A method, device and system for measuring the thickness of thin plates in a rolling mill

Abstract

The present invention relates to the technical field of thickness measurement, and in particular to a method, device and system for measuring the thickness of a thin plate of a rolling mill. The device comprises two linear motors and a distance sensor above and below the thin plate, and obtains the observed thickness at each position according to the data output by the distance sensor, obtains the vibration characteristics according to the observed thickness, filters the observed thickness according to the vibration characteristics and obtains the thickness measurement error; at the same time, obtains the position error of the linear motor, decomposes the thickness measurement error into a motion interference amount F1 and a measurement interference amount F2 according to the position error, and obtains the measured thickness according to F1 and F2. The present invention further improves the accuracy of thickness measurement during the hot rolling process of the thin plate.

Description

Technical Field

[0001] The invention relates to the technical field of thickness measurement, and in particular to a method, device and system for measuring the thickness of a thin plate of a rolling mill. Background Art

[0002] The mill thin plate is a kind of smelting plate. The slab is hot rolled to the required thickness by the rolling mill. The thickness range is generally specified as 1.2 to 6 mm. In order to ensure that the thickness of the thin plate meets the requirements, it is necessary to use a measuring instrument to measure its thickness. The existing measuring instruments or measuring methods have a measurement accuracy that fully meets the requirements (for example, the accuracy can reach 0.2 mm). This measurement accuracy can only be used to detect whether the thickness of the finished thin plate meets the requirements. However, for the hot rolling process before the finished thin plate, in order to ensure the flatness of the finished thin plate, it is necessary to perform thickness detection at each hot rolling stage during the hot rolling process. The purpose of this thickness detection is no longer to determine whether the finished thin plate meets the requirements, but to determine whether the semi-finished thin plate has uneven thickness during the hot rolling process (for example, whether there are wrinkles). Even if there is a slight uneven thickness, it may cause the final thin plate product to not meet the requirements (that is, rolling waste). Therefore, the thickness detection during the hot rolling process needs to have a higher measurement accuracy.

[0003] Further consideration is that in order to comprehensively detect whether the thickness of the semi-finished thin plate during the hot rolling process is uniform, it is necessary to scan and measure the thickness of each position of the thin plate. Errors will be introduced during the scanning process. At the same time, the accuracy of existing measuring instruments may not be sufficient (compared to measuring the thickness of the finished thin plate), and the high temperature and vibration during the hot rolling process interfere with the measuring instrument and the measuring process, resulting in the inability to directly measure the thickness with higher accuracy. Summary of the invention

[0004] In order to solve the above problems, the present invention provides a method, device and system for measuring the thickness of a thin plate of a rolling mill.

[0005] A rolling mill thin plate thickness measurement method, device and system of the present invention adopts the following technical solutions:

[0006] An embodiment of the present invention provides a rolling mill thin plate thickness measuring device, comprising: an upper working roll, a lower working roll, and a thin plate, the device also comprising a distance sensor and a linear motor, wherein there are two distance sensors, which are symmetrically distributed above and below the thin plate respectively; there are also two linear motors, which are symmetrically distributed above and below the thin plate respectively; the distance sensor is fixed on a mover of the linear motor, and the distance sensor is vertically facing the thin plate;

[0007] When measuring thickness, the movers of the two linear motors reciprocate synchronously. A displacement sensor is installed on the linear motor, and the displacement sensor is used to obtain the position of the mover on the thin plate. For the distance sensors above and below each position on the thin plate, the processor measures the thickness of each position of the thin plate by reading the data output by the distance sensor.

[0008] An embodiment of the present invention further provides a method for measuring the thickness of a thin plate in a rolling mill, the method comprising the following steps:

[0009] For the data output by the two distance sensors above and below each position on the thin plate, the observed thickness at each position is obtained according to the data output by the two distance sensors, which is recorded as A;

[0010] The vibration characteristic Z of the thin plate is obtained according to the change frequency of the observed thickness at different positions within a preset time period, [A, Z] is used as the state vector, and the observed thickness at each position within the preset time period is filtered using the Kalman filter algorithm to obtain the filtering result at each position, and the absolute value of the difference between the observed thickness at each position and the filtering result at each position is used as the thickness measurement error at each position;

[0011] Perform Kalman filtering on each position of each distance sensor within a preset time period to obtain a position error of each position;

[0012] Obtaining the matching of the thickness measurement error and the position error at all positions, and decomposing the thickness measurement error at all positions into a motion disturbance amount F1 and a measurement disturbance amount F2 according to the matching; wherein the sum of F1 and F2 is positively correlated with the thickness measurement error at all positions, and the difference between F1 and F2 is positively correlated with the matching;

[0013] Taking [A, Z, F1, F2] as the state vector, the Kalman filter algorithm is used to re-filter the observed thickness at each position within the preset time period to obtain a new filtering result for each position. The new filtering result is used as the measured thickness of the thin plate at each position.

[0014] Preferably, the step of obtaining the observed thickness at each position according to the data output by the two distance sensors comprises the following steps:

[0015] The sum of the data output by the two distance sensors is obtained, which is recorded as D1, and the distance between the two distance sensors is obtained, which is recorded as D2; D2-D1 is used as the observed thickness at each position.

[0016] Preferably, the step of obtaining the matching of thickness measurement errors and position errors at all positions comprises the following steps:

[0017] The thickness measurement error at each position within a preset time period constitutes a thickness measurement error sequence, and the position error at each position within the preset time period constitutes a position error sequence. The absolute value of the Pearson correlation coefficient between the thickness measurement error sequence and the position error sequence is used as the position measurement relationship of each position. The position measurement relationships of all positions are clustered by mean shift to obtain several categories. The sum of the position measurement relationships in the category with the largest density is obtained, which is recorded as Q1. The ratio of Q1 to the sum of the position measurement relationships of all positions is recorded as the matching.

[0018] Preferably, the step of decomposing the thickness measurement errors at all positions into motion interference F1 and measurement interference F2 according to the matching includes the following specific steps:

[0019] The matching is recorded as x, and the mean of all thickness measurement errors at all positions is recorded as y;

[0020] Solving Systems of Equations , we get F1 and F2, where , are the preset parameters.

[0021] Preferably, the step of obtaining the vibration characteristic Z of the thin plate according to the change frequency of the observed thickness at different positions within a preset time period includes the following specific steps:

[0022] The observed thickness at each position within a preset time period constitutes an observed thickness sequence. The frequency spectrum of the observed thickness sequence is obtained using the Fourier transform algorithm and recorded as the observed spectrum at each position. The observed spectra at all positions are clustered by mean shift to obtain several categories, among which the spectrum in the category with the largest density is retained, and the spectrum in other categories is set to 0. The retained spectrum is transformed into the time domain using the inverse Fourier transform algorithm to obtain a filtered sequence of the observed thickness sequence, and the difference between the maximum and minimum values ​​of the filtered sequence is used as the vibration feature Z.

[0023] Preferably, the specific steps of obtaining the filtering result are as follows:

[0024] Any state vector among the state vector [A, Z] and the state vector [A, Z, F1, F2] is denoted as Q, the observed thickness at each position within a preset time period constitutes an observed thickness sequence, and the observed thickness sequence is filtered using a Kalman filter algorithm. The filtering process includes several prediction processes and update processes for the state vector, the state vector before the t-th update process is denoted as Q(t), the average of Q(t) and the state vector Q is denoted as Q1(t), Q1(t) is updated for the t-th update process, and the state vector after the t-th update process is obtained, denoted as Q2(t), Q2(t) is used to perform the t+1-th prediction process and update process, and so on, until the filtering is completed to obtain the filtering result.

[0025] Preferably, the Kalman filtering is performed on each position of each distance sensor within a preset time period to obtain the position error of each position, and the specific steps include the following:

[0026] The position of each distance sensor within a preset time period constitutes a position sequence of each distance sensor; the position sequence of each distance sensor is filtered using a Kalman filter algorithm to obtain a position filtering result of each distance sensor; the absolute value of the difference between each position in the position sequence and each position in the position filtering result is recorded as the motion error of each distance sensor at each position; the maximum value of the motion errors of all distance sensors at the same position is recorded as the position error of each position.

[0027] Preferably, the distance between the two distance sensors comprises the following specific steps:

[0028] A plate with a thickness of D0 is placed between the two distance sensors, and the data output by the two distance sensors are read. The sum of the output data is recorded as d0, and the distance between the two distance sensors is d0+D0.

[0029] An embodiment of the present invention also provides a rolling mill thin plate thickness measurement system, the system includes the above-mentioned rolling mill thin plate thickness measurement device and the above-mentioned rolling mill thin plate thickness measurement method, and the system also includes an alarm system, the alarm system is used to: when the difference in thickness measured at all positions is greater than a preset threshold, an alarm is issued.

[0030] The beneficial effects of the technical solution of the present invention are:

[0031] The present invention uses the vibration characteristic Z of the thin plate as a state to filter the observed thickness, obtain the thickness measurement error, reduce the interference caused by the vibration, and further decompose the thickness measurement error into the motion interference F1 and the measurement interference F2 according to the matching of the position error of the linear motor and the thickness measurement error, so that F1 and F2 can describe the error caused by the linear motor movement in the high temperature environment of hot rolling, as well as the measurement error of the distance sensor itself. Further, Z, F1 and F2 are used as states and the observed thickness is re-filtered, so that the thickness measurement result obtained finally can remove different error interferences in a targeted manner. The accuracy of the measurement result is improved, and the problem of incomplete and non-targeted error removal in conventional filtering algorithms is avoided. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0033] Figure 1 A front view of a thin plate thickness measuring device for a rolling mill provided by one embodiment of the present invention;

[0034] Figure 2 A top view of a thin plate thickness measuring device for a rolling mill provided by one embodiment of the present invention;

[0035] Figure 3 A flow chart of the steps of a method for measuring the thickness of a thin plate in a rolling mill provided in one embodiment of the present invention.

[0036] In the figure: 1. Upper working roller; 2. Lower working roller; 3. Thin plate; 4. Distance sensor; 5. Linear motor. DETAILED DESCRIPTION

[0037] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of a rolling mill thin plate thickness measurement method, device and system proposed by the present invention, its specific implementation, structure, features and effects in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0038] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0039] The specific scheme of a rolling mill thin plate thickness measurement method, device and system provided by the present invention is described in detail below with reference to the accompanying drawings.

[0040] Embodiment 1:

[0041] See also Figure 1 and Figure 2 , which shows a rolling mill thin plate thickness measuring device provided by the first embodiment of the present invention, comprising: an upper working roll 1, a lower working roll 2, and a thin plate 3. The thin plate 3 in this embodiment is made of metal. After being hot-rolled by the upper working roll 1 and the lower working roll 2, the thin plate 3 will undergo ductile deformation, so that the thickness of the thin plate 3 becomes smaller.

[0042] It should be noted that the thin plate 3 during hot rolling may have the following problems: different wear in the width direction of the plate, different deformation degrees, uneven temperature and defects, which may easily cause the thickness of the thin plate 3 to be uneven in the width direction, and waves and wrinkles may easily appear during rolling, and even cause scrap.

[0043] This embodiment requires real-time thickness detection of the thin plate 3 during hot rolling, so the thickness measuring device also includes: a distance sensor 4 and a linear motor 5, wherein there are two distance sensors 4, which are symmetrically distributed above and below the thin plate 3; there are also two linear motors 5, which are symmetrically distributed above and below the thin plate 3; the distance sensor 4 is fixed on the mover of the linear motor 5, and the distance sensor 4 is vertically facing the thin plate 3.

[0044] When measuring the thickness, the movers of the two linear motors 5 reciprocate synchronously, and a displacement sensor is installed on the linear motor 5. The displacement sensor is used to locate the position of the mover on the stator, and the projection point of the position on the thin plate 3 is recorded as the position of the mover on the thin plate 3. For the distance sensors 4 above and below each position on the thin plate 3, the processor measures the thickness of each position on the thin plate 3 by reading the data output by the distance sensors 4.

[0045] The distance sensor 4 used in this embodiment is a laser distance sensor. This embodiment does not limit the specific type of the distance sensor 4. Other types of distance sensors 4 may be used in other embodiments, such as an ultrasonic distance sensor or a depth camera.

[0046] The data output by the distance sensor 4 indicates the distances from the upper and lower surfaces of the thin plate 3 to the upper and lower distance sensors 4 respectively.

[0047] The working principle of the above-mentioned measuring device is: during the hot rolling process, the two linear motors 5 are started, and the movers of the two linear motors 5 move synchronously, and the movement process is carried out in a reciprocating cycle. In this embodiment, every time the thin plate 3 is hot-rolled forward 0.3 cm, the movers of the two linear motors 5 reciprocate for one cycle. In other embodiments, the reciprocating cycle can be set to other values, which are not specifically limited in this embodiment.

[0048] During the movement, the distance sensor 4 collects data every 0.2 seconds (and obtains the position of the mover on the thin plate 3 every 0.2 seconds). The processor records the data output by the upper and lower distance sensors 4 at each moment, as well as the position of the mover (that is, the position of the mover on the thin plate 3). Based on the output data of different positions on the thin plate 3 at different moments, the thickness of each position on the thin plate 3 is measured.

[0049] Embodiment 2:

[0050] This embodiment provides a method for measuring the thickness of a thin plate in a rolling mill based on the first embodiment. Figure 3 As shown, the method includes:

[0051] Step S001, obtaining the observed thickness at each position on the thin plate.

[0052] For any position on the thin plate 3, the mover of the linear motor 5 will pass through the position, that is, the projection point of the mover on the thin plate 3 is the position. At this time, the upper and lower distance sensors 4 at the position both output a data, which represents the distance between the upper and lower distance sensors 4 and the upper and lower surfaces of the thin plate 3. The observed thickness at the position is obtained based on the two data output by the upper and lower distance sensors 4, which is recorded as A.

[0053] As an example, the observed thickness at the position is obtained according to the two data output by the upper and lower distance sensors 4, including the following method:

[0054] For any position on the thin plate 3, the sum of the data output by the two upper and lower distance sensors 4 at the position is recorded as D1, and the distance between the two distance sensors 4 is obtained, recorded as D2; D2-D1 is used as the observed thickness at each position, recorded as A.

[0055] It should be noted that D2 represents the effective distance between the two distance sensors 4 , and the effective distance refers to the distance between the photosensitive units of the two distance sensors 4 .

[0056] It should be noted that the distance between the photosensitive units of the two distance sensors 4 is difficult to measure because the photosensitive units are generally packaged inside the distance sensor 4. This embodiment provides a method for determining the distance between the two distance sensors 4:

[0057] Before hot rolling, a standard plate with known thickness is placed between the two distance sensors 4, and its thickness is recorded as D0. At this time, the data output by the two distance sensors 4 are read, and the sum of the two data is recorded as d0. Then the distance between the two distance sensors 4 is d0+D0.

[0058] So far, whenever the two distance sensors 4 pass through any position on the thin plate 3, the observed thickness at that position can be obtained.

[0059] It should be noted that the observed thickness cannot be directly used as the measurement result of the thickness of the thin plate 3. Because the thin plate 3 will vibrate during the hot rolling process, and the high temperature environment on the surface of the thin plate 3 also affects the measurement accuracy of the distance sensor 4. At the same time, there are errors in the movement of the linear motor 5 in the hot rolling environment, which leads to errors in the observed thickness. In some usage scenarios, this error in the observed thickness can be ignored, such as judging whether the thickness of the thin plate 3 is qualified; however, in this embodiment, it is necessary to detect whether there is uneven thickness based on the thickness of each position on the thin plate 3 to avoid the accumulation and amplification of this uneven thickness in the subsequent hot rolling process, resulting in scrap. Therefore, this embodiment needs to reduce the error of the measurement result as much as possible, so the observed thickness obtained above cannot be used as the final measurement result.

[0060] Step S002, obtain the vibration characteristic Z of the thin plate according to the change frequency of the observed thickness at different positions within a preset time period, use [A, Z] as the state vector, and use the Kalman filter algorithm to filter the observed thickness at each position within the preset time period to obtain the filtering result at each position.

[0061] For the current moment in the hot rolling process, the preset time period T1 before the current moment (including the current moment) is obtained. Each position on the thin plate 3 is passed by the movers of the upper and lower linear motors 5 multiple times within the time period T1. Therefore, each position within the time period T1 corresponds to multiple observed thicknesses. These observed thicknesses constitute a time series sequence, which is recorded as the observed thickness sequence for each position.

[0062] In this embodiment, the time length of the time period T1 is equal to 30 times the motion period of the mover of the linear motor 5. In other embodiments, the time length of the time period T1 can be set to other values, which are not specifically limited in this embodiment.

[0063] Further, the vibration characteristic Z of the thin plate 3 is obtained according to the change frequency of the observed thickness sequence at different positions within the preset time period T1. The vibration characteristic Z describes the vibration of the thin plate 3, which is inevitable during the hot rolling process. The vibration will cause more errors in the observed thickness; it should be noted that this error is mainly reflected in the observed thickness between different positions. For example, when there is no vibration, the difference in the observed thickness between two adjacent positions is 0.5mm, but when there is vibration, the difference in the observed thickness between two adjacent positions is 0.2mm or 0.8mm. If the motion error of the linear motor 5 is taken into account (that is, the two linear motors 5 do not reach a certain position at the same time), the vibration condition will also cause errors in the observed thickness at each position. Therefore, it is necessary to obtain the vibration condition of the thin plate 3.

[0064] As a preferred example, the method for obtaining the vibration feature Z includes:

[0065] The frequency spectrum of the observed thickness sequence at each position is obtained using the Fourier transform algorithm, which is recorded as the observed spectrum at each position. The observed spectrum represents the frequencies contained in the observed thickness sequence and the amplitude of each frequency (the two-dimensional coordinate points composed of frequency and amplitude are recorded as spectrum points).

[0066] The observed spectra at all positions (that is, the spectrum points on the observed spectra at all positions) are clustered by mean shift to obtain several categories, among which the spectrum in the category with the largest density is retained, and the spectrum in other categories is set to 0 (that is, the amplitude corresponding to the frequency is set to 0). The retained spectrum at each position is transformed to the time domain using the inverse Fourier transform algorithm to obtain the filtered sequence of the observed thickness sequence at each position (abbreviated as the vibration filter sequence at each position), and the difference between the maximum and minimum values ​​of the vibration filter sequence at all positions is taken as the vibration feature Z.

[0067] The category with the highest density refers to the category containing the largest number of frequencies, which describes that most positions on the thin plate 3 have the same or similar frequencies, and this frequency is used to represent the overall vibration frequency of the thin plate 3. The vibration filter sequence obtained above represents the change in the observed thickness caused by the vibration of the thin plate 3, that is, the interference caused by the vibration.

[0068] The above-mentioned Fourier transform algorithm and mean shift clustering algorithm are well-known technologies and will not be described in detail in this embodiment.

[0069] The vibration characteristic Z obtained above describes the vibration condition of the thin plate 3, and the observed thickness at each position on the thin plate 3 is affected by the vibration characteristic Z. A conventional implementation method is to subtract the vibration characteristic Z from the observed thickness at each position to obtain the observed thickness after removing the influence of the vibration condition, thereby reducing the error of the observed thickness. However, this method is not optimal because: on the one hand, the acquisition process of the vibration characteristic Z is obtained based on the frequency of change of the observed thickness at all positions, and can only roughly evaluate the vibration condition of the thin plate 3; on the other hand, the observed thickness obtained above not only includes the error caused by the vibration condition, but also includes other errors (such as the error of the distance sensor 4 at high temperature and the error caused by the movement of the linear motor 5). This error causes the vibration characteristic Z to be unable to accurately describe the actual vibration condition.

[0070] Based on this, this embodiment regards the vibration feature Z as a state that can affect the thickness measurement result of the thin plate 3, that is, the thickness measurement result of the thin plate 3 is affected by the vibration feature Z to a certain extent.

[0071] Therefore, further, [A, Z] is used as the state vector, and the Kalman filter algorithm is used to filter the observed thickness at each position within the preset time period T1 to obtain the filtering result at each position. The filtering result at each position describes the observed thickness after removing the influence of the vibration. Compared with the conventional implementation method mentioned above, this method is more accurate and reliable; the reason is that: on the one hand, Z is used as a dimension in the state vector, so that the filtering result combines the influence of the vibration characteristics, and on the other hand, the Kalman filter algorithm is used to avoid: when Z is directly used in the conventional implementation method mentioned above, the error problem caused by the inaccurate calculation of Z (that is, to avoid introducing errors in the calculation model for calculating Z).

[0072] It should also be noted that if the Kalman filter algorithm is directly used to filter the observed thickness at each position within a preset time period to obtain the filtering result at each position, and Z is not used as a dimension in the state vector, the filtering result will still contain errors caused by vibration, and the filtering result will only reduce noise and errors, but cannot further reduce errors in a certain aspect (such as vibration errors) in a targeted manner.

[0073] Step S003, taking the absolute value of the difference between the observed thickness at each position and the filtering result at each position as the thickness measurement error at each position; performing Kalman filtering on each position of each distance sensor 4 within a preset time period to obtain the position error of each position.

[0074] The above filtering result describes the observed thickness after removing the influence of vibration, and the thickness measurement error at each position describes the possible error size of the observed thickness A when the influence of vibration is not considered.

[0075] It should be noted that the observed thickness at each of the above-mentioned positions and the filtering results at each position are time series within the time period T1, so the thickness measurement error at each position is also a time series within the time period T1. In other words, the thickness measurement error at each position represents: within the time period T1, when the distance sensor 4 passes through each position on the thin plate 3, after removing the influence of vibration, the observed thickness at that position still has an error.

[0076] The position of each distance sensor 4 in the time period T1 is obtained. In this embodiment, only the position in the width direction of the thin plate 3 is considered. These positions constitute a position sequence, which is recorded as the position sequence of each distance sensor 4, and describes the position of each distance sensor 4 at different times during the movement process; the position sequence includes the error of each distance sensor 4 during the movement (such as the speed error of the mover of the linear motor 5, the acceleration error, and the error caused by the unstable operation of the linear motor 5 in a high temperature environment, etc., which may cause the upper and lower linear motors 5 to work asynchronously);

[0077] The position sequence of each distance sensor 4 is filtered using the Kalman filter algorithm to obtain a position filtering result of the distance sensor 4. The absolute value of the difference between each position in the position sequence and each position in the position filtering result is recorded as the motion error of each distance sensor 4 at each position.

[0078] The motion error of each distance sensor 4 at each position describes the error caused by the motion of the linear motor 5 when each distance sensor 4 is at each position on the thin plate 3. For the upper and lower distance sensors 4, the maximum value of the motion errors of the two distance sensors 4 at the same position is recorded as the position error of the same position.

[0079] At this point, the position error of each position is obtained by combining the two distance sensors 4 .

[0080] Step S004: Obtain the matching of the thickness measurement errors and position errors at all positions, and decompose the thickness measurement errors at all positions into a motion disturbance amount F1 and a measurement disturbance amount F2 according to the matching.

[0081] The thickness measurement error at each position is obtained in the time period T1, which describes that when the distance sensor 4 is at each position on the thin plate 3 in the time period T1, after removing the influence of vibration, the error still exists in the observed thickness.

[0082] At the same time, the position error of each position is also obtained, which describes the error caused by the movement of the linear motor 5 when the distance sensor 4 is at each position on the thin plate 3 during the time period T1.

[0083] Considering that the thickness measurement error at each position mainly includes two aspects, one is the measurement error of the distance sensor 4 itself in the high temperature environment; the other is the error caused by the movement of the linear motor 5 in the high temperature environment.

[0084] This embodiment obtains the matching of the thickness measurement error and the position error at all positions. The matching is used to describe whether the distribution law or change trend between the position error and the thickness measurement error has great similarity or consistency. If there is a more obvious similarity or consistency, the matching is greater, indicating that the thickness measurement error is mainly caused by the error generated by the movement of the linear motor 5; on the contrary, if there is no obvious similarity or consistency, the matching is smaller, indicating that the thickness measurement error is not caused by the error generated by the movement of the linear motor 5.

[0085] As an example, obtaining the matching of thickness measurement errors and position errors at all positions includes the following steps:

[0086] The thickness measurement error of each position within the preset time period T1 constitutes a thickness measurement error sequence, and the position error of each position within the preset time period T1 constitutes a position error sequence. The absolute value of the Pearson correlation coefficient between the thickness measurement error sequence and the position error sequence is used as the position measurement relationship of each position. The position measurement relationships of all positions are clustered by mean shift to obtain several categories. The sum of the position measurement relationships in the category with the largest density is obtained, which is recorded as Q1. The sum q1 of the position measurement relationships of all positions is obtained, and the ratio of Q1 to q1 is recorded as matching.

[0087] The position measurement relationship of each position described above describes the consistency (or correlation) of the changing trend between the position error and the thickness measurement error at the same position. The category with the highest density refers to the category containing the largest number of position measurement relationships. The larger Q1 is, the more obvious the consistency is between the position error and the thickness measurement error at most positions on the sheet 3, indicating that the thickness measurement error is mainly caused by the error generated by the movement of the linear motor 5.

[0088] Furthermore, the thickness measurement errors at all positions are decomposed into a motion disturbance amount F1 and a measurement disturbance amount F2 according to the matching.

[0089] F1 represents the error caused by the movement of the linear motor 5 in a high temperature environment, and F2 represents the measurement error of the distance sensor 4 itself in a high temperature environment.

[0090] As an example, the steps for obtaining F1 and F2 are:

[0091] The matching is recorded as x, and the mean of all thickness measurement errors at all positions within the preset time period T1 is recorded as y;

[0092] Solving Systems of Equations , we get F1 and F2, where , is a preset parameter. , This is described as an example. Other embodiments may be set to other values. This embodiment does not specifically limit this. It only needs to be less than or equal to 1 and greater than 0.

[0093] in The error caused by the movement of the linear motor 5 and the measurement error of the distance sensor 4 itself are combined to form the thickness measurement error. The role of w1 is to take into account that there are many reasons for the thickness measurement error in the hot rolling process, which may include not only the movement error of the linear motor 5 and the measurement error of the distance sensor 4 itself. Therefore Only with Positively correlated, but may be less than or equal to .

[0094] It represents the relative size (or difference) of the motion error of the linear motor 5 and the measurement error of the distance sensor 4 itself, where When the larger the value, the more obvious the motion error of the linear motor 5 is, so the difference, that is, It is positively correlated with x.

[0095] Step S005, taking [A, Z, F1, F2] as the state vector, using the Kalman filter algorithm to re-filter the observed thickness at each position within the preset time period to obtain a new filtering result for each position, and the new filtering result is used as the measured thickness of each position of the thin plate 3.

[0096] The motion disturbance amount F1 and the measurement disturbance amount F2 are obtained above, which respectively describe the error caused by the motion of the linear motor 5 in the high temperature environment and the measurement error of the distance sensor 4 itself in the high temperature environment.

[0097] A conventional implementation method is to subtract F1 and F2 from the filtering result obtained in step S002 to obtain a measurement result that does not contain the above two errors. This method is not accurate enough because: the acquisition process of F1 and F2 is obtained based on the trend relationship between the thickness measurement error and the position error (see step S004 for details), and the process of obtaining F1 and F2 may also have model errors. Therefore, F1 and F2 can only roughly describe the distribution of errors, or the actual size of the above two errors is only significantly correlated with F1 and F2, but F1 and F2 cannot be regarded as real errors. Therefore, this conventional implementation method is not optimal.

[0098] In this embodiment, [A, Z, F1, F2] is used as the state vector, and the Kalman filter algorithm is used to re-filter the observed thickness at each position within the preset time period T1 to obtain a new filtering result for each position.

[0099] In this embodiment, F1 and F2 are used as two dimensions in the state vector, that is, as states that affect the actual thickness measurement result. These two states are: the state of movement of the linear motor 5 in a high temperature environment (in this embodiment, the error generated by the movement of the linear motor 5 is used to represent it), and the measurement state of the distance sensor 4 itself in a high temperature environment (in this embodiment, the measurement error of the distance sensor 4 itself is used to represent it); the introduction of these two states, on the one hand, makes the new filtering result take into account the movement state of the linear motor 5 and the measurement state of the distance sensor 4 itself, and these two states have a direct impact on the measurement result (or the observed thickness is determined by these two states), and on the other hand, avoids the interference of the calculation error existing in these two states themselves (that is, the error existing in the calculation model or calculation process involved in step S004), and avoids: although the interference of the movement of the linear motor 5 and the measurement error of the distance sensor 4 are eliminated, the error of the calculation model is introduced, which leads to the problem that the final thickness measurement result is still not accurate enough.

[0100] As an example, in step S002: [A, Z] is used as a state vector, and the Kalman filter algorithm is used to filter the observed thickness at each position within the preset time period T1 to obtain a filtering result at each position, and in step S005: [A, Z, F1, F2] is used as a state vector, and the Kalman filter algorithm is used to re-filter the observed thickness A at each position within the preset time period T1 to obtain a new filtering result at each position, including the method of:

[0101] Any state vector among [A, Z] and [A, Z, F1, F2] is recorded as state vector Q;

[0102] First of all, it should be explained that the Kalman filter algorithm is an algorithm based on state filtering. Before filtering begins, an initial state vector needs to be set. The conventional setting method is to manually specify the dimension of the state vector, and then randomly assign an initial state vector. In this embodiment, the state vector Q is used as the initial state vector of the Kalman filter algorithm, and the observed thickness at each position within the preset time period T1 is filtered (or re-filtered).

[0103] In other embodiments, several additional dimensions (eg, 3 dimensions) may be added to the state vector Q, and the values ​​of these dimensions are randomly initialized.

[0104] As another example, the method included is:

[0105] As known from the Kalman filtering algorithm, the Kalman filtering algorithm includes a prediction process and an update process. The prediction process refers to predicting the state vector at the current moment from the state vector at the previous moment. The update process refers to updating the predicted state vector at the current moment, and then predicting the state vector at the next moment based on the updated state vector, and then updating the predicted state vector at the next moment, and so on, iteratively executing the prediction process and the update process until the filtering ends.

[0106] If only the state vector Q is used as the initial state vector, it will cause some state information in some dimensions of the state vector Q to be lost during some prediction processes and update processes. Although it can still reduce some noise and errors in the filtering result, it cannot further specifically reduce the errors in a certain aspect (such as vibration error, the error generated by the linear motor 5 moving in a high-temperature environment, and the measurement error of the distance sensor 4 itself in a high-temperature environment), resulting in an inaccurate filtering result.

[0107] In this embodiment, the state vector before the t-th update process is denoted as Q(t), the mean value of Q(t) and the state vector Q is denoted as Q1(t), the t-th update process is performed on Q1(t) to obtain the state vector after the t-th update process, denoted as Q2(t), and Q2(t) is used for the (t + 1)-th prediction process and update process, and so on, until the filtering ends to obtain the filtering result of the observed thickness sequence.

[0108] Another example is given as follows. The method included is: only when t < t0 (in this embodiment, t0 = 15 is used as an example for description), the above-mentioned: the mean value of Q(t) and the state vector Q is denoted as Q1(t), and the t-th update process is performed on Q1(t). When t is greater than or equal to t0, the following process is no longer executed: the mean value of Q(t) and the state vector Q is denoted as Q1(t), and the t-th update process is performed on Q1(t). At this time: the value of the state vector before the t-th update process is denoted as Q(t), the t-th update process is performed on Q(t) to obtain the value of the state vector after the t-th update process, denoted as Q2(t), and Q2(t) is used for the (t + 1)-th prediction process and update process, and so on, until the filtering ends to obtain the filtering result of the observed thickness sequence.

[0109] This example takes more into account the specific process of the Kalman filtering algorithm and avoids the situation where the errors existing in the state vector Q are consistently introduced into the Kalman filtering algorithm.

[0110] It should be noted that in all embodiments of the present invention, considering that the Kalman filtering algorithm has a relatively fast calculation speed, the Kalman filtering algorithm is used. In the specific implementation of all embodiments of the present invention, the Kalman filtering algorithm can be replaced with the extended Kalman filtering algorithm or the unscented Kalman filter.

[0111] The above process allows the vibration error, the error caused by the movement of the linear motor 5 in a high temperature environment, and the measurement error of the distance sensor 4 itself in a high temperature environment to be taken into account each time the prediction process and the update process are executed. By combining the interference and influence of these errors on the observed thickness, the filtering result (and the new filtering result) at each position can be targeted to remove the error as much as possible, making the new filtering result more reliable.

[0112] The new filtering result is used as the thickness measured at each position of the thin plate 3.

[0113] It can be seen from steps S001 to S005 that in this embodiment, the observed thickness at each position in a preset time period (i.e., time period T1) is filtered by the Kalman filter algorithm to obtain a new filtering result at each position. Subsequently, a new filtering result at each position is obtained according to the method of steps S001 to S005 after each time period T1.

[0114] If the time period before the current moment is less than T1, the observed thickness at each position before the current moment is used as the measured thickness.

[0115] The measurement of the thickness at each position on the thin plate 3 during the hot rolling process is now completed.

[0116] Embodiment three:

[0117] The difference between this embodiment and the second embodiment is that in step S004, A method for obtaining , specifically comprising:

[0118] For the filtering result of each position obtained in step S002, the filtering result is regarded as an observed thickness sequence. Based on the observed thickness sequence, the vibration feature Z acquisition method in step S002 is used to obtain the vibration feature corresponding to the observed thickness sequence again, which is recorded as Z1. As , among which When it is less than 0.1, Set to 0.1.

[0119] The larger the Z1 is, the more obvious the vibration interference error is in the filtering result of each position obtained in step S002. The smaller the value, the The value of is relatively small, that is, the thickness measurement error at this time contains a more obvious vibration interference error, so the two errors corresponding to F1 and F2 are relatively small. When Z1 is smaller, it means that there is no obvious vibration interference error in the filtering result of each position obtained in step S002. The larger the value, the greater the The value of is relatively large, that is, the thickness measurement error at this time does not contain obvious vibration interference error, so the two errors corresponding to F1 and F2 are relatively large.

[0120] Embodiment 4:

[0121] This embodiment provides a rolling mill thin plate thickness measurement system, including a rolling mill thin plate thickness measurement device in embodiment 1, and a rolling mill thin plate thickness measurement method in embodiments 2 and 3, wherein the rolling mill thin plate thickness measurement method in embodiments 2 and 3 is a computer program running in a processor. It also includes an alarm system, which obtains the maximum and minimum values ​​of the measured thickness at all positions in the time period T1. When the difference between the maximum and minimum values ​​is greater than the preset threshold value th, an alarm signal is issued, indicating that there is uneven thickness on the surface of the thin plate 3 during the hot rolling process, and the relevant personnel are notified to handle it, such as manually adjusting the hot rolling temperature, or adjusting the rotation speed of the upper working roll 1 and the lower working roll 2. In this way, the uneven thickness is prevented from accumulating in the subsequent hot rolling process and becoming more serious or uncontrollable, thereby avoiding rolling waste. This embodiment is described by taking th=0.1mm as an example. In other embodiments, th can be set to other values, which are not specifically limited in this embodiment.

[0122] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for measuring the thickness of a thin plate of a rolling mill, using a device for measuring the thickness of a thin plate of a rolling mill, comprising: An upper working roll (1), a lower working roll (2), and a thin plate (3), characterized in that the device further comprises a distance sensor (4) and a linear motor (5), wherein the distance sensors (4) are two and are symmetrically distributed above and below the thin plate (3); the linear motor (5) is also two and is symmetrically distributed above and below the thin plate (3); the distance sensor (4) is fixed on a mover of the linear motor (5), and the distance sensor (4) is perpendicular to the thin plate (3); When measuring the thickness, the movers of the two linear motors (5) reciprocate synchronously, and a displacement sensor is installed on the linear motor (5), wherein the displacement sensor is used to obtain the position of the mover on the thin plate (3); for the distance sensors (4) above and below each position on the thin plate (3), the processor measures the thickness of each position of the thin plate (3) by reading the data output by the distance sensors (4); The method is executed in a processor, and comprises the following steps: For data output by two distance sensors (4) above and below each position on the thin plate (3), an observed thickness at each position is obtained based on the data output by the two distance sensors (4), which is recorded as A; The vibration characteristic Z of the thin plate (3) is obtained according to the change frequency of the observed thickness at different positions within a preset time period, [A, Z] is used as a state vector, and the observed thickness at each position within the preset time period is filtered using a Kalman filter algorithm to obtain a filtering result at each position, and the absolute value of the difference between the observed thickness at each position and the filtering result at each position is used as the thickness measurement error at each position; Performing Kalman filtering on each position of each distance sensor (4) within a preset time period to obtain a position error at each position; Obtaining the matching of the thickness measurement error and the position error at all positions, and decomposing the thickness measurement error at all positions into a motion disturbance amount F1 and a measurement disturbance amount F2 according to the matching; wherein the sum of F1 and F2 is positively correlated with the thickness measurement error at all positions, and the difference between F1 and F2 is positively correlated with the matching; [A, Z, F1, F2] is used as the state vector, and the Kalman filter algorithm is used to re-filter the observed thickness at each position within the preset time period to obtain a new filtering result at each position, and the new filtering result is used as the thickness measured at each position of the thin plate (3); The step of obtaining the vibration characteristic Z of the thin plate (3) according to the change frequency of the observed thickness at different positions within a preset time period comprises the following specific steps: The observed thickness at each position within a preset time period constitutes an observed thickness sequence. The spectrum of the observed thickness sequence is obtained using the Fourier transform algorithm and recorded as the observed spectrum at each position. The observed spectra at all positions are clustered by mean shift to obtain several categories, among which the spectrum in the category with the largest density is retained, and the spectrum in other categories is set to 0. The retained spectrum is transformed into the time domain using the inverse Fourier transform algorithm to obtain a filtered sequence of the observed thickness sequence, and the difference between the maximum value and the minimum value of the filtered sequence is used as the vibration feature Z; The specific steps for obtaining the filtering result are as follows: Any state vector among the state vector [A, Z] and the state vector [A, Z, F1, F2] is denoted as Q, the observed thickness at each position within a preset time period constitutes an observed thickness sequence, and the observed thickness sequence is filtered using a Kalman filter algorithm, wherein the filtering process includes several prediction processes and update processes of the state vector, the state vector before the t-th update process is denoted as Q(t), the average of Q(t) and the state vector Q is denoted as Q1(t), Q1(t) is updated for the t-th update process, and the state vector after the t-th update process is obtained, denoted as Q2(t), Q2(t) is used to perform the t+1-th prediction process and update process, and so on, until the filtering is completed to obtain the filtering result; The Kalman filter is performed on each position of each distance sensor (4) within a preset time period to obtain a position error of each position, and the specific steps include the following: The position of each distance sensor (4) within a preset time period constitutes a position sequence of each distance sensor (4); the position sequence of each distance sensor (4) is filtered using a Kalman filter algorithm to obtain a position filtering result of each distance sensor (4); the absolute value of the difference between each position in the position sequence and each position in the position filtering result is recorded as the motion error of each distance sensor (4) at each position; and the maximum value of the motion errors of all distance sensors (4) at the same position is recorded as the position error of each position.

2. A method for measuring the thickness of a rolling mill sheet according to claim 1, characterized in that: The step of obtaining the observed thickness at each position according to the data output by the two distance sensors (4) comprises the following steps: The sum of the data output by the two distance sensors (4) is obtained, recorded as D1, and the distance between the two distance sensors (4) is obtained, recorded as D2; D2-D1 is used as the observed thickness at each position.

3. A method for measuring the thickness of a rolling mill sheet according to claim 1, characterized in that: The step of obtaining the matching of thickness measurement errors and position errors at all positions includes: The thickness measurement error at each position within a preset time period constitutes a thickness measurement error sequence, and the position error at each position within the preset time period constitutes a position error sequence. The absolute value of the Pearson correlation coefficient between the thickness measurement error sequence and the position error sequence is used as the position measurement relationship of each position. The position measurement relationships of all positions are clustered by mean shift to obtain several categories. The sum of the position measurement relationships in the category with the largest density is obtained, which is recorded as Q1. The ratio of Q1 to the sum of the position measurement relationships of all positions is recorded as the matching.

4. A method for measuring the thickness of a rolling mill sheet according to claim 1, characterized in that: The specific steps of decomposing the thickness measurement errors at all positions into motion disturbance F1 and measurement disturbance F2 according to the matching are as follows: The matching is recorded as x, and the mean of all thickness measurement errors at all positions is recorded as y; Solving Systems of Equations , we get F1 and F2, where , are the preset parameters.

5. A method for measuring the thickness of a rolling mill sheet according to claim 2, characterized in that: The distance between the two distance sensors (4) comprises the following specific steps: A plate with a thickness of D0 is placed between the two distance sensors (4), and the data output by the two distance sensors (4) are read. The sum of the output data is recorded as d0, and the distance between the two distance sensors (4) is d0+D0.

6. A rolling mill thin plate thickness measurement system, characterized in that: The system is used to execute a rolling mill thin plate thickness measurement method as described in any one of claims 1-5, and the system also includes an alarm system, which is used to: when the difference in thickness measured at all positions is greater than a preset threshold, an alarm is issued.

Images