CN115631138A

CN115631138A - Zirconium alloy plate laser cutting quality monitoring method and device

Info

Publication number: CN115631138A
Application number: CN202211186427.6A
Authority: CN
Inventors: 计效园; 王泽明; 涂先猛; 罗建东; 吴楚澔; 王伟; 侯明君; 王治国; 秦应雄; 潘喆
Original assignee: Huazhong University of Science and Technology; Nuclear Power Institute of China
Current assignee: Huazhong University of Science and Technology; Nuclear Power Institute of China
Priority date: 2022-09-27
Filing date: 2022-09-27
Publication date: 2023-01-20

Abstract

The invention discloses a method and a device for monitoring the laser cutting quality of a zirconium alloy plate, and belongs to the field of laser cutting. The method comprises the steps of erecting a linear array CMOS camera below the side of a laser cutting section, obtaining image information of a cutting surface and a bottom surface of a cut sheet through the linear array CMOS camera after the laser cutting of the zirconium alloy sheet is completed, correcting the image of the bottom surface without correcting the photographed section through a computer image correction and image defect identification and classification program, fusing the image into a picture, carrying out graying and grid division processing, and rapidly identifying, classifying, marking, early warning, judging and evaluating defects. The invention adopts a linear array CMOS camera to fuse the cutting section and the bottom surface of the plate into one picture, and simultaneously carries out online monitoring detection on three types of defects, compared with the existing zirconium alloy plate laser cutting quality detection, the invention develops the online monitoring detection method and the device with high function integration level by using lower software and hardware cost, and improves the reliability of detection and the product quality.

Description

Zirconium alloy plate laser cutting quality monitoring method and device

Technical Field

The invention belongs to the technical field of laser cutting, and particularly relates to a method and a device for monitoring the laser cutting quality of a zirconium alloy plate.

Background

The working principle of laser cutting is that the laser generated by a laser is guided by an optical device, focused on the surface of a material to melt and vaporize the material, and meanwhile, the melted material is blown away by compressed gas coaxial with the laser beam, and the laser beam and the material move relatively along a certain track, so that a cut with a certain shape is formed. The laser cutting technology can be used for processing metal and non-metal materials, can greatly reduce the processing time, reduce the processing cost, improve the quality of workpieces, and can be widely applied to cutting of metal plates in production. The zirconium alloy has good plasticity, can be made into pipes, plates and the like, and is mainly used in the fields of petroleum and nuclear technology.

At present, the lancing environment in the laser cutting process is complex, and parameters of laser cutting such as laser power, cutting speed, gas flow, defocusing amount and the like are not properly selected, so that a plurality of defects such as slag adhering, splashing, excessive tangent plane roughness and the like are easily generated during cutting, key quality indexes of products of laser cutting are unqualified, the service performance of the finally assembled products is seriously influenced, and even serious safety accidents are caused. However, the quality detection of the traditional laser cutting product depends on manual detection, and due to fatigue of workers, misjudgment and misjudgment are possible, a lot of manpower is consumed, and the reliability of the detection result needs to be improved. In addition, the traditional laser cutting does not have a detection device for detecting the defects of slag adhering, splashing, standard exceeding of surface roughness and the like, the products produced by the laser cutting of the zirconium alloy plate cannot be evaluated in a short time, and the products can only be cooled and then manually detected whether the slag adhering, the splashing and the surface roughness reach the standard or not, so that a large amount of time is required for evaluating the cutting quality of the laser cutting products. Therefore, it is necessary to develop a highly functional integrated method and device for monitoring and detecting the laser cutting quality on line.

Disclosure of Invention

Aiming at the defects of the prior art, the invention aims to provide a method and a device for monitoring the laser cutting quality of a zirconium alloy plate, and aims to solve the problems of saving a large amount of manpower and improving the detection reliability of a laser cutting product.

In order to achieve the above object, in a first aspect, the present invention provides a method for monitoring laser cutting quality of a zirconium alloy plate, the method comprising:

s1, obtaining a picture containing a cutting surface and a bottom surface of a zirconium alloy plate after laser cutting, correcting the bottom surface of the plate, fusing the corrected picture with an uncorrected cutting section picture into a picture, and carrying out graying and meshing on the picture;

s2, calculating the ratio of the distribution standard deviation and the distribution root mean square height value of the gray level histogram of each grid, respectively comparing the ratio with a threshold value corresponding to the maximum acceptable roughness, wherein if the ratio exceeds the threshold value, the grid has the defect that the surface roughness exceeds the standard, otherwise, the grid does not have the defect that the surface roughness exceeds the standard;

s3, inputting the gray-scale pictures of each grid into a convolutional neural network to obtain the probability of slag adhering defects and splashing defects of each grid, and calibrating the defects according to the probability;

and S4, evaluating the cutting quality of the plate by integrating the number and the area of three defects of overproof surface roughness, slag adhering and splashing.

Preferably, in step S1, the bottom surface correction is performed on the picture, and the method includes:

1) Identifying an intersection line of the bottom surface and the cutting surface based on perspective transformation and extending the intersection line to the edge of the picture;

2) Carrying out perspective transformation correction on the picture below the obtained extension line;

3) And fusing the corrected lower picture with the cutting surface along the extension line to obtain a final correction picture.

Preferably, the horizontal distance between the shooting device and the cutting surface is obtained, and according to the trigonometric function relation, the acute included angle between the vertical plane of the axis of the shooting device and the bottom surface of the plate is obtained and used as the bottom surface correction angle.

Preferably, before each grid grayed picture is input to the convolutional neural network, a selective search algorithm is adopted to perform preliminary screening of suspicious grid pictures.

Preferably, in step S3, regarding the obtained defect determination result, if the maximum probability exceeds 95%, it is determined that the defect exists, and a red frame mark is directly added to the region, and if the maximum probability is 75% to 95%, it is determined that the defect is suspected to exist, and a yellow frame mark is added to the region, and if the maximum probability is 75% or less, no mark is added, and finally, the region is displayed on a computer interface by a graphic representation, so that the region is easy to distinguish.

Preferably, the method further comprises any one of the following processing modes:

the first processing mode is as follows: carrying out three-color lamp early warning and buzzer early warning on a visual display interface pop window and linear array CMOS monitoring system; wherein, the red early warning light is on to indicate that the defect problem is serious, the defect quantity or size exceeds the normal range, and the buzzer sends out continuous warning sound; the yellow early warning lamp is on to indicate that the quality of the cutting surface and the quality of the bottom surface are both in a normal range, but the grade is not good, and the buzzer gives an intermittent warning sound; the green early warning lamp is on to indicate that the defect condition is normal, the score is excellent, and no warning sound exists;

the second treatment method comprises the following steps: and carrying out quantity statistics, size statistics and defect density statistics on the detected defects, comparing with a standard, and judging whether the defects reach the standard or not.

In order to achieve the above object, in a second aspect, the present invention provides a zirconium alloy plate laser cutting quality monitoring device, including: the device comprises a shooting device, a guide rail, a support connecting device and an image processing and analyzing module;

the guide rail is positioned below the side of the laser cutting surface and is parallel to the x axis of the laser cutting, and is used for driving the shooting device to move smoothly without obstacles and ensuring that the shooting device can simultaneously scan the full view of the cutting surface and the bottom surface of the laser cutting workpiece;

the shooting device is fixed on the guide rail through the supporting and connecting device and used for shooting pictures containing the bottom surface of the cut sheet and a cutting surface after the laser cutting of the zirconium alloy sheet is finished according to the control signal and uploading the pictures to the image processing and analyzing module;

the image processing and analyzing module is used for processing and analyzing the shot picture by the method of the first aspect to obtain the defect position and type and the cutting quality of the zirconium alloy plate.

Preferably, the elevation angle of the photographing device is 45 ° to 65 °.

Preferably, the device further comprises a protective sheet;

the screening glass is installed in the lens surface of shooting device for according to control signal, close at the laser cutting in-process, with isolated lens and the external world of shooting the device, open at shooting device scanning collection in-process, with the collection of assurance image.

Preferably, the apparatus further comprises: a micro ventilation device;

the miniature ventilation device is arranged on the periphery of a lens of the shooting device and used for blowing out gas in the scanning and collecting process of the shooting device according to a control signal so as to clear residual smoke dust in the visual angle of the shooting device.

Generally, compared with the prior art, the technical scheme conceived by the invention has the following beneficial effects:

(1) The invention provides a zirconium alloy plate laser cutting quality monitoring method, which obtains and analyzes image information of the bottom of a cut plate and the cut surface through one-time shooting, reduces shooting and analyzing times, and reduces time spent on quality monitoring. Correcting the image by a computer correction program, correcting only the bottom part while keeping the cut surface as a side view, fusing the two parts into a final correction image, carrying out region division and selection on the image, classifying and judging various defects of the image by a convolutional neural network, displaying image defect information on a computer interface, evaluating the cutting quality of the laser-cut zirconium alloy plate, quickly acquiring the defect position and type of the laser-cut zirconium alloy plate, feeding the defect information back to a laser cutting system, adjusting laser cutting parameters, and monitoring and detecting the quality of the laser-cut zirconium plate.

(2) The invention provides a zirconium alloy plate laser cutting quality monitoring device, which is characterized in that a linear array CMOS camera and a motion auxiliary and protection device thereof are additionally arranged below the side of a laser cutting seam on the basis of the existing laser cutting quality, monitoring and detection of three defects on two surfaces by one picture are realized by means of a computer image processing technology, and the whole processing equipment is not greatly modified, so that a large amount of manpower is saved, the detection reliability is improved, and a large amount of cost is saved in a long-term view.

Drawings

Fig. 1 is a flow chart of a laser cutting quality monitoring method for a zirconium alloy plate provided by the invention.

Fig. 2 is a schematic view of a camera mounting structure according to an embodiment of the invention.

Fig. 3 is a schematic view of the bottom surface correction according to the embodiment of the present invention.

Fig. 4 is a schematic diagram of a defect marking result according to an embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

Fig. 1 is a flow chart of a laser cutting quality monitoring method for a zirconium alloy plate provided by the invention. As shown in fig. 1, the method includes:

step S1: the camera is mounted at an angle.

Fig. 2 is a schematic view of a camera mounting structure according to an embodiment of the invention. As shown in fig. 2, a linear array CMOS camera is mounted on the bottom side of the zirconium plate, and the camera is covered with a protective glass to prepare for laser cutting of the zirconium plate. The main viewing direction of the linear array CMOS camera is perpendicular to the intersection line of the cutting surface 1 and the bottom surface 2, the elevation angle set by the linear array CMOS camera is optimal within 45-65 degrees, the slag hanging and rough surface defect outline of the cutting surface is highlighted in the angle range, meanwhile, the bottom surface information of a plate with larger size can be covered, and the accurate identification and classification of the defects by a following neural network model are facilitated.

And a guide rail is arranged below the laser cutting surface side and used for moving the linear array CMOS camera, and the guide rail is arranged parallel to the x axis of laser cutting. The linear array CMOS camera is connected with the guide rail through the supporting device, so that the linear array CMOS camera can move on the guide rail without obstacles and stably and cannot be damaged by laser.

Step S2: and (5) laser cutting the zirconium alloy plate.

And setting parameters of laser cutting, setting parameters such as cutting power, cutting rate and air pressure, clamping the zirconium plate, and performing laser cutting.

And step S3: and acquiring and reconstructing linear array CMOS images.

After laser cutting is finished, after a smoke exhaust system absorbs most of smoke dust generated in the cutting process, a protection plate of a photographic device is opened, a ventilation system is started, residual smoke dust in the visual angle of a linear array CMOS camera is dispersed, the linear array CMOS camera is started, and the linear array CMOS camera simultaneously collects optical images of a cutting surface and the bottom surface of a cut piece along a pre-installed guide rail.

The lens that can completely cut off photographic means with the screening glass of linear array CMOS camera through external member and screening glass and external in the laser cutting process, through motor traction soft rope open the screening glass when the scanning of linear array CMOS camera is collected information, does not disturb photographic means's image information's collection, the screening glass is glass piece, rubber piece, and the screening glass shape adopts the rectangle.

A small ventilation and smoke-expelling system is arranged around the linear array CMOS camera, so that residual smoke dust in the visual angle of the photographic device is removed in the image acquisition process of the photographic device, and the image information acquired by photographing is not interfered by the residual smoke dust.

The linear array CMOS camera is driven by a servo motor to move to the cutting tail end on a guide rail along the cutting track direction from the cutting start end of the plate, and the reconstruction of the image information of the whole cutting surface and the bottom surface of the plate is completed. The reconstructed cutting surface and the reconstructed bottom surface are not the front view of the cutting surface and the bottom surface, and the presented image is formed by connecting two trapezoids.

And step S4: and (4) bottom surface image correction and section image fusion.

Because the camera is placed below the side of the cutting seam, the bottom surface of the collected image and the cutting surface are side views with angles, the image needs to be corrected, the cutting surface is not corrected, the bottom surface of the plate is corrected, and then the two parts are fused to obtain a corrected image.

S4.1: correcting the bottom image of the plate: the correction process adopts perspective transformation based on the improvement of the geometric intersection line of the bottom surface and the cut surface: the method comprises the steps of firstly identifying the intersection line of a bottom surface and a cutting surface and extending the intersection line to the edge of a picture, and carrying out perspective transformation correction on the picture below the obtained extension line based on the perspective transformation. Because the roughness value needs to be paid attention to in the detection of cutting plane, correct and cause the influence to the detection of roughness easily, and the cutting plane is corrected and is not influenced greatly to the detection of hanging the sediment, so can choose not to correct the cutting plane.

In the specific correction calculation, if a point (x, y, 1) on the image before correction is selected, which corresponds to a point (x ', y', 1) on the image after correction, and the transformation matrix is H, the transformation relationship is as follows:

because the width of the bottom surface of the cut zirconium plate is far greater than the thickness of the plate, the bottom surface occupies most of the detection area, and the placing angle of the camera needs to be more deviated to the bottom surface for imaging. After the guide rail is installed, the height from the camera to the bottom surface of the workpiece to be machined can be determined, the guide rail is parallel to the x axis of laser cutting, and after the machining program is loaded, the horizontal distance from the camera to the cutting surface can be obtained. Fig. 3 is a schematic view of the bottom surface correction according to the embodiment of the present invention. As shown in fig. 3, an included angle between the axis of the camera and the intersection line of the cutting surface and the bottom surface, which is the angle θ for bottom surface correction, can be obtained according to the trigonometric function relationship.

S4.2: fusing the bottom surface of the plate with the cut surface image: the corrected bottom surface image and an image directly shot by a linear array CMOS camera are placed in one image, the bottom surface of the plate is distinguished from the cutting surface through a boundary line of the bottom surface and the cutting surface, the corrected bottom surface image is rectangular, the image of the cutting surface which is not corrected is trapezoidal, the top edge of the rectangle is overlapped with the bottom edge of the trapezoid, and the reconstructed dotted line on the overlapped line is visually distinguished.

Step S5: defect identification classification, early warning, marking, judging and evaluation.

In the identification and classification of defects, based on the zirconium alloy plate laser cutting quality monitoring method and device, an image input into a computer is converted into a gray image through gray processing, wherein the RGB three-color image is subjected to gray processing by adopting but not limited to a weighted average method to obtain a gray image with a pixel value in a range of [0, 255 ]. The algorithm used was I (x, y) = a × I _ R (x, y) + B × I _ G (x, y) + c × I _ B (x, y), where a + B + c =1.

S5.1: and (3) defect identification and classification: for the detection of the surface roughness, the surface roughness of a cut surface is continuous in a large range, the surface roughness of a workpiece is related to the standard deviation and the root mean square of the distribution of an image gray histogram, the specific surface roughness is increased along with the increase of the ratio of the distribution standard deviation and the distribution root mean square height value of the image gray histogram, the ratio corresponding to each metal is different, the gray histogram is analyzed by shooting the surface of the zirconium alloy cut by normal laser, the distribution standard deviation S and the distribution root mean square height value H of the gray histogram are obtained, a threshold value is determined by calculating the ratio S/H, 100 graphs are selected for testing and determining the threshold value, and in the test result of the 100 graphs, the corresponding S/H maximum value is selected as the threshold value in the data of the roughness in the normal range. And cutting the corrected picture into a picture analysis gray histogram of a small area, judging whether the surface roughness of the area exceeds the standard or not according to a threshold value corresponding to the maximum acceptable roughness obtained by testing in advance, and judging that the surface roughness exceeds the standard if the calculated value is larger than the threshold value. In this embodiment, the standard for the laser-cut surface roughness value of the zirconium alloy sheet is Ra 3.2.

For detecting slag adhering and splashing, firstly, dividing a picture into a plurality of grids with side length of 84 multiplied by 84 pixels, screening the areas, and selecting all suspicious areas, wherein a selective search algorithm is selected, and the specific steps are as follows: firstly, an input picture is divided through double-threshold initialization, a target area is recommended, the similarity between adjacent sub-areas in the picture is calculated from the aspects of texture features, overlapping and the like of adjacent areas, and finally, a smaller number of target areas are obtained by continuously combining the adjacent similar areas, so that the target search range is reduced.

The obtained suspicious region is cut into a plurality of pictures of 42 multiplied by 42 pixels, the defects shot by a linear array CMOS camera are in the level of 0.1mm, the 42 multiplied by 42 pixel pictures can include the defects, a convolutional neural network is applied to input the 42 multiplied by 42 pixel pictures at the cut position, the characteristics of the defect position are extracted through 4-6 layers of convolution, pooling, activation and other processing, the transition between a convolution layer and a full connection layer is realized through flattening processing and is input into 3 layers of full connection, and finally a Softmax activation function is used to output the probability value for judging that the images have various types of defects.

The training of the neural network is obtained from a cut product, the cut product is photographed and input, the input data are 1000 pictures with 42 multiplied by 42 pixels with slag adhering defects and 1000 pictures with 42 multiplied by 42 pixels with splashing defects, 80% of the obtained data are used for the training of the CNN convolution neural network, the other 20% of the obtained data are used for prediction, and when the accuracy of the predicted value reaches more than 99%, the trained neural network can be considered to be reliable and can be used for actual production.

S5.2: defect early warning: the visual display interface pop-up window and the linear array CMOS monitoring system are subjected to three-color lamp early warning and buzzer warning sound early warning. The red early warning lamp is on to indicate that the defect problem is serious, the number or the size of the defects exceeds a normal range, and the buzzer sends out continuous warning sound; the yellow early warning lamp is on to indicate that the quality of the cutting surface and the quality of the bottom surface are both in a normal range, but the grade is not good, and the buzzer gives an intermittent warning sound; the green warning lamp is on to indicate that the defect condition is normal, and the score is excellent without warning sound.

S5.3: marking defects: the area marked by the defect area is displayed through a computer interface, and the defect area is colored in red. For the obtained defect judgment result, a red frame mark is directly added to the region with the maximum probability of more than 95%, a yellow frame mark is added to the region with the maximum probability of 75-95%, and the region with the maximum probability of less than 75% is obtained. And finally, the computer interface is convenient to distinguish when being displayed through a diagram. Fig. 4 is a schematic diagram of a defect marking result according to an embodiment of the present invention. As shown in FIG. 4, the small circles indicate the slag adhering defect, the ellipses indicate the splashing defect, and the boxes indicate the roughness exceeding defect.

S5.4: and (3) defect judgment: and comparing with the standard to see whether the single defect reaches the standard or not. Counting the number, size and defect density of the detected defects, wherein the number of the adhered slag cannot exceed 4, and the density cannot exceed 2mm ² /dm ² The amount of splashes can not exceed 10 and the density can not exceed 3mm ² /dm ² The surface roughness exceeding standard area can not exceed 12mm ² 。

S5.5: and (3) defect evaluation: combining the area of an actually measured size and a surface roughness standard exceeding area, and carrying out comprehensive evaluation on the plate material quality by using quality indexes such as the size, the quantity and the like of slag adhering and splashing: according to key indexes such as surface roughness, slag adhering and splashing, the comprehensive quality score is obtained through weighted average: the quality evaluation standards of the slag adhering, splashing and surface roughness are set as 60 points, the defect-free quality evaluation standard is set as full 100 points, the proportion of each quality index data of one defect in the defect score is equal, each index score is 60 points according to the standard value, the defect-free quality index data is 100 points, the scores are uniformly distributed along with the quality data, the total quality score is calculated by weighting the three defect scores according to the proportion of 1: 1, and the quality is sequentially divided into 4 grades: the grade of 90 minutes or more is excellent, the grade of 75 minutes or more is good, the grade of 60 minutes or more is acceptable, and the grade of 60 minutes or less is unacceptable. And for a certain item of data, the standard value is exceeded, and the data is classified as unqualified.

Furthermore, after the evaluation of the laser cutting product is finished, the detected problems can be fed back to the laser cutting system, the laser cutting parameters can be adjusted in time, the product quality is improved, and the purpose of monitoring the product quality is achieved.

It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.

Claims

1. A method for monitoring the laser cutting quality of a zirconium alloy plate is characterized by comprising the following steps:

s1, obtaining a picture containing a cutting surface and a bottom surface of a zirconium alloy plate after laser cutting, correcting the bottom surface of the plate, fusing the corrected picture and an uncorrected cut section picture into a picture, and carrying out graying and grid division on the picture;

s2, calculating the ratio of the distribution standard deviation and the distribution root mean square height value of the gray level histogram of each grid, and comparing the ratio with a threshold corresponding to the maximum acceptable roughness respectively, wherein if the ratio exceeds the threshold, the grid has the defect that the surface roughness exceeds the standard, otherwise, the grid does not have the defect that the surface roughness exceeds the standard;

2. The method as claimed in claim 1, wherein the step S1 of performing the bottom surface correction on the picture comprises:

3. The method of claim 2, wherein the horizontal distance of the camera from the cut surface is obtained, and the acute angle between the vertical plane of the camera axis and the bottom surface of the plate is obtained as the bottom surface correction angle according to a trigonometric function relationship.

4. A method as claimed in any one of claims 1 to 3, wherein a selective search algorithm is used to perform a preliminary screening of suspect grid pictures before each grid greyed picture is input to the convolutional neural network.

5. The method according to claim 1, wherein in step S3, regarding the obtained defect determination result, if the maximum probability exceeds 95%, the defect is determined to be present, and a red frame mark is directly added to the region, and if the maximum probability is 75% to 95%, the defect is determined to be present, and a yellow frame mark is added to the region, and if the maximum probability is 75% or less, the region is not marked, and finally, the region is displayed on a computer interface by illustration, so that the region is easy to distinguish.

6. The method of claim 1, further comprising any of the following:

the second processing mode: and carrying out quantity statistics, size statistics and defect density statistics on the detected defects, comparing with a standard, and judging whether the defects reach the standard or not.

7. The utility model provides a zirconium alloy panel laser cutting quality monitoring device which characterized in that, the device includes: the device comprises a shooting device, a guide rail, a support connecting device and an image processing and analyzing module;

the image processing and analyzing module is used for processing and analyzing the shot picture by the method according to any one of claims 1 to 6 to obtain the defect position and type and the cutting quality of the zirconium alloy plate.

8. The apparatus of claim 7, wherein the elevation angle of the camera is 45 ° to 65 °.

9. The device of claim 7, further comprising a protective sheet;

10. The apparatus of claim 7, wherein the apparatus further comprises: a micro ventilation device;