along the depth direction. This represents the time interval between two adjacent image samples.

[0176] Applying a first-order approximation to the above constant brightness constraint, we can obtain the pixel motion constraint equation: .

[0177] In the formula: grayscale field The first-order partial derivative with respect to the vertical axis. grayscale field The first partial derivative with respect to the coordinates in the depth direction. grayscale field The first-order partial derivative with respect to the sampling time. This represents the pixel's vertical movement speed. This represents the pixel movement speed along the depth direction.

[0178] The pixel displacement between adjacent sampling times is represented as: .

[0179] .

[0180] In the formula: This refers to the pixel number of the schlieren image. For the first The vertical pixel displacement of each pixel. For the first The pixel displacement along the depth direction. For the first The speed at which each pixel moves along the vertical direction. For the first The pixel movement speed along the depth direction of each pixel.

[0181] S32. Determine the airflow diffusion displacement based on the conversion relationship between pixel displacement and actual airflow diffusion displacement. Specifically, by introducing a conversion coefficient between pixel motion and airflow diffusion motion through calibration experiments, the actual diffusion displacement of the airflow temperature field location point in the longitudinal direction and the actual diffusion displacement of the airflow temperature field location point in the depth direction can be solved.

[0182] .

[0183] .

[0184] In the formula: For the first Each pixel corresponds to the actual longitudinal diffusion displacement of the airflow. This refers to the pixel number of the schlieren image. This is the conversion coefficient between the longitudinal pixel displacement and the actual airflow diffusion displacement. For the first The vertical pixel displacement of each pixel. For the first Each pixel corresponds to the actual diffusion displacement of the airflow in the depth direction. This is the conversion coefficient between pixel displacement in the depth direction and actual airflow diffusion displacement. For the first The pixel displacement along the depth direction.

[0185] Further calculations were performed on the average longitudinal diffusion rate of the front region of the diffusion airflow: .

[0186] In the formula: It represents the average diffusion rate in the longitudinal direction of the leading edge region of the diffused airflow. This represents the number of pixels involved in the calculation within the front region of the diffused airflow. It is the set of pixels in the front region of the diffused airflow.

[0187] The average diffusion rate in the depth direction of the front region of the diffusion flow is: .

[0188] In the formula: This represents the average diffusion rate of the front region of the diffused airflow in the depth direction.

[0189] S33. Determine the position coordinates of the transient airflow diffusion front end based on the airflow diffusion displacement, and mark the gridded position where the transient airflow diffusion front end is located.

[0190] The position coordinates of the front end of the diffused airflow in the longitudinal and depth directions are as follows: .

[0191] .

[0192] In the formula: The coordinates of the longitudinal position of the front end of the diffused airflow. These are the initial longitudinal spatial coordinates of the pixel at the front end of the diffused airflow. It is the set of pixels in the front region of the diffused airflow. This indicates taking the maximum value in the set. The coordinates of the front end of the diffused airflow in the depth direction. The initial spatial coordinates in the depth direction corresponding to the pixel at the front end of the diffused airflow.

[0193] The above calculations can obtain the overall displacement flow field of the diffusion airflow, the average diffusion rate of the diffusion front end, and the coordinate information of the location point of the diffusion front end, and mark the gridded location of the diffusion front end.

[0194] S4. Determine the average temperature characteristic value of the transient airflow diffusion front end based on the temperature information in the grid set where the transient airflow diffusion front end is located. Preferably, S4 specifically includes S41 to S42.

[0195] S41. Determine the grid set where the diffusion front end is located based on the location coordinate information, and extract the average temperature characteristics of each grid region in the grid set where the diffusion front end is located.

[0196] S42. Calculate the average temperature characteristic value of the transient airflow diffusion front end based on the average temperature characteristic value of each grid region.

[0197] .

[0198] In the formula: This represents the average temperature characteristic value of the transient airflow diffusion front end. This represents the number of grids in the grid set containing the diffusion front end. This refers to the set of grids containing the diffusion front end. For the first Line 1 The mean temperature characteristics of the grid region.

[0199] By comprehensively utilizing the average temperature characteristic values ​​of different grid regions of the transient airflow temperature field in the high-temperature operation area obtained by S2 calculation, and the grid position information of the transient airflow diffusion front extracted in step S3, the average temperature characteristic values ​​of the transient airflow temperature field diffusion front and its diffusion position coordinate information of different grid regions are identified in computer 107.

[0200] The mean value of the diffusion front-end temperature characteristic obtained from this and their corresponding position coordinates It serves as a high-temperature early warning and monitoring indicator for adaptive linkage control of the bottom-level air supply unit and the top-level exhaust unit.

[0201] S5. Compare the average temperature characteristic value of the transient airflow diffusion front end with the preset temperature warning threshold, and generate supply and exhaust ventilation control instructions and alarm instructions based on the number of personnel working in the heat source monitoring area. Preferably, the first temperature warning threshold is set to 31℃, the second temperature warning threshold is set to 33℃, the third temperature warning threshold is set to 35℃, and the preset stop threshold is set to 26℃.

[0202] S5 specifically includes S51 to S53.

[0203] S51. When the average temperature characteristic value of the transient airflow diffusion front end exceeds 31°C but does not reach 33°C, if the number of operators is less than 10, a first speed gear ventilation operation command is generated; if the number of operators exceeds 10, a second speed gear ventilation operation command is generated.

[0204] S52. When the average temperature characteristic value of the transient airflow diffusion front end exceeds 33°C but does not reach 35°C, if the number of operators is less than 10, a second speed gear ventilation operation command is generated; if the number of operators exceeds 10, a third speed gear ventilation operation command is generated.

[0205] S53. When the average temperature characteristic value of the transient airflow diffusion front end exceeds 35°C, a third speed gear ventilation operation command is generated.

[0206] This invention is applicable to actual working conditions based on the average temperature characteristic value of the nonlinear diffusion front end of the airflow temperature field in different directions in indoor high-temperature work areas. It adaptively links and controls the indoor air supply and exhaust devices of buildings to achieve the goal of low-energy consumption, precise control, and improved indoor working environment quality.

[0207] Specifically, the temperature warning thresholds are pre-set to 35℃, 33℃, and 31℃ in the computer's built-in temperature field neural network calculation software and high temperature early warning monitoring platform (taking medium labor intensity in high temperature operations as an example), corresponding to three control levels: high, medium, and low (GBZ / T 229.3-2025).

[0208] The computer monitors and extracts the average transient high-temperature characteristic value and its coordinate location information of the diffusion front of the airflow temperature field in different grid high-temperature operation areas in real time, according to steps S2 to S4.

[0209] When the average value of the transient airflow high temperature characteristic exceeds the above-mentioned temperature warning threshold, the number of people working is determined by the infrared camera 401 installed indoors. The start / stop information is issued to the communication network signal receiving devices of the normally closed ventilation gate 102, the bottom air supply unit 110, and the top exhaust unit 101 through the 4G / 5G communication device of the single-chip microcomputer control platform 108 connected to the computer 107. According to the warning response level, the corresponding control unit spindle speed control level is adaptively started online, and a negative pressure air suction operating environment for air supply and exhaust is quickly formed in the vertical air supply flue, so as to realize the rapid diversion and discharge of high temperature airflow from the room.

[0210] At the same time, the infrared camera captures thermal images of the workers in the video recording. Using AI-assisted facial recognition technology, it determines how many workers are currently in the factory environment and activates a buzzer to sound an alarm according to the following rules: When the temperature in the working environment exceeds 31℃ but does not reach 33℃, if the number of workers is less than 10, the microcontroller on the floor where the high-temperature environment is located will send a low-level ventilation operation command. If the number of workers exceeds 10, a medium-level ventilation operation command will be activated.

[0211] When the temperature in the working environment exceeds 33℃ but does not reach the maximum limit of 35℃, if the number of workers is less than 10, the microcontroller sends a medium-range ventilation operation command; if the number of workers exceeds 10, a high-range ventilation operation command is activated.

[0212] When the temperature in the working environment exceeds 35℃, regardless of the number of people working, the microcontroller will prioritize sending high-level ventilation operation commands.

[0213] S6. Control the normally closed ventilation damper, the bottom air supply unit and the top exhaust unit to operate in conjunction according to the air supply and exhaust control command, so that the indoor airflow in the heat source monitoring area is discharged through the vertical ventilation flue, and the air supply and exhaust operation is stopped when the average temperature characteristic value of the transient airflow diffusion front end is lower than the preset stop threshold.

[0214] Preferably, S6 specifically includes S61 to S62.

[0215] S61. During the start-up air supply and exhaust phase, continue to collect background schlieren images under the disturbance state of the airflow temperature field, and continuously calculate the average value of the transient airflow diffusion front temperature characteristic value and the location coordinate information.

[0216] S62. When the average temperature characteristic value of the transient airflow diffusion front end is lower than 26°C, control commands are generated to close the normally closed ventilation gate, stop the bottom air supply unit, and stop the top exhaust unit.

[0217] Specifically, during the initial air supply and exhaust phase, the first black-and-white industrial camera 105 continues to capture 10 background schlieren patterns with high-temperature airflow temperature field disturbances every minute.

[0218] Combining steps S2 to S4, the temperature characteristic value and its mean value of the airflow diffusion front end of the temperature field are calculated in real time, as well as the location information of the airflow diffusion front end is calculated in real time. The obtained results of the location and mean value of the temperature field of the airflow diffusion front end are fed back to the computer temperature measurement and high temperature early warning platform, and compared with the high temperature early warning threshold. The comparison results are used to adaptively control the indoor air supply and exhaust device.

[0219] When the average temperature at the indoor airflow diffusion front drops below the minimum warning threshold of 26°C, a stop message is issued by the communication network signal receiving device connected to the computer, controlling the normally closed ventilation damper to close, and the air supply and exhaust units to stop operating and return to normal working status.

[0220] This invention utilizes background schlieren, schlieren imaging technology, and neural network methods to achieve quantitative, visual, and large-field-of-view rapid and accurate measurement of the transient airflow temperature gradient distribution characteristics in high-temperature work areas indoors of buildings, and to calibrate and monitor nonlinear airflow diffusion characteristics. Furthermore, based on the actual working conditions of high-temperature early warning at the airflow diffusion front end, it adaptively and in conjunction with the building's indoor air supply and exhaust devices to improve the engineering practicality and reliability of precisely controlling the diversion and exhaust of high-temperature airflow from the work area.

[0221] This embodiment introduces a working platform and neural network method for measuring the light deflection displacement field under transient airflow temperature field disturbance in the background schlieren measurement state, so as to realize the rapid and accurate measurement of the transient airflow temperature gradient distribution characteristics in a large field of view in indoor high-temperature work areas.

[0222] Using experimental measurement data of transient airflow temperature field, a neural network model with good robustness, strong generalization ability and high prediction accuracy is trained. Combined with the partial dependency graph analysis method, a mapping relationship model between the large field of view light deflection displacement field and the airflow temperature characteristics of the working area is established, and the characteristic value of temperature gradient distribution is calculated.

[0223] The control method in this embodiment has high accuracy in calculating the characteristic value of transient airflow temperature gradient distribution with a large field of view, and its calculation efficiency can be 3 times higher than that of classical optical or acoustic temperature measurement methods. It better meets the practical needs of engineering applications for real-time measurement of airflow temperature gradient with a large field of view and dynamic monitoring of airflow diffusion high temperature early warning in indoor high-temperature work areas.

[0224] This embodiment introduces a schlieren imaging monitoring platform to monitor airflow diffusion characteristics, enabling visualized monitoring of transient airflow nonlinear diffusion characteristics and quantifying the temperature gradient distribution characteristics at the diffusion front. By utilizing the quantified temperature gradient distribution characteristics of the temperature field and the coordinate information of the schlieren imaging position at the airflow diffusion front, the temperature characteristic values ​​and their average values ​​of the transient airflow temperature field diffusion front are quantitatively calibrated, serving as important monitoring indicators for the high-temperature warning threshold of the adaptive linkage control indoor air supply and exhaust system.

[0225] This embodiment introduces a negative pressure indoor air supply and exhaust system that uses pressurized air supply at the bottom of the building and exhaust air diversion at the top. Combined with a high-temperature early warning response signal based on the average temperature characteristic value of the airflow diffusion front end, the indoor air supply and exhaust system is adaptively and remotely controlled. This achieves real-time diversion and exhaust of high-temperature airflow from the room with low energy consumption, improving the quality of the working environment and ensuring the health and safety of workers.

[0226] This embodiment introduces an algorithm to correct the measurement error of the light deflection displacement field of the background schlieren pattern. Based on the measured amplitude of the displacement field of each background schlieren pattern mosaic panel, the calculation results of the light deflection displacement field of the background pattern in different grid areas are averaged and normalized to correct the noise error in the displacement field measurement of different background schlieren pattern mosaic panels.

Claims

1. A method for measuring temperature gradients based on schlieren and neural networks, characterized in that, Include: S1. Background schlieren images and airflow diffusion schlieren images are acquired by using the first background schlieren temperature measurement platform and the second schlieren diffusion imaging platform arranged in the heat source monitoring area. S2. Extract the light deflection displacement features from the background schlieren image, input them into the trained BP neural network model, and calculate the gridded temperature field information and temperature gradient distribution information of the heat source monitoring area; wherein, the neural network model is trained using the light deflection displacement features and the temperature field training dataset, and the temperature field and temperature gradient are calculated using the light deflection displacement features and the neural network model. S3. Extract the position coordinate information of the transient airflow diffusion front end from the airflow diffusion schlieren image, and map the position coordinate information to the gridded temperature field information to determine the grid set where the transient airflow diffusion front end is located; S4. Determine the mean temperature characteristic value of the transient airflow diffusion front end based on the temperature information and temperature gradient information in the grid set; S5. Compare the average temperature characteristic value of the transient airflow diffusion front end with the preset temperature warning threshold, and generate supply and exhaust ventilation control instructions and alarm instructions in combination with the number of workers in the heat source monitoring area; S6. Control the normally closed ventilation damper, the bottom air supply unit and the top exhaust unit to operate in conjunction according to the air supply and exhaust control command, so that the high-temperature indoor airflow in the heat source monitoring area is discharged through the vertical air supply flue, and the air supply and exhaust operation is stopped when the average temperature characteristic value at the front end of the transient airflow diffusion is lower than the preset stop threshold.

2. The method for measuring temperature gradients based on schlieren and neural networks according to claim 1, characterized in that, S1 includes: The first background schlieren temperature measurement platform includes a background schlieren pattern splicing board, a first black and white industrial camera and an LED light source. The first black and white industrial camera is coaxially arranged with the center position of the background schlieren pattern splicing board. The LED light source is a stable, controllable and uniformly illuminated dot matrix distributed LED light source. The second schlieren diffusion imaging platform includes a second monochrome industrial camera, a concave spherical mirror, a knife edge, and a white LED point light source. The optical axis of the second monochrome industrial camera is coaxially arranged with the principal optical axis of the concave spherical mirror. The knife edge and the white LED point light source are arranged in parallel, and the knife edge is located at the radius of curvature of the coaxial optical path of the concave spherical mirror. The imaging direction of the first background schlieren thermometry platform is perpendicular to the imaging direction of the second schlieren diffusion imaging platform. In the absence of airflow and temperature field disturbance, the first black and white industrial camera is used to acquire static background schlieren images, and the reference noise coordinates are determined based on multiple static background schlieren images. Under the condition of airflow temperature field disturbance, the first black and white industrial camera is used to acquire background schlieren images, and the second black and white industrial camera is used to simultaneously acquire transient airflow dynamic diffusion schlieren images after reflection by a concave spherical mirror and refraction by airflow temperature field disturbance.

3. The method for measuring temperature gradients based on schlieren and neural networks according to claim 1, characterized in that, The steps for extracting the light deflection displacement feature in S2 include: The pixel deflection displacement is determined based on the pixel coordinate difference of the same reference noise point before and after the airflow temperature field disturbance. ； In the formula: For the first The pixel offset displacement vector of each reference noise point; The reference noise number; For the first The pixel offset displacement of each reference noise point in the horizontal direction of the image; For the first The pixel offset displacement of each reference noise point in the vertical direction of the image; For the first Each reference noise point at the sampling time The horizontal pixel coordinates of the image; For the first The horizontal reference pixel coordinates of the image under undisturbed airflow conditions; For the first Each reference noise point at the sampling time The vertical pixel coordinates of the image; For the first The vertical reference pixel coordinates of the image under undisturbed airflow conditions; The image sampling time; The pixel deflection displacement is converted into the actual light deflection displacement based on the pixel equivalent. ； In the formula: For the first The actual light deflection displacement vector corresponding to each reference noise point; For the first The actual lateral light deflection displacement corresponding to each reference noise point; For the first The actual longitudinal light deflection displacement corresponding to each reference noise point; In pixel equivalents; Based on the actual light deflection displacement and the distance between the heat source monitoring area and the background schlieren pattern splicing board, calculate the lateral light deflection angle and the longitudinal light deflection angle. ； ； In the formula: For the first The lateral light deflection angle corresponding to each reference noise point; For the first The longitudinal light deflection angle corresponding to each reference noise point; This refers to the distance between the on-site heat source monitoring area and the background mosaic pattern splicing board.

4. The method for measuring temperature gradients based on schlieren and neural networks according to claim 1, characterized in that, In S2, the steps of inputting the light deflection displacement features into the trained BP neural network model to obtain the gridded temperature field information of the heat source monitoring area include: The extracted light deflection displacement features are divided into grids, averaged, and normalized, and then input into the trained BP neural network model to output temperature field feature values ​​of different grid regions. The mean temperature feature value of each grid region is calculated based on the temperature field feature values ​​of different grid regions. The steps in S2 for training the neural network model used to obtain gridded temperature field information include: The training dataset is divided into sample datasets corresponding to gridded regions according to their spatial distribution. The calculation results of the light deflection displacement field of the background schlieren pattern in the boundary region of adjacent grids are averaged, and the light deflection displacement field data in different grid regions are normalized. The training dataset of the light deflection displacement field and the gridded airflow temperature field after mean and normalization processing is input into the double hidden layer BP neural network model for training, and a neural network model is obtained to characterize the nonlinear mapping relationship between the light deflection displacement field and the temperature field. The single-layer neural network output variable of the double-hidden-layer BP neural network model satisfies: ； ； In the formula: This refers to the layer number of the neural network. The neuron number of the current layer; This refers to the sequence number of the neuron in the previous layer; For the first Layer Weighted inputs to each neuron; For the first The number of neurons in a layer; For the first Layer The first neuron to the second Layer The connection weights of each neuron; For the first Layer The output value of each neuron; For the first Layer Threshold bias of each neuron; For the first Layer The output value of each neuron; For the first The node activation function of the layer.

5. The method for measuring temperature gradients based on schlieren and neural networks according to claim 2, characterized in that, The steps to obtain temperature gradient distribution information in S2 include: The target sample feature dataset is selected from the light deflection displacement feature dataset after mean and normalization preprocessing. The target sample feature dataset and the remaining sample feature datasets are used together to form the complete input feature data of the neural network model. Then, the target sample feature dataset is input into the trained neural network model to obtain the predicted output of the airflow temperature field feature value dataset and calculate the expected output. The desired output is characterized by a partial dependency function, which satisfies the following: ； In the formula: For partially dependent functions; This is a dataset of 7 randomly selected target sample features; This refers to the feature dataset of all samples other than the target sample feature dataset. For the remaining sample feature dataset The mathematical expectation; For the trained neural network temperature field prediction model; Based on the discontinuous and nonlinear distribution characteristics of the temperature field features, the marginal expectation estimation algorithm of the target features is used to calculate the sample estimate of the partial dependency function; ； In the formula: The sample estimates of the partially dependent functions are stored as a set of temperature feature values ​​with corresponding grid region identifiers; The number of samples in the remaining sample feature dataset; The sample number in the remaining sample feature dataset; For the remaining sample feature dataset, the first Group sample characteristics; Based on the average temperature characteristics of adjacent grid regions, calculate the characteristic values ​​of the horizontal temperature gradient, the characteristic values ​​of the vertical temperature gradient, and the characteristic value of the combined temperature gradient. ； ； ； In the formula: For the first Line 1 The characteristic value of the temperature gradient in the horizontal direction of the column grid region; The row number of the grid region; The column number of the grid region; For the first Line 1 The characteristic value of the temperature gradient in the vertical direction of the column grid region; For the first Line 1 The comprehensive temperature gradient characteristic value of the column grid region; For the first Line 1 Mean temperature characteristics of the grid region; For the first Line 1 The average temperature characteristics of adjacent grid regions in a column; For the first Line 1 The average temperature characteristics of adjacent grid regions.

6. The method for measuring temperature gradients based on schlieren and neural networks according to claim 3, characterized in that, S2 calculates the temperature field based on the light deflection displacement feature and forms a training dataset, which includes: Calculate the gas refractive index within the heat source monitoring area based on the lateral light deflection angle and the longitudinal light deflection angle; ； In the formula: Location within the heat source monitoring area The gas refractive index at that location; For a two-dimensional Laplace operator in space; The horizontal coordinate in space; The vertical coordinate in space; For environmental refractive index; The effective thickness of the airflow temperature field along the optical path; For position The angle of refraction of lateral light rays at that location; For position The longitudinal light refracting angle at that location; This is a partial derivative operation; Calculate the gas flow density based on the gas refractive index; ； In the formula: For the first Airflow density at the location of the heat source monitoring area corresponding to each reference noise point; For the first The gas refractive index at the location of the heat source monitoring area corresponding to each reference noise point; It is the Glaston-Dale constant; Calculate the thermodynamic temperature based on the airflow density; ； ； ； In the formula: For the first The gas pressure at the location of the heat source monitoring area corresponding to each reference noise point; For the first The gas volume at the location of the heat source monitoring area corresponding to each benchmark noise point; For the first The amount of gaseous substance at the location of the heat source monitoring area corresponding to each reference noise point; The molar gas constant; For the first Thermodynamic temperature at the location of the heat source monitoring area corresponding to each reference noise point; The molar mass of air; The light deflection displacement feature vectors and thermodynamic temperatures corresponding to multiple benchmark noise points are combined to form a training dataset. ； ； In the formula: The training dataset is obtained from background schlieren measurement; For the first The light deflection displacement feature vector corresponding to each reference noise point; This represents the total number of reference noise points used in the calculation.

7. The method for measuring temperature gradients based on schlieren and neural networks according to claim 1, characterized in that, S3 includes: Optical flow analysis is performed on the airflow diffusion schlieren images at adjacent sampling times to determine the pixel displacement of pixels in the schlieren images; The optical flow analysis employs a constant brightness constraint: ； In the formula: Sampling time In the location of the schlieren image The grayscale value at that location; The vertical coordinate in space; The coordinates are in the spatial depth direction; This represents the vertical pixel displacement of a pixel in a schlieren image. This represents the pixel displacement of a pixel in a schlieren image along the depth direction. The time interval between two adjacent image samples; The airflow diffusion displacement is determined based on the conversion relationship between pixel displacement and actual airflow diffusion displacement; ； ； In the formula: For the first Each pixel corresponds to the actual longitudinal diffusion displacement of the airflow; The pixel number of the schlieren image; This is the conversion coefficient between the vertical pixel displacement and the actual airflow diffusion displacement; For the first The pixel displacement along the vertical direction; For the first Each pixel corresponds to the actual diffusion displacement of the airflow in the depth direction; This is the conversion coefficient between pixel displacement in the depth direction and actual airflow diffusion displacement; For the first The pixel displacement of each pixel along the depth direction; The position coordinates of the transient airflow diffusion front are determined based on the airflow diffusion displacement, and the gridded position of the transient airflow diffusion front is marked. The location coordinate information satisfies: ； ； In the formula: The coordinates of the longitudinal position of the front end of the diffused airflow; The initial vertical spatial coordinates of the pixel at the front end of the diffused airflow; It is the set of pixels in the leading region of the diffused airflow; This indicates retrieving the maximum value from the set. The coordinates of the leading edge of the diffused airflow in the depth direction; The initial spatial coordinates in the depth direction corresponding to the pixel at the front end of the diffused airflow.

8. The method for measuring temperature gradients based on schlieren and neural networks according to any one of claims 1 to 7, characterized in that, in, High temperature in indoor airflow refers to a temperature greater than or equal to 35°C. S4 specifically includes: The grid set where the diffusion front end is located is determined based on the location coordinate information, and the average temperature characteristics of each grid region in the grid set where the diffusion front end is located are extracted. The average temperature characteristic value of the transient airflow diffusion front end is calculated based on the average temperature characteristic value of each grid region; ； In the formula: This represents the average temperature characteristic value at the transient airflow diffusion front end. This refers to the number of grid cells in the grid set containing the diffusion front end. The set of grids containing the diffusion front end; For the first Line 1 The mean temperature characteristics of the grid region.

9. The method for measuring temperature gradients based on schlieren and neural networks according to any one of claims 1 to 7, characterized in that, S5 specifically includes: Set the first temperature warning threshold to 31℃, the second temperature warning threshold to 33℃, the third temperature warning threshold to 35℃, and the preset stop threshold to 26℃. When the average temperature characteristic value of the transient airflow diffusion front end exceeds 31°C but does not reach 33°C, if the number of operators is less than 10, a first speed gear ventilation operation command is generated; if the number of operators exceeds 10, a second speed gear ventilation operation command is generated. When the average temperature characteristic value of the transient airflow diffusion front end exceeds 33°C but does not reach 35°C, if the number of operators is less than 10, a second speed gear ventilation operation command is generated; if the number of operators exceeds 10, a third speed gear ventilation operation command is generated. When the average temperature characteristic value of the transient airflow diffusion front end exceeds 35°C, a third speed gear ventilation operation command is generated. S6 specifically includes: During the initial air supply and exhaust phase, background schlieren images under the disturbance state of the airflow temperature field are continuously acquired, and the mean value of the transient airflow diffusion front temperature feature value and the location coordinate information are continuously calculated. When the average temperature characteristic value of the transient airflow diffusion front end is below 26°C, control commands are generated to close the normally closed ventilation gate, stop the bottom air supply unit, and stop the top exhaust unit.

10. A supply and exhaust ventilation control system based on schlieren and neural network for measuring temperature gradients, characterized in that, Used to perform the method for measuring temperature gradients based on schlieren and neural networks as described in any one of claims 1 to 9; The control system includes: The first background schlieren temperature measurement platform is used to acquire background schlieren images before and after airflow temperature field disturbance; The second schlieren diffusion imaging platform is used to acquire airflow diffusion schlieren images; The computing platform is used to generate gridded temperature field information and temperature gradient distribution information based on the background schlieren image, extract the position coordinate information of the transient airflow diffusion front end based on the airflow diffusion schlieren image, and determine the mean value of the temperature feature value of the transient airflow diffusion front end. Personnel detection device, used to detect the number of workers in the heat source monitoring area; The control platform is used to generate air supply and exhaust control commands and alarm commands based on the average temperature characteristic value of the transient airflow diffusion front end and the number of operators; A supply and exhaust ventilation actuator is used to exhaust indoor airflow through a vertical supply and exhaust duct in response to the supply and exhaust ventilation control command.

Images