Rod and wire rolling force measuring method and device based on laser vision

Abstract

The invention relates to a rod and wire rolling force measuring method and device based on laser vision, and the method comprises the following steps: simulating rolling of rods and wires with different shaft diameters by adopting different rolling forces; in the simulation process, multiple sets of oil pressure and rolling force data of oil film bearings used for supporting rolling rollers are collected, and a visual recognition method is used for collecting multiple sets of shaft diameter size data before and after rod and wire rolling; taking oil pressure and shaft diameter size data as input, taking rolling force data as output, and establishing a CNN-based parameter mapping neural network model; the multiple sets of oil pressure, shaft diameter size and rolling force data serve as a training set, a CNN-based parameter mapping neural network model is trained, and a network model used for predicting rolling force is obtained; in the actual rolling process, the oil pressure of the oil film bearing and the shaft diameter size of the rod and wire before and after rolling are monitored, and the rolling force is predicted through the network model used for predicting the rolling force.

Description

Technical Field

[0001] The invention relates to the technical field of rolling, and in particular to a method and a device for measuring the rolling force of a bar or wire rod based on laser vision. Background Art

[0002] The steel industry refers to the industry that produces pig iron, steel, steel products, industrial pure iron and ferroalloys. It is one of the basic industries of all industrialized countries in the world. In 2021, my country's crude steel output exceeded 1 billion tons, of which the output of bars and wires generally accounted for 40%-50%, and was widely used in bearings, springs, welding rods, mechanical equipment, building materials, etc. Bar and wire rolling equipment is a rolling mill specially used for rolling bars and wires. Due to the small cross-sectional size of the wire, the total elongation coefficient is large in the process of rolling from billet to finished product, and a large amount of hot rolling process equipment is required. It is generally composed of 21 to 28 continuous rolling mills, with a maximum production speed of 140m / s and a bar and wire temperature of nearly 900℃. During the rolling process, the vertical size of the rolling force will affect the quality of the finished bars and wires. Too small rolling force will lead to poor surface quality and uneven deformation of the finished product; too large rolling force will increase rolling waste, additional energy and material consumption. Therefore, rolling force monitoring during the rolling process is of great significance to ensure product quality and process adjustment.

[0003] Existing plate and strip rolling equipment can directly monitor the rolling force by installing a dedicated measurement system, such as the rolling force measurement system of ABB and KELK, which directly measures the rolling force by arranging pressure sensors between the bearing seat, frame and nut. However, most existing bar and wire rolling equipment is limited by the roll box structure and high-temperature and high-speed rolling conditions, making it difficult to directly measure the rolling force by arranging sensors. More and more researchers are paying attention to predicting the rolling force through model construction to achieve indirect measurement. The traditional rolling force prediction model calculates the rolling force through material theory combined with the key parameters of related sensors. However, since some parameters are difficult to obtain, there is a deviation between the calculated results and the actual results; the method of using a neural network model to predict the rolling force requires obtaining the measured rolling force during the network training process, which makes model construction difficult.

[0004] For example, a Chinese patent with publication number CN111400928B discloses a rolling force compensation method and device based on multivariate regression. The method is based on a rolling force calculation model and historical rolling data, and iteratively calculates the rolling force compensation coefficient through a multivariate linear regression model to achieve rolling force compensation;

[0005] The Chinese patent with publication number CN112588840B discloses a rolling force compensation method and system suitable for a cold rolling mill. The method calculates additional tension to compensate for the tension in the speed increase and decrease stages of the cold rolling mill by using the equipment parameters and rolling process parameters of the cold rolling mill. This type of method is generally applicable to the rolling process of plates and strips. The accuracy of the model also needs to be verified. The measuring device is difficult to adapt to the harsh environment of high speed and high temperature, and it is difficult to achieve accurate measurement of the rolling force of bars and wires. Summary of the invention

[0006] The technical problem to be solved by the present invention is to provide a method and device for measuring the rolling force of rods and wires based on laser vision in view of the above shortcomings. The present invention collects relevant data through a simulation test device, and constructs a model related to the rolling force based on the relevant data, thereby realizing indirect measurement of the rolling force of rods and wires.

[0007] In order to solve the above technical problems, the present invention adopts the following technical solutions:

[0008] A method for measuring rolling force of a bar or wire rod based on laser vision comprises the following steps:

[0009] Use different rolling forces to simulate rolling bars and wires with different shaft diameters;

[0010] During the simulation, multiple sets of oil pressure and rolling force data of the oil film bearings supporting the rolling rolls were collected, and multiple sets of shaft diameter dimension data before and after the rolling of the bars and wires were collected using the visual recognition method.

[0011] Taking the oil pressure and shaft diameter data as input and the rolling force data as output, a CNN-based parameter mapping neural network model is established;

[0012] Using multiple sets of oil pressure, shaft diameter and rolling force data as training sets, a CNN-based parameter mapping neural network model was trained to obtain a network model for predicting rolling force.

[0013] During the actual rolling process, the oil pressure of the oil film bearing and the shaft diameter size before and after the rolling of the rod and wire are monitored, and the rolling force is predicted by a network model used to predict the rolling force.

[0014] Furthermore, the process of collecting the shaft diameter dimensions of the rod and wire before and after rolling by visual recognition method includes calibration, image correction and light profile recognition.

[0015] Furthermore, the calibration and image correction include the following steps:

[0016] The intersection of the black and white grid on the calibration plate is used as the calibration reference point;

[0017] Using the geometric change matrix, the calibration reference points on the calibration plate are matched one-to-one with the calibration reference points in the image captured by the camera;

[0018] The geometric transformation matrix is ​​solved by using the above multiple groups of corresponding calibration reference points to obtain the geometric change matrix of camera calibration.

[0019] Furthermore, the light profile recognition includes the following steps:

[0020] A laser is used to illuminate a calibration plate, and the rods and wires before and after rolling are located between the laser and the calibration plate, so as to form a projection of the contours of the rods and wires before and after rolling on the calibration plate;

[0021] The calibration plate is photographed by a camera to collect the contour projection images of the rods and wires before and after rolling, and the contour dimensions of the rods and wires before and after rolling in the contour projection images are calculated using Hough transformation;

[0022] The contour dimensions are mapped through the geometric change matrix of camera calibration to obtain the actual contour dimensions, thereby obtaining the shaft diameter dimensions before and after rolling of bars and wires.

[0023] Furthermore, the CNN-based parameter mapping neural network model includes three ResNet structures, and the ResNet structure is composed of two convolutional layers and two pooling layers alternately combined.

[0024] A bar and wire rolling force measuring device based on laser vision comprises a rolling simulation unit, a visual recognition unit and a data acquisition unit.

[0025] The rolling simulation unit comprises a roll box, a gear box and a pressure mechanism, the roll box comprises a first roll shaft and a second roll shaft, the gear box is externally connected to a power unit, the gear box is used to transmit power to the first roll shaft, and the pressure mechanism is used to apply pressure to the second roll shaft toward the first roll shaft;

[0026] The visual recognition unit is used to collect the contour images of the rods and wires before and after rolling, and obtain the shaft diameter dimensions of the rods and wires before and after rolling according to the contour images;

[0027] The data acquisition unit is used to collect the oil pressure of the oil film bearing on the first roller shaft, and the data acquisition unit is also used to collect the pressure applied by the pressure mechanism to the second roller shaft.

[0028] Furthermore, the visual recognition unit includes a laser, a camera and a calibration plate, the laser is used to emit laser to the calibration plate, the rolled rods and wires are located between the laser and the calibration plate, and the camera is used to collect a contour image on the calibration plate.

[0029] Furthermore, the laser propagation direction of the laser is perpendicular to the calibration plate;

[0030] The camera obliquely photographs the calibration plate.

[0031] Furthermore, the data acquisition device includes an oil pressure sensor and a pressure sensor, the oil pressure sensor is used to detect the oil pressure of the oil film bearing, and the pressure sensor is used to monitor the pressure output by the pressurizing mechanism.

[0032] After adopting the above technical solution, the present invention has the following advantages compared with the prior art:

[0033] (1) The present invention collects the shaft diameter of the rod and wire before and after rolling by visual recognition, and this collection method is not affected by high-speed and high-temperature working conditions;

[0034] (2) After obtaining relevant data of the rolling process, the present invention realizes the acquisition of rolling force through a neural network model, which can be used for quality monitoring and process adjustment.

[0035] The present invention is described in detail below with reference to the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 The general layout diagram of the measuring device;

[0037] Figure 2 It is a partial structural diagram of the measuring device;

[0038] Figure 3 It is a data collection schematic diagram of the measuring device;

[0039] Figure 4 It is a structural diagram of the visual recognition unit;

[0040] Figure 5 This is the schematic diagram of the visual recognition unit;

[0041] Figure 6 Schematic diagram of the structure of the parameter mapping neural network model based on CNN.

[0042] In the accompanying drawings, the components represented by the reference numerals are listed as follows:

[0043] 1. Roller box; 11. First roller shaft; 12. Second roller shaft; 2. Gear box; 3. Pressurizing mechanism; 4. Oil film bearing; 51. Laser; 52. Camera; 53. Calibration plate; 61. Oil pressure sensor; 62. Pressure sensor; 7. Power unit; 8. Rods and wires. DETAILED DESCRIPTION

[0044] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0045] In the description of the present invention, it should be noted that the terms "center", "up", "down", "left", "right", "vertical", "horizontal", "inside", "outside", "clockwise" and "counterclockwise" etc. indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as a limitation on the present invention.

[0046] Embodiment 1:

[0047] A method for measuring rolling force of bars and wires based on laser vision comprises the following steps:

[0048] S1: Build a bar and wire rolling force measuring device based on laser vision, which includes a rolling simulation unit, a visual recognition unit and a data acquisition unit;

[0049] S2: A laser vision-based bar and wire rolling force measurement device is used to simulate the rolling process of bars and wires;

[0050] S3: During the simulation, the average oil pressure P of the upper oil film bearing 4 of the first roller 11 is collected by the data acquisition unit. 1m and the average oil pressure P of the lower oil film bearing 2m , and the rolling force F output by the pressurizing mechanism 3, and the shaft diameter dimension d1 before rolling and the shaft diameter dimension d2 after rolling of the rod and wire are collected by the visual recognition unit;

[0051] S4, repeat steps S2 to S3 several times, and change the rolling force F output by the pressure mechanism 3 and the axial diameter of the rods and wires to be rolled to obtain multiple groups of P 1m , P 2m , F, d1 and d2 data;

[0052] S5. 1m , P 2m , d1 and d2 as input, F as output, and establish a CNN-based parameter mapping neural network model;

[0053] S6, the multiple groups P obtained in step S4 1m , P 2m , F, d1 and d2 data are used as training sets to train the parameter mapping neural network model based on CNN, and a network model for predicting rolling force is obtained;

[0054] S7. During the actual rolling process, the input / output shaft diameters of the bars and wires and the oil pressure data of the upper / lower oil film bearings are monitored in real time, and the rolling force is predicted by a network model for predicting the rolling force.

[0055] like Figure 6 As shown in the figure, it is the structural diagram of the parameter mapping neural network model of CNN, where conv is the convolution operation, max_pooling is the pooling operation, concat is the tensor concatenation operation, batch_norm operation is the batch normalization operation, dense is the tensor direct connection operation, and fl atten is the expansion operation of converting a two-dimensional tensor to a one-dimensional array.

[0056] In this embodiment, the parameter mapping neural network model of CNN is mainly composed of three ResNet structures. The three ResNet structures use the idea of ​​hierarchical extraction. Each ResNet structure is responsible for information extraction at the corresponding level. The ResNet structure is formed by alternating two convolutional layers and two pooling layers, and is mainly used for hierarchical abstraction of data in the current ResNet structure.

[0057] The processing flow of the CNN parameter mapping neural network model includes the following steps:

[0058] (a1) combining the input data into a set of two-dimensional data and inputting the two-dimensional data;

[0059] (a2) abstracting the data in the current ResNet structure hierarchically through the ResNet structure, and performing tensor splicing of the top-level data and output data of the current ResNet structure;

[0060] (a3) Repeat step (2) three times, and perform a batch normalization operation between each ResNet structure to normalize and connect the previous and next ResNet structures;

[0061] (a4) After completing the feature extraction of the three ResNet structures, two batch normalization processes and tensor connection are performed, and then a two-dimensional tensor expansion process is performed and the result is output.

[0062] The process of collecting the shaft diameter dimension d1 before rolling and the shaft diameter dimension d2 after rolling of the rod and wire by the visual recognition unit includes calibration, image correction and light profile recognition, which specifically includes the following steps:

[0063] (b1) The intersection of the black and white grid on the calibration plate is used as the calibration reference point;

[0064] (b2) using the geometric change matrix, the calibration reference points on the calibration plate are matched one-to-one with the calibration reference points in the image captured by the camera;

[0065] (b3) using the corresponding sets of calibration reference points in step (b2) to solve the geometric transformation matrix to obtain the geometric change matrix of camera calibration;

[0066] (b4) using a laser to illuminate a calibration plate, with the rod and wire before and after rolling being located between the laser and the calibration plate, so as to form a contour projection of the rod and wire before and after rolling on the calibration plate;

[0067] (b5) using a camera to shoot the calibration plate, collecting the contour projection images of the rods and wires before and after rolling, and using Hough transformation to calculate the contour dimensions of the rods and wires before and after rolling in the contour projection images;

[0068] (b6) The contour dimensions are mapped through the geometric change matrix of camera calibration to obtain the actual contour dimensions, thereby obtaining the shaft diameter dimensions before and after the rod and wire rolling.

[0069] The average oil pressure P of the upper oil film bearing 4 1m and the average oil pressure P of the lower oil film bearing 2m In order to obtain the data, when setting the data acquisition unit, multiple oil pressure sensors 61 are evenly arranged on the oil film bearing. The multiple oil pressure sensors will measure and obtain multiple oil pressure data. The multiple oil pressure data are averaged, and the average value finally obtained is the average oil pressure of the corresponding oil film bearing.

[0070] Embodiment 2:

[0071] This embodiment describes a measuring device for implementing the method in the first embodiment.

[0072] A bar and wire rolling force measuring device based on laser vision comprises a rolling simulation unit, a visual recognition unit and a data acquisition unit, wherein:

[0073] The rolling simulation unit includes a roll box 1, a gear box 2 and a pressure mechanism 3. The roll box 1 includes a first roller 11 and a second roller 12. The gear box 2 is externally connected to a power unit 7. The power unit 7 is a motor. The gear box 2 is used to transmit power to the roll box 1, thereby driving the first roller 11 to rotate. The pressure mechanism 3 is used to apply pressure (i.e., rolling force) to the second roller 12 toward the first roller 11. The pressure mechanism 3 can be a hydraulic pressure mechanism. The rolled rod 8 is arranged between the first roller 11 and the second roller 12.

[0074] The visual recognition unit includes a laser 51, a camera 52 and a calibration plate 53. The calibration plate 53 is arranged on the roller box 1. A black and white checkerboard image is arranged on the calibration plate 53. The propagation direction of the laser emitted by the laser 51 is perpendicular to the surface of the calibration plate 53. The rolled rod 8 is located between the laser 51 and the calibration plate, so that the contour of the rod 8 is projected onto the calibration plate 53. The camera 52 is used to obliquely shoot the calibration plate 53 at a certain angle, so as to collect the contour projection image of the rod 8.

[0075] A set of visual recognition units is disposed on both sides of the first roller 11 and the second roller 12, that is, a set of visual recognition units is disposed at the input and output ends of the rods and wires, respectively, for collecting projection images before and after the rods and wires are rolled.

[0076] The data acquisition unit includes an oil pressure sensor 61 and a pressure sensor 62. A group of oil film bearings 4 and ball bearings are provided on the upper and lower parts of the first roller 11. Multiple oil pressure sensors 61 are evenly arranged on the oil film bearing 4. After collecting the data of the oil pressure sensor 61, the average value of all the oil pressure sensors 61 is taken; the pressure sensor 62 is arranged at the rear end of the pressurizing mechanism 3, and the pressure sensor 62 is used to measure the pressure output by the pressurizing mechanism 3 (i.e., rolling force).

[0077] It also includes a data acquisition card and an industrial computer. The oil pressure sensor 61, the pressure sensor 62 and the camera 52 are all connected to the data acquisition card, and the data acquisition card is connected to the industrial computer. The data acquisition card is used to transmit the oil pressure data of the oil film bearing measured by the oil pressure sensor 61, the rolling force data measured by the pressure sensor 62 and the contour projection image taken by the camera to the industrial computer, and then the industrial computer processes and stores the data.

[0078] The above is an example of the best implementation of the present invention, and the parts not described in detail are common knowledge of ordinary technicians in the field. The protection scope of the present invention shall be based on the content of the claims, and any equivalent transformation based on the technical enlightenment of the present invention is also within the protection scope of the present invention.

Claims

1. A method for measuring rolling force of bars and wires based on laser vision, characterized in that: The following steps are involved: Use different rolling forces to simulate rolling bars and wires with different shaft diameters; During the simulation, multiple sets of oil pressure and rolling force data of the oil film bearings supporting the rolling rolls were collected, and multiple sets of shaft diameter dimension data before and after the rolling of the bars and wires were collected using visual recognition methods. Taking the oil pressure and shaft diameter data as input and the rolling force data as output, a CNN-based parameter mapping neural network model is established; Using multiple sets of oil pressure, shaft diameter and rolling force data as training sets, a CNN-based parameter mapping neural network model was trained to obtain a network model for predicting rolling force. During the actual rolling process, the oil pressure of the oil film bearing and the shaft diameter size before and after the rolling of the rod and wire are monitored, and the rolling force is predicted by the network model used to predict the rolling force.

2. The method for measuring rolling force of bars and wires based on laser vision according to claim 1, characterized in that: The process of collecting the shaft diameter dimensions of bars and wires before and after rolling by visual recognition method includes calibration, image correction and light profile recognition.

3. The method for measuring rolling force of bars and wires based on laser vision according to claim 2, characterized in that: The calibration and image correction comprises the following steps: The intersection of the black and white grid on the calibration plate is used as the calibration reference point; Using the geometric change matrix, the calibration reference points on the calibration plate are matched one-to-one with the calibration reference points in the image captured by the camera; The geometric transformation matrix is ​​solved by using the above multiple groups of corresponding calibration reference points to obtain the geometric change matrix of camera calibration.

4. The method for measuring rolling force of bars and wires based on laser vision according to claim 3, characterized in that: The light profile recognition comprises the following steps: A laser is used to illuminate a calibration plate, and the rods and wires before and after rolling are located between the laser and the calibration plate, so as to form a projection of the contours of the rods and wires before and after rolling on the calibration plate; The calibration plate is photographed by a camera to collect the contour projection images of the rods and wires before and after rolling, and the contour dimensions of the rods and wires before and after rolling in the contour projection images are calculated using Hough transformation; The contour dimensions are mapped through the geometric change matrix of camera calibration to obtain the actual contour dimensions, thereby obtaining the shaft diameter dimensions before and after rolling of bars and wires.

5. The method for measuring rolling force of bars and wires based on laser vision according to claim 1, characterized in that: The CNN-based parameter mapping neural network model includes three ResNet structures, and the ResNet structure is composed of two convolutional layers and two pooling layers alternately combined.

6. A bar and wire rolling force measuring device based on laser vision, characterized in that: Including rolling simulation unit, visual recognition unit and data acquisition unit, The rolling simulation unit comprises a roller box (1), a gear box (2) and a pressure mechanism (3); the roller box (1) comprises a first roller shaft (11) and a second roller shaft (12); the gear box (2) is externally connected to a power unit (7); the gear box (2) is used to transmit power to the first roller shaft (11); and the pressure mechanism (3) is used to apply pressure to the second roller shaft (12) toward the first roller shaft (11); The visual recognition unit is used to collect the contour images of the rods and wires before and after rolling, and obtain the shaft diameter dimensions of the rods and wires before and after rolling according to the contour images; The data acquisition unit is used to collect the oil pressure of the oil film bearing (4) on the first roller (11), and the data acquisition unit is also used to collect the pressure applied by the pressure mechanism (3) to the second roller (12).

7. The laser vision-based bar and wire rolling force measuring device according to claim 1, characterized in that: The visual recognition unit comprises a laser (51), a camera (52) and a calibration plate (53); the laser (51) is used to emit laser light to the calibration plate (53); the rolled rod or wire is located between the laser (51) and the calibration plate (53); and the camera (52) is used to collect a contour image on the calibration plate (53).

8. The laser vision-based bar and wire rolling force measuring device according to claim 1, characterized in that: The laser propagation direction of the laser (51) is perpendicular to the calibration plate (53); The camera (52) obliquely photographs the calibration plate (53).

9. The laser vision-based bar and wire rolling force measuring device according to claim 1, characterized in that: The data acquisition device comprises an oil pressure sensor (61) and a pressure sensor (62), wherein the oil pressure sensor (61) is used to detect the oil pressure of the oil film bearing (4), and the pressure sensor (62) is used to monitor the pressure output by the pressurizing mechanism (3).

Images