Research On Detection Method Of Plate Defects In Plate Heat Exchanger

Plate heat exchanger is a device for heat transfer through a series of corrugated metal plates.In the process of production of plate heat exchanger plate,the surface of the plate is defective due to gravity stamping.When the plate is put into use,the defect position will not be able to withstand the pressure after water injection and crack,and there are potential safety hazards.At present,manual sampling inspection and penetration inspection are commonly used to detect defects on plate heat exchangers,but such methods are inefficient and affected by subjective factors of inspectors.With the deep exploration of machine vision application,according to the defect characteristics of plate surface of plate heat exchanger and based on the understanding of the defect area,designed plate heat exchanger plate defect detection method based on machine vision,and completed the automatic detection of plate surface defects of plate heat exchanger.First,the ROI regions are divided according to the causes of defects Locating the ROI area can greatly reduce the workload of computer and increase the detection accuracy.The classical edge detection algorithm is used to test,and the detection effects of Roberts,Prewitt,Sobel,Canny and Shenjun operator in the extraction of bright and dark boundaries are compared.In the end,it is determined that Shenjun operator has the best effect in the detection of bright and dark region boundaries,and the dark region is located according to the bright and dark region boundaries.Then,defect detection is carried out in the positioned ROI area.According to the relationship between the amplitude of concave and convex line segment and defects,three defect detection algorithms are designed:(1)multi-scale convex segment absolute amplitude histogram defect detection method.In the first place,the image is preprocessed by concave-convex line segment shaping algorithms at three scales.The absolute amplitude histogram of convex line segment was used to perform threshold segmentation on the preprocessed image to obtain the defect detection results at three scales,and then the defect detection results were fused.Finally,feature extraction was carried out to remove reflective position and rough position error detection.(2)Defect detection algorithm of concave-convex line segment amplitude combined with defect location.First of all,the image is preprocessed by gaussian down-sampling and then Gaussian filtering.Then,the detection template is designed according to the defect location characteristics and amplitude characteristics.(3)Defect detection method of concave-convex line segment amplitude combined with 3 times standard deviation.Above all,the convex segment amplitude expansion method was used to recalculate the convex segment absolute amplitude,and then the defect was extracted by the principle of triple standard deviation twice.In the end,reflection error detection,and the approximate collinear contour combination operator was used to remove the rough position error detection.The above three algorithms can accurately detect the location of defects and have great robustness.The second defect detection algorithm has the highest detection efficiency and only produces reflective error detection.For reflective error detection,it is better to use the concaveconvex surface algorithm to remove false detection through experimental tests.For rough error detection,it is better to use approximate contour collinear operator to remove false detection through experimental tests.The image processing algorithm based on machine vision provides a new solution for plate detection and has certain practical value.