A visual navigation system and operation method of a heat exchanger automatic cleaning device

Abstract

The present invention discloses a visual navigation system for an automatic cleaning device for a heat exchanger, including an industrial camera, a PC, and a communication interface. The industrial camera is connected to the network port of the PC through a Gige interface, and the PC is connected to the spray gun actuator of the automatic cleaning device through a USB interface and a USB to RS-485 interface. The operation method of the navigation system includes the following steps: the industrial camera obtains the front image of the heat exchanger tube bundle; image distortion correction and image enhancement; the improved YOLOv3 target detection algorithm is used to detect the pipe mouth and obtain the pixel coordinates of the center of the pipe mouth; the coordinate conversion formula is used to convert the pixel coordinates of the center of the pipe mouth into actual coordinates; the orthogonal path method is used to plan the cleaning operation path; and the cleaning device automatically performs the cleaning operation according to the planned path. The present invention can obtain the position of the heat exchanger pipe mouth in real time and accurately, with high efficiency, can effectively reduce human labor, and realize the fully automated cleaning operation of the heat exchanger.

Description

Technical Field

[0001] The invention relates to the field of machine vision, and in particular to a visual navigation system of an automatic cleaning device for a heat exchanger and an operating method thereof. Background Art

[0002] A heat exchanger is a heat exchange device that transfers part of the heat of a high-temperature fluid to a low-temperature fluid. As an important process equipment, it plays an increasingly important role in industrial fields such as petroleum, chemical industry, atomic energy, and building materials. The heat transfer surface inside the heat exchanger will form sediments during use, and the dirt will block the channel cross section, resulting in a decrease in heat transfer efficiency, and even potential safety hazards such as explosions caused by corrosion of equipment. Therefore, relevant management departments in my country require regular cleaning and plugging of heat exchanger equipment. At present, the mainstream heat exchanger cleaning method is high-pressure water jet cleaning technology, which is low-cost and pollution-free. However, in actual applications, workers are mostly required to use high-pressure water guns to forcefully flush the dirt on the heat exchanger tube bundle, which has low cleaning efficiency, high labor intensity and high risk. However, the existing heat exchanger cleaning devices are mostly semi-automatic structures, which can only achieve automatic flushing. The identification and positioning of the heat exchanger nozzles still require operators to operate nearby. The overall operating efficiency is restricted by manpower, which is easy to cause missed cleaning, and does not fundamentally solve the problem of personnel safety.

[0003] With the advent of the Industrial 4.0 era, machine vision technology has been continuously updated and iterated, and has been widely used in the guidance, detection, measurement and identification links of the intelligent manufacturing process, making machine intelligence a reality. Therefore, the target detection technology based on machine vision can be applied to the identification and positioning of the heat exchanger nozzle, significantly improving the accuracy and efficiency, and realizing the full automation of the heat exchanger cleaning device. Summary of the invention

[0004] In order to solve the problems described in the background technology, the present invention provides a visual navigation system for an automatic cleaning device of a heat exchanger and an operating method thereof. First, the front image of the heat exchanger tube bundle captured by a camera is preprocessed, and then the pipe opening in the image is identified. Thereafter, the actual position of the pipe opening is calculated through a coordinate conversion formula, and finally, a cleaning path is planned and the planned path is sent to the spray gun actuator of the automatic cleaning device of the heat exchanger.

[0005] In order to achieve the above object, the present invention provides the following technical solutions:

[0006] A visual navigation system for an automatic cleaning device for a heat exchanger comprises: an industrial camera, a PC, and a communication interface, wherein the industrial camera is connected to a network port of the PC via a Gige interface, and the PC is connected to a spray gun actuator of the automatic cleaning device via a USB interface and a USB to RS-485 interface.

[0007] A method for operating a visual navigation system of an automatic cleaning device for a heat exchanger comprises the following steps:

[0008] Step 1) using an industrial camera to obtain a front image of the heat exchanger tube bundle;

[0009] Step 2) preprocessing the acquired frontal image, including distortion correction and image enhancement;

[0010] Step 3) Use the improved YOLOv3 target detection algorithm to perform nozzle detection;

[0011] Step 4) according to the camera imaging principle, the coordinate conversion formula is used to convert the pixel coordinates of the center of the nozzle into actual coordinates;

[0012] Step 5) planning the cleaning operation path;

[0013] Step 6) Send the planned path to the execution agency for subsequent operations.

[0014] As a preferred embodiment, the industrial camera in step 1) is a CCD camera or a CMOS camera, and the length and width of the pixels are changed to equal sizes when acquiring the image.

[0015] As a preferred embodiment, the camera intrinsic parameters used in the image distortion correction and coordinate transformation process in step 2) and step 4) are obtained through camera calibration.

[0016] As a preferred embodiment, the nozzle identification in step 3) uses a target detection algorithm based on YOLOv3, and is improved according to the specific conditions of the heat exchanger cleaning operation.

[0017] As a preferred embodiment, the improvement is specifically as follows: the backbone network DarkNet-53 of the YOLOv3 algorithm is replaced with a lightweight network MobileNet v2, which greatly reduces the amount of calculation and parameters of the model to improve the detection speed; the input size of the network is adjusted to 416×416; the anchor frame is re-clustered and analyzed using a self-made data set, and it is decided to use 9 anchor frames for prediction, and the sizes of the anchor frames are: 10×10, 15×15, 22×23, 36×36, 55×55, 62×63, 79×78, 95×96, 115×116; MobileNet The feature maps of the 55th, 116th and 150th layers of the v2 network are used for detection, and their feature map sizes are 52×52×192, 26×26×576 and 13×13×1280 respectively; the obtained feature maps are upsampled to increase the size and fused with the corresponding layers in the network, and the 9 anchor boxes are sorted from large to small and evenly distributed to 3 feature maps according to the principle of obtaining larger anchor boxes with deeper feature maps; finally, non-maximum suppression is used to remove duplicate bounding boxes to obtain the final pipe mouth detection result; compared with the traditional target detection algorithm, the pipe mouth recognition method has greatly improved in accuracy and real-time performance.

[0018] As a preferred embodiment, the improved YOLOv3 target detection algorithm in step 3) needs to be trained in advance using a data set.

[0019] As a preferred embodiment, the path planning method in step 5) is an orthogonal path method, which is completed in a pixel coordinate system. The algorithm flow is as follows:

[0020] Step 51) Calculate the average value of the lengths of all bounding boxes as the diameter of the nozzle, denoted as d;

[0021] Step 52) Find the smallest ordinate value among all the nozzle center coordinates, denoted as v min ;

[0022] Step 53) The vertical coordinate of the center of the other nozzle is marked as v i , if v i -v min <k·d, then these circles can be considered to be in the same row, k is a threshold;

[0023] Step 54) Sort the circles obtained in the previous step according to the horizontal coordinates, with odd-numbered rows in order and even-numbered rows in reverse order;

[0024] Step 55) Repeat steps 2 to 4 for the remaining nozzles;

[0025] Step 56) After all management is processed, the obtained sorting is the cleaning path;

[0026] Step 57) Finally, the planned cleaning path is sent to the actuator via the communication interface to complete the cleaning operation.

[0027] Beneficial effects: The navigation system of the present invention can obtain the position of the heat exchanger pipe mouth in real time and accurately, with high efficiency, can effectively reduce manual labor, and realize the fully automated cleaning operation of the heat exchanger. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 A schematic diagram of the flow chart of the visual navigation system for the heat exchanger cleaning device provided in an embodiment of the present invention.

[0029] Figure 2 It is a schematic diagram of the structure of the pipe orifice recognition module of the visual navigation system.

[0030] Figure 3 Schematic diagram for visualizing the results of the visual navigation system.

[0031] Figure 4 This is a schematic structural diagram of a heat exchanger cleaning device provided in an embodiment of the present invention.

[0032] The symbols in the figure represent: 1. Base; 2. Front and rear moving guide rails; 3. Vertical moving guide rails; 4. Left and right moving guide rails; 5. Sliding mounting base; 6. Motor base; 7. Front and rear drive motors; 8. Vertical drive motors; 9. Left and right drive motors; 10. Coupling; 11. Drive shaft; 12. Slider. DETAILED DESCRIPTION

[0033] The following is a detailed description of an embodiment of the present invention in conjunction with the accompanying drawings: This embodiment is implemented on the premise of the technical solution of the present invention, and a detailed implementation method and a specific operation process are given, but the protection scope of the present invention is not limited to the following embodiments.

[0034] In order to enable those skilled in the art to better understand the solutions of the present invention, the embodiments of the present invention are described below in conjunction with the accompanying drawings.

[0035] Please refer to Figures 1 to 4 , Figure 1 A schematic diagram of a flow chart of a visual navigation system for a heat exchanger cleaning device provided in an embodiment of the present invention;

[0036] Figure 2 A network diagram for orifice identification of the visual navigation system;

[0037] Figure 3 Visualizing the results of the visual navigation system;

[0038] Figure 4 A heat exchanger cleaning device is provided in an embodiment of the present invention. Figure 4It includes a base 1; a front and rear moving guide rail 2; a vertical moving guide rail 2; a left and right moving guide rail 4; a sliding mounting base 4; a motor base 6; a front and rear driving motor 7; a vertical driving motor 8; a left and right driving motor 9; a coupling 10; a transmission shaft 11; and a slider 12.

[0039] The embodiment of the present invention provides a visual navigation system for an automatic cleaning device of a heat exchanger, as shown in the attached manual. Figure 1 As shown, it mainly includes the following steps:

[0040] Step 1) The industrial camera acquires the front image of the heat exchanger tube bundle. The camera should be facing the heat exchanger tube bundle surface, and the depth of field is measured and recorded;

[0041] Step 2) preprocessing the image, including distortion correction and image enhancement. The parameters required for the distortion correction need to be obtained by camera calibration experiments in advance, and filtering is used to enhance the image quality;

[0042] Step 3) Use the improved YOLOv3 target detection algorithm to detect the nozzle. The network structure diagram of the improved YOLOv3 target detection algorithm is as follows: Figure 2 As shown in the figure; after the image is subjected to the convolution, pooling, and full connection operations of the MobileNet v2 network, the feature maps of the 55th, 116th, and 150th layers of the network are selected for detection, and the feature map sizes are 52×52×192, 26×26×576, and 13×13×1280, respectively; the obtained feature maps are upsampled to increase the size and fused with the corresponding layers in the network; the 9 anchor boxes obtained by clustering the data set are sorted from large to small and evenly distributed to 3 feature maps for detection according to the principle of obtaining larger anchor boxes from deeper feature maps; finally, non-maximum suppression is used to remove duplicate bounding boxes in the detection results to obtain the final nozzle detection results;

[0043] Step 4) According to the camera imaging principle, the coordinate conversion formula is used to convert the pixel coordinates of the center of the nozzle into actual coordinates. The conversion is sequentially performed from the pixel coordinate system to the image coordinate system, from the image coordinate system to the camera coordinate system, and from the camera coordinate system to the world coordinate system. The parameters required for the coordinate conversion need to be obtained in advance through a camera calibration experiment;

[0044] Step 5) Perform cleaning operation path planning. The path planning process first determines the nozzles located in the same row, and then determines the final cleaning path according to the principle of odd-numbered rows in order and even-numbered rows in reverse order. The nozzle positioning and path planning results are shown in Figure 5. Figure 3 As shown;

[0045] Step 6) The system converts the planned path into a command and sends it to the motor driver to control the cleaning device to automatically perform the operation. The movement speed of the slide and the cleaning time of a single nozzle can be adjusted; the cleaning device is as follows Figure 4 As shown, it is a three-axis gantry slide. The slide body includes: two front and rear moving guide rails, two vertical moving guide rails, one left and right moving guide rail, a water gun fixing structure, etc. The guide rails in each direction are equipped with motors to control their movement. When the cleaning device operates automatically, it will traverse each pipe opening in turn according to the cleaning path.

[0046] The above shows and describes the basic principles and main features of the present invention and the advantages of the present invention. It should be understood by those skilled in the art that the present invention is not limited to the above embodiments. The above embodiments and descriptions are only for explaining the principles of the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention to be protected. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. A method for operating a visual navigation system of an automatic cleaning device for a heat exchanger, characterized in that: The following steps are involved: Step 1) using an industrial camera to obtain a front image of the heat exchanger tube bundle; Step 2) preprocessing the acquired frontal image, including distortion correction and image enhancement; Step 3) Use the improved YOLOv3 target detection algorithm to perform nozzle detection; Step 4) according to the camera imaging principle, the coordinate conversion formula is used to convert the pixel coordinates of the center of the nozzle into actual coordinates; Step 5) planning the cleaning operation path; Step 6) Send the planned path to the execution agency for subsequent operations; The visual navigation system of the automatic cleaning device of the heat exchanger includes: an industrial camera, a PC, and a communication interface. The industrial camera is connected to the network port of the PC through a Gige interface, and the PC is connected to the spray gun actuator of the automatic cleaning device through a USB interface and a USB to RS-485 interface.

2. The method for operating a visual navigation system of an automatic cleaning device for a heat exchanger according to claim 1, characterized in that: The industrial camera in step 1) is a CCD camera or a CMOS camera, and the length and width of the pixels are changed to equal sizes when acquiring the image.

3. The method for operating a visual navigation system of an automatic cleaning device for a heat exchanger according to claim 1, characterized in that: The camera intrinsic parameters used in the image distortion correction and coordinate transformation process in step 2) and step 4) are obtained through camera calibration.

4. The method for operating a visual navigation system of an automatic cleaning device for a heat exchanger according to claim 1, characterized in that: The pipe orifice identification in step 3) uses a target detection algorithm based on YO LOv3, and is improved according to the specific conditions of the heat exchanger cleaning operation.

5. The method for operating a visual navigation system of an automatic cleaning device for a heat exchanger according to claim 4, characterized in that: The improvements are as follows: the backbone network DarkNet-53 of the YOLOv3 algorithm is replaced with a lightweight network MobileNet v2, which greatly reduces the amount of calculation and parameters of the model to improve the detection speed; the input size of the network is adjusted to 416×416; the anchor boxes are re-clustered using a self-made data set, and it is decided to use 9 anchor boxes for prediction, with the sizes of 10×10, 15×15, 22×23, 36×36, 55×55, 62×63, 79×78, 95×96, and 115×116 respectively; MobileNet The feature maps of the 55th, 116th and 150th layers of the v2 network are used for detection, and their feature map sizes are 52×52×192, 26×26×576 and 13×13×1280 respectively; the obtained feature maps are upsampled to increase the size and fused with the corresponding layers in the network, and the 9 anchor boxes are sorted from large to small and evenly distributed to 3 feature maps according to the principle of obtaining larger anchor boxes with deeper feature maps; finally, non-maximum suppression is used to remove duplicate bounding boxes to obtain the final pipe mouth detection result; compared with the traditional target detection algorithm, the pipe mouth recognition method has greatly improved in accuracy and real-time performance.

6. The method for operating a visual navigation system of an automatic cleaning device for a heat exchanger according to claim 1, characterized in that: The improved YOLOv3 target detection algorithm in step 3) needs to be trained in advance using a data set.

7. The method for operating a visual navigation system of an automatic cleaning device for a heat exchanger according to claim 1, characterized in that: The path planning method in step 5) is an orthogonal path method, which is completed in a pixel coordinate system. The algorithm flow is as follows: Step 51) Calculate the average value of the lengths of all bounding boxes as the diameter of the nozzle, denoted as d; Step 52) Find the smallest ordinate value among all the nozzle center coordinates, denoted as v min ; Step 53) The vertical coordinate of the center of the other nozzle is marked as v i , if v i -v min <k·d, then these circles can be considered to be in the same row, k is a threshold; Step 54) Sort the circles obtained in the previous step according to the horizontal coordinates, with odd-numbered rows in order and even-numbered rows in reverse order; Step 55) Repeat steps 52 to 54 for the remaining nozzles; Step 56) After all management is processed, the obtained sorting is the cleaning path; Step 57) Finally, the planned cleaning path is sent to the actuator via the communication interface to complete the cleaning operation.

Images