Rapid testing method of terminal block based on data analysis

Abstract

The invention relates to the technical field of electric variable measurement, and in particular to a method for quickly testing a terminal block based on data analysis, comprising: collecting no-load current data and load current data of all terminal blocks, and constructing a no-load current curve and a load current curve of each terminal block; constructing a relative difference distance acquisition model, and acquiring the relative difference distance between any two terminal blocks; clustering the relative difference distances between all any two terminal blocks to acquire unqualified terminal blocks, and completing terminal block detection; when determining the relative difference distances for clustering terminal blocks in the cluster, the invention comprehensively considers the stability and abnormality of each terminal block under no-load and load conditions, thereby improving the accuracy of relative difference distance measurement and the accuracy of terminal block testing.

Description

Technical Field

[0001] The invention relates to the technical field of electric variable measurement, and in particular to a fast testing method for a terminal block based on data analysis. Background Art

[0002] In the power industry, especially in the field of industrial and commercial electricity, the reliability and performance of terminal blocks are crucial. With the continuous development of industrial automation and digitalization, the demand for rapid quality testing of terminal blocks is increasing. Terminal blocks are usually used to connect different components in the power system, such as cables and wires, and their safety is directly related to the safety of the electrical system. If the terminal blocks are not firmly connected, have poor contact or have other safety hazards, it may cause safety accidents such as failure, short circuit, electric shock, etc. of electrical equipment, and even cause fire or explosion in serious cases.

[0003] The K-means clustering method is used to cluster the relative difference distances of each terminal row to screen out unqualified terminal rows. When determining the relative difference distances for clustering between terminal rows, it is not considered that the current in the circuit of a qualified terminal row will also fluctuate, while the current in the circuit of an unqualified terminal row will also be stable. As a result, the calculated relative difference distance between the two terminal rows is inaccurate, which in turn causes errors in the clustering results, affecting the accuracy of the detection of the terminal rows. Summary of the invention

[0004] The present invention provides a rapid testing method for terminal blocks based on data analysis to solve the problem that in the existing method of using K-means clustering to cluster the relative difference distances of each terminal block, when determining the relative difference distances for clustering the terminal blocks, the situation in which normal current fluctuations occur in qualified terminal blocks and a stable state occurs in unqualified terminal blocks is not fully considered, thereby affecting the accuracy of the detection result.

[0005] The data analysis-based rapid testing method for terminal blocks of the present invention adopts the following technical solutions:

[0006] An embodiment of the present invention provides a method for quickly testing a terminal block based on data analysis, the method comprising the following steps:

[0007] Collect no-load current data and load current data of all terminal blocks, with the sampling time as the horizontal axis and the no-load current data and load current data as the vertical axis, and construct a no-load current curve and a load current curve for each terminal block;

[0008] A relative difference distance acquisition model is constructed, and the no-load current curve and the load current curve of any two terminal blocks are input into the relative difference distance acquisition model to obtain the relative difference distance of any two terminal blocks; the relative difference distances of all any two terminal blocks are clustered to obtain unqualified terminal blocks, and the terminal block detection is completed;

[0009] The specific contents of the relative difference distance acquisition model are as follows: decompose each input current curve to obtain all IMF components of each current curve, obtain the fluctuation value of each IMF component of each current curve according to the period of each IMF component of each current curve and the current data at the end point of each period of each IMF component; obtain the current stability evaluation of each current curve according to the number of IMF components of each current curve and the fluctuation value of each IMF component;

[0010] The high point spike abnormality of each current curve is obtained according to the time from the maximum point of each current curve to the circuit access, and the difference between the current data of the maximum point and the adjacent data points; the low point interruption abnormality of each current curve is obtained according to the current data size of the minimum point of each current curve, and the slope of the connection line formed by the minimum point and the adjacent data points; the current abnormality evaluation of each current curve is obtained according to the high point spike abnormality and the low point interruption abnormality of each current curve;

[0011] According to the current stability evaluation and the current anomaly evaluation of each current curve of the two terminal blocks, the relative difference distance of the two terminal blocks is obtained.

[0012] Furthermore, the input current curves are decomposed to obtain all IMF components of the current curves, and the fluctuation value of each IMF component of each current curve is obtained according to the period of each IMF component of each current curve and the current data at the end point of each period of each IMF component, including the specific method of:

[0013] All no-load current curves are decomposed by the EMD algorithm to obtain all IMF components of each no-load current curve and arrange them in descending order from high frequency to low frequency according to the IMF components to obtain the descending sequence of the IMF components of each no-load current curve;

[0014] For the vth IMF component in the descending sequence of the IMF components of the i-th no-load current curve, obtain all cycles on the vth IMF component of the i-th no-load current curve. The cycle length of each cycle is the sampling time interval from the first current data to the last current data in the cycle. The volatility coefficient of the vth IMF component of the i-th no-load current curve is The calculation method is:

[0015] Where: represents the current data corresponding to the left endpoint of the kth cycle of the vth IMF component of the i-th no-load current curve, represents the current data corresponding to the right endpoint of the kth cycle of the vth IMF component of the i-th no-load current curve, represents the period length of the kth period of the vth IMF component of the ith no-load current curve, represents the mean value of all cycle lengths of the vth IMF component of the i-th no-load current curve; It represents the standard deviation function;

[0016] The sequence of volatility coefficients of all IMF components of the i-th no-load current curve according to the order of the descending sequence of IMF components is recorded as the volatility coefficient sequence of the i-th no-load current curve, and each volatility performance in the volatility performance sequence of the i-th no-load current curve is linearly normalized to obtain the volatility value of each volatility performance, and the sequence composed of the volatility values ​​of all volatility performances is recorded as the volatility value sequence of the i-th no-load current curve.

[0017] Furthermore, the current stability evaluation of each current curve is obtained according to the number of IMF components of each current curve and the fluctuation value of each IMF component, including the specific method of:

[0018] Current stability evaluation of the i-th no-load current curve The calculation method is:

[0019] Where: represents the number of IMF components of the i-th no-load current curve, represents the sequence number of the vth IMF component of the i-th no-load current curve in the descending sequence of IMF components, Represents the fluctuation value of the vth IMF component in the fluctuation value sequence of the i-th no-load current curve.

[0020] Furthermore, the method of obtaining the high point peak abnormality of each current curve according to the time from the maximum point of each current curve to the access circuit and the difference in current data between the maximum point and the adjacent data points includes: presetting the semi-neighborhood length U, taking each maximum point on each no-load current curve as the center, taking U data points adjacent to the sampling time on the left and right sides respectively, and recording the interval formed by the sampling time of the Uth data point on the left side of each maximum point to the sampling time of the Uth data point on the right side as the neighborhood of each maximum point; obtaining all data points in the neighborhood of each maximum point; the no-load neighborhood current difference of the gth maximum point in the ith no-load current curve The calculation method is:

[0021] Where: represents the current data of the gth maximum point in the i-th no-load current curve, Indicates the sampling time of the g-th maximum point in the i-th no-load current curve, and represents the time from the g-th maximum point in the i-th no-load current curve to the circuit connection. represents the current data of the gth maximum point in the i-th no-load current curve, represents the current data of the ath data point to the right of the gth maximum point in the i-th no-load current curve, represents the current data of the ath data point to the left of the gth maximum point in the i-th no-load current curve, represents the hyperparameter, It represents the absolute value function;

[0022] The calculation method of the high point peak abnormality of the i-th no-load current curve is: the sampling time of each maximum point in the i-th no-load current curve and the no-load neighborhood current difference of each maximum point in the i-th no-load current curve are respectively multiplied to obtain the product, and the products of all the maximum points in the i-th no-load current curve are summed, and then the summation result is normalized using the Sigmoid function, and the normalized result is recorded as the high point peak abnormality of the i-th no-load current curve.

[0023] Furthermore, the method of obtaining the low-point interruption abnormality of each current curve according to the current data size of the minimum point of each current curve and the slope of the line formed by the minimum point and the adjacent data points includes: presetting the semi-neighborhood length U, taking each minimum point on each no-load current curve as the center, taking U data points adjacent to the sampling time on the left and right sides respectively, recording the interval formed by the sampling time of the Uth data point on the left side of each minimum point to the sampling time of the Uth data point on the right side as the neighborhood of each minimum point, and obtaining all data points in the neighborhood of each minimum point; the low-point interruption abnormality of the i-th no-load current curve The calculation method is:

[0024] Where: represents the number of minimum points in the i-th no-load current curve, represents the sampling time of the hth minimum point in the i-th no-load current curve, represents the current data of the hth minimum point in the i-th no-load current curve, It represents the slope of the line connecting the hth minimum point in the i-th no-load current curve and the bth data point on the left side of the minimum point, It represents the slope of the line connecting the hth minimum point in the i-th no-load current curve and the bth data point on the right side of the minimum point, represents the absolute value function, Represents the Sigmoid function.

[0025] Furthermore, the current anomaly evaluation of each current curve is obtained according to the high point peak anomaly and the low point current interruption anomaly of each current curve, including the specific method of:

[0026] The average of the high point peak abnormality and the low point interruption abnormality of the no-load current curve is recorded as the current abnormality evaluation of the no-load current curve.

[0027] Furthermore, the relative difference distance between the two terminal blocks is obtained according to the current stability evaluation and the current anomaly evaluation of each current curve of the two terminal blocks, including the specific method of:

[0028] For the i-th terminal row and the j-th terminal row, calculate the square of the no-load current difference between the i-th terminal row and the j-th terminal row, and the square of the load current difference between the i-th terminal row and the j-th terminal row, and sum the two square results, and take the square root of the sum result as the relative difference distance between the i-th terminal row and the j-th terminal row.

[0029] Furthermore, the no-load current difference includes a specific acquisition method as follows:

[0030] Subtract the current data of each data point on the i-th no-load current curve from the j-th no-load current curve to obtain the difference, add up the results of the difference, and multiply them by the sum of the no-load adjustment coefficient and 1. The product obtained is the no-load current difference between the i-th terminal block and the j-th terminal block.

[0031] Furthermore, the no-load adjustment coefficient includes a specific acquisition method as follows:

[0032] Take the maximum value in the current anomaly evaluation of the i-th no-load current curve and the j-th no-load current curve as the numerator, take the minimum value in the current stability evaluation of the i-th no-load current curve and the j-th no-load current curve, and take the sum of the minimum value and the hyperparameter as the denominator; the result of comparing the numerator and the denominator is recorded as the no-load adjustment coefficient between the i-th terminal block and the j-th terminal block.

[0033] Furthermore, the clustering of the relative difference distances between any two terminal blocks to obtain unqualified terminal blocks includes the following specific methods:

[0034] K-means clustering is used to cluster the relative difference distances between every two terminal blocks, the number of clusters is a preset number of clusters, and the preset number of clusters is obtained.

[0035] The relative difference distance in each cluster is obtained by statistics as the frequency of each terminal row. In the cluster, if there is a terminal row whose frequency is greater than the total frequency of all terminal rows than the preset abnormal threshold, the cluster is recorded as an abnormal cluster, and all terminal rows in the abnormal cluster whose frequency is greater than the preset abnormal threshold are recorded as unqualified terminal rows.

[0036] The beneficial effects of the technical solution of the present invention are as follows: the present invention collects the current data of the terminal block in no-load and load states to establish a current curve, covers the terminal block in different working states, and improves the reliability of the detection result; when analyzing the stable current performance of the terminal block, the IMF component is obtained by decomposing the current curve, and the current curve is decomposed to analyze the potential periodicity and noise fluctuations therein, and the stability of the current curve is measured according to the periodic distribution of the IMF component and the periodic endpoint current data. Compared with directly measuring the stability of the current curve by statistics such as the variance of the current data, the stability obtained by the IMF component has higher accuracy; when analyzing the abnormal current performance of the terminal block, it is divided into current spike anomalies and The current interruption anomaly is analyzed separately, and the normal current fluctuation that occurs when the terminal block is just connected to the circuit is taken into account to improve the accuracy of the calculation of the abnormal degree of the current curve; when determining the relative difference distance for clustering the terminal blocks in the K-means clustering, the stability and abnormality of the two terminal blocks under no-load and load conditions are used to adjust the relative difference calculated by the current data difference, so that when obtaining the relative difference distance, it fully covers the situation where the qualified terminal block has normal current fluctuations and the unqualified terminal block is in a stable state, and accurately measures the relative difference distance between the two terminal blocks, thereby obtaining a more accurate clustering result and improving the accuracy of the terminal block test. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0038] Figure 1 The present invention is a flowchart of the steps of the method for quickly testing a terminal block based on data analysis. DETAILED DESCRIPTION

[0039] In order to further explain the technical means and effects adopted by the present invention to achieve the predetermined invention purpose, the following is a detailed description of the specific implementation method, structure, features and effects of the terminal block rapid testing method based on data analysis proposed by the present invention in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" does not necessarily refer to the same embodiment. In addition, specific features, structures or characteristics in one or more embodiments may be combined in any suitable form.

[0040] Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

[0041] The specific scheme of the terminal block rapid testing method based on data analysis provided by the present invention is described in detail below with reference to the accompanying drawings.

[0042] See also Figure 1 , which shows a flowchart of a method for rapid testing of a terminal block based on data analysis provided by an embodiment of the present invention, the method comprising the following steps:

[0043] Step S001: Collect no-load current data and load current data of all terminal blocks, and construct a no-load current curve and a load current curve of each terminal block.

[0044] The purpose of this embodiment is to assign different weights to the current data of the terminal row under no-load and load according to the stability and mutation of the current data of each terminal row under no-load and load, so as to calculate the distance measurement between every two terminal rows. Therefore, the current data of each terminal row is first collected; the current data of all terminal rows under no-load state is collected by using a current tester, and the current tester collects current data once every 0.05 seconds as a sampling moment, and the current data of all sampling moments within 30 seconds is obtained. The number of all current data collected for a terminal row under no-load state is recorded as n. It should be noted that the moment when the sampling moment is 0 is the moment when the terminal row is connected to the circuit; with the sampling moment as the horizontal axis and the current data as the vertical axis, the no-load current curve of the terminal row is established, and there are n data points on each no-load current curve; similarly, the no-load current curve of each terminal row is obtained, and the current data of all terminal rows under load state are collected by using the current tester to construct the load current curve of each terminal row.

[0045] Step S002: decompose each input current curve to obtain all IMF components of each current curve, obtain the fluctuation value of each IMF component of each current curve according to the period of each IMF component of each current curve and the current data at the end point of each period of each IMF component; obtain the current stability evaluation of each current curve according to the number of IMF components of each current curve and the fluctuation value of each IMF component.

[0046] It should be noted that the use of the EMD (Empirical Mode Decomposition) algorithm can decompose all current curves including the no-load current curve and the load current curve into multiple components. Each IMF component represents an essential feature or frequency component in the current curve. Therefore, the composition structure of the current curve can be better understood, and the potential periodicity, trend and noise components in the current curve can be discovered to better analyze the stability of the current curve. The no-load current curve is decomposed by the EMD algorithm to obtain several IMF components of each no-load current curve, and the current stability evaluation of the no-load current curve is calculated according to the periodic characteristics and current data exhibited by the IMF components.

[0047] Specifically, all no-load current curves are decomposed by the EMD algorithm, all IMF components of each no-load current curve are obtained, and the IMF components are arranged in descending order from high frequency to low frequency to obtain the descending sequence of the IMF components of each no-load current curve; for the i-th no-load current curve, the number of IMF components obtained by the EMD algorithm for the i-th no-load current curve is recorded as ; For the vth IMF component in the descending sequence of the IMF components of the i-th no-load current curve, obtain all the maximum points and all the minimum points on the vth IMF component of the i-th no-load current curve, and record the current data including the first maximum point and the first minimum point, as well as the current data between the first maximum point and the first minimum point as the first current cycle of the vth IMF component of the i-th no-load current curve; record the current data including the first minimum point and the second maximum point, as well as the current data between the first minimum point and the second maximum point as the vth IMF component of the i-th no-load current curve. The second current cycle of the MF component; the current data including the second maximum and the second minimum points, as well as the current data between the second maximum point and the second minimum point are recorded as the third current cycle of the vth IMF component of the i-th no-load current curve, and so on, to obtain all cycles of the vth IMF component of the i-th no-load current curve; similarly, to obtain all cycles of all IMF components of all no-load current curves; the cycle length of each cycle is the sampling time interval from the first current data to the last current data in the cycle; the volatility coefficient of the vth IMF component of the i-th no-load current curve is The calculation method is:

[0048] Where: represents the current data corresponding to the left endpoint of the kth cycle of the vth IMF component of the i-th no-load current curve, represents the current data corresponding to the right endpoint of the kth cycle of the vth IMF component of the i-th no-load current curve, represents the period length of the kth period of the vth IMF component of the ith no-load current curve, represents the mean value of all cycle lengths of the vth IMF component of the i-th no-load current curve; Represents the standard deviation function.

[0049] It should be noted that represents the amplitude of the kth cycle of the vth IMF of the i-th no-load current curve; represents the cycle length deviation of the kth cycle of the vth IMF component of the i-th no-load current curve; It means taking the standard deviation of the product of the amplitude of all cycles of the vth IMF component of the i-th no-load current curve and the deviation of the cycle length. The larger the standard deviation is, the larger the volatility coefficient of the vth IMF component of the i-th no-load current curve is.

[0050] The sequence of volatility coefficients of all IMF components of the i-th no-load current curve according to the order of the descending sequence of IMF components is recorded as the volatility coefficient sequence of the i-th no-load current curve, and each volatility performance in the volatility performance sequence of the i-th no-load current curve is linearly normalized to obtain the volatility value of each volatility performance, and the sequence composed of the volatility values ​​of all volatility performances is recorded as the volatility value sequence of the i-th no-load current curve; similarly, the volatility value sequence of all no-load current curves is obtained.

[0051] Furthermore, the fluctuation value of the vth IMF component in the fluctuation value sequence of the i-th no-load current curve is recorded as , current stability evaluation of the i-th no-load current curve The calculation method is:

[0052] Where: represents the number of IMF components of the i-th no-load current curve, represents the number of the vth IMF component of the i-th no-load current curve in the descending sequence of IMF components; it should be noted that for higher frequency IMF components, The smaller the value of is, the more likely noise will appear in the higher-frequency IMF components. The value of is used as the weight, and the higher frequency IMF components are given smaller weights; It represents the fluctuation value of the vth IMF component of the i-th no-load current curve. The larger the value is, the smaller the current stability evaluation of the i-th no-load current curve is.

[0053] Similarly, obtain the current stability evaluation of all no-load current curves;

[0054] Step S003, obtaining the high point peak abnormality of each current curve according to the time from the maximum point of each current curve to the circuit connection, and the difference in current data between the maximum point and the adjacent data points; obtaining the low point interruption abnormality of each current curve according to the current data size of the minimum point of each current curve, and the slope of the connecting line formed by the minimum point and the adjacent data points; obtaining the current abnormality evaluation of each current curve according to the high point peak abnormality and the low point interruption abnormality of each current curve.

[0055] It should be noted that a sudden increase in current may occur in the circuit, which is manifested as a rapid increase and rapid drop in current data in a short period of time on all current curves including the no-load current curve and the load current curve. When the terminal block is just connected to the circuit, the current may fluctuate to a certain extent at first because it takes a certain response time to enter a stable state. However, the fluctuation caused by entering a stable state will not affect the current spike abnormality of the quality of the terminal block. Therefore, when analyzing whether the terminal block has a current spike abnormality, the time when the fluctuation occurs should be combined for analysis. The earlier the fluctuation occurs, the more likely it is that it is a normal fluctuation caused by connecting to the circuit.

[0056] Specifically, taking each maximum point on each no-load current curve as the center, U data points adjacent to the sampling time are taken on the left and right sides respectively, and the interval formed by the sampling time of the Uth data point on the left side of each maximum point to the sampling time of the Uth data point on the right side is recorded as the neighborhood of each maximum point, and U is recorded as the half-neighborhood length; all data points in the neighborhood of each maximum point are obtained; this embodiment is described by taking U=10 as an example, and other embodiments may take other values. This embodiment is not specifically limited. It should be noted that if the number of data points on the left or right side of a certain maximum point is less than U, as many data points as there are; the no-load neighborhood current difference of the gth maximum point in the i-th no-load current curve is The calculation method is:

[0057] Where: represents the current data of the gth maximum point in the i-th no-load current curve, represents the half-neighborhood length, represents the sampling time of the gth maximum point in the i-th no-load current curve, represents the current data of the gth maximum point in the i-th no-load current curve, represents the current data of the ath data point to the right of the gth maximum point in the i-th no-load current curve, represents the current data of the ath data point to the left of the gth maximum point in the i-th no-load current curve, Represents a hyperparameter. In this embodiment, y Other embodiments may be set to other values, which are not specifically limited in this embodiment. To avoid the situation where the denominator is 0, represents the absolute value function; it should be noted that, It represents the difference between the current data mean of the gth maximum point in the i-th no-load current curve and the ath maximum point on its left and right sides. The larger the value, the greater the degree of current surge at the gth maximum point in the i-th no-load current curve. It represents the difference in current data between the ath maximum point on the left and right sides of the gth maximum point in the i-th no-load current curve. The smaller the value, the faster the current data falls back after the current surge at the gth maximum point, and the more likely it is that a current spike abnormality occurs. Similarly, the difference in no-load neighborhood current of all maximum points in the i-th no-load current curve is obtained. The peak abnormality of the high point of the i-th no-load current curve The calculation method is:

[0058] Where: represents the number of maximum points in the i-th no-load current curve, represents the sampling time of the gth maximum point in the i-th no-load current curve, represents the no-load neighborhood current difference of the g-th maximum point in the i-th no-load current curve, Represents the Sigmoid function; it should be noted that, The smaller the value, the more likely the fluctuation on the current curve is the normal fluctuation that occurs when the terminal block is just connected to the circuit. The product of the sampling time of the g-th maximum point in the i-th no-load current curve and the current difference in its neighborhood is used to measure the possibility of a current spike abnormality at the g-th maximum point in the i-th no-load current curve. It represents the sum of the possibilities of current spike anomaly occurring at all maximum points in the i-th no-load current curve, and the result after normalization of this value is taken as the high point spike anomaly degree of the i-th no-load current curve;

[0059] Get the peak abnormality of all no-load current curves.

[0060] It should be further explained that the circuit may experience current interruption anomalies due to certain faults, equipment damage, disconnection, etc., which may be manifested on all current curves including the no-load current curve and the load current curve as current data being 0 or very close to 0 for a period of time; the more data points the current is close to 0, the greater the probability of a current interruption anomaly occurring in the current curve.

[0061] Specifically, taking each minimum point on each no-load current curve as the center, U data points adjacent to the sampling time are taken on the left and right sides respectively, and the interval formed by the sampling time of the Uth data point on the left side of each minimum point to the sampling time of the Uth data point on the right side is recorded as the neighborhood of each minimum point, and all data points in the neighborhood of each minimum point are obtained; it should be noted that if the number of data points on the left or right side of a certain minimum point is less than U, as many data points as there are; the low point current interruption abnormality of the i-th no-load current curve is The calculation method is:

[0062] Where: represents the number of minimum points in the i-th no-load current curve, represents the sampling time of the hth minimum point in the i-th no-load current curve, represents the current data of the hth minimum point in the i-th no-load current curve, represents the half-neighborhood length, It represents the slope of the line connecting the hth minimum point in the i-th no-load current curve and the bth data point on the left side of the minimum point, It represents the slope of the line connecting the hth minimum point in the i-th no-load current curve and the bth data point on the right side of the minimum point, represents the absolute value function, Represents the Sigmoid function; it should be noted that, and The closer the value is to 0, the more the hth minimum point in the i-th no-load current curve conforms to the current data performance of the current interruption abnormality, and thus the low point of the i-th no-load current curve has a greater degree of abnormal current interruption; The smaller the value of , the greater the abnormality of the low point current interruption of the i-th no-load current curve;

[0063] Similarly, obtain the low-point interruption abnormality of all no-load current curves.

[0064] Furthermore, the average of the high point peak abnormality and the low point interruption abnormality of the no-load current curve is recorded as the current abnormality evaluation of the no-load current curve, and the current abnormality evaluation of all no-load current curves is obtained.

[0065] Step S004: Obtain the relative difference distance between the two terminal blocks according to the current stability evaluation and the current anomaly evaluation of the current curves of the two terminal blocks.

[0066] After obtaining the current stability evaluation and current abnormality evaluation of all no-load current curves, calculate the no-load adjustment coefficient between every two terminal blocks; the no-load adjustment coefficient between the i-th terminal block and the j-th terminal block The calculation method is:

[0067] Where: represents the current anomaly evaluation of the i-th no-load current curve, represents the current anomaly evaluation of the jth no-load current curve, represents the current stability evaluation of the i-th no-load current curve, represents the current stability evaluation of the jth no-load current curve, Represents a hyperparameter. In this embodiment, Other embodiments may be set to other values, which are not specifically limited in this embodiment. Used to avoid the situation where the denominator is 0; represents the maximum value function, represents the minimum value function; it should be noted that, It means taking the largest one of the current abnormality evaluations of the i-th no-load current curve and the j-th no-load current curve, It means taking the smallest one of the current stability evaluations of the i-th no-load current curve and the j-th no-load current curve. The larger the maximum value in the current abnormality evaluation of the i-th no-load current curve and the j-th no-load current curve, the larger the no-load adjustment coefficient between the i-th terminal block and the j-th terminal block, and the smaller the minimum value in the current stability evaluation of the i-th no-load current curve and the j-th no-load current curve, the larger the no-load adjustment coefficient between the i-th terminal block and the j-th terminal block; the larger the no-load adjustment coefficient, the greater the possibility of relative abnormality between the i-th terminal block and the j-th terminal block.

[0068] Obtain the no-load adjustment coefficient between every two terminal blocks, and adjust the difference in current data under no-load state between every two detection terminal blocks; the no-load current difference between the i-th terminal block and the j-th terminal block The calculation method is:

[0069] Where: It represents the no-load adjustment coefficient between the i-th terminal block and the j-th terminal block. Indicates the number of data points on the no-load current curve, represents the current data of the pth data point in the i-th no-load current curve, represents the current data of the pth data point in the jth no-load current curve; it should be noted that, Indicates the difference between the current data of the i-th terminal block and the j-th terminal block in the no-load state. The larger the value, the greater the difference in no-load current between the i-th terminal block and the j-th terminal block. Indicates the degree of adjustment of the difference in the current data. The larger the value, the greater the difference in no-load current between the i-th terminal row and the j-th terminal row. Similarly, the difference in no-load current between every two terminal rows is obtained.

[0070] Furthermore, according to each load current curve, all IMF components of each load current curve are obtained; according to all IMF components of each load current curve, the current stability evaluation and the current anomaly evaluation of each load current curve are obtained; according to the current stability evaluation and the current anomaly evaluation of each load current curve, the load adjustment coefficient of each two terminal blocks is obtained; according to the current data of each load current curve and the load adjustment coefficient of each two terminal blocks, the load current difference between each two terminal blocks is obtained.

[0071] It should be noted that the load current difference and no-load current difference of the qualified terminal row under no-load and load conditions are in a stable state, so the load current difference and no-load current difference between each qualified terminal row and each other qualified terminal row do not fluctuate much. Therefore, this embodiment obtains the relative difference distance between two terminal rows through the load current difference and no-load current difference between each terminal row and each other terminal row. The relative difference distance of the qualified terminal row fluctuates less, and the distance between the unqualified terminal row and the other terminal rows is larger or smaller, which is then used for clustering to obtain abnormal terminal rows.

[0072] Specifically, the relative difference distance between the i-th terminal row and the j-th terminal row The calculation method is:

[0073] Where: It represents the difference in no-load current between the i-th terminal block and the j-th terminal block. Indicates the load current difference between the i-th terminal block and the j-th terminal block; it should be noted that, in this embodiment, the no-load adjustment coefficient between every two terminal blocks is obtained based on the current stability evaluation and current anomaly evaluation of each no-load current curve; on the basis of comparing the current data difference between every two terminal blocks, the current data difference is adjusted using the no-load adjustment coefficient, thereby obtaining a more accurate relative difference distance between every two terminal blocks, so that the clustering results obtained by subsequent clustering are more accurate, thereby improving the detection quality of the terminal blocks. Step S005, cluster the relative difference distances of all any two terminal blocks to obtain unqualified terminal blocks, and complete the terminal block test.

[0074] It should be noted that since the relative difference distances between qualified terminal blocks are concentrated in several fixed ranges, if the quality of a terminal block is unqualified, the relative difference distance between the terminal block and each qualified terminal block is larger than the relative difference distance between two qualified terminal blocks. When K-means clustering is used to cluster the relative difference distances between every two terminal blocks to obtain several clustering results, the relative difference distances between the terminal block and each qualified terminal block will be clustered into the same cluster, and the cluster will not contain the relative difference distances between any two qualified terminal blocks, and the cluster will be recorded as an abnormal cluster; and since in actual production, the defective rate is usually 3%, the number of unqualified terminal blocks in a production batch is far less than the qualified terminal blocks; therefore, when the relative difference distances in a cluster are all the relative difference distances between one or a few terminal blocks and other terminal blocks, the cluster will be recorded as an abnormal cluster.

[0075] Specifically, the relative difference distance between each two terminal blocks is clustered using K-means clustering, the number of clusters for clustering is a preset number of clusters, and a preset number of clusters of clusters is obtained. In this embodiment, the preset number of clusters is described as 8, and other embodiments may be set to other values, which are not specifically limited in this embodiment;

[0076] The relative difference distance in each cluster is obtained by counting the frequency of each terminal row. As an example, if a cluster is "the relative difference distance between the first terminal row and the second terminal row, the relative difference distance between the first terminal row and the third terminal row, the relative difference distance between the first terminal row and the fourth terminal row, and the relative difference distance between the second terminal row and the fourth terminal row", then the frequency of the first terminal row in the cluster is 3, the frequency of the second terminal row is 2, and the frequency of the third terminal row is 1. The frequency of the first terminal row is 1, and the frequency of the 4th terminal row is 2; in a cluster, if there is a terminal row whose frequency ratio to the total frequency of all terminal rows is greater than a preset abnormal threshold, the cluster is recorded as an abnormal cluster, and all terminal rows in the abnormal cluster whose frequencies are greater than the preset abnormal threshold are recorded as unqualified terminal rows. In this embodiment, the preset abnormal threshold is 50%, and other embodiments can be set to other values, which are not specifically limited in this embodiment; all unqualified terminal rows in the terminal row are obtained to complete the rapid detection of the terminal rows.

[0077] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the principles of the present invention should be included in the protection scope of the present invention.

Claims

1. Rapid testing method of terminal blocks based on data analysis, It is characterized in that The method comprises the following steps: Collect no-load current data and load current data of all terminal blocks, with the sampling time as the horizontal axis and the no-load current data and load current data as the vertical axis, and construct a no-load current curve and a load current curve for each terminal block; A relative difference distance acquisition model is constructed, and the no-load current curve and the load current curve of any two terminal blocks are input into the relative difference distance acquisition model to obtain the relative difference distance of any two terminal blocks; the relative difference distances of all any two terminal blocks are clustered to obtain unqualified terminal blocks, and the terminal block detection is completed; The specific contents of the relative difference distance acquisition model are as follows: decompose each input current curve to obtain all IMF components of each current curve, obtain the fluctuation value of each IMF component of each current curve according to the period of each IMF component of each current curve and the current data at the end point of each period of each IMF component; obtain the current stability evaluation of each current curve according to the number of IMF components of each current curve and the fluctuation value of each IMF component; The high point spike abnormality of each current curve is obtained according to the time from the maximum point of each current curve to the circuit access, and the difference between the current data of the maximum point and the adjacent data points; the low point interruption abnormality of each current curve is obtained according to the current data size of the minimum point of each current curve, and the slope of the connection line formed by the minimum point and the adjacent data points; the current abnormality evaluation of each current curve is obtained according to the high point spike abnormality and the low point interruption abnormality of each current curve; According to the current stability evaluation and the current anomaly evaluation of each current curve of the two terminal blocks, the relative difference distance of the two terminal blocks is obtained; The method of decomposing each input current curve to obtain all IMF components of each current curve, and obtaining the fluctuation value of each IMF component of each current curve according to the period of each IMF component of each current curve and the current data at the end point of each period of each IMF component, includes the following specific methods: All no-load current curves are decomposed by the EMD algorithm to obtain all IMF components of each no-load current curve and arrange them in descending order from high frequency to low frequency according to the IMF components to obtain the descending sequence of the IMF components of each no-load current curve; For the vth IMF component in the descending sequence of the IMF components of the i-th no-load current curve, obtain all cycles on the vth IMF component of the i-th no-load current curve. The cycle length of each cycle is the sampling time interval from the first current data to the last current data in the cycle. The volatility coefficient of the vth IMF component of the i-th no-load current curve is The calculation method is: Where: represents the current data corresponding to the left endpoint of the kth cycle of the vth IMF component of the i-th no-load current curve, represents the current data corresponding to the right endpoint of the kth cycle of the vth IMF component of the i-th no-load current curve, represents the period length of the kth period of the vth IMF component of the ith no-load current curve, represents the mean value of all cycle lengths of the vth IMF component of the i-th no-load current curve; It represents the standard deviation function; The sequence of the volatility coefficients of all IMF components of the i-th no-load current curve according to the order of the descending sequence of the IMF components is recorded as the volatility coefficient sequence of the i-th no-load current curve, each volatility performance in the volatility performance sequence of the i-th no-load current curve is linearly normalized to obtain the fluctuation value of each volatility performance, and the sequence composed of the fluctuation values ​​of all volatility performances is recorded as the fluctuation value sequence of the i-th no-load current curve; the current stability evaluation of each current curve is obtained according to the number of IMF components of each current curve and the fluctuation value of each IMF component, including the specific method as follows: Current stability evaluation of the i-th no-load current curve The calculation method is: Where: represents the number of IMF components of the i-th no-load current curve, represents the sequence number of the vth IMF component of the i-th no-load current curve in the descending sequence of IMF components, represents the fluctuation value of the vth IMF component in the fluctuation value sequence of the i-th no-load current curve; The specific method of obtaining the high point peak abnormality of each current curve according to the time from the maximum point of each current curve to the circuit connection and the difference between the current data of the maximum point and the adjacent data point is as follows: The semi-neighborhood length U is preset, and each maximum point on each no-load current curve is taken as the center. U data points adjacent to each other at the sampling time are taken on the left and right sides respectively. The interval formed by the sampling time of the Uth data point on the left side of each maximum point to the sampling time of the Uth data point on the right side is recorded as the neighborhood of each maximum point; all data points in the neighborhood of each maximum point are obtained; the no-load neighborhood current difference of the gth maximum point in the i-th no-load current curve The calculation method is: Where: represents the current data of the gth maximum point in the i-th no-load current curve, Indicates the sampling time of the g-th maximum point in the i-th no-load current curve, and represents the time from the g-th maximum point in the i-th no-load current curve to the circuit connection. represents the current data of the gth maximum point in the i-th no-load current curve, represents the current data of the ath data point to the right of the gth maximum point in the i-th no-load current curve, represents the current data of the ath data point to the left of the gth maximum point in the i-th no-load current curve, represents the hyperparameter, represents an absolute value function; the calculation method of the high point spike abnormality of the ith no-load current curve is: the sampling time of each maximum point in the ith no-load current curve and the no-load neighborhood current difference of each maximum point in the ith no-load current curve are respectively multiplied to obtain products, and the products of all maximum points in the ith no-load current curve are summed, and then the summation result is normalized using the Sigmoid function, and the normalized result is recorded as the high point spike abnormality of the ith no-load current curve; the low point interruption abnormality of each current curve is obtained according to the current data size of the minimum point of each current curve and the slope of the line formed by the minimum point and the adjacent data point, including the specific method: The length of the semi-neighborhood is preset as U. With each minimum point on each no-load current curve as the center, U data points adjacent to each other at the sampling time are taken on both sides. The interval formed by the sampling time of the Uth data point on the left side of each minimum point to the sampling time of the Uth data point on the right side is recorded as the neighborhood of each minimum point, and all data points in the neighborhood of each minimum point are obtained; the low point interruption abnormality of the i-th no-load current curve is calculated. The calculation method is: Where: represents the number of minimum points in the i-th no-load current curve, represents the sampling time of the hth minimum point in the i-th no-load current curve, represents the current data of the hth minimum point in the i-th no-load current curve, It represents the slope of the line connecting the hth minimum point in the i-th no-load current curve and the bth data point on the left side of the minimum point, It represents the slope of the line connecting the hth minimum point in the i-th no-load current curve and the bth data point on the right side of the minimum point, represents the absolute value function, Represents the Sigmoid function; The specific method of obtaining the current anomaly evaluation of each current curve according to the high point peak anomaly and the low point current interruption anomaly of each current curve is as follows: The average of the high point peak abnormality and the low point interruption abnormality of the no-load current curve is recorded as the current abnormality evaluation of the no-load current curve; The specific method of obtaining the relative difference distance between the two terminal blocks according to the current stability evaluation and the current anomaly evaluation of each current curve of the two terminal blocks is as follows: For the i-th terminal row and the j-th terminal row, calculate the square of the no-load current difference between the i-th terminal row and the j-th terminal row, and the square of the load current difference between the i-th terminal row and the j-th terminal row, and sum the two square results, and take the square root of the sum result as the relative difference distance between the i-th terminal row and the j-th terminal row.

2. According to the data analysis-based rapid testing method for terminal blocks of claim 1, It is characterized in that The no-load current difference includes a specific acquisition method as follows: Subtract the current data of each data point on the i-th no-load current curve from the j-th no-load current curve to obtain the difference, add up the results of the difference, and multiply them by the sum of the no-load adjustment coefficient and 1. The product obtained is the no-load current difference between the i-th terminal block and the j-th terminal block.

3. According to the data analysis-based rapid testing method for terminal blocks in claim 2, It is characterized in that The no-load adjustment coefficient includes a specific method for obtaining: Take the maximum value in the current anomaly evaluation of the i-th no-load current curve and the j-th no-load current curve as the numerator, take the minimum value in the current stability evaluation of the i-th no-load current curve and the j-th no-load current curve, and take the sum of the minimum value and the hyperparameter as the denominator; the result of comparing the numerator and the denominator is recorded as the no-load adjustment coefficient between the i-th terminal block and the j-th terminal block.

4. According to the data analysis-based rapid testing method for terminal blocks in claim 1, It is characterized in that The specific method of clustering the relative difference distances between all arbitrary two terminal blocks to obtain unqualified terminal blocks includes: K-means clustering is used to cluster the relative difference distances between every two terminal blocks, the number of clusters is a preset number of clusters, and the preset number of clusters is obtained. The relative difference distance in each cluster is obtained by statistics as the frequency of each terminal row. In the cluster, if there is a terminal row whose frequency is greater than the total frequency of all terminal rows than the preset abnormal threshold, the cluster is recorded as an abnormal cluster, and all terminal rows in the abnormal cluster whose frequency is greater than the preset abnormal threshold are recorded as unqualified terminal rows.

Images