Method for realizing nondestructive testing of fuse by equivalence class division method

Abstract

The invention discloses a method for realizing nondestructive testing of a fuse through an equivalence class division method, and relates to the technical field of power electronics. The method for realizing the nondestructive testing of the fuses through the equivalence class division method comprises the following steps: 1) testing all to-be-detected samples of the fuses by using small current, and collecting data; 2) initializing the collected data, giving a clustering category number, setting an iteration stop threshold, initializing a clustering center matrix, and setting an iteration counter b to be zero; 3) calculating or updating a membership matrix U of the data; (4) updating the clustering center matrix V of the data, and (5) stopping and outputting the membership matrix U and the clustering center matrix V if the variation dV of the clustering center matrix is smaller than a threshold value, otherwise, adding one to the counter, and repeating the step (4). The fuse detection method can avoid damage to the sample to be detected and obtain reliable data of sample quality.

Description

Technical Field

[0001] The invention relates to the technical field of power electronics, and in particular to a method for implementing nondestructive testing of fuses using an equivalence class division method. Background Art

[0002] In the process of implementing the existing technology, the short-circuit breaking test of the fuse is destructive, and the fuse is unusable after use, so the fuse can only be sampled and inspected, and it is impossible to comprehensively inspect the quality of the produced fuse products. It is urgent to adopt non-destructive testing methods to avoid damaging the samples and obtain reliable data on the sample quality so as to apply it to the quality control in the production process of fuses. Summary of the invention

[0003] In view of the shortcomings of the prior art, the present invention provides an equivalence class division method to realize non-destructive testing of fuses, which solves the problem that the short-circuit breaking test of the fuse is lossy and the fuse is unusable after use, so only sampling inspection of the fuse can be carried out, and comprehensive quality inspection of the produced fuse products cannot be carried out.

[0004] To achieve the above objectives, the present invention is implemented through the following technical solutions: A non-destructive testing method for fuses using an equivalence class division method comprises the following steps: 1) Test all fuse samples to be tested with a small current and collect data; 2) Initialize the collected data; 3) Calculate or update the data membership matrix U; 4) Update the cluster center matrix V of the data; 5) If the change in the cluster center matrix dV is less than the threshold, stop and output the membership matrix U and the cluster center matrix V, otherwise increase the counter by one and repeat step 4; 6) Based on the clustering results, randomly select individual data in each category, perform high current short-circuit tests, and output the results.

[0005] Preferably, the method of testing all fuse samples to be tested with a small current and collecting data refers to using an appropriate small current that does not damage the fuse, energizing the fuse for a period of time, and collecting data on changes in voltage, current, and temperature over time.

[0006] Preferably, the initialization includes giving the number of cluster categories, setting an iteration stop threshold, initializing the cluster center matrix, and setting the iteration counter b to zero.

[0007] Preferably, the calculation or update of the membership matrix U of the data is performed by using a fuzzy c-means algorithm according to the following formula: ij ; ; Among them, v i is the cluster center matrix, x j For data, Represents the Euclidean distance.

[0008] Preferably, the cluster center matrix V of the updated data is calculated using the fuzzy c-means algorithm according to the following formula: i ]: ; Among them, n is the number of data.

[0009] Preferably, the high current short circuit test, the output result refers to a short circuit high current breaking test on the fuse, judging whether the fuse is working normally according to the breaking time and breaking current, and outputting it as the quality evaluation result of all samples in this cluster.

[0010] An equivalence class division method is used to implement a nondestructive testing method for fuses. A small number of fuse samples in the equivalence class need to undergo high-current destructive testing, while the majority of the remaining samples undergo nondestructive low-current testing.

[0011] The present invention provides a method for implementing nondestructive testing of fuses using an equivalence class division method. It has the following beneficial effects: This equivalence class division method realizes the non-destructive testing method of fuses. A small current test is performed on the fuse, data is collected, and equivalence classes are divided. By testing a single fuse in an equivalence class, the conditions of fuses in all classes can be inferred. By adopting the equivalence class division method to realize the non-destructive testing method of fuses, damage to samples can be avoided, and reliable data on sample quality can be obtained. This method can be applied to quality control in the production process of fuses. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 The workflow diagram of the non-destructive testing method for fuses implemented by the equivalence class partitioning method provided in this application; Figure 2 A temperature variation diagram of a small current test result over time of a nondestructive testing method for a fuse using the equivalence class partitioning method provided in an embodiment of the present application; Figure 3 A current variation over time diagram of a small current test result of a nondestructive testing method for a fuse implemented by the equivalence class partitioning method provided in an embodiment of the present application; Figure 4 A voltage variation over time diagram of a small current test result of a nondestructive testing method for a fuse implemented by the equivalence class division method provided in an embodiment of the present application; Figure 5 A clustering diagram of a non-destructive testing method for fuses implemented by the equivalence class division method provided in an embodiment of the present application; Figure 6 A high current test result diagram of a nondestructive testing method for fuses implemented using the equivalence class partitioning method provided in an embodiment of the present application. DETAILED DESCRIPTION

[0013] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0014] Example 1, please refer to Figure 1 The present invention provides a technical solution: a method for implementing nondestructive testing of fuses using an equivalence class partitioning method, comprising the following steps: 1) Test all fuse samples to be tested with a small current and collect data; 2) Initialize the collected data; 3) Calculate or update the data membership matrix U; 4) Update the cluster center matrix V of the data; 5) If the change in the cluster center matrix dV is less than the threshold, stop and output the membership matrix U and the cluster center matrix V, otherwise increase the counter by one and repeat step 4; 6) Based on the clustering results, randomly select individual data in each category, perform high current short-circuit tests, and output the results.

[0015] It should be further explained that, in the specific implementation process, the implementation process of the non-destructive testing method of the fuse using the equivalence class partitioning method includes the following steps: Step 1. Small current test and data acquisition: Use 5% to 20% of the rated current of the fuse, such as a rated current of 10A, and a test current of 0.5 to 2A to ensure that the fuse action is not triggered. Collect data on voltage (accuracy ±0.1V), current (accuracy ±0.05A), and temperature (monitored by infrared thermal imager, accuracy ±1°C) changing with time in real time. Use Keysight N6705B for constant current power supply, NIPXIe-6363 for high-precision DAQ, or data acquisition card, and FLIRA315 for temperature sensor to output dynamic response data sets for each fuse, including voltage fluctuation curve and temperature rise rate.

[0016] Step 2. Data initialization: Set the number of cluster categories, that is, set the number of categories based on historical data or experience to 5 categories, corresponding to the quality levels of excellent, good, medium, poor, and unqualified, and generate the initial cluster center: randomly select the data vectors of 5 samples as the initial cluster center matrix V (0), set the threshold, that is, set the iteration stop threshold dV threshold =0.01, counter b=0.

[0017] Step 3. Calculate the membership matrix U: The fuzzy c-means algorithm formula is: ; where x j is the jth sample data, v i is the i-th cluster center, Represents the Euclidean distance.

[0018] The Euclidean distance is calculated as: ; where d ij The Euclidean distance between the jth sample and the i-th cluster center, D: data dimension, including three parameters: voltage, current, and temperature, D=3; x j，d : The d-dimensional data of the j-th sample, such as voltage value; v i，d : The d-th dimension value of the cluster center of the ith class; thereby outputting the membership matrix U, which represents the probability that each sample belongs to each cluster.

[0019] Step 4. Update the cluster center matrix V: The update formula is: ; where n is the number of data, such as 1000 fuse samples; that is, the total number of samples, v i : Cluster center vector of the i-th class, u ij : The membership degree of the jth sample to the i-th class, x j : The complete data vector of the jth sample, including voltage, current, and temperature; the cluster center of each category is recalculated by weighted average to complete the matrix update.

[0020] Step 5. Iteration termination judgment: Change calculation formula: ; The termination condition is: if dV<0.01, output the final U and V; otherwise b=b+1, return to step 4.

[0021] Step 6. High current short-circuit verification test: Set sampling rules, that is, randomly select 1 sample from each cluster, and select 5 samples when there are 5 categories; apply rated short-circuit current, such as 10kA, for 10ms, and record the breaking time, arc energy, etc. If a certain type of sample passes the test, such as the breaking time ≤20ms, then all samples of this type are judged to be qualified.

[0022] All fuse samples to be tested are tested with a small current. Data collection refers to the use of a small current that does not damage the fuse, powering on the fuse for a period of time, and collecting data on voltage, current, and temperature changes over time. It should be further explained that in the specific implementation process: the current range is 5% to 20% of the rated current, such as the test current of a 10A fuse is 0.5 to 2A; the power-on time is 3 to 10 minutes, which can be adjusted according to the thermal capacity characteristics of the fuse; among them, voltage / current sampling is collected in real time by a high-precision sensor with a sampling frequency of ≥1kHz, and the temperature is monitored by an infrared thermal imager to record the surface temperature distribution every 5 seconds. The data storage format is a CSV file containing three columns of data: timestamp, voltage, current, and temperature.

[0023] Initialization includes setting the number of cluster categories, setting the iteration stop threshold, initializing the cluster center matrix, and setting the iteration counter b to zero. It should be further explained that in the specific implementation process: the number of categories is set based on the historical quality distribution or empirical value, which is 5 categories. The threshold is set by experimentally verifying the balance between convergence speed and accuracy: dV threshold =0.01.

[0024] During the initialization of cluster centers, for the random method, the data vectors of 5 samples are randomly selected as the initial centers. For the K-means++ optimization, an improved algorithm is used to avoid the initial centers being too close. The counter is initialized with b=0 to record the number of iterations.

[0025] Calculate or update the membership matrix U of the data, and use the fuzzy c-means algorithm according to the following calculation formula to calculate the matrix U=u ij :It needs to be further explained that, in the specific implementation process, the matrix calculation formula is: ; Among them, v i is the cluster center matrix, x j For data, Right now is the Euclidean distance, indicating that sample x j With cluster center v i The similarity.

[0026] The calculation process includes: S41. For each sample x j , and calculate its distance to all cluster centers.

[0027] S42. Generate membership matrix U according to the formula, matrix element u ij ∈[0,1], and the sum of each row is 1.

[0028] Update the cluster center matrix V of the data and use the fuzzy c-means algorithm according to the following calculation formula to calculate the cluster center matrix V=vi :It needs to be further explained that in the specific implementation process: The formula is analyzed as follows: ; Where n is the number of data, that is, the total number of samples, u ij is the element of membership matrix, and each cluster center is recalculated by weighted average to complete the matrix update.

[0029] The calculation process is as follows: First, for each cluster center v i , calculate the weighted average of all samples. The updated v i Reflects the optimal representative point of the current cluster.

[0030] The output result of the high-current short-circuit test refers to the short-circuit and high-current breaking test of the fuse. The fuse is judged whether it is working normally according to the breaking time and breaking current, and is output as the quality evaluation result of all samples in this cluster. It should be further explained that in the specific implementation process: in the test parameters, the current intensity is the rated short-circuit current of 10kA, which lasts until the fuse is operated. The monitoring indicators are: breaking time ≤20ms, arc energy ≤50J. If the sample passes the test, all the clusters to which it belongs are marked as "qualified". If it fails, all samples in the cluster are marked as "unqualified" and need to be re-inspected or scrapped. A quality report is then generated, including the test results and processing suggestions for each cluster.

[0031] A small number of fuse samples in the equivalent class need to undergo high current lossy testing, while the majority of the remaining samples undergo non-destructive low current testing. It should be further explained that in the specific implementation process: 1 sample is taken from each class, that is, 5 samples are taken from 5 classes, accounting for 0.5% of the total. The 995 samples that were not sampled were only tested for low current, retaining full functionality.

[0032] The equivalence class partitioning method realizes the non-destructive testing method of the fuse. It requires a small number of fuse samples in the equivalence class to undergo high-current lossy testing, and the remaining majority of samples are non-destructive low-current testing. The non-destructive testing method of the fuse is realized by adopting the equivalence class partitioning method. The basic principle of this technical solution is: product testing is one of the most important means of product quality assurance. However, as a disposable electrical appliance, it is impossible to perform short-circuit current testing on all fuses, so it is difficult to effectively control the quality of fuse products during the production process. Equivalence class partitioning is to represent the entire data set by selecting an appropriate data subset, that is, it divides all possible data, that is, valid or invalid, into several equivalence classes, selects certain representative values ​​from each equivalence class for testing, and uses relatively few fuse test cases to cover more possible data to discover more product defects. The equivalence class partitioning method can greatly reduce repetitiveness and greatly improve test efficiency.

[0033] Embodiment 2, as Figure 5As shown in the figure, according to the equivalence classification method, nondestructive testing of fuses is realized. Here, 17 fuse samples need to be nondestructively tested and divided into 4 equivalence classes. Figure 1 The steps shown perform the detection: 1) Use a small current to test all fuse samples to be tested and collect data. Performing a small current test on the fuse samples to be tested can generate 17 sets of result data, including the changes in voltage, current, and temperature over time. Figure 2 , Figure 3 and Figure 4 The figures are respectively the small current test results of fuse 1, which are the temperature change over time, the current change over time and the voltage change over time; 2) Initialize the collected data. Initialization means organizing the data into a suitable format to facilitate processing by the algorithm program; 3) Calculate or update the data membership matrix U according to the following formula: ; Among them, v i is the cluster center matrix, x j For data, Represents the Euclidean distance.

[0034] 4) Update the cluster center matrix V of the data according to the following formula: ; Among them, n is the number of data, that is, the total number of samples, u ij is the element of membership matrix, and each cluster center is recalculated by weighted average to complete the matrix update.

[0035] 5) Execute the algorithm. If the change in the cluster center matrix dV is less than the threshold, stop and output the membership matrix U and the cluster center matrix V. Otherwise, increase the counter by one and repeat step 4).

[0036] After this step is executed, the clustering results can be obtained, such as Figure 5 As shown, 4 clustering results are obtained for 17 fuse samples, among which fuse 1 is classified into equivalence class 1.

[0037] 6) According to the clustering results, select fuse 1 in category 1 and perform a high current short circuit test. The output result is as follows: Figure 6 As shown. Figure 6 As a result, it can be seen that fuse 1 can be successfully opened and the quality is qualified. Therefore, correspondingly, the fuses of category 1 all have qualified breaking capacity. So far, the non-destructive testing of all fuse samples in equivalence class 1 is completed.

[0038] In this regard, this method implements a nondestructive testing method for fuses through an equivalence class partitioning method, which can avoid damaging the sample and obtain reliable data on the sample quality.

[0039] Among them, according to Figure 4 As a result, we can see that this category of fuses can be successfully opened and are of qualified quality. Therefore, the non-destructive testing method of fuses can be implemented through the equivalence class division method, which can avoid damaging the sample and obtain reliable data on the sample quality.

[0040] Embodiment 3, a nondestructive testing method for fuses based on the equivalence class partitioning method, when used, its technical implementation process is as follows: First, a small current test is performed on all fuse samples to be tested, using 5% to 20% of the rated current, and the test current is 0.5 to 2A when the rated current is 10A. The data of voltage, current and temperature changes over time are collected in real time through a constant current power supply, a high-precision voltage / current sensor and an infrared thermal imager. The sampling frequency is not less than 1kHz to ensure that the dynamic response curve, including voltage fluctuations and temperature rise rates, is obtained without damaging the fuse. After the data collection is completed, the data is initialized and the number of clustering categories is set according to historical experience or preset rules. When there are 5 categories, the corresponding quality levels are excellent, good, medium, poor, and unqualified, and the initial cluster center matrix V is randomly selected. (0) , and set the iteration stop threshold dV threshold =0.01 and iteration counter b=0.

[0041] Subsequently, the fuzzy c-means algorithm is used for cluster analysis. Specifically, by calculating each sample data point x j With each cluster center v i The Euclidean distance of ||x j -v i ||, and based on the formula: ; Generate a membership matrix U, which represents the probability distribution of each sample belonging to each cluster. After completing the membership calculation, use the formula: ; Update the cluster center matrix V, and determine whether the iteration is terminated by calculating the Euclidean distance change dV between the two cluster center matrices. If dV<0.01, output the final membership matrix and cluster center matrix; otherwise, increment the counter b and repeatedly update the cluster center until the convergence condition is met.

[0042] After clustering, a sample is randomly selected from each category for a high-current short-circuit test. The rated short-circuit current of 10kA is applied for 10ms, and key parameters such as breaking time, maximum breaking current and arc energy are recorded. If the breaking time of a certain type of sample is ≤20ms and the arc energy is ≤50J, the quality of all fuses in the cluster is determined to be qualified; otherwise, it is determined to be unqualified, and the entire type of samples needs to be re-inspected or scrapped. This method combines small current testing with fuzzy clustering analysis. Only 0.5% of the samples, that is, 5 samples, need to be destructively tested to infer the quality of all samples, which significantly reduces the detection cost. Data verification shows that compared with the traditional sampling method, its prediction consistency rate is 98%, and the clustering results are intuitively displayed through a three-dimensional scatter plot, that is, with voltage stability, temperature rise rate, and current fluctuation as the coordinate axes, which is convenient for rapid differentiation of quality levels.

[0043] In addition, the K-means++ algorithm can be used to optimize the selection of initial cluster centers during the initialization process to avoid local optimal problems caused by random initialization; the test data is bound to the fuse through a unique ID to achieve full life cycle quality traceability. This method can be integrated into an automated production line to adjust clustering parameters in real time to adapt to the characteristics of different batches of products, further improving detection efficiency and adaptability. Finally, through the combination of equivalence class division and representative sampling verification, the efficiency and reliability of non-destructive testing of fuses are achieved, providing an innovative solution for the quality control of disposable electronic components.

[0044] In this case, the low-current nondestructive test combined with fuzzy cluster analysis successfully achieved efficient evaluation of fuse quality, greatly reducing the number of destructive test samples while ensuring detection accuracy. This method can be extended to other quality inspection scenarios for disposable electronic components and has certain economic and technical value.

[0045] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the statement "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0046] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for implementing nondestructive testing of fuses using an equivalence class partitioning method, characterized in that: The following steps are involved: 1) Test all fuse samples to be tested with a small current and collect data; 2) Initialize the collected data; 3) Calculate or update the data membership matrix U; 4) Update the cluster center matrix V of the data; 5) If the change in the cluster center matrix dV is less than the threshold, stop and output the membership matrix U and the cluster center matrix V, otherwise increase the counter by one and repeat step 4; 6) Based on the clustering results, randomly select individual data in each category, perform high current short-circuit tests, and output the results.

2. The method for implementing nondestructive testing of fuses using an equivalence class partitioning method according to claim 1 is characterized in that: The method of testing all fuse samples to be tested with a small current and collecting data refers to using a small current that does not damage the fuse, energizing the fuse for a period of time, and collecting data on changes in voltage, current, and temperature over time.

3. The method for implementing nondestructive testing of fuses using an equivalence class partitioning method according to claim 2 is characterized in that: The initialization includes giving the number of cluster categories, setting an iteration stop threshold, initializing the cluster center matrix, and setting the iteration counter b to zero.

4. The method for implementing nondestructive testing of fuses using an equivalence class partitioning method according to claim 3 is characterized in that: The calculation or update of the membership matrix U of the data adopts the fuzzy c-means algorithm according to the following calculation formula: ij : ； Among them, v i is the cluster center matrix, x j For data, Represents the Euclidean distance.

5. The method for implementing nondestructive testing of fuses using an equivalence class partitioning method according to claim 4 is characterized in that: The cluster center matrix V of the updated data is calculated using the fuzzy c-means algorithm according to the following calculation formula: i : ； Among them, n is the number of data.

6. The method for implementing nondestructive testing of fuses using an equivalence class partitioning method according to claim 5 is characterized in that: The high current short circuit test outputs a result of a short circuit high current breaking test on the fuse, and judges whether the fuse is working normally based on the breaking time and breaking current, and outputs the result as the quality evaluation result of all samples in this cluster.

7. The method for implementing nondestructive testing of fuses using an equivalence class partitioning method according to claim 6 is characterized in that: A small number of fuse samples in the equivalence class need to undergo high-current destructive testing, while the majority of the remaining samples require non-destructive low-current testing.

Images