Steel coil position automatic alignment method based on two-time internal defect detection data

Abstract

The invention discloses a steel coil position automatic alignment method based on two-time internal defect detection data. The method comprises the following steps: S1, judging whether a steel coil is overturned or not and performing data overturning processing; s2, searching similar feature points of the defect data detected twice, and aligning the defect data detected twice according to the relative offset between the similar feature points by taking the defect data detected last time as a reference; and S3, dividing the aligned defect data into a plurality of voltage intervals, dividing the defect data in each voltage interval into grids, and calculating the similarity of the number of the defect data in the grids. According to the method, based on internal defect data detected by internal instrument equipment, a position alignment method is found out from disordered large and small defects, and similarity evaluation is carried out on two times of detection data, so that the repeatability of the internal instrument detection equipment is evaluated.

Description

Technical Field

[0001] The present invention relates to a cold-rolled thin plate steel coil intelligent manufacturing technology, and more specifically, to a steel coil position automatic alignment method based on two undercut defect detection data. Background Art

[0002] The internal inclusion defect (i.e., internal defect) of cold-rolled thin steel coil is one of the key qualities of the steel coil. The internal defect meter is an important device for detecting the quality of internal inclusions in thin strip steel. It is generally installed in the cold rolling finishing unit. The test results can be used to determine the quality of the steel coil, block unqualified products in time, and reduce user complaints and quality objections. The data output by the internal defect meter detection equipment includes the defect position (i.e., position X in the length direction of the steel coil and position Y in the width direction) and the defect size (expressed by voltage value, the larger the voltage value, the larger the defect, generally between 0-12V). The stability of the internal defect meter equipment state affects the accuracy of the detection, and the evaluation method of the stability of the detection equipment state is to use the same steel coil for multiple tests and evaluate the repeatability accuracy of the equipment through the similarity of the test results. In the past, qualitative evaluation was performed by manual visual observation. With the development of intelligence and automation, the traditional manual observation method is often inefficient and has many unfavorable factors such as errors caused by human factors. Therefore, the research and application of intelligent similarity evaluation methods have become increasingly important.

[0003] At present, there are many literatures on the control of internal inclusion defects at home and abroad, but there are almost no patents and literature on the research of internal inclusion instrument equipment, especially the processing methods of internal inclusion instrument detection data. No relevant patents were found through searching at home and abroad. Summary of the invention

[0004] In view of the defects existing in the prior art, the purpose of the present invention is to provide a method for automatic alignment of steel coil position based on two internal undercut defect detection data, which is based on the internal undercut defect data detected by the internal undercut instrument equipment, finds a position alignment method among the messy large and small defects, and performs similarity evaluation on the two detection data to evaluate the repeatability of the internal undercut instrument detection equipment.

[0005] To achieve the above object, the present invention adopts the following technical solutions:

[0006] A method for automatically aligning the position of a steel coil based on two undercut defect detection data comprises the following steps:

[0007] S1. Determine whether the steel coil is turned over and process the data;

[0008] S2, finding similar feature points of the defect data of the two tests, and aligning the defect data of the two tests based on the defect data of the previous test according to the relative offset between the similar feature points;

[0009] S3, dividing the aligned defect data into several voltage intervals, dividing the defect data in each voltage interval into grids again, and calculating the similarity of the number of defect data in the grids.

[0010] Preferably, the step S1 specifically includes the following steps:

[0011] S11, judging whether the steel coil is turned over, if the difference between the two detection times is an odd number, the steel coil is turned over, and if it is an even number, the steel coil is not turned over;

[0012] S12. If the steel coil is turned over during the two inspections, the defect data of the latter inspection is processed; if it is not turned over, the defect data is not processed.

[0013] Preferably, the step S2 specifically includes the following steps:

[0014] S21, respectively calculating the X-coordinate difference data set between two points of defect data with larger values ​​in the defect data detected twice;

[0015] S22, determining a threshold value A;

[0016] S23, screening the data set according to the distance difference between two points of defect data detected in the previous time and the distance difference between two points of defect data detected in the next time not exceeding the threshold value A;

[0017] S24, using the Y coordinate to eliminate pseudo similar points in the defect data set;

[0018] S25, obtaining similar feature points of two detections;

[0019] S26. Align the defect data using similar feature points and obtain the head shear force and tail shear amount verified last time.

[0020] Preferably, the step S3 specifically includes the following steps:

[0021] S31, dividing the grid;

[0022] S32, calculating the similarity evaluation values ​​under two detection voltage intervals;

[0023] S33, calculating the similarity evaluation values ​​under two detection voltage intervals;

[0024] S34. Calculate the comprehensive similarity evaluation.

[0025] The present invention provides a method for automatically aligning the position of a steel coil based on two internal undercut defect detection data, and on this basis evaluates the similarity of the two internal undercut defect detection results to evaluate the accuracy of the equipment. The present invention has the advantages of intelligence and automation, reduces the errors of manual processing, alignment and observation, and reduces the workload. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 It is a schematic diagram of the process of the automatic alignment method of the steel coil position of the present invention;

[0027] Figure 2 1 is a schematic diagram of the length direction data flipping in step S12 of the automatic steel coil position alignment method of the present invention;

[0028] Figure 3 It is a schematic diagram of data flipping in the width direction in step S12 of the method for automatic alignment of the steel coil position of the present invention;

[0029] Figure 4 It is a schematic diagram of position alignment in step S2 of the automatic position alignment method of steel coils of the present invention;

[0030] Figure 5 It is a schematic diagram of the data structure of step S21 in the method for automatic alignment of steel coil positions of the present invention;

[0031] Figure 6 It is a schematic diagram of the data structure of step S22 in the method for automatic alignment of steel coil positions of the present invention;

[0032] Figure 7 It is a schematic diagram of the dichotomy of step S23 in the method for automatic alignment of the steel coil position of the present invention;

[0033] Figure 8 1 is a schematic diagram comparing the binary search value and the true value (with decimal point) in step S23 of the automatic steel coil position alignment method of the present invention;

[0034] Fig. 9 It is a schematic diagram comparing the binary search value and the true value (without the decimal point) of step S23 in the automatic alignment method of the steel coil position of the present invention;

[0035] Fig.10 This is a schematic diagram of the first pseudo similarity point in step S24 of the method for automatic alignment of steel coil positions of the present invention;

[0036] Fig.11 It is a schematic diagram of a second pseudo similar point in step S24 of the method for automatic alignment of steel coil positions of the present invention;

[0037] Fig.12 It is a schematic diagram of a third pseudo similar point in step S25 of the method for automatic alignment of steel coil positions of the present invention;

[0038] Fig.13 It is a schematic diagram of a fourth pseudo similar point in step S25 of the method for automatic alignment of steel coil positions of the present invention;

[0039] Fig.14is a schematic diagram of step S31 in the method for automatic alignment of steel coil positions of the present invention;

[0040] Fig.15 It is a schematic diagram of detecting the similarity of corresponding grids twice in step S32 of the automatic alignment method for steel coil positions of the present invention;

[0041] Fig.16 1 is a schematic diagram of scattered points of two detections of original point defects in Example 1 of the method for automatic alignment of steel coil positions of the present invention, (a) is the first detection spectrum, and (b) is the second detection spectrum;

[0042] Fig.17 Schematic diagram of similar feature points found in Example 1 of the method for automatic alignment of steel coil positions of the present invention, (a) is a similar feature point found in the first detection, and (b) is a similar feature point found in the second detection;

[0043] Fig.18 1 is a schematic diagram of the steel coil position automatic alignment method after alignment in Example 1 of the present invention, (a) is a spectrum after the first detection alignment, (b) is a spectrum after the second detection alignment;

[0044] Fig.19 Schematic diagram of scatter points in the voltage range of 2-3V in Example 1 of the method for automatic alignment of steel coil positions of the present invention, (a) is a scatter point spectrum in the voltage range of 2-3V after the first detection alignment, (b) is a scatter point spectrum in the voltage range of 2-3V after the second detection alignment;

[0045] Fig. 20 It is a schematic diagram of detecting the similarity of corresponding grids twice in Example 1 of the method for automatic alignment of steel coil positions of the present invention;

[0046] Fig.21 1 is a schematic diagram of scattered points of detecting original point defects twice in Example 2 of the method for automatic alignment of steel coil positions of the present invention, (a) is the first detection spectrum, and (b) is the second detection spectrum;

[0047] Fig. 22 Schematic diagram of similar feature points found in Example 2 of the method for automatic alignment of steel coil positions of the present invention, (a) is a similar feature point found in the first detection, and (b) is a similar feature point found in the second detection;

[0048] Fig.23 1 is a schematic diagram of the steel coil position automatic alignment method after alignment in Example 2 of the present invention, (a) is a spectrum after the first detection alignment, (b) is a spectrum after the second detection alignment;

[0049] Fig.24Schematic diagram of scatter points in the voltage range of 2-3V in Example 2 of the method for automatic alignment of steel coil positions of the present invention, (a) is a scatter point spectrum in the voltage range of 2-3V after the first detection alignment, (b) is a scatter point spectrum in the voltage range of 2-3V after the second detection alignment;

[0050] Fig.25 It is a schematic diagram of two detections of the similarity of corresponding grids in Example 2 of the method for automatic alignment of steel coil positions of the present invention. DETAILED DESCRIPTION

[0051] In order to better understand the above technical solution of the present invention, the technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0052] Since the head and tail lengths of the steel coil will be sheared during the two inspections, the lengths of the two inspections will change, and the collected shear length data is often inaccurate. In addition, the head and tail positions and width positions of two adjacent inspections will be flipped, so position alignment is a prerequisite for similarity evaluation. The present invention provides an automatic position alignment method for steel coils based on two internal under-defect inspection data, and on this basis, a similarity evaluation is performed on the repeatability of any two inspection results of the steel coil.

[0053] The implementation of the automatic coil position alignment method of the present invention involves the following data items:

[0054] (1) Defect data of internal shorting instrument

[0055] Defect position: Defect_x (X coordinate) in the length direction, Defect_y (Y coordinate) in the width direction, and output voltage Defect_V reflecting the size of the defect.

[0056] (2) Basic information data of the test volume

[0057] Detect the roll length Coil_len and width Coil_wide.

[0058] Combination Figure 1 As shown, the method for automatic alignment of steel coil positions of the present invention specifically comprises the following steps:

[0059] S1. Determine whether the steel coil is turned over and process the data, specifically including the following steps:

[0060] S11. Determine whether the steel coil is turned over:

[0061] When the steel coil is produced on the unit, it must pass through the unwinder at the entrance and the coiler at the exit. After one pass of production, the head and tail are reversed and the head and tail are cut and welded. Therefore, when the next inspection is performed, the head, tail and width are flipped compared to the previous time, and the length is changed. During the third inspection, the head, tail and width are flipped again based on the second inspection, but the head, tail and width directions are the same compared to the first inspection, and no flipping is required. Therefore, before position alignment and similarity evaluation, it is necessary to determine whether the two comparative inspections need to be flipped.

[0062] Whether the difference between the two detection times is an odd number or an even number is used to determine whether the two detections are flipped. If it is an odd number, it is flipped, and if it is an even number, it is not flipped. Assuming that Defect_data1 corresponds to the first detection and Defect_data2 corresponds to the second detection, the difference between the two detection times is only 2-1=1, which is an odd number, so it is flipped; if the two detection times are even numbers, there is no need to flip the data. For example, for the first and third detections, except for the change in length, the length direction and width direction are the same.

[0063] S12. Data processing:

[0064] Assume that the defect data set of the previous detection is Defect_data1, and the defect data set of the next detection is Defect_data2. According to step 1.1, determine whether the steel coils of the two detections are flipped, and process the data of the next detection to facilitate data alignment between the two subsequent detections.

[0065] If there is no rollover, the data does not need to be processed.

[0066] If flipping is performed, the defect data of the next detection is flipped according to the length Coil_len2 of the next detection coil and the centerline position coordinates of the strip width (such as Figure 2 and Figure 3 As shown), define:

[0067] Defect_x i =Coil_len2-Defect_x i

[0068] Defect_y i = mid_line-(Defect_y i -mid_line)

[0069] in:

[0070] Defect_x i is the X coordinate of the ith defect,

[0071] Defect_y i is the Y coordinate of the i-th defect,

[0072] mid_line is the Y coordinate of the center line of the strip width.

[0073] The flipping process does not change the relative distance between any two defect points and does not affect the subsequent search for similar features.

[0074] S2. Position alignment:

[0075] Find similar feature points of the two inspections, and align the two defect inspection data based on the relative offset between the similar features and the previous inspection. Figure 4 As shown in the figure, the black solid points are similar feature points automatically found by the model. One of the black solid points is used for data alignment, that is, the X-coordinate data and Y-coordinate data in the defect data of the next detection are moved as a whole to align with the position of the previous detection data. After alignment, the redundant part in the X-coordinate direction of the previous detection is deleted, and the defect X-coordinate and Y-coordinate data need to be deleted together. Generally, no shearing is performed in the width direction, and the width does not change. Alignment is based on the center line. Specifically, the following steps are included:

[0076] S21. Calculate the X-coordinate difference data set between two points of the defect data with a large value (such as ≥4V) in the defect data detected twice:

[0077] Take out the defect data Defect_data1_4 and Defect_data2_4 with voltage value ≥ 4V from the two defect data sets Defect_data1 and Defect_data2, calculate the absolute value len_sub1 and len_sub2 of the X coordinate difference in the defect data Defect_data1_4 and Defect_data2_4 respectively, and add len_sub1 and len_sub2 to the lists len_sub1_list and len_sub2_list respectively, and record which two defect points the difference is obtained from. The data structure is as follows: Figure 5 shown.

[0078] Taking the previous inspection as an example, the calculation is shown in formula (1), where n is the total number of defects with a voltage value ≥ 4V, and x represents the number of defect points. When x=1, the X coordinate of the first defect point needs to be subtracted from the X coordinates of the second defect point and the subsequent defect points in turn to obtain a set of data; when x=2, because the difference between the first point and the second point has been calculated when x=1, the X coordinate of the second defect point needs to be subtracted from the X coordinates of the third defect point and the subsequent defect points in turn to obtain a set of data; then the differences from x=3 to x=n-1 are calculated in turn, and a total of n-1 sets of data are obtained.

[0079] len_sub1 n-x =|Defect_x1 x -Defect_x1 i+x |x=1,i=1,2,3……nx

[0080] len_sub1 n-x =|Defect_x1 x -Defect_x1 i+x |x=2,i=1,2,3……nx

[0081] …

[0082] len_sub1 n-x =|Defect_x1 x -Defect_x1 i+x |x=n-1,i=1,2,3……nx (1)

[0083] After the calculation is completed, the calculated n-1 groups of data are added to len_sub1_list, and the next test is performed on all len_sub2 n-x Add to len_sub2_list.

[0084] S22, find the point where the distance difference between the two points detected in the previous time is similar to the distance difference between the two points detected in the next time:

[0085] First, take out the first data in len_sub1_list, and search for data in len_sub2_list whose absolute difference with the data does not exceed A (to eliminate the error in the length direction detection). If found, put the first data in len_sub1_list into the new llen_sub1_list, and put the data found in len_sub2_list into the new len_sub2_list. The data structure is as follows: Figure 6 shown.

[0086] The program framework is shown in Table 1

[0087] Table 1 Similarity finding program framework

[0088]

[0089] S23. Method for determining threshold value A:

[0090] The larger the threshold A is set, the more points will be found, but the accuracy will decrease accordingly. Therefore, it is necessary to minimize A while finding similar feature points. Because this problem is monotonic and will not fall into the local optimum, the binary search method is used to find the optimal solution and the algorithm is designed based on this.

[0091] Bisection method: For a function y=f(x) that is continuous on the interval [a, b] and f(a)·f(b)<0, the method of obtaining the approximate value of the zero point by continuously dividing the interval where the zero point of the function f(x) is located into two, so that the two endpoints of the interval gradually approach the zero point, is called the bisection method, such as Figure 7 shown.

[0092] The core algorithm principle program framework is shown in Table 2

[0093] Table 2 Core algorithm principle program framework

[0094]

[0095]

[0096] Test the algorithm: The difference between the optimal solution found by the algorithm and the actual solution is as follows: Figure 8 As shown in the figure, it can be seen that the optimal solution is basically consistent with the actual solution, such as Fig. 9 This algorithm has the advantages of high accuracy and fast calculation efficiency.

[0097] S24. Use the Y coordinate to remove pseudo similar points in the defect data set:

[0098] The two defect points from which the data in len_sub1_list and len_sub2_list are obtained by subtracting have been recorded before. Now the data in len_sub1_list and len_sub2_list will be screened again.

[0099] (1) Fig.10In the case where the difference between the X coordinates of the two pairs of points is the same, but the difference between the Y coordinates is not equal and the difference is large, in order to avoid this situation being misjudged as the same point pair and matching successfully, the difference between the Y coordinates of the two defect points corresponding to the first data in len_sub1_list is compared with the difference between the Y coordinates of the two defect points corresponding to the first data in len_sub2_list. If the absolute value of the difference is less than B (the recommended value is 1-2mm), the first data in len_sub1_list and the first data in len_sub2_list are retained, and the second data in len_sub1_list and the second data in len_sub2_list are compared in the same way as above, and so on. If the difference is not less than B, the first data in len_sub1_list and the first data in len_sub2_list are eliminated, and the second data in len_sub1_list and the second data in len_sub2_list are compared in the same way as above, and so on.

[0100] (2) Fig.11 In the two tests, although the difference between the X coordinates and the Y coordinates match, the Y coordinates do not match and the positions are far apart. In order to prevent this situation from being misjudged as a similar point, the larger Y coordinate value of the two defect points corresponding to the first data in len_sub1_list is compared with the larger Y coordinate value of the two defect points corresponding to the first data in len_sub2_l ist for difference, and then the smaller Y coordinate value of the two defect points corresponding to the first data in len_sub1_list is compared with the smaller Y coordinate value of the two defect points corresponding to the first data in len_sub2_l ist for difference. If the absolute values ​​of the two differences are both less than C (to eliminate errors, the recommended value of C is 1-4mm), the first data in len_sub1_list and the first data in len_sub2_list are retained, and the second data in len_sub1_list and the second data in len_sub2_list are compared in the same way, and so on. If the absolute values ​​of the two differences are both greater than or equal to C, then eliminate the first data in len_sub1_list and the first data in len_sub2_list, and continue to compare the second data in len_sub1_list and the second data in len_sub2_list, using the same method as above, and so on.

[0101] S25. Eliminate pseudo similarities during data alignment:

[0102] Because the corresponding similar points in len_sub1_list and len_sub2_list are found in pairs, some pseudo-similar points can be eliminated by using the change of one point to verify whether the other point has changed accordingly.

[0103] The data in len_sub1_list and len_sub2_list are obtained by the absolute value of the difference in the X coordinates of the two defect points in the respective detection data. The two defect points (Defect_x2_1 i , Defect_y2_1 i ),(Defect_x2_2 i , Defect_y2_2 i ) The two defect points (Defect_x1_1) corresponding to the i-th data in len_sub1_list i , Defect_y1_1 i ),(Defect_x1_2 i , Defect_y1_2 i ) as the basis for data alignment and define:

[0104] len_max i =max(Defect_x2_1 i ,Defect_x2_2 i )-max(Defect_x1_1 i ,Defect_x1_2 i )

[0106] len_min i =min(Defect_x2_1 i ,Defect_x2_2 i )-min(Defect_x1_1 i ,Defect_x1_2 i )

[0108] wide_max i =max(Defect_y2_1 i ,Defect_y2_2 i )-max(Defect_y1_1 i ,Defect_y1_2 i )

[0110] (1) Fig.12In the example, the relative coordinate differences of the two pairs of points in X and Y are the same, but the distribution positions in the X coordinate direction are very different. In fact, they are not similar points that can be matched. In order to avoid this situation being misjudged as similar points and successfully matched, it is necessary to determine whether |len_maxi| is greater than ΔL (ΔL is the absolute value of the difference between the lengths of the two detection rolls) when i=1, 2...n (n is the number of data in len_sub1_list or len_sub2_list). If not, the i-th data in len_sub1_list and the i-th data in len_sub2_list are discarded. If so, the i-th data in len_sub1_list and the i-th data in len_sub2_list are retained.

[0111] (2) Fig.13 In the figure, the X coordinates of the two defect points in the left figure are similar to those of the two defect points in the right figure, but the Y coordinates are reversed. In order to prevent this situation from being misjudged as similar points, the black solid defect point in the right figure below is moved to the corresponding position of the black solid defect point in the left figure, and its displacement is recorded. The black hollow defect in the right figure is moved by the same displacement amount to determine whether the position of the black hollow defect in the right figure is similar to that of the black hollow defect in the left figure after displacement. The calculation process is as follows:

[0112] len_max i =Defect_x2_s i -Defect_x1_s i

[0113] wide_max i =Defect_y2_s i -Defect_y1_s i

[0114] Defect_x2_k i =Defect_x2_k i -len_max

[0115] Defect_y2_k i =Defect_y2_k i -wide_max

[0116] When i=1, 2...n (n is the number of data in len_sub1_list or len_sub2_list), judge in sequence:

[0117] |Defect_x1_k i -Defect_x2_k i |==|len_sub1_i|-|len_sub2_i|and

[0118] |Defect_y1_k i -Defect_y2_k i |<D (Suggested value: 1 - 2 mm)

[0119] Whether it holds.

[0120] len_sub1_i represents the i-th data in len_sub1_list, and len_sub2_i represents the i-th data in len_sub2_list.

[0121] If it does not hold, the i-th data in len_sub1_list and len_sub2_list will be excluded, and so on.

[0122] Finally, new len_sub1_list and len_sub2_list are obtained, and the two defect points corresponding to the data at the corresponding index positions are the similar features of the two detections.

[0123] Take the two defect points corresponding to the first data in len_sub1_list and the two defect points corresponding to the first data in len_sub2_list as the similar features of the two detections. len_max is the offset in the X coordinate direction of the latter detection relative to the former detection. Use len_max and the length Coil_len2 of the latter detection coil to align the data of the latter detection and the former detection.

[0124] S26. Align the data and calculate the head and tail shearing amounts of the previous detection:

[0125] The similar feature points have been obtained previously. Using the offset between the similar feature points, the data of the two detections are aligned. The part where the length of the previous detection exceeds the length of the latter detection after alignment is the head shearing amount and the tail shearing amount.

[0126] First, mark the starting position 0 and the ending position Coil_len2 of the length of the latter detection coil, and add points (0, 0, -1) and (Coil_len2, 0, -1) to the defect data of the latter detection coil.

[0127] The length data (new) in the defect data of the latter detection = the length position in the defect data of the latter detection - len_max.

[0128] The length data (Defect_x2) in the defect data (new) of the last detection and the length data (Defect_x1) in the defect data of the previous detection need to meet the conditions (the point (0, 0, -1), (Coil_len2, 0, -1) becomes the point (-len_max, 0, -1), (Coil_len2-len_max, 0, -1), the head and tail of the steel coil are removed and the defect data changes accordingly):

[0129] -len_max <Defect_x1<Coi l_len2-len_max

[0130] -len_max <Defect_x2<Coi l_len2-len_max

[0131] Coil_len2: The length of the coil to be detected next time.

[0132] The defect data (new) is screened according to the above conditions to obtain new defect data.

[0133] Then use the new defect data to get the final aligned defect data:

[0134] The X coordinate in the previous defect detection data = the X coordinate in the previous defect detection data + len_max

[0135] X coordinate in the next defect detection data = X coordinate in the next defect detection data + len_max

[0136] Y coordinate in the next defect detection data = Y coordinate in the next defect detection data - wide_max

[0137] The head and tail shear amount of the previous test:

[0138] Head clipping amount: -len_max

[0139] Tail clipping amount: Coil_len1-(Coil_len2-len_max).

[0140] S3. Similarity evaluation:

[0141] The aligned defect data is divided into several voltage intervals, the defect data in each voltage interval is further divided into grids and the similarity of the number of defect data in the grids is calculated. Specifically, the following steps are included:

[0142] S31. Grid division:

[0143] In the length direction, because the length data has been aligned, the grids are divided equally to the right based on the length 0 of the steel coil (taking 100m as an example); in the width direction, the grids are divided equally from the center line of the strip width to both sides (taking 25mm as an example). Fig.14 shown.

[0144] S32, calculating the similarity evaluation values ​​under two detection voltage intervals:

[0145] Calculate the similarity evaluation value of each voltage interval, divide the voltage into several interval levels, and perform similarity evaluation on the defects in each voltage interval. First, calculate the difference percentage of the grid at the corresponding index position of the two tests in each voltage interval. The calculation logic is shown in Table 3:

[0146] Table 3 Grid difference percentage calculation logic

[0147]

[0148] Set the allowable similarity error range according to the actual situation (for example, if the difference is less than 5%, the similarity is 100%, if the difference is 5-10%, the similarity is 90%, if the difference is 10-15%, the similarity is 85%, and so on). Fig.15 shown.

[0149] After calculating the similarity of each grid in each voltage interval, the sum is divided by the total number of grids to obtain the average value, which is the similarity evaluation value under the corresponding voltage interval.

[0150] S33, calculate the comprehensive similarity evaluation:

[0151] The weighted average of the similarity evaluations for each voltage interval.

[0152] The internal defect detector is an instrument used to detect the internal inclusion defects of steel coils. The similarity of two tests can be used to evaluate the repeatability accuracy of the equipment. Since the length of the steel coils tested twice changes due to the need to cut the head and tail during the production of the cold rolling finishing unit, the position coordinates of the defects on the steel coils will also change accordingly. In addition, the head, tail and width positions of the steel coils tested twice will be flipped, so position alignment is the prerequisite for quantitative similarity evaluation. The present invention is based on the automatic alignment method of the steel coil position based on the data of two internal defect detections, and on this basis, the similarity of the two internal defect detection results is evaluated to evaluate the equipment accuracy.

[0153] Example 1

[0154] Taking the finishing unit of a cold rolling mill as an example, the unit is equipped with an internal shorting instrument. The same steel coil is tested twice in succession. Each time it is tested, it has to re-enter the unit and undergo head and tail shearing. The length changes, and the head, tail and width directions are reversed. The lengths of the steel coils tested twice are: 2348000mm and 2265000mm.

[0155] The original defect map is as follows Fig.16 as shown in (a) and (b).

[0156] Determine whether the two tests have been flipped. According to the number of interval tests between the two tests, determine whether the second test has been flipped. In this case, the two tests are the first test and the second test, and 2-1=1 is an odd number, so the latter test has been flipped, and the steel coil data needs to be flipped.

[0157] Turn over the steel coil for the second test:

[0158] X-coordinate direction: X-coordinate of defect data after flipping = length of steel coil detected for the second time (2265000mm) - X-coordinate of defect data before flipping

[0159] Y coordinate direction: Y coordinate of defect data after flipping = Y coordinate of strip width centerline (235mm) - (Y coordinate of defect data before flipping - Y coordinate of strip width centerline (235mm))

[0160] Find similar feature points:

[0161] The first inspection defect data: (1642713, 277, 7.2), (1903817, 205, 8.2)······

[0162] Second inspection defect data: (1591788, 274, 7.2), (1852810, 202, 8.2)······

[0163] 1) Filter the defect data with ≥4V in turn

[0164] In the first detection, the voltage of the defect data (1642713, 277, 7.2) is 7.2V and is retained. The defect data in the second detection is processed in the same way.

[0165] 2) Calculate the absolute value of the difference between the X coordinates of any two points in the first and second inspection defect data;

[0166] For example, (1642713, 277, 7.2), (1903817, 205, 8.2), |1642713-1903817|=261104;

[0167] Put into the list len_sub1_list = [261104];

[0168] (1591788, 274, 7.2), (1852810, 202, 8.2), |1591788 - 1852810| = 261022. Put into the list len_sub2_list = [261022].

[0169] 3) Use the bisection method to find the threshold A

[0170] Set the initial point start = 0, end = 300, and successively take the median to find the minimum value A that can find similar feature points. Finally, the optimal solution is found to be 82.

[0171] 4) Screen out the points where the difference in the distance between two points in the first detection and the difference in the distance between two points in the second detection is not greater than A

[0172] For example, |261104 - 261022| <= A = 82, retain 261104 in len_sub1_list and 261022 in len_sub2_list. Make corresponding judgments for other points in len_sub1_list and len_sub2_list.

[0173] 5) Use the Y - coordinate data to eliminate false similar points

[0174] For example, the Y - coordinates of the two points corresponding to 261104 in len_sub1_list are 277 and 205, 277 - 205 = 72; the Y - coordinates of the two points corresponding to 261022 in len_sub2_list are 274 and 202, 274 - 202 = 72. |72 - 72| < B = 1, and |277 - 274| < C = 4, |205 - 202| < C = 4.

[0175] Retain 261104 in len_sub1_list and 261022 in len_sub2_list. Make corresponding judgments for other points in len_sub1_list and len_sub2_list.

[0176] 6) Eliminate false similar points during the data alignment process

[0177] a. For example, the maximum value of the X - coordinates of the two points corresponding to 261104 in len_sub1_list is 1903817; the maximum value of the X - coordinates of the two points corresponding to 261022 in len_sub2_list is 1852810.

[0178] |1852810 - 1903817| = 51007 < ΔL = |2348000 - 2265000| = 83000

[0179] Retain 261104 in len_sub1_list and 261022 in len_sub2_list. Make corresponding judgments for other points in len_sub1_list and len_sub2_list.

[0180] b. 1852810 - 1903817 = -51007, 202 - 205 = -3;

[0181] 1591788 - (-51007) = 1642795, 274 - (-3) = 277;

[0182] |1642713 - 1642795| = |261104 - 261022| and |277 - 277| < D = 1

[0183] Retain 261104 in len_sub1_list and 261022 in len_sub2_list. Make corresponding judgments for other points in len_sub1_list and len_sub2_list.

[0184] 261022 in len_sub1_list and 261022 in len_sub2_list meet all the screening conditions, and their corresponding defect points are similar feature points. Data alignment can be achieved using the offset between similar feature points.

[0185] The X coordinates of the defect data in the second inspection are all -(-51007) and are aligned with the first inspection in the length direction. Then, data with X coordinates between -(-51007) and 2348000 - (-51007) are selected after removing the head and tail, and other defects are eliminated.

[0186] The X coordinate in the first detected defect data = the X coordinate in the first detected defect data + (-51007)

[0187] The X coordinate in the second detected defect data = the X coordinate in the second detected defect data + (-51007)

[0188] The Y coordinate in the second detected defect data = the Y coordinate in the second detected defect data - (-3)

[0189] The head and tail shearing amounts in the previous inspection:

[0190] The head shearing amount: -(-51007) = 51007

[0191] The tail shearing amount: 2348000 - (2265000 - (-51007)) = 31993

[0192] The similar feature points found are as follows Fig.17 As shown in (a) and (b), the circled points are similar feature points determined and output by the model. According to the offset between the two similar feature points detected, the two maps can be aligned and the head and tail shearing amount of the steel coil can be obtained.

[0193] The scatter plot after data alignment is as follows Fig.18 as shown in (a) and (b).

[0194] After the defect data is aligned, the similarity of the two tests is evaluated. In this example, the detection voltage range of the internal short circuit instrument equipment is 0-12V, and the voltage intervals are divided into 0-1.2V, 1.2-1.5V, 1.5-1.8V, 1.8-2V, 2-3V, 3-4V, 4-5V, and 5-12V.

[0195] Here we take the similarity evaluation in the 2-3V voltage range as an example. Fig.19 as shown in (a) and (b).

[0196] The grids are divided into intervals of 100m in length and 25mm in width, and the similarity of each corresponding grid is evaluated, such as Fig. 20 shown.

[0197] The similarity in the 2-3V voltage range is the weighted average of a single grid: 88.5%

[0198] Similarly, the similarity evaluations for the voltage ranges 0-1.2V, 1.2-1.5V, 1.5-1.8V, 1.8-2V, 2-3V, 3-4V, 4-5V, and 5-12V are 63%, 77.5%, 88.8%, 95.7%, 94.9%, 97.6%, and 98.7%.

[0199] The comprehensive similarity evaluation is the weighted average of the similarities of each voltage interval: 88.1%.

[0200] Example 2

[0201] Take a cold rolling mill finishing unit as an example. The unit is equipped with an internal underrun meter. The same steel coil is tested three times in a row. Each time it is tested, it has to re-enter the unit, undergo head and tail shearing, change its length, and reverse the head, tail and width directions. Take the first and third tests:

[0202] The lengths of the steel coils tested twice are: 2441000mm and 2348000mm.

[0203] The original defect map is as follows Fig.21 as shown in (a) and (b).

[0204] Determine whether the two tests have been flipped. According to the number of interval tests between the two tests, determine whether the third test has been flipped. In this case, the two tests are the first test and the third test, and 3-1=1 is an even number, so the latter test has not been flipped, and there is no need to flip the steel coil data.

[0205] Find similar feature points:

[0206] The first inspection defect data: (1695461, 276, 7.2), (1985994, 122, 8.2)······

[0207] The third inspection defect data: (1642713, 276, 7.2), (1933249, 122, 8.2)······

[0208] 1) Filter the defect data with ≥4V in turn

[0209] The voltage of the defect data (1695461, 276, 7.2) in the first detection is 7.2V and is retained. The defect data in the third detection is processed in the same way.

[0210] 2) Calculate the absolute value of the difference between the X coordinates of any two points in the first and third inspection defect data

[0211] For example, (1695461, 276, 7.2), (1985994, 122, 8.2), |1695461-1985994|=290533;

[0212] Put into the list len_sub1_list = [290533];

[0213] (1642713, 276, 7.2), (1933249, 122, 8.2), |1642713-1933249|=290536;

[0214] Put into the list len_sub2_list = [290536].

[0215] 3) Binary search for threshold A

[0216] Set the initial point start = 0, end = 300, take the median value in turn to find the minimum value A that can find similar feature points, and finally find the optimal solution of 3.

[0217] 4) Select points where the difference in the distance between two points in the first detection and the difference in the distance between two points in the third detection is not greater than A, such as |290533 - 290536| <= 3. Retain 290533 in len_sub1_list and 290536 in len_sub2_list. Make corresponding judgments for other points in len_sub1_list and len_sub2_list.

[0218] 5) Use the Y - coordinate data to eliminate pseudo - similar points

[0219] For example, the Y - coordinates of the two points corresponding to 290533 in len_sub1_list are 276 and 122, and 276 - 122 = 154; the Y - coordinates of the two points corresponding to 290536 in len_sub2_list are 276 and 122, and 276 - 122 = 154. |154 - 154| < B = 1, and |276 - 276| < C = 1, |122 - 122| < C = 1. Retain 290533 in len_sub1_list and 290536 in len_sub2_list. Make corresponding judgments for other points in len_sub1_list and len_sub2_list.

[0220] 6) Eliminate pseudo - similar points during the data alignment process

[0221] a. For example, the maximum value of the X - coordinates of the two points corresponding to 290533 in len_sub1_list is 1985994; the maximum value of the X - coordinates of the two points corresponding to 290536 in len_sub2_list is 1933249.

[0222] |1933249 - 1985994| = 52745 < ΔL = |2441000 - 2348000| = 93000

[0223] Retain 290533 in len_sub1_list and 290536 in len_sub2_list. Make corresponding judgments for other points in len_sub1_list and len_sub2_list.

[0224] b. 1933249 - 1985994 = - 52745, 122 - 122 = 0;

[0225] 1642713 - (-52745) = 1695458, 276 - 0 = 276;

[0226] |1695458 - 1695461| = |290533 - 290536| and |276 - 276| < D = 1

[0227] Keep 290533 in len_sub1_list and 290536 in len_sub2_list. Make corresponding judgments on other points of len_sub1_list and len_sub2_list.

[0228] 290533 in len_sub1_list and 290536 in len_sub2_list meet all the screening conditions, and their corresponding defect points are similar feature points. Data alignment can be achieved by using the offset between similar feature points.

[0229] The X coordinates of the defective data in the third inspection are all -(-52745) and aligned in the length direction with the first inspection. Then the data with X coordinates between -(-52745) and 2348000-(-52745) are filtered out after removing the head and tail, and all other defects are eliminated.

[0230] X coordinate in the first defect detection data = X coordinate in the first defect detection data + (-52745)

[0231] The X coordinate of the third detection defect data = the X coordinate of the third detection defect data + (-52745)

[0232] The Y coordinate of the third defect detection data = the Y coordinate of the third defect detection data - 0

[0233] The head and tail shear amount of the previous test:

[0234] Head shearing amount: -(-52745) = 52745

[0235] Tail shearing amount: 2441000-(2348000-(-52745))=40255

[0236] The similar feature points found are as follows Fig. 22 As shown in (a) and (b), the circled points are similar feature points determined and output by the model. According to the offset between the two similar feature points detected, the two maps can be aligned and the head and tail shearing amount of the steel coil can be obtained.

[0237] The scatter plot after data alignment is as follows Fig.23 as shown in (a) and (b).

[0238] After the defect data is aligned, the similarity of the two tests is evaluated. In this example, the detection voltage range of the internal short circuit instrument equipment is 0-12V, and the voltage intervals are divided into 0-1.2V, 1.2-1.5V, 1.5-1.8V, 1.8-2V, 2-3V, 3-4V, 4-5V, and 5-12V.

[0239] Here we take the similarity evaluation in the 2-3V voltage range as an example. Fig.24 as shown in (a) and (b).

[0240] The grids are divided into intervals of 100m in length and 25mm in width, and the similarity of each corresponding grid is evaluated, such as Fig.25 shown.

[0241] The similarity in the 2-3V voltage range is the weighted average of a single grid: 85%

[0242] Similarly, the similarity evaluations for the voltage ranges 0-1.2V, 1.2-1.5V, 1.5-1.8V, 1.8-2V, 2-3V, 3-4V, 4-5V, and 5-12V are 74%, 80.5%, 84.7%, 93.2%, 94.7%, 97.1%, and 98.8% respectively.

[0243] The comprehensive similarity evaluation is the weighted average of the similarities of each voltage interval: 88.5%.

[0244] Those skilled in the art should recognize that the above embodiments are only used to illustrate the present invention, and are not intended to limit the present invention. As long as they are within the spirit of the present invention, any changes or modifications to the above embodiments will fall within the scope of the claims of the present invention.

Claims

1. A method for automatic alignment of steel coil position based on two undercut defect detection data, characterized in that: The following steps are involved: S1. Determine whether the steel coil is turned over and process the data; S2, finding similar feature points of the defect data of the two tests, and aligning the defect data of the two tests based on the defect data of the previous test according to the relative offset between the similar feature points; S3, dividing the aligned defect data into several voltage intervals, dividing the defect data in each voltage interval into grids again, and calculating the similarity of the number of defect data in the grids.

2. The method for automatic alignment of steel coil position based on two undercut defect detection data according to claim 1 is characterized in that: The step S1 specifically includes the following steps: S11, judging whether the steel coil is turned over, if the difference between the two detection times is an odd number, the steel coil is turned over, and if it is an even number, the steel coil is not turned over; S12. If the steel coil is turned over during the two inspections, the defect data of the latter inspection is processed; if it is not turned over, the defect data is not processed.

3. The method for automatic alignment of steel coil position based on two undercut defect detection data according to claim 2 is characterized in that: The step S2 specifically includes the following steps: S21, respectively calculating the X-coordinate difference data set between two points of defect data with larger values ​​in the defect data detected twice; S22, determining a threshold value A; S23, screening the data set according to the distance difference between two points of defect data detected in the previous time and the distance difference between two points of defect data detected in the next time not exceeding the threshold value A; S24, using the Y coordinate to eliminate pseudo similar points in the defect data set; S25, obtaining similar feature points of two detections; S26. Align the defect data using similar feature points and obtain the head shear force and tail shear amount verified last time.

4. The method for automatic alignment of steel coil position based on two undercut defect detection data according to claim 3 is characterized in that: The step S3 specifically comprises the following steps: S31, dividing the grid; S32, calculating the similarity evaluation values ​​under two detection voltage intervals; S33, calculating the similarity evaluation values ​​under two detection voltage intervals; S34. Calculate the comprehensive similarity evaluation.

Images