Method for monitoring surface defects and geometry and device for its implementation

Abstract

FIELD: production of linear products.SUBSTANCE: production of linear products by drawing, rolling, extrusion, twisting, layering, methods and devices for contactless monitoring of surface defects. Claimed device comprises a horseshoe-shaped housing mounted on a telescopic stand with the ability to be folded out of the working area by means of a hinge with light sources, video cameras and a laser Doppler velocimeter installed in it, a switch, a computer, a switch for connecting to a corporate network, a light and sound signaling unit, a connection unit with a process line, a connection unit with peripheral devices in the form of a printer and a marking machine, a mouse, a keyboard, a touch screen and software for collecting, processing and displaying information.EFFECT: increase in the accuracy of defect detection, determination of the geometric dimensions of the body, monitoring of the surface of the working body from all sides in a plane perpendicular to the direction of movement of the working body, creation of the possibility of introducing and removing the control and measuring system from the measurement zone without stopping the technological process and determining the coordinates of the detected defect along the length of the inspected working body.16 cl, 6 dwg

Description

[0001] The invention relates to the field of production of linear products by drawing, rolling, extrusion, twisting, layering, in particular, to methods and devices for contactless monitoring of surface defects, and can be used in optical monitoring and control of the technological process of production of cable, fiber, rod, wire rod, pipe, optical fiber, wire, rope, twisted bundle, electrodes, linear products made of polymers, wood, wood-polymer composites, metals, rubber, glass, composite materials.

[0002] A system for optical inspection of surface defects using a video camera and an artificial neural network is known from the prior art. (News of universities. North Caucasus region. Technical sciences. 2020. No. 1, pp. 33-40, dated 03.02.2020).

[0003] This control method is based on the use of a video camera that films the surface of the working fluid and generates a video stream that is analyzed using a trained artificial neural network that identifies defects on its surface.

[0004] The system contains a video camera and a computing system for collecting, processing, and displaying information. All system elements are fixed relative to the moving surface of the working body.

[0005] The specified method and device, consisting of monitoring the quality of the surface of the working fluid, make it possible to more successfully combine optical control and intelligent analysis of the video stream using a trained artificial neural network into a single control and measuring complex than in known technical solutions.

[0006] However, this method and system do not allow determining the geometric dimensions of the working fluid, monitoring its surface from all sides in a plane perpendicular to the direction of movement of the working fluid during its vibrations relative to the video camera, and introducing the control and measuring system into the measurement zone and also removing it as necessary without interrupting the technological process.

[0007] The closest comparable device to the claimed device is the optical inspection device disclosed in Russian Patent No. 2716465, published March 11, 2020. This device comprises at least one video camera, at least two light sources, and a computing system for analyzing the video stream from the video camera using a trained artificial neural network. The video camera and light sources are fixed relative to each other on a fixed frame of the device and are positioned relative to each other at a certain angle in a plane perpendicular to the direction of motion of the working fluid. The device also comprises a computing system for collecting, processing, and displaying information.The disadvantage of this device is a significant reduction in accuracy in the case where the working fluid undergoes oscillatory movements, and also does not have the ability to determine the geometric dimensions of the working fluid, control its surface from all sides in a plane perpendicular to the direction of movement of the working fluid during oscillations relative to the video camera, and also does not have the ability to introduce the control and measuring system into the measurement zone and remove it as needed without interrupting the technological process.

[0008] The closest to the claimed method is the method disclosed in Russian Patent No. 2716465, published March 11, 2020, for monitoring the surface of a working fluid, based on video recording of the surface of the working fluid illuminated by two light sources, followed by processing the video stream using an artificial neural network to identify defects in the filmed surface.

[0009] This method also has the disadvantages typical of the device intended for its implementation: the measurement accuracy drops significantly when the working fluid oscillates, the method is not intended for video recording of the surface of the working fluid from all sides in a plane perpendicular to the direction of movement of the working fluid, the method cannot be used to determine the geometric dimensions of the working fluid and is not applicable for the case when it is necessary to introduce and remove the control and measuring system into and from the measurement zone without stopping the technological process.

[0010] This circumstance gives rise to a technical problem associated with inaccurate determination of defects, the impossibility of determining the geometric dimensions of the body, the impossibility of monitoring the surface of the working body from all sides in a plane perpendicular to the direction of movement of the working body, the impossibility of introducing and removing the control and measuring system from the measurement zone without stopping the technological process.

[0011] The technical result that the claimed group of inventions is aimed at achieving is an increase in the accuracy of defect detection, determination of the geometric dimensions of the body, monitoring of the surface of the working body from all sides in a plane perpendicular to the direction of movement of the working body, creating the possibility of introducing and removing the control and measuring system from the measurement zone without stopping the technological process and determining the coordinates of the detected defect along the length of the inspected working body.

[0012] The technical result is achieved in that, unlike the known device for monitoring a surface, the system is a horseshoe-shaped housing having a hinged connection with a fixed stand, allowing it to be swung out of the working area with at least one laser Doppler velocimeter installed at a certain angle in a plane perpendicular to the direction of movement of the working fluid on a frame inside the sealed horseshoe-shaped housing, at least two video cameras directed into the working area, as well as at least two light sources directed into the working area and illuminating it, wherein the outputs of the laser Doppler velocimeter, video cameras and light sources are connected to a computing system for control, collection, processing and display of information.

[0013] The working body can be made in the form of a cable, fiber, rod, wire rod, pipe, optical fiber, wire, rope, twisted bundle, electrodes, molded products made of polymers, wood, wood-polymer composites, metals, rubber, glass, composite materials.

[0014] It is preferable to use at least four video cameras and at least four light sources installed at 90 degrees and one laser Doppler velocimeter.

[0015] The technical result is also achieved by the fact that, unlike the known method, the video stream from all video cameras and the laser Doppler velocimeter is received and processed in such a way as to, with the help of a trained artificial neural network, determine surface defects on all sides of the working body in a plane perpendicular to the direction of its movement, find the location of the working body in the working zone and determine its geometric dimensions, as well as determine the coordinate of the detected defect on the working body along its length.

[0016] The location of the working fluid in the working gap and its geometric dimensions are determined using pairs of video cameras installed at an angle of 90 degrees in a plane perpendicular to the direction of movement of the working fluid and directed at the workpiece by processing and comparing the video stream from each of them with the selection of the contours of the working fluid.

[0017] The coordinate of the defect on the working body along its length is determined by obtaining the value of the working body velocity using a laser Doppler velocimeter installed inside the horseshoe-shaped housing so that it is directed at the working body and continuously measures its velocity.

[0018] It is preferable that the trained artificial neural network is an artificial intelligence system implementing a classifier based on a convolutional neural network, a classifier based on a deep convolutional neural network, a classifier based on a residual convolutional neural network, wherein the type of defect and whether the defect is false or true are determined by inputting said set of images into the trained artificial intelligence classifier, and wherein the artificial intelligence classifier is pre-trained on a plurality of training sets of images of different defects indicating the types of defects.

[0019] It is preferable to display the measurement results in digital and graphical form on the computer screen.

[0020] In the claimed group of inventions, an increase in the accuracy of control is achieved by combining images from several video cameras and synthesizing video streams from these cameras in such a way as to avoid errors associated with the working fluid moving away from the optimal position for video filming during oscillatory motion due to the overlapping of images of the same section of the surface of the working fluid with sections that are filmed by different video cameras from different angles, filming the entire surface of the working fluid in a plane perpendicular to the direction of movement of the working fluid is carried out by overlapping sections of the surface filming obtained from different video cameras, the determination of the geometric dimensions of the working fluid is carried out by analyzing video images from different cameras with the allocation of the contours of the working fluid and synthesizing the video image of these contours from different cameras taking into account the location of these cameras relative to the working fluid,The introduction and removal of the measuring system from the measurement zone is carried out by tilting it using a hinged mount to a fixed telescopic stand, and the determination of the coordinates of the detected defect along the length of the working body is carried out by feedback with the production line.

[0021] The claimed group of inventions is illustrated by the following drawings.

[0022] Fig. 1 shows schematically the design of the device, where the following are depicted: 1-4 - light sources installed inside the horseshoe-shaped housing in a plane perpendicular to the direction of movement of the working fluid and directed at the working fluid; 5-8 - video cameras installed in pairs at right angles to each other on a frame inside the horseshoe-shaped housing in a plane perpendicular to the direction of movement of the working fluid and directed at the working fluid; 9 - horseshoe-shaped housing; 10 - working fluid; 11 - fastening ears, 12 - hinge; 13 - telescopic stand; 14 - laser Doppler velocimeter installed on a frame inside the horseshoe-shaped housing in a plane perpendicular to the direction of movement of the working fluid and directed at the working fluid; 15 - working area of ​​the system; a) front view, b) left view; c) right view.

[0023] Fig. 2 shows a block diagram of the device, which shows: light sources - 1-4, video cameras - 5-8, laser Doppler velocimeter - 14, switch - 16, computer - 17, switch for connection to the corporate network - 18, light and sound signaling unit - 19, unit for connection to the process line - 20, unit for connection to peripheral devices in the form of a printer, marker - 21, mouse and keyboard manipulator - 22, touch screen - 23.

[0024] Fig. 3 shows a geometric diagram that is used to calculate the dimensions of a working fluid with a circular cross-section, where X and Y are the directions of the absolute fixed coordinate system, M Х and M Y -video camera matrices, D X and D Y - diameters of the working fluid displayed on the matrices of video cameras M Х and M Y accordingly, X L and Y L - distances from the origin of the absolute coordinate system to the video camera matrices M Х and MY in the X and Y directions respectively, X С and Y C - distances from the origin of the absolute coordinate system to the center of the working fluid in the X and Y directions, respectively, D RX and D RY - actual diameters of the working fluid, measured in the X and Y directions, respectively.

[0025] Fig. 4 shows a block diagram describing the operation of the system.

[0026] Fig. 5 shows the result of the system’s work on searching for defects in the armor of an oil-submersible cable during its production on a production line: a) the external appearance of the installation; b) the result of the artificial neural network’s work, where the detected defects are marked with squares.

[0027] Fig. 6 shows the result of the operation of the system for searching for surface defects and marking a rubber hose during its production on a production line: a) the external appearance of the installation; b) the result of the operation of an artificial neural network, where the detected defects are marked with squares.

[0028] The claimed method and device are implemented as follows. The device for monitoring surface defects and geometry comprises: light sources - 1-4; video cameras - 5-8; horseshoe-shaped body - 9; fastening ears - 11, hinge - 12; telescopic stand - 13; laser Doppler velocimeter - 14; working area - 15 (Fig. 1). Telescopic stand 13 can raise and lower horseshoe-shaped body 9 in the vertical direction; hinge 12 allows to tilt horseshoe-shaped body 9 to the left from the area of ​​the working body 10; horseshoe-shaped body 9 is connected to telescopic stand 13 by means of fastening ears 11 through a bolted connection and hinge 12; arrows indicate the directions of rotation of hinge 12 and movement of telescopic stand 13. The working body 10 is passed through the working area 15.

[0029] Light sources 1-4, video cameras 5-8 and laser Doppler velocimeter 14 are mounted in the same plane perpendicular to the direction of motion of the working fluid on a frame inside the horseshoe-shaped housing 9 and directed at the working fluid 10 so that light sources 1-4 uniformly illuminate the working fluid 10, and cameras 5-8 completely capture the image of the working fluid 10. For the operation of light sources 1-4, video cameras 5-8 and laser Doppler velocimeter 14, openings are provided in the horseshoe-shaped housing, which are covered with transparent glass, so that the interior of the horseshoe-shaped housing and the light sources 1-4, video cameras 5-8 and laser Doppler velocimeter 14 installed therein do not have direct contact with the environment. Telescopic stand 13 is rigidly fixed to the floor by means of an anchor connection. A horseshoe-shaped body 9 is attached to the rack through ears 11, attached to the hinge 12 by means of a bolted connection.With the help of the telescopic stand 13, the device is positioned in such a way that the working fluid 10 passes approximately in the middle of the working zone 15. In this case, the system can be removed from the working fluid location area without removing it from the process line by lifting it upward with the help of the telescopic stand 13 and tilting it to the side with the help of the hinge 12.

[0030] Light sources 1-4, … N L , video cameras 5-8, … N C and a laser Doppler velocimeter 14 are connected to a switch 16, which transmits data to a computer 17. Connected to the computer are: a switch for connecting to a corporate network - 18, a light and sound signaling unit - 19, a unit for connecting to a process line - 20, a unit for connecting to peripheral devices in the form of a printer and marker - 21, a mouse and keyboard - 22, a touch screen - 23. Additional items included in the description: camera N C and light source N Lindicate the possibility of installing additional cameras and light sources in the system, the number of which is determined by the cross-sectional geometry of the specific working fluid and the characteristics of its surface. These cameras and light sources are installed in the same plane, perpendicular to the direction of movement of the working fluid, as the plane in which light sources 1-4 and video cameras 5-8 are installed, within the horseshoe-shaped housing so that they are directed at the working fluid (Fig. 2).

[0031] Light source intensity 1-4, … N L is installed by the system software and controlled by the computer on which the system software is installed, ensuring the execution of its functions. Images from cameras 5-8, ... N CThe data corresponding to the current moment t are sent via switch 16 to computer 17, where they are analyzed using software containing a trained artificial neural network to detect defects and calculate the geometric dimensions of the working fluid. The velocity of the working fluid through the working gap is measured using laser Doppler velocimeter 14, directed at the working fluid, and transmitted to the system software installed on the computer.The software transmits processed data to the corporate network via switch 18, controls audible and visual alarms 19 in the event of a defect being detected or the working fluid's dimensions exceeding the set values, and issues control commands to process line 20 depending on its type and purpose, such as commands to change the operating mode or a command to stop the line. It also receives data from process line 20, such as the working fluid's speed, process line parameters, etc. The software also controls printer and marking device 21 for printing labels or applying markings to the working fluid as it moves along the process line. The computer is controlled using a mouse, keyboard 22, and touchscreen display 23, which displays information about the software's operation, including the results of working fluid size measurements and surface defect monitoring.

[0032] The method using the claimed device is implemented as follows. The working fluid is moved through the working gap 15, while defects are manually photographed using video cameras 5-8. This allows the working fluid to be photographed using video cameras 5-8 and an artificial neural network to be trained to identify a given set of defects. At the same time, the system software automatically adjusts the working fluid illumination by controlling the operating modes of light sources 1-4 (Fig. 1). Simultaneously, the measuring part of the system is calibrated to determine the dimensions of a specific type of working fluid. For example, for a working fluid with a circular cross-section, the type of functions F is determined during the calibration process. X and F Y , connecting the following quantities: D RX =F X (X L , Y L , X C , Y c , D X , D Y ), D RY =F Y (X L , YL , X C , Y c , D x , D Y ) provided that the video camera matrices M X ,M Y are located at right angles parallel to the axes of the absolute coordinate system X, Y, respectively (Fig. 3).

[0033] After training the neural network and calibrating, the system is put into operation (Fig. 4). The first step, at time t, is to receive images from all video cameras. The second step is to obtain the velocity of the working body measured by a laser Doppler velocimeter. The third step is to calculate the current coordinate of the working body along its length using the formula: L(t)=L(t - Δt)+V D Δt, where L(t) is the coordinate of the working body along the length calculated by the system, corresponding to the moment of time t, Δt is the time step corresponding to the time step between measurements, V D- the velocity of the working fluid obtained from a laser Doppler velocimeter. The fourth step involves analyzing the images received from the video cameras and identifying surface defects of the working fluid using software based on a trained artificial neural network. The fifth step involves analyzing the images of the working fluid obtained from the video cameras at time t and calculating its cross-sectional dimensions.If a defect is detected or the dimensions of the workpiece exceed the set limits, the software displays an image of the defect, the dimensions of the workpiece, and the length coordinates of the workpiece, activates an audible and visual alarm, transmits a preset set of commands to the process line, and, depending on the scenario defined for the specific process line and process, either stops the process line or adjusts its operation to eliminate the detected defect or the workpiece dimensions exceed the set limits and proceeds to the first step. If a surface defect is detected by the software but is not in the defect database, the system prompts the operator to enter the defect into the defect database for classification and additional neural network training and continues with the next iteration of surface inspection and workpiece dimension measurement, i.e., to the first step.

[0034] The claimed method can be experimentally confirmed by the following examples.

[0035] Example 1.

[0036] Dynamic tests were conducted on the system to detect defects in oil-submersible cable during the armoring process on the production line.

[0037] This cable was passed through the system and analyzed for surface defects and armor geometry violations.

[0038] The system has identified areas on the cable section that it has classified as defects. These areas are outlined in Figure 5.

[0039] Example 2.

[0040] Dynamic tests were conducted on the system to detect defects in rubber hose during production on the production line.

[0041] This rubber hose was passed through the system and analyzed for surface defects and markings.

[0042] The system identified areas on the hose that it classified as defects; these areas are outlined in Figure 6.

[0043] The claimed method and device can find wide application in control and measuring systems for monitoring and controlling the technological process in the production of linear products by drawing, rolling, extrusion, twisting, layering during optical monitoring and control of the technological process of producing cable, fiber, rod, wire rod, pipe, optical fiber, wire, cable, twisted bundle, electrodes, linear products made of polymers, wood, wood-polymer composites, metals, rubber, glass, composite materials, etc.

[0044] List of reference designations

[0045] 1, 2, 3, 4 - light sources;

[0046] 5, 6, 7, 8 - video cameras:

[0047] 9 - horseshoe-shaped body;

[0048] 10 - working fluid;

[0049] 11 - fastening ears;

[0050] 12 - hinge;

[0051] 13 - telescopic stand;

[0052] 14 - Laser Doppler velocimeter;

[0053] 15 - system working area;

[0054] 16 - switch;

[0055] 17 - Computer;

[0056] 18 - switch for connecting to a corporate network:

[0057] 19 - light and sound signaling unit;

[0058] 20 - connection block with the process line;

[0059] 21 - connection unit with peripheral devices in the form of a printer, marker;

[0060] 22 - mouse and keyboard manipulator;

[0061] 23 - touch display.

Claims

1. A device for monitoring defects and surface geometry, characterized in that it contains a horseshoe-shaped body, in the center of which a cutout is made for moving the working fluid through it, a telescopic stand, a hinge, fastening ears for attaching the horseshoe-shaped body to the telescopic stand, light sources, video cameras, a laser Doppler velocimeter, fixed relative to each other on a frame inside the horseshoe-shaped body, a switch, a computer, a switch for connecting to a corporate network, a light and sound signaling unit, a connection unit with a process line, a connection unit with peripheral devices in the form of a printer and a marker, a mouse manipulator, a keyboard, a touch screen, wherein the horseshoe-shaped body on the telescopic stand is made with the possibility of its movement from the working area without removing the working fluid from the working area, and the light sources, video cameras,the laser Doppler velocimeter is rigidly mounted on a frame inside a horseshoe-shaped housing in a plane perpendicular to the direction of movement of the working fluid and directed at the working fluid, while the outputs of the light sources, video cameras, and the laser Doppler velocimeter are connected to a switch, and the switch is connected to a computer on which software with a trained artificial neural network is installed, and the computer itself has a connection to the switch for connection to a corporate network, a light and sound signaling unit, a connection unit with a process line, a connection unit with peripheral devices in the form of a printer and marker, a mouse manipulator, a keyboard, a touch screen in such a way that the working fluid is always located in the working area of ​​the light sources, video cameras, and the laser Doppler velocimeter.

2. The device according to paragraph 1, characterized in that the working body is made in the form of linear products manufactured by drawing, rolling, extrusion, twisting, layering in the form of cable, fiber, rod, wire rod, pipe, optical fiber, wire, cable, twisted bundle, electrodes, linear products made of polymers, wood, wood-polymer composites, metals, rubber, glass, composite materials.

3. The device according to paragraph 1, characterized in that the light sources, video cameras and laser Doppler velocimeter are mounted on a frame inside a horseshoe-shaped housing in a plane perpendicular to the direction of movement of the working fluid and are directed at the working fluid, and for their operation, openings protected by glass are made in the horseshoe-shaped housing.

4. The device according to paragraph 1, characterized in that the horseshoe-shaped body is connected to the telescopic stand by means of ears, a bolted connection and a hinge.

5. The device according to paragraph 1, characterized in that the telescopic stand is rigidly fixed to the floor using an anchor connection.

6. The device according to paragraph 1, characterized in that at least two light sources, at least two video cameras, and at least one laser Doppler velocimeter are installed on the frame inside the horseshoe-shaped housing in a plane perpendicular to the direction of movement of the working fluid, installed in such a way that the working fluid is always in their working zone.

7. The device according to paragraph 1, characterized in that the computer is equipped with software containing a trained artificial neural network, algorithms for determining the dimensions of the working body and controlling light sources.

8. A method for monitoring defects and surface geometry, characterized in that it is based on passing a working fluid through a working gap in a horseshoe-shaped housing, in which, in a plane perpendicular to the direction of movement of the working fluid, video cameras are mounted on a frame, directed at the working fluid, filming it and transmitting its image to a computer through a switch, light sources controlled by a computer, directed at the working fluid and illuminating it, a laser Doppler velocimeter directed at the working fluid, measuring and transmitting through a switch to a computer the speed of its movement, and when receiving images and the speed of movement of the working fluid, software installed on the computer, using a trained artificial neural network, determines the presence of defects on the surface of the working fluid,determines, using calculations and image analysis, the coordinates of defects along the length of the working body and the dimensions of the working body and controls the switch for connection to the corporate network, the light and sound signaling unit, the unit for connection to the process line, the unit for connection to peripheral devices in the form of a printer and marker, the mouse manipulator, the keyboard and the touch screen.

9. The method according to paragraph 8, characterized in that the location and size of the working fluid in the working gap are determined by analyzing pairs of images coming from pairs of video cameras located at right angles.

10. The method according to paragraph 8, characterized in that the measurement results are displayed in digital and graphic form on the touch screen of the computer.

11. The method according to paragraph 8, characterized in that the illumination of the working fluid in the working gap is regulated using software that controls light sources directed at the working fluid.

12. The method according to paragraph 8, characterized in that the images received from the video cameras are analyzed using a trained artificial neural network to identify defects in the surface of the working body.

13. The method according to paragraph 8, characterized in that the coordinate of the defect along the length of the working body is determined by software installed on the computer and calculating this value based on measurements of the speed of movement of the working body using a laser Doppler velocimeter installed in a horseshoe-shaped housing and directed at the working body.

14. The method according to paragraph 8, characterized in that when a defect is detected, the software installed on the computer turns on an audible and visual alarm, displays the defect and its coordinate along the length of the working body on a touch screen, and transmits an established set of commands to the process line.

15. The method according to paragraph 8, characterized in that when the dimensions of the working fluid go beyond the specified limits, the software installed on the computer turns on an audible and visual alarm, displays the working fluid and its coordinate along the length of the working fluid on a touch screen, and transmits a set of commands to the process line.

16. The method according to paragraph 8, characterized in that if there is no defect in the database of defects of the artificial neural network, the software sends an image of this defect to the database in automatic mode or offers to train the artificial neural network to recognize this defect in manual mode.

Images