Research On Image Recognition And Positioning Method Of Heat Exchanger Pipe Mouth Based On U-Net

Heat exchanger is an important heat exchange equipment widely used in industrial production and energy fields.During usage,the interior and exterior of heat exchanger pipes gradually accumulate dirt and deposits,which decrease the heat conduction efficiency of the heat exchanger and affect its normal operation.Currently,the mechanical cleaning used for heat exchanger cleaning relies on manual operation,which is inefficient and requires frequent maintenance.To improve the automation and efficiency of heat exchanger cleaning,this paper focuses on shell-and-tube heat exchangers and proposes using deep learning in conjunction with robotic cleaning of heat exchanger pipes.The specific research work conducted in this paper is as follows:1.Research on image feature recognition algorithm for heat exchanger tube inlet.Based on the requirements of feature extraction for heat exchanger tube inlet images in this paper,two approaches,namely traditional image processing and deep learning,are proposed for the recognition of heat exchanger tube inlet images.Firstly,the image enhancement,edge detection,threshold segmentation,and feature extraction processes are performed using the MATLAB platform to obtain the feature images of the heat exchanger tube inlet.Secondly,based on the U-Net model,the network depth is reduced,and depth separable modules are added to the upsampling process,while multi-scale mixed dilated convolutions are incorporated into the downsampling process.The trained model is then used to obtain the feature images of the tube inlet.Experimental results demonstrate that deep learning not only improves the prediction speed compared to traditional image processing methods but also enhances the recognition accuracy.2.Research on image target localization algorithms for heat exchangers.Based on the characteristics of heat exchanger pipe openings,the Hough transform algorithm is used to extract the coordinates of the heat exchanger pipe openings and to locate the position of the robot spray gun.Camera calibration using Zhang’s calibration method is performed to obtain camera intrinsic and extrinsic parameters,as well as distortion coefficients.These parameters are used to transform the obtained pixel coordinates into world coordinates,and the cleaning process and cleaning path are explained.3.Design of the visual recognition and localization system for heat exchanger cleaning robots.The requirements for the heat exchanger pipe opening image recognition and localization system are described.Based on the system requirements analysis,the necessary hardware and software design is completed.The system interface is built using Py QT5 in Py Charm software,and the design and development of various functional modules are carried out.4.Verification of the functionality of the visual recognition and localization system for heat exchanger cleaning robots.Experimental tests are conducted according to the cleaning process to verify the functionality of camera calibration,image pipe opening recognition,pipe opening coordinates,and robot coordinate localization.The test results validate the effectiveness and feasibility of the system.