Anomaly screening method and apparatus for a bolt tightening process

Abstract

The application aims to provide an abnormality screening method and device for a bolt tightening process. Compared with the prior art, the application acquires a quantitative curve of the bolt tightening process; determines a similarity value of the bolt tightening process and a preset standard tightening process according to the quantitative curve and a preset standard line through a Fréchet distance, wherein the preset standard line is a quantitative curve of a standard bolt tightening process; and determines whether the bolt tightening process is abnormal according to a preset similarity threshold and the similarity value. In this way, a fast abnormality screening method with low cost is realized, and the method is stable in the scene of variable frequency acquisition of bolt tightening data.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to an anomaly screening technique for the bolt tightening process. Background Technology

[0002] Existing methods for detecting abnormal bolt tightening typically involve model training. However, model training is costly in terms of both time and application. Furthermore, since abnormal bolt tightening is rare in actual production processes, it is difficult to provide sufficient abnormal data for model training, resulting in high training time costs.

[0003] Meanwhile, since bolt tightening is usually done by frequency conversion sampling, existing methods for detecting bolt tightening anomalies perform poorly in frequency conversion sampling scenarios. Summary of the Invention

[0004] The purpose of this application is to provide a method and device for screening anomalies in the bolt tightening process.

[0005] According to one aspect of this application, a method for anomaly screening in a bolt tightening process is provided, wherein the method includes:

[0006] Obtain a quantitative curve of the bolt tightening process;

[0007] The similarity value between the bolt tightening process and the preset standard tightening process is determined by the Fréchet distance based on the quantization curve and the preset standard line, wherein the preset standard line is the quantization curve of the standard bolt tightening process;

[0008] The bolt tightening process is determined to be abnormal based on a preset similarity threshold and the similarity value.

[0009] Furthermore, the bolt tightening process has a unique resulting torque, wherein, after obtaining the quantization curve of the bolt tightening process, the process further includes:

[0010] Based on the resulting torque, a target segment before reaching the resulting torque is extracted from the quantization curve;

[0011] The step of determining the similarity value between the bolt tightening process and the preset standard tightening process based on the quantization curve and the preset standard line using the Fréchet distance includes:

[0012] The similarity value between the bolt tightening process and the preset standard tightening process is determined by the Fréchet distance based on the target segment and the standard segment, wherein the standard segment is the segment in the preset standard line before reaching its corresponding result torque.

[0013] Furthermore, the quantization curve includes a torque-time curve, a preset fitting threshold, and a fitting ratio, wherein the step of extracting a target segment from the quantization curve before reaching the result torque based on the result torque includes:

[0014] Determine the first point (x1, y1) corresponding to the resulting torque y1 on the torque-time curve;

[0015] The second point (x2, y2) on the time-torque curve is set according to the fitting ratio p, where y2 / y1 = p;

[0016] Determine the fitted straight line based on the first point and the second point;

[0017] If the third point (x3, y3) on the fitted line and the fourth point (x3, y4) on the torque-time curve satisfy |y3-y4|=a, the fourth point is taken as the starting point of the target segment, where a is the fitting threshold.

[0018] The segment of the torque-time curve located between the starting point and the first point is taken as the target segment.

[0019] Furthermore, the quantification curve also includes a torque angle curve, wherein determining the similarity value between the bolt tightening process and the preset standard tightening process based on the target segment and the standard segment using the Fréchet distance includes:

[0020] Determine the torque angle segment corresponding to the target segment in the torque angle curve;

[0021] The similarity between the torque angle segment and the standard torque angle segment is calculated based on the Fréchet distance to determine the similarity value between the bolt tightening process and the preset standard tightening process. The standard torque angle segment is the corresponding segment of the standard segment in the torque angle curve corresponding to the preset standard line.

[0022] Further, the calculation of the similarity between the torque angle segment and the standard torque angle segment based on the Fréchet distance includes:

[0023] The similarity between the torque angle segment and the standard torque angle segment is calculated in a non-recursive manner based on the Fréchet distance.

[0024] Furthermore, both the torque angle segment and the standard torque angle segment consist of several sampling points containing torque and angle. The calculation of the similarity between the torque angle segment and the standard torque angle segment using a non-recursive method based on the Fréchet distance includes:

[0025] Construct a distance matrix between the sampling points of the torque angle segment and the standard torque angle segment;

[0026] The sampling point matching sequence between the torque angle segment and the standard torque angle segment is determined based on the distance matrix;

[0027] The distance between each pair of matched torque angle segment sampling points and the sampling points of the standard torque angle segment is calculated based on the sampling point matching sequence.

[0028] The maximum value of the distance is used as the similarity between the torque angle segment and the standard torque angle segment.

[0029] Furthermore, the construction of the distance matrix between the sampling points of the torque angle segment and the standard torque angle segment includes:

[0030] Construct the distance matrix D = []m×n, where m and n are the number of sampling points for the torque angle segment A and the standard torque angle segment B, respectively.

[0031] D[0,0] = dist(A0,B0),

[0032] D[i,0]=dist(A i ,B0)+D[i-1,0], where i∈[1,m-1],

[0033] D[0,j]=dist(A0,B j )+D[0,j-1], where j∈[1,n-1],

[0034] D[i,j]=dist(A i B j )+min(D[i-1,j],D[i,j-1],D[i-1,j-1]), where

[0035] i∈[1,m-1], j∈[1,n-1],

[0036] Among them, dist(A p B q ) is the Euclidean distance between the p-th sampling point of the torque angle segment and the q-th sampling point of the standard torque angle segment.

[0037] Further, determining the sampling point matching sequence between the torque angle segment and the standard torque angle segment based on the distance matrix includes:

[0038] Using D[m-1,n-1] as the starting point of the sampling point matching sequence, determine the minimum value D[a,b] among its three adjacent matrix units, and add D[a,b] to the sampling point matching sequence;

[0039] If a≠0 and b≠0, then repeat the above steps starting from D[a,b] until D[a,b]=D[0,0];

[0040] If a = 0 or b = 0, then all matrix units including D[0,0] on the straight path from D[a,b] to D[0,0] are sequentially added to the sampling point matching sequence.

[0041] Further, the step of calculating the distance value between each pair of matched torque angle segmented sampling points and the standard torque angle segmented sampling points based on the sampling point matching sequence includes:

[0042] Traverse the sampling point matching sequence, and for each traversed D[a,b], calculate the Euclidean distance between the a-th sampling point of the torque angle segment and the b-th sampling point of the standard torque angle segment.

[0043] According to another aspect of this application, a computer-readable medium is also provided, on which computer-readable instructions are stored, which can be executed by a processor to perform the operation as described above.

[0044] According to another aspect of this application, an anomaly screening device for a bolt tightening process is also provided, wherein the device comprises:

[0045] One or more processors; and

[0046] A memory storing computer-readable instructions, which, when executed, cause the processor to perform the operations described above.

[0047] Compared with existing technologies, this application obtains a quantitative curve of the bolt tightening process; based on the quantitative curve and a preset standard line, it determines the similarity value between the bolt tightening process and a preset standard tightening process using Fréchet distance, wherein the preset standard line is the quantitative curve of a standard bolt tightening process; and it determines whether the bolt tightening process is abnormal based on a preset similarity threshold and the similarity value. This method achieves a rapid anomaly screening approach with low application cost, and it performs stably in scenarios involving frequency conversion acquisition of bolt tightening data. Attached Figure Description

[0048] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0049] Figure 1 A flowchart of an anomaly screening method for a bolt tightening process according to one aspect of this application is shown.

[0050] Figure 2 A flowchart of an anomaly screening method for a bolt tightening process according to a preferred embodiment of this application is shown.

[0051] Figure 3 A flowchart of an anomaly screening method for a bolt tightening process according to another preferred embodiment of this application is shown;

[0052] Figure 4(a) shows the results of Fréchet distance in a frequency conversion acquisition scenario;

[0053] Figure 4(b) shows the results of DTW (Dynamic Time Warping) distance in a frequency conversion acquisition scenario.

[0054] The same or similar reference numerals in the accompanying drawings represent the same or similar parts. Detailed Implementation

[0055] The present invention will now be described in further detail with reference to the accompanying drawings.

[0056] In a typical configuration of this application, the terminal, the device of the service network, and the trusted party all include one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0057] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0058] Computer-readable media include both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include non-transitory computer-readable media, such as modulated data signals and carrier waves.

[0059] Bolt tightening, as a mechanical connection process, is widely used in assembly. With the development of automation technology, manufacturing enterprises are increasingly using automatic tightening guns to complete the tightening process. Compared with manual tightening, the operation of automatic tightening guns requires more stringent control, including ensuring reasonable torque and angle changes between the automatic tightening gun and the bolt during the tightening process, thereby ensuring the stability and high reliability of the entire tightening process. This application provides a method for screening tightening anomalies in scenarios where automatic tightening guns are used for bolt tightening. This method performs anomaly screening based on tightening data output by the automatic tightening gun controller. Specifically, based on the torque and angle data output by the automatic tightening gun controller, and combined with the time information of the collected data, a quantitative curve of the tightening process is constructed, and then anomalies are screened based on the constructed quantitative curve.

[0060] To further illustrate the technical means adopted and the effects achieved in this application, the technical solution of this application will be clearly and completely described below in conjunction with the accompanying drawings and preferred embodiments.

[0061] Figure 1 This application illustrates an anomaly screening method for a bolt tightening process, comprising:

[0062] S11 obtains the quantitative curve of the bolt tightening process;

[0063] S12 determines the similarity value between the bolt tightening process and the preset standard tightening process based on the quantization curve and the preset standard line through the Fréchet distance, wherein the preset standard line is the quantization curve of the standard bolt tightening process;

[0064] S13 determines whether the bolt tightening process is abnormal based on the preset similarity threshold and the similarity value.

[0065] In this embodiment, in step S11, a quantitative curve of the bolt tightening process is obtained; in step S12, the similarity value between the bolt tightening process and the preset standard tightening process is determined by the Fréchet distance based on the quantitative curve and the preset standard line.

[0066] Here, the preset standard line is a quantitative curve of the standard bolt tightening process. Specifically, the preset standard line is a quantitative curve of the standard tightening process of the same specification bolts according to the same tightening strategy as the tightening process to be judged, in accordance with the operation requirements of the automatic tightening gun. The preset standard line can be manually selected from multiple quantitative curves, or it can be constructed from the output data of the automatic tightening gun controller of a standard tightening process.

[0067] Here, the similarity between the quantization curve to be judged and the preset standard line is calculated using the Fréchet distance. Since the Fréchet distance aligns the two curves during the process of calculating the similarity and finding the matching trajectory, it eliminates the phase difference between the quantization curve to be judged and the preset standard line. Compared with other lockstep similarity calculation methods such as Euclidean distance, it achieves effective calculation of curve similarity.

[0068] Meanwhile, in actual scenarios where automatic bolt tightening guns are used, the data acquisition frequency during the tightening process cannot be guaranteed to be stable due to unstable changes in data processing volume and network fluctuations and delays. Frequency conversion acquisition is typically used. In contrast, the curve similarity calculation method of DTW distance, which equals the sum of the Euclidean distances of all paired points between two defined curves, results in a larger calculation result due to the higher sampling frequency and the more paired points on the curves for the same duration. This is unsuitable for frequency conversion sampling scenarios. Therefore, even if the DTW algorithm is used to flexibly measure the quantized curve to be judged against the preset standard line, the stability of the calculation result cannot be guaranteed. The Fréchet distance, on the other hand, is the maximum Euclidean distance of all paired points between two defined curves, and this calculation method is far more stable than the DTW algorithm in the aforementioned frequency conversion acquisition environment. Figure 4 compares the results of Fréchet distance and DTW algorithm at different sampling intervals. Figures 4(a) and 4(b) show the calculation results of Fréchet distance and DTW algorithm for the same batch curve in scenarios with sampling intervals of 6ms, 12ms and 18ms, respectively. It can be seen that the calculation results of Fréchet distance are basically the same, while the calculation results of DTW algorithm are very different.

[0069] Preferably, determining the similarity value between the bolt tightening process and the preset standard tightening process based on the quantization curve and the preset standard line using the Fréchet distance includes: calculating the similarity between the quantization curve and the preset standard line using a non-recursive method based on the Fréchet distance. Here, the non-recursive calculation method, compared to the recursive method, requires less computation and is faster, effectively improving the speed of anomaly screening.

[0070] Preferably, both the quantization curve and the preset standard line are composed of a plurality of sampling points containing torque and angle. The step of calculating the similarity between the quantization curve and the preset standard line using a non-recursive method based on the Fréchet distance includes: constructing a distance matrix of the sampling points of the quantization curve and the preset standard line; determining a sampling point matching sequence between the quantization curve and the preset standard line based on the distance matrix; calculating the distance value between each pair of matched sampling points of the quantization curve and the preset standard line based on the sampling point matching sequence; and using the maximum value of the distance value as the similarity between the quantization curve and the preset standard line.

[0071] Specifically, the distance matrix D = []m×n is constructed, where m and n are the number of sampling points of the quantization curve A and the preset standard line B, respectively:

[0072] D[0,0] = dist(A0,B0),

[0073] D[i,0]=dist(A i ,B0)+D[i-1,0], where i∈[1,m-1],

[0074] D[0,j]=dist(A0,B j )+D[0,j-1], where j∈[1,n-1],

[0075] D[i,j]=dist(A i B j )+min(D[i-1,j],D[i,j-1],D[i-1,j-1]), where

[0076] i∈[1,m-1], j∈[1,n-1],

[0077] Among them, dist(A p B q ) represents the Euclidean distance between the p-th sampling point of the quantization curve and the q-th sampling point of the preset standard line.

[0078] Here, the distance between each point in two sampling point sequences is determined using a non-recursive matrix value calculation method. This method has a time complexity of O(n), which is significantly lower than the O(n) time complexity of the recursive method. 2 The non-recursive method has a lower time complexity than the recursive method, making it significantly faster for calculating long sequences of sampling points. For example, in practical applications, quantization curves typically have 500-2000 sampling points, and the corresponding length of the input sampling point sequence is also between 500-2000 sequence units. In this scenario, the non-recursive method effectively improves the computation speed.

[0079] In the process of determining the sampling point matching sequence of the quantization curve and the preset standard line based on the distance matrix, D[m-1,n-1] is taken as the starting point of the sampling point matching sequence, the minimum value D[a,b] of its three adjacent matrix units is determined, and D[a,b] is added to the sampling point matching sequence.

[0080] If a≠0 and b≠0, then repeat the above steps starting from D[a,b] until D[a,b]=D[0,0];

[0081] If a = 0 or b = 0, then all matrix units including D[0,0] on the straight path from D[a,b] to D[0,0] are sequentially added to the sampling point matching sequence.

[0082] In this way, a path is determined in the distance matrix from the starting point of the two sampling point sequences to the ending point of the two sampling point sequences (i.e., from D[m-1,n-1] to D[0,0] in the distance matrix), and the sum of the distance element values ​​on this path is minimized.

[0083] Traversing the sampling point matching sequence determined by the above method, for each traversed D[a,b], the Euclidean distance between the a-th sampling point of the quantization curve and the b-th sampling point of the preset standard line is calculated, and the maximum value of the distance is taken as the similarity between the quantization curve and the preset standard line. Here, the Fréchet distance between the quantization curve and the preset standard line is determined in this way.

[0084] Furthermore, in step S13, it is determined whether the bolt tightening process is abnormal based on a preset similarity threshold and the similarity value.

[0085] The above method identifies quantization curves with low similarity to preset standard lines. The bolt tightening process corresponding to the quantization curve is identified as an abnormal tightening process. This method enables rapid screening of abnormal bolt tightening processes. On the one hand, the screening results can be directly used to guide production. On the other hand, in scenarios requiring precise anomaly detection, the rapid screening results can be used as training data to establish a supervised learning model and achieve high accuracy in anomaly detection.

[0086] In another preferred embodiment, such as Figure 2 As shown, where, Figure 2 Steps S21 and S24 in Figure 1 Steps S11 and S13 in the embodiments are the same or substantially the same, and therefore will not be repeated here, but are included here by reference only. Steps S22 and S23 include: the bolt tightening process has a unique result torque; after obtaining the quantization curve of the bolt tightening process, a target segment before reaching the result torque is extracted from the quantization curve according to the result torque; the similarity value between the bolt tightening process and the preset standard tightening process is determined by Fréchet distance based on the target segment and the standard segment, wherein the standard segment is the segment in the preset standard line before reaching its corresponding result torque.

[0087] Here, the resulting torque refers to the final torque actually reached with reference to the standard tightening torque corresponding to bolts of different specifications. This resulting torque value is close to the standard tightening torque; therefore, each quantitative curve constructed from tightening process data has a unique corresponding resulting torque. Since torque-based or torque-angle-based tightening strategies are often used in actual tightening processes, under a step-by-step tightening strategy, the final tightening action that reaches the resulting torque plays a decisive role in the tightening quality. The one or more tightening actions before this play a negligible role in the tightening quality. Therefore, to improve anomaly screening efficiency and reduce data computation, only the final tightening action that reaches the resulting torque in the tightening process to be judged is extracted. The similarity between this action and the final tightening action that reaches the resulting torque in the standard tightening process is analyzed, and anomaly screening is performed based on this similarity.

[0088] In this embodiment, in step S22, the bolt tightening process has a unique result torque. After obtaining the quantization curve of the bolt tightening process, the target segment before reaching the result torque is extracted from the quantization curve according to the result torque.

[0089] Specifically, the quantization curve includes a torque-time curve, a preset fitting threshold and fitting ratio, and a first point (x1, y1) corresponding to the result torque y1 on the torque-time curve; a second point (x2, y2) on the time-torque curve is set according to the fitting ratio p, where y2 / y1 = p; a fitting straight line is determined based on the first point and the second point; if a third point (x3, y3) on the fitting straight line and a fourth point (x3, y4) on the torque-time curve satisfy |y3-y4| = a, the fourth point is taken as the starting point of the target segment, where a is the fitting threshold; the curve segment on the torque-time curve located between the starting point and the first point is taken as the target segment.

[0090] Here, a method is provided to extract the final tightening action that reaches the result torque from a complete quantitative curve of the tightening process. Since both the torque method and the torque angle method involve gradually tightening until the standard tightening torque corresponding to the bolt is reached, adjacent tightening steps can be segmented based on the sudden drop in torque in the torque-time curve.

[0091] Specifically, first, the first point on the torque-time curve corresponding to the arrival of the final torque is determined. A fitting ratio is pre-set based on experience; this ratio indicates that after a certain torque value, the tightening process will no longer be segmented, meaning that this tightening operation must be the final tightening action to reach the final torque. For example, setting the fitting ratio to 80% means that if the measured torque is greater than or equal to 80% of the final torque, then the corresponding tightening action must be the final tightening action to reach the final torque. Second, a second point corresponding to this fitting ratio is determined on the torque-time curve. Then, a straight line is fitted based on the first and second points. Accordingly, the sampling points corresponding to the final tightening action to reach the final torque should be discretely distributed around this fitted straight line. Based on experience, a fitting threshold is pre-set. This threshold indicates the acceptable discrete distance. For example, if the torque difference between the third point on the fitted line and the fourth point on the torque-time curve at the same time is greater than the fitting threshold, it indicates that the fourth point does not belong to the final tightening action to reach the final torque. If the torque difference between the third point on the fitted line and the fourth point on the torque-time curve at the same time is equal to the fitting threshold, it indicates that the fourth point is the starting point of the final tightening action to reach the final torque. Finally, the final tightening action is extracted from the complete quantization curve using the above method.

[0092] Correspondingly, the preset standard line is also extracted in the same way as described above for the final tightening action, and the extracted curve is used as the standard segment.

[0093] Furthermore, in this embodiment, in step S23, the similarity value between the bolt tightening process and the preset standard tightening process is determined by the Fréchet distance based on the target segment and the standard segment, wherein the standard segment is the segment in the preset standard line before reaching its corresponding result torque.

[0094] Here, the similarity between the target segment corresponding to the final tightening action extracted from the quantization curve to be judged and the standard segment extracted from the preset standard line is calculated using Fréchet distance. In this embodiment, torque variation is used as the main anomaly screening factor.

[0095] In another preferred embodiment, see Figure 3 ,in, Figure 3 Step S35 and Figure 1 Step S13 in the embodiments is the same or substantially the same, and therefore will not be repeated here, but is included here by reference only. Steps S31 to S34 include: obtaining a quantitative curve of the bolt tightening process, the quantitative curve including a torque-time curve and a torque-angle curve; the bolt tightening process has a unique result torque, and the target segment corresponding to the last tightening action to reach the result torque is extracted according to the torque-time curve; the torque angle segment corresponding to the target segment is determined in the torque-angle curve; the similarity between the torque-angle segment and the standard torque-angle segment is calculated according to the Fréchet distance to determine the similarity value between the bolt tightening process and the preset standard tightening process, where the standard torque-angle segment is the corresponding segment of the standard segment in the torque-angle curve corresponding to the preset standard line.

[0096] Specifically, in this embodiment, the quantification curve includes a torque-time curve and a torque-angle curve. After extracting the target segment corresponding to the final tightening action that reaches the result torque based on the torque-time curve, the torque-angle segment corresponding to the target segment is determined in the torque-angle curve. The similarity between the torque-angle segment and the standard torque-angle segment is calculated based on the Fréchet distance to determine the similarity value between the bolt tightening process and the preset standard tightening process. Here, the standard torque-angle segment is the corresponding segment of the standard segment in the torque-angle curve corresponding to the preset standard line.

[0097] Here, the final tightening action that reaches the result torque is first extracted based on the torque-time curve. Then, the angle factor is introduced, and the changes in torque and angle are used together as factors for anomaly screening. Compared with the case that only considers the torque factor, this scheme considers more dimensions and the detection results are more accurate.

[0098] Specifically, in this embodiment, the quantitative curve of the tightening process to be determined includes a torque-time curve and a torque-angle curve. Correspondingly, the preset standard line of the standard tightening process includes a standard torque-time curve and a standard torque-angle curve. First, the target segment corresponding to the final tightening action is extracted from the torque-time curve to be judged. Correspondingly, the standard segment corresponding to the final tightening action is extracted from the standard torque-time curve. Second, based on the time information of the target segment, the torque angle segment corresponding to the target segment is determined in the torque angle curve to be judged. It should be clarified that although the torque angle curve does not directly display the time information, it still uses time information as the basis for the arrangement of sampling points. Therefore, the time information is implicit information in the torque angle curve. Based on this, the torque angle segment corresponding to the target segment can be determined based on the time information of the target segment. Correspondingly, the standard torque angle segment corresponding to the standard segment is determined in the standard torque angle curve based on the time information of the standard segment. Finally, the Fréchet distance between the torque angle segment to be judged and the standard torque angle segment is calculated to determine the similarity between the bolt tightening process to be judged and the standard tightening process, thereby screening for abnormal tightening processes.

[0099] Specifically, the similarity between the torque angle segments and the standard torque angle segments is calculated using a non-recursive method based on the Fréchet distance. This non-recursive calculation method, compared to recursive methods, requires less computation and is faster, effectively improving the speed of anomaly screening.

[0100] Here, both the torque angle segment and the standard torque angle segment consist of several sampling points containing torque and angle. In calculating the similarity between the torque angle segment and the standard torque angle segment, a distance matrix is ​​constructed between the sampling points of the torque angle segment and the standard torque angle segment. Based on the distance matrix, a matching sequence of sampling points between the torque angle segment and the standard torque angle segment is determined. Based on the matching sequence, the distance value between the sampling points of each matched torque angle segment and the standard torque angle segment is calculated. The maximum value of the distance is taken as the similarity between the torque angle segment and the standard torque angle segment.

[0101] Specifically, the distance matrix D = []m×n is constructed, where m and n are the number of sampling points for torque angle segment A and standard torque angle segment B, respectively:

[0102] D[0,0] = dist(A0,B0),

[0103] D[i,0]=dist(A i ,B0)+D[i-1,0], where i∈[1,m-1],

[0104] D[0,j]=dist(A0,Bj )+D[0,j-1], where j∈[1,n-1],

[0105] D[i,j]=dist(A i B j )+min(D[i-1,j],D[i,j-1],D[i-1,j-1]), where

[0106] i∈[1,m-1], j∈[1,n-1],

[0107] Among them, dist(A p B q ) represents the Euclidean distance between the p-th sampling point of the torque angle segment and the q-th sampling point of the standard torque angle segment.

[0108] Here, the distance between each point in two sampling point sequences is determined using a non-recursive matrix value calculation method. This method has a time complexity of O(n), which is significantly lower than the O(n) time complexity of the recursive method. 2 The non-recursive method has a lower time complexity than the recursive method, making it significantly faster for calculating long sequences of sampling points. For example, in practical applications, quantization curves typically have 500-2000 sampling points, and the corresponding length of the input sampling point sequence is also between 500-2000 sequence units. In this scenario, the non-recursive method effectively improves the computation speed.

[0109] In the process of determining the sampling point matching sequence of torque angle segment and standard torque angle segment based on the distance matrix, D[m-1,n-1] is taken as the starting point of the sampling point matching sequence, the minimum value D[a,b] of its three adjacent matrix units is determined, and D[a,b] is added to the sampling point matching sequence.

[0110] If a≠0 and b≠0, then repeat the above steps starting from D[a,b] until D[a,b]=D[0,0];

[0111] If a = 0 or b = 0, then all matrix units including D[0,0] on the straight path from D[a,b] to D[0,0] are sequentially added to the sampling point matching sequence.

[0112] In this way, a path is determined in the distance matrix from the starting point of the two sampling point sequences to the ending point of the two sampling point sequences (i.e., from D[m-1,n-1] to D[0,0] in the distance matrix), and the sum of the distance element values ​​on this path is minimized.

[0113] Traversing the sampling point matching sequence determined by the above method, for each traversed D[a,b], the Euclidean distance between the a-th sampling point of the torque angle segment and the b-th sampling point of the standard torque angle segment is calculated, and the maximum value of the distance is taken as the similarity between the torque angle segment and the standard torque angle segment. Here, the Fréchet distance between the torque angle segment and the standard torque angle segment is determined in this way.

[0114] Compared with existing technologies, this application obtains a quantitative curve of the bolt tightening process; based on the quantitative curve and a preset standard line, it determines the similarity value between the bolt tightening process and a preset standard tightening process using Fréchet distance, wherein the preset standard line is the quantitative curve of a standard bolt tightening process; and it determines whether the bolt tightening process is abnormal based on a preset similarity threshold and the similarity value. This method achieves a rapid anomaly screening approach with low application cost, and it performs stably in scenarios involving frequency conversion acquisition of bolt tightening data.

[0115] Furthermore, embodiments of this application also provide a computer-readable medium having computer-readable instructions stored thereon, which can be executed by a processor to implement the aforementioned method.

[0116] This application embodiment also provides an anomaly screening device for the bolt tightening process, wherein the device includes:

[0117] One or more processors; and

[0118] A memory storing computer-readable instructions, which, when executed, cause the processor to perform the operations of the aforementioned method.

[0119] For example, when executed, computer-readable instructions cause the one or more processors to: acquire a quantized curve of the bolt tightening process;

[0120] The similarity value between the bolt tightening process and the preset standard tightening process is determined by Fréchet distance based on the quantization curve and the preset standard line, wherein the preset standard line is the quantization curve of the standard bolt tightening process;

[0121] The bolt tightening process is determined to be abnormal based on a preset similarity threshold and the similarity value.

[0122] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the invention. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be embraced within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims. Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices recited in the apparatus claims may also be implemented by a single unit or device in software or hardware. The terms "first," "second," etc., are used to indicate names and do not indicate any particular order.

Claims

1. An anomaly screening method for bolt tightening process, wherein, The method includes: A quantitative curve of the bolt tightening process is obtained, wherein the bolt tightening process has a unique result torque, the quantitative curve includes a torque-time curve, a preset fitting threshold and fitting ratio, and a target segment before reaching the result torque is extracted from the quantitative curve based on the result torque, including: Determine the first point (x1, y1) corresponding to the resulting torque y1 on the torque-time curve; set the second point (x2, y2) on the torque-time curve according to the fitting ratio p, where y2 / y1=p; determine the fitting line based on the first point and the second point; if the third point (x3, y3) on the fitting line and the fourth point (x3, y4) on the torque-time curve satisfy |y3-y4|=a, take the fourth point as the starting point of the target segment, where a is the fitting threshold; take the curve segment on the torque-time curve located between the starting point and the first point as the target segment; The similarity value between the bolt tightening process and the preset standard tightening process is determined by Fréchet distance based on the target segment and the standard segment. The standard segment is the segment in the preset standard line before reaching its corresponding result torque. The preset standard line is a quantization curve of the standard bolt tightening process. The bolt tightening process is determined to be abnormal based on a preset similarity threshold and the similarity value.

2. The method according to claim 1, wherein the quantization curve further includes a torque angle curve, wherein, The step of determining the similarity value between the bolt tightening process and the preset standard tightening process based on the target segment and the standard segment using the Fréchet distance includes: Determine the torque angle segment corresponding to the target segment in the torque angle curve; The similarity between the torque angle segment and the standard torque angle segment is calculated based on the Fréchet distance to determine the similarity value between the bolt tightening process and the preset standard tightening process. The standard torque angle segment is the corresponding segment of the standard segment in the torque angle curve corresponding to the preset standard line.

3. The method according to claim 2, wherein, The calculation of the similarity between the torque angle segment and the standard torque angle segment based on the Fréchet distance includes: The similarity between the torque angle segment and the standard torque angle segment is calculated in a non-recursive manner based on the Fréchet distance.

4. The method according to claim 3, wherein both the torque angle segment and the standard torque angle segment are composed of a plurality of sampling points containing torque and angle, wherein, The calculation of the similarity between the torque angle segment and the standard torque angle segment using a non-recursive method based on the Fréchet distance includes: Construct a distance matrix between the sampling points of the torque angle segment and the standard torque angle segment; The sampling point matching sequence between the torque angle segment and the standard torque angle segment is determined based on the distance matrix; The distance between each pair of matched torque angle segment sampling points and the sampling points of the standard torque angle segment is calculated based on the sampling point matching sequence. The maximum value of the distance is used as the similarity between the torque angle segment and the standard torque angle segment.

5. The method according to claim 4, wherein, The process of constructing the distance matrix between the sampling points of the torque angle segment and the standard torque angle segment includes: Construct the distance matrix D = [ ]m×n, where m and n are the number of sampling points for the torque angle segment A and the standard torque angle segment B, respectively. [0,0] = ， [i,0] = + D[i-1,0], inside , [0,j] = + D[0,j-1], where , [i,j] = ,in , , in, The Euclidean distance between the p-th sampling point of the torque angle segment and the q-th sampling point of the standard torque angle segment.

6. The method according to claim 5, wherein, The step of determining the sampling point matching sequence between the torque angle segment and the standard torque angle segment based on the distance matrix includes: Using D[m-1,n-1] as the starting point of the sampling point matching sequence, determine the minimum value D[a,b] among its three adjacent matrix units, and add D[a,b] to the sampling point matching sequence; If a≠0 and b≠0, then repeat the above steps starting from D[a,b] until D[a,b]=D[0,0]; If a=0 or b=0, then all matrix units including D[0,0] on the straight path from D[a,b] to D[0,0] are sequentially added to the sampling point matching sequence.

7. The method according to claim 6, wherein, The step of calculating the distance between each pair of matched torque angle segmented sampling points and the standard torque angle segmented sampling points based on the sampling point matching sequence includes: Traverse the sampling point matching sequence, and for each traversed D[a,b], calculate the Euclidean distance between the a-th sampling point of the torque angle segment and the b-th sampling point of the standard torque angle segment.

8. A computer-readable medium having stored thereon computer-readable instructions that can be executed by a processor to implement the method as described in any one of claims 1 to 7.

9. An anomaly screening device for the bolt tightening process, wherein, The device includes: One or more processors; and A memory storing computer-readable instructions, which, when executed, cause the processor to perform the operations of the method as described in any one of claims 1 to 7.

Images