A method for on-line adjustment of rolling mill roll gap based on laser radar

Abstract

This invention discloses an online adjustment method for mill roll gap based on lidar. It includes at least the following steps: Step S400: Real-time slicing based on the normal plane of the discharge direction to obtain end-face point cloud data; Step S500: Extracting adjacent multi-frame point cloud slices and fitting the sliced ​​point clouds into feature circles or feature polygons using orthogonal distance regression; Step S600: The control center calculates the difference between the target size of the rolled material and the size of the feature shape. If it deviates from a preset range, proceed to step S700; if it is within the preset range, proceed to step S800; Step S700: Start the hydraulic motor and adjust the mill roll gap according to the offset value. Once the rolled material size returns to the preset range, proceed to step S800; Step S800: Brake the output shaft of the hydraulic motor to eliminate over-adjustment caused by rotational inertia and roll jump during operation. This invention obtains the mill roll gap adjustment direction by comparing the preset size with its actual processing size deviation, significantly improving the efficiency of roll gap adjustment.

Description

Technical Field

[0001] This invention relates to a method for online adjustment of rolling mill roll gap based on lidar. Background Technology

[0002] The rolling mill's reduction mechanism is typically driven by a hydraulic motor, which, through a worm gear and gear transmission, drives four gear sleeves to rotate simultaneously. These gear sleeves then rotate the mill's tie rod, which, via a threaded connection, moves the adjusting nut, thereby causing the rolls to move up and down to adjust the mill's roll gap.

[0003] With the technological development in the steel smelting field, traditional roll gap adjustment requires production to be stopped and manual adjustment is required. The adjustment process is cumbersome and the labor cost is high. The efficiency of traditional roll gap adjustment is insufficient.

[0004] Currently, some rolling mills are equipped with automated gauge control (AGC) systems. These systems use encoders and other sensors to measure the roll gap based on the hydraulic bearing angle, thus enabling continuous feedforward control of the roll gap. However, depending on the mill model, the speed ratio of the reduction device is generally between 40 and 80, resulting in very low measurement accuracy on the low-speed tie rod shaft. Furthermore, the roll gap is affected by multiple factors such as temperature and material properties during mill operation, thus current automated control systems have limitations. Summary of the Invention

[0005] To overcome the above-mentioned shortcomings, the purpose of this invention is to provide a high-precision, high-efficiency, and low-cost online adjustment method and system for mill roll gap based on lidar.

[0006] To achieve the above objectives, the online adjustment method for mill roll gap based on lidar of the present invention includes at least the following steps:

[0007] Step S400: Slice the surface in real time according to the normal plane of the discharge direction to obtain the end face point cloud data;

[0008] Step S500: Extract point cloud slices from adjacent multi-frames and fit the sliced ​​point cloud into feature circles or feature polygons using orthogonal distance regression.

[0009] Step S600: The control center calculates the difference between the target size of the rolled material and the size of the feature graphic. If it deviates from the preset range, proceed to step S700; if it is within the preset range, proceed to step S800.

[0010] Step S700: Start the hydraulic motor and adjust the mill roll gap according to the offset value. Once the rolled material size returns to the preset range, proceed to step S800.

[0011] Step S800: Brake the output shaft of the hydraulic motor to eliminate over-adjustment caused by rotational inertia and roll jump during operation.

[0012] Furthermore, the following steps are included before step S400:

[0013] Step S100: Install more than two lidar sensors on both sides of the discharge roller conveyor;

[0014] Step S200: Initialize the lidar and calibrate its extrinsic parameters through matrix transformation;

[0015] Step S300: Perform point cloud segmentation and downsampling processing on the radar point cloud data;

[0016] Further, step S200 includes:

[0017] Step S201: Synchronize the data timestamps of the reference radar and the radar to be calibrated;

[0018] Step S202: Match the point cloud data of the two above. Through IPC registration, namely geometric structure consistency constraint and nearest neighbor matching algorithm, the matching error is reduced by iterating through multiple frames of data.

[0019] Step S203: Until the algorithm converges and satisfies the rigidity and invariance property of the calibration matrix (six parallel lines), the final calibration matrix is ​​calculated using the consensus algorithm.

[0020] Further, step S300 includes:

[0021] Step S301: Determine the lidar angle, remove the point cloud from the non-rolled material direction of the two lidars, and segment the point cloud according to the angle;

[0022] Step S302: Divide the entire point cloud into voxels, and then calculate the centroid of the non-empty voxel to replace all points within the voxel, thereby achieving downsampling of the point cloud and reducing the amount of calculation by the controller. To ensure that too much rolling material surface parameter data is not lost, downsampling should be minimized.

[0023] Step S303: Confirm the point cloud coordinates of fixed objects such as the rolling mill, the ground, and the walls;

[0024] Step S304: Since the position of the lidar is fixed, point cloud segmentation can be performed based on relative coordinates to remove irrelevant objects.

[0025] Further, step S302 includes:

[0026] The formula for downsampling using physical fitness and mental techniques is as follows:

[0027]

[0028] Where: V is a known non-empty voxel containing m points, P centroid (xcentroid ,y centroid ,z centroid ) represents the centroid coordinates of the voxel.

[0029] Further, step S400 includes:

[0030] Step S401: Cluster the point cloud of rolled material using the KMeans algorithm;

[0031] Step S402: Approximate the rolled material as an irregular long strip object, and use the maximum inscribed sphere sequence method to find a series of maximum inscribed spheres inside the object through iteration. The line connecting the centers of these spheres is the center line.

[0032] Step S403: For non-circular rolled materials, the ridge-first tracing method is adopted. The distance from the points on the ridge line to all points on the model boundary is equal, and the continuous path on the ridge line is the center line of the object.

[0033] Step S404: Select an appropriate interval and cut slices from the center line normal plane.

[0034] Further, step S401 includes:

[0035] Point cloud clustering is performed using the KMeans algorithm, minimizing the loss function J. The loss function formula is as follows:

[0036]

[0037] Where: the three-dimensional dataset x1, x2, ..., x N ,x n ∈R D Cluster center μ k ,k=1,...,k,k represents the number of clusters, which needs to be set manually in KMeans; binary variable γ nk ∈{0,1}, where 1 indicates that a point in three dimensions belongs to class k, and 0 does not belong to class k.

[0038] To ensure that J is minimized, the EM algorithm is required, which consists of E-steps and M-steps, as shown in the following formula;

[0039]

[0040] In step E, μ is fixed. k By caring about r nk To reduce J. Because the 3D point cloud data are independent of each other, we can optimize for each point n individually, and r nk The value can only be 0 or 1. Therefore, only n needs to be allocated. th The data points to the nearest cluster center, so ||x n -μ kThe minimum value is 2.

[0041] In step M, r is fixed. nk J is a variable related to μ k If J is a quadratic function, then the minimum value can be found by differentiation. k Solve for the first derivative:

[0042]

[0043] therefore:

[0044]

[0045] At this time, through μ k The formula shows that the center point of a class is equal to the average value of the data points belonging to that class.

[0046] Further, step S500 includes:

[0047] Step S501: To reduce random errors, multiple frames of slice data should be extracted simultaneously;

[0048] Step S502: Combine and project the multi-frame point cloud data along the centerline;

[0049] Step S503: Perform polygon fitting based on the target rolled material shape and point cloud coordinate data;

[0050] Step S504: Calculate the rolled material dimensions using the point cloud coordinates of the target polygon;

[0051] Furthermore, the specific method for fitting the polygon in step S503 is as follows:

[0052] Taking the fitting of the circumscribed polygon of a common circular rolled material as an example, the formula is as follows:

[0053] (xa) 2 +(yb) 2 =r 2

[0054] Where (a,b) are the coordinates of the center of the circle, and r is the radius of the circle.

[0055] To find the parameters a, b, r that minimize the sum of the squared distances from all sample points to the center of the circle, the following is an example:

[0056]

[0057] Then, taking the partial derivatives with respect to a, b, and r, and setting them equal to zero, we obtain the following equation:

[0058]

[0059]

[0060] The center of the circle (a, b) is estimated using the mean of all point coordinates, and then the radius is estimated based on the distance of each point to this initial center.

[0061]

[0062] Alternatively, numerical optimization methods can be used to solve the above nonlinear equations to obtain more accurate fitting results based on actual production requirements.

[0063] Further, step S600 includes:

[0064] Step S601: Calculate the actual dimensions of the rolled material using the previous step;

[0065] Step S602: Set the target polygon size and warning size range for the rolled material;

[0066] Step S603: Compare the difference between the target polygon size and the actual dimensions of the rolled material in this process;

[0067] Step S604: If the rolled material size deviates from the boundary size, proceed to step S700; otherwise, proceed to step S800.

[0068] Further, step S700 includes:

[0069] Step S701: When the rolled material size is greater than the warning value, a feedback signal is sent to the control center to reduce the roll gap;

[0070] Step S702: When the rolled material size is less than the warning value, a feedback signal is sent to the control center to increase the roll gap;

[0071] Step S703: When the rolled material size returns to the warning range, proceed to step S800;

[0072] Step S800 includes:

[0073] Step S801: Proceed to this step when the above steps have been completed or the rolled material dimensions have remained within the warning range;

[0074] Step S802: Send a signal through the control center to lock the output shaft of the hydraulic motor.

[0075] To achieve the above objectives, the present invention provides a laser radar-based online adjustment system for mill roll gap, comprising a laser radar, a computing control center, and a hydraulic motor with braking function.

[0076] The lidar is installed at the exit of the rolling mill roller table, away from the rolling mill body and the rolled material, while the computing and control center is installed at the rolling mill. Both are equipped with heat insulation and buffering measures to reduce the impact of temperature vibration on the sensor and control center.

[0077] The lidar and hydraulic motor are both connected to the computing and control center.

[0078] The present invention has the following advantages:

[0079] (1) This invention uses pre-set multiple lidar sensors for external parameter matching to ensure data consistency, thereby accurately acquiring data and determining the size of the rolled material. A high-precision three-dimensional model and size of the rolled material are obtained using KMeans point cloud clustering. The actual processing size of the rolled material is obtained by fitting the characteristic polygons of the rolled material, and the size of the rolled material is monitored by setting a pre-set size warning line. The deviation between the pre-set size and the actual processing size is then compared to obtain the direction of the mill roll gap adjustment, and continuous monitoring continues until the rolled material size returns to the standard. Compared with traditional manual adjustment or feedforward automatic control methods, this method uses automatic feedback control, which does not control the roll gap size but monitors the actual production size. This ignores the influence of environmental factors such as temperature and vibration on the roll gap, ensuring the accuracy and robustness of the monitoring system, and significantly improving the efficiency of roll gap adjustment, thereby increasing production efficiency and reducing costs.

[0080] (2) This invention adopts a control method and system based on radar sensors. Apart from the sensors and controllers, no additional rolling mill hardware is required, which can facilitate and quickly improve traditional rolling mills. At the same time, the algorithm also has the characteristics of high precision and high efficiency. Overall, it is low in cost and good in effect, and is suitable for widespread use. Attached Figure Description

[0081] Figure 1 This is a flowchart illustrating the automatic roll gap control method of the present invention.

[0082] Figure 2 This is a schematic diagram illustrating the principle of sensor extrinsic parameter calibration in this invention;

[0083] Figure 3 This is a schematic diagram of the basic structure of the traveling device in this invention.

[0084] Attached diagrams: 1. LiDAR; 2. Hydraulic motor; 3. Simplified diagram of rolling mill; 4. Computational control center. Detailed Implementation

[0085] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0086] In the description of this invention, it should be understood that the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0087] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0088] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "joining" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0089] As shown in the figure, the online adjustment system for mill roll gap based on lidar of the present invention includes lidar 1, calculation and control center 4, and hydraulic motor 2 with braking function;

[0090] The lidar 1 is installed at the exit of the rolling mill roller table, away from the rolling mill body 3 and the rolling material area, and the computing and control center 4 is installed away from the rolling mill. Both are equipped with heat insulation and buffer measures to reduce the impact of temperature vibration on the sensors and the control center. The lidar and hydraulic motor are both connected to the computing and control center.

[0091] Example 1

[0092] As shown in the figure, the online adjustment method for rolling mill roll gap based on lidar in this embodiment includes the following steps:

[0093] Step S100: Install more than two lidar sensors on both sides of the discharge roller conveyor. To prevent sensor interference, they should be installed in a staggered manner.

[0094] Step S200: Initialize the lidar and calibrate its extrinsic parameters through matrix transformation;

[0095] Step S300: Perform point cloud segmentation and downsampling processing on the radar point cloud data;

[0096] Step S400: Cluster and identify the point cloud data of the rolled material, and slice it according to the normal plane of the discharge direction to obtain the end face point cloud data;

[0097] Step S500: Extract point cloud slices from adjacent multi-frames and fit the sliced ​​point cloud into feature circles or feature polygons using orthogonal distance regression.

[0098] Step S600: The control center calculates the difference between the target size of the rolled material and the size of the feature graphic. If it deviates from the preset range, proceed to step S700; if it is within the preset range, proceed to step S800.

[0099] Step S700: Start the hydraulic motor and adjust the mill roll gap according to the offset value. If the rolled material size returns to the preset range, proceed to step S800.

[0100] Step S800: Brake the output shaft of the hydraulic motor to eliminate over-adjustment caused by rotational inertia and roll jump during operation.

[0101] The above technical solution determines the parameters and relative positions of the lidar, performs point cloud clustering on the rolled material within a set range, and determines its size based on its appearance parameters. The difference between the target size and the finished rolled material is calculated and fed back to the calculation center. Once it is determined that the rolled material size deviates from the predetermined size range, the hydraulic motor can be adjusted according to the difference to restore the rolled material size to the target range. Simultaneously, a hydraulic motor braking mechanism is added. When the rolled material size is within the target range, the hydraulic motor is braked to reduce the impact of the motor's rotational inertia on the system.

[0102] Example 2

[0103] As a further improvement to the above embodiment, step S200 includes:

[0104] Step S201: Synchronize the data timestamps of the reference radar and the radar to be calibrated;

[0105] Step S202: Match the point cloud data of the two above. Through IPC registration, namely geometric structure consistency constraint and nearest neighbor matching algorithm, the matching error is reduced by iterating through multiple frames of data.

[0106] Step S203: Until the algorithm converges and satisfies the rigidity and invariance property of the calibration matrix (six parallel lines), the final calibration matrix is ​​calculated using the consensus algorithm.

[0107] Through the above technical solution and algorithm, the point cloud data of dual lidar can be fused to obtain a new point cloud set, thereby ensuring the accuracy of subsequent rolling material size detection and reducing errors.

[0108] Example 3

[0109] As a further improvement to the above embodiments, step S300 includes:

[0110] Step S301: Determine the lidar angle, remove the point cloud from the non-rolled material direction of the two lidars, and segment the point cloud according to the angle;

[0111] Step S302: Divide the entire point cloud into voxels, and then calculate the centroid of the non-empty voxel to replace all points within the voxel, thereby achieving downsampling of the point cloud and reducing the amount of calculation by the controller. To ensure that too much rolling material surface parameter data is not lost, downsampling should be minimized.

[0112] Step S303: Confirm the point cloud coordinates of fixed objects such as the rolling mill, the ground, and the walls;

[0113] Step S304: Since the position of the lidar is fixed, point cloud segmentation can be performed based on relative coordinates to remove irrelevant objects;

[0114] By employing the above technical solution, redundant and useless data is removed, reducing the impact of environmental point clouds on the rolled material. Subsequent steps should only acquire the target rolled material point cloud data for calculation. This reduces the computational load on the central computing unit, improving the processing speed and efficiency of subsequent steps.

[0115] Example 4

[0116] As a further improvement to the above embodiment, step S302 includes:

[0117] The formula for downsampling using physical fitness and mental techniques is as follows:

[0118]

[0119] Where: V is a known non-empty voxel containing m points, P centroid (x centroid ,y centroid ,z centroid () represents the coordinates of the centroid of the voxel:

[0120] Example 5

[0121] As a further improvement to the above embodiment, step S400 includes:

[0122] Step S401: Cluster the point cloud of rolled material using the KMeans algorithm;

[0123] Step S402: The rolled material can be approximated as an irregular long strip. The maximum inscribed sphere sequence method is used to find a series of maximum inscribed spheres inside the object through iteration. The line connecting the centers of these spheres is the center line.

[0124] Step S403: For non-circular rolled materials, the ridge-first tracing method is adopted. The distance from the points on the ridge line to all points on the model boundary is equal, and the continuous path on the ridge line is the center line of the object.

[0125] Step S404: Select an appropriate interval and cut slices from the surface normal to the center line;

[0126] Using the above technical solution, point cloud clustering algorithms can further segment the point cloud to obtain a high-precision 3D point cloud map of the rolled material, thus providing information on the shape and size of the rolled material. Furthermore, the centerline of the rolled material is obtained through centerline extraction, leading to more accurate measurements of the rolled material slice shape and size.

[0127] Example 6

[0128] As a further improvement to the above embodiment, step S401 includes:

[0129] Point cloud clustering is performed using the KMeans algorithm, minimizing the loss function J. The loss function formula is as follows:

[0130]

[0131] Where: the three-dimensional dataset x1, x2, ..., x N ,x n ∈R D Cluster center μ k ,k=1,…,k,k represents the number of clusters, which needs to be set manually in KMeans; binary variable γ nk ∈{0,1}, where 1 indicates that a point in three dimensions belongs to class k, and 0 does not belong to class k.

[0132] To ensure that J is minimized, the EM algorithm is required, which consists of E-steps and M-steps, as shown in the following formula;

[0133]

[0134] In step E, μ is fixed. k By caring about r nk To reduce J. Because the 3D point cloud data are independent of each other, we can optimize for each point n individually, and r nk The value can only be 0 or 1. Therefore, only n needs to be allocated. th The data points to the nearest cluster center, so ||x n -μ k The minimum value is 2.

[0135] In step M, r is fixed. nk J is a variable related to μ k If J is a quadratic function, then the minimum value can be found by differentiation.k Solve for the first derivative:

[0136]

[0137] therefore:

[0138]

[0139] At this time, through μ k The formula shows that the center point of a class is equal to the average value of the data points belonging to that class.

[0140] The steps can be summarized as follows: a K is given manually. Here is a trick about the starting center point: any K points in the original data can be selected as the starting center point to speed up the algorithm.

[0141] Step E: To determine which class each point belongs to, given n points and k centers, we need to find the distance from each of the n points to the k centers.

[0142] M-step: After knowing that each point has been assigned to a class, update the position of the center point of that class;

[0143] Iterate through the E-step and M-step until the algorithm converges.

[0144] Example 7

[0145] As a further improvement to the above embodiment, step S500 includes:

[0146] Step S501: To reduce random errors, multiple frames of slice data should be extracted simultaneously;

[0147] Step S502: Combine and project the multi-frame point cloud data along the centerline;

[0148] Step S503: Perform polygon fitting based on the target rolled material shape and point cloud coordinate data;

[0149] Step S504: Calculate the rolled material dimensions using the point cloud coordinates of the target polygon;

[0150] By employing the above technical solution, precise outer contour coordinates of the rolled material can be obtained through slicing and projection overlay. Then, the target rolled material shape is further fitted into a target feature polygon to obtain the accurate true size of the rolled material, ensuring detection accuracy and efficiency.

[0151] Example 8

[0152] As a further improvement to the above embodiment, the specific method for fitting the polygon in step S503 is as follows:

[0153] Taking the fitting of the circumscribed polygon of a common circular rolled material as an example, the formula is as follows:

[0154] (xa) 2 +(yb) 2 =r 2

[0155] Where (a,b) are the coordinates of the center of the circle, and r is the radius of the circle.

[0156] To find the parameters a, b, r that minimize the sum of the squared distances from all sample points to the center of the circle, the following is an example:

[0157]

[0158] Then, taking the partial derivatives with respect to a, b, and r, and setting them equal to zero, we obtain the following equation:

[0159]

[0160] The center of the circle (a, b) is estimated using the mean of all point coordinates, and then the radius is estimated based on the distance of each point to this initial center.

[0161]

[0162]

[0163] Alternatively, numerical optimization methods can be used to solve the above nonlinear equations to obtain more accurate fitting results based on actual production requirements.

[0164] Example 9

[0165] As a further improvement to the above embodiments, step S600 includes:

[0166] Step S601: Calculate the actual dimensions of the rolled material using the previous step;

[0167] Step S602: Set the target polygon size and warning size range for the rolled material;

[0168] Step S603: Compare the difference between the target polygon size and the actual dimensions of the rolled material in this process;

[0169] Step S604: If the rolled material size deviates from the boundary size, proceed to step S700; otherwise, proceed to step S800.

[0170] The above technical solution uses a method of setting a warning range for the dimensions of rolled materials to determine the dimensional deviations in rolled materials processing, provide early warnings for production, and enable rapid adjustments to ensure the dimensional accuracy of rolled material production.

[0171] Example 10

[0172] As a further improvement to the above embodiments, step S700 includes:

[0173] Step S701: When the rolled material size is greater than the warning value, a feedback signal is sent to the control center to reduce the roll gap;

[0174] Step S702: When the rolled material size is less than the warning value, a feedback signal is sent to the control center to increase the roll gap;

[0175] Step S703: When the rolled material size returns to the warning range, proceed to step S800;

[0176] The above technical solution can adjust the dimensions of the rolled material in subsequent processing, so that incorrect production dimensions can be restored to correct ones. Since a warning line range smaller than the tolerance can be set, the rolled material can still be guaranteed to be within the tolerance range, ensuring the quality of product processing.

[0177] Example 11

[0178] As a further improvement to the above embodiments, step S800 includes:

[0179] Step S801: Proceed to this step when the above steps have been completed or the rolled material dimensions have remained within the warning range;

[0180] Step S802: Send a signal through the control center to lock the output shaft of the hydraulic motor;

[0181] The above technical solutions can reduce over-adjustment caused by rotational inertia or vibration during production, thus ensuring the accuracy of rolled material production.

[0182] The present invention has been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the embodiments described above. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention. Many other changes and modifications made without departing from the concept and scope of the present invention should be considered within the scope of protection of the present invention.

[0183] In the description of this specification, specific features, structures, materials, or characteristics may be combined in any suitable manner in one or more embodiments or examples.

[0184] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for online adjustment of rolling mill roll gap based on lidar, characterized in that: It should include at least the following steps: Step S400: Slice the surface in real time according to the normal plane of the discharge direction to obtain the end face point cloud data; Step S500: Extract point cloud slices from adjacent multi-frames and fit the sliced ​​point cloud into feature circles or feature polygons using orthogonal distance regression. Step S600: The control center calculates the difference between the target size of the rolled material and the size of the feature graphic. If it deviates from the preset range, proceed to step S700; if it is within the preset range, proceed to step S800. Step S700: Start the hydraulic motor and adjust the mill roll gap according to the offset value. Once the rolled material size returns to the preset range, proceed to step S800. Step S800: Brake the output shaft of the hydraulic motor to eliminate over-adjustment caused by rotational inertia and roll jump during operation; the following steps are included before step S400: Step S100: Install more than two lidar sensors on both sides of the discharge roller conveyor; Step S200: Initialize the lidar and calibrate its extrinsic parameters through matrix transformation; Step S300: Perform point cloud segmentation and downsampling processing on the radar point cloud data; Step S300 includes: Step S301: Determine the lidar angle, remove the point cloud from the non-rolled material direction of the two lidars, and segment the point cloud according to the angle; Step S302: Divide the entire point cloud into voxels, and then calculate the centroid of the non-empty voxel to replace all points within the voxel, thereby achieving downsampling of the point cloud and reducing the amount of calculation by the controller. To ensure that too much rolling material surface parameter data is not lost, downsampling should be minimized. Step S303: Confirm the point cloud coordinates of fixed objects such as the rolling mill, the ground, and the walls; Step S304: Since the position of the lidar is fixed, point cloud segmentation can be performed based on relative coordinates to remove irrelevant objects.

2. The online adjustment method for mill roll gap based on lidar as described in claim 1, characterized in that: Step S200 includes: Step S201: Synchronize the data timestamps of the reference radar and the radar to be calibrated; Step S202: Match the point cloud data of the two above. Through IPC registration, namely geometric structure consistency constraint and nearest neighbor matching algorithm, the matching error is reduced by iterating through multiple frames of data. Step S203: Until the algorithm converges and satisfies the rigidity and invariance property of the calibration matrix (six parallel lines), the final calibration matrix is ​​calculated using the consensus algorithm.

3. The online adjustment method for mill roll gap based on lidar as described in claim 1, characterized in that: Step S302 includes: The formula for downsampling using physical fitness and mental techniques is as follows: Where: V is a known non-empty voxel containing m points, P centroid (x centroid ,y centroid ,z centroid ) represents the centroid coordinates of the voxel.

4. The online adjustment method for mill roll gap based on lidar as described in claim 1, characterized in that: Step S400 includes: Step S401: Cluster the point cloud of rolled material using the KMeans algorithm; Step S402: Approximate the rolled material as an irregular long strip object, and use the maximum inscribed sphere sequence method to find a series of maximum inscribed spheres inside the object through iteration. The line connecting the centers of these spheres is the center line. Step S403: For non-circular rolled materials, the ridge-first tracing method is adopted. The distance from the points on the ridge line to all points on the model boundary is equal, and the continuous path on the ridge line is the center line of the object. Step S404: Select an appropriate interval and cut slices from the center line normal plane.

5. The online adjustment method for mill roll gap based on lidar as described in claim 4, characterized in that: Step S401 includes: Point cloud clustering is performed using the KMeans algorithm, minimizing the loss function J. The loss function formula is as follows: Where: a three-dimensional dataset x1, x2, ..., x N ,x n ∈R D Cluster center μ k ,k=1,...,k,k represents the number of clusters, which needs to be set manually in KMeans; binary variable γ nk ∈{0,1}, where 1 indicates that a point in three dimensions belongs to class k, and 0 does not belong to class k; To ensure that J is minimized, the EM algorithm is required, which consists of E-steps and M-steps, as shown in the following formula; In step E, μ is fixed. k By caring about r nk To reduce J; because the 3D point cloud data are independent of each other, we can optimize for each point n individually, and r nk The values ​​can only be 0 or 1; therefore, only n needs to be allocated. th The data points to the nearest cluster center, so ||x n -μ k The minimum value is 2. In step M, r is fixed. nk J is a variable related to μ k If J is a quadratic function, then the minimum value can be found by differentiation; for J with respect to μ k Solve for the first derivative: therefore: At this time, through μ k The formula shows that the center point of a class is equal to the average value of the data points belonging to that class.

6. The online adjustment method for mill roll gap based on lidar as described in claim 1, characterized in that: Step S500 includes: Step S501: To reduce random errors, multiple frames of slice data should be extracted simultaneously; Step S502: Combine and project the multi-frame point cloud data along the centerline; Step S503: Perform polygon fitting based on the target rolled material shape and point cloud coordinate data; Step S504: Calculate the rolled material dimensions using the point cloud coordinates of the target polygon.

7. The online adjustment method for mill roll gap based on lidar as described in claim 6, characterized in that: The specific method for polygon fitting in step S503 is as follows: Taking the fitting of the circumscribed polygon of a common circular rolled material as an example, the formula is as follows: (x-a) 2 +(y-b) 2 =r 2 Where (a,b) are the coordinates of the center of the circle, and r is the radius of the circle; To find the parameters a, b, r that minimize the sum of the squared distances from all sample points to the center of the circle, the following is an example: Then, taking the partial derivatives with respect to a, b, and r, and setting them equal to zero, we obtain the following equation: The center of the circle (a, b) is estimated using the mean of all point coordinates, and then the radius is estimated based on the distance of each point to this initial center. Alternatively, numerical optimization methods can be used to solve the above nonlinear equations to obtain more accurate fitting results based on actual production requirements.

8. The online adjustment method for mill roll gap based on lidar as described in claim 1, characterized in that: Step S600 includes: Step S601: Calculate the actual dimensions of the rolled material using the previous step; Step S602: Set the target polygon size and warning size range for the rolled material; Step S603: Compare the difference between the target polygon size and the actual dimensions of the rolled material in this process; Step S604: If the rolled material size deviates from the boundary size, proceed to step S700; otherwise, proceed to step S800.

9. The online adjustment method for mill roll gap based on lidar as described in claim 1, characterized in that: Step S700 includes: Step S701: When the rolled material size is greater than the warning value, a feedback signal is sent to the control center to reduce the roll gap; Step S702: When the rolled material size is less than the warning value, a feedback signal is sent to the control center to increase the roll gap; Step S703: When the rolled material size returns to the warning range, proceed to step S800; Step S800 includes: Step S801: Proceed to this step when the above steps have been completed or the rolled material dimensions have remained within the warning range; Step S802: Send a signal through the control center to lock the output shaft of the hydraulic motor.

Images