A method and device for detecting vulcanization uniformity of composite insulator

Abstract

The present invention relates to the field of data processing, and in particular to a method and device for detecting the vulcanization uniformity of a composite insulator. The method comprises the following steps: obtaining an NMR spectrum of an insulator after vulcanization; performing a first-layer decomposition process on the NMR spectrum using a wavelet filtering algorithm to obtain a detail coefficient; distinguishing the noise detail coefficient from the signal detail coefficient in the approximate coefficient; performing a second-layer decomposition process on the approximate coefficient to obtain a detail coefficient; calculating the signal purity; in response to the signal purity being less than a preset purity threshold, screening out the noise detail coefficient in the detail coefficient obtained by the second-layer decomposition process; continuing to decompose and calculate the signal purity, and in response to the signal purity being not less than the preset purity threshold, stopping the decomposition; reconstructing the denoised NMR spectrum after correcting and adjusting the noise detail coefficient; so as to realize the vulcanization uniformity detection. Thus, the accuracy of the vulcanization uniformity detection is improved by improving the accuracy of denoising.

Description

Technical Field

[0001] The present invention relates to the field of data processing, and in particular to a method and a device for detecting the vulcanization uniformity of a composite insulator. Background Art

[0002] In order to improve the insulation performance and mechanical strength of insulators, they need to be vulcanized before leaving the factory. Uneven vulcanization will lead to a decrease in electrical performance, and even safety hazards in serious cases. Therefore, after the insulators are vulcanized, some insulators will be extracted to test their vulcanization uniformity.

[0003] The cross-link density detection method is often used for vulcanization uniformity detection. There are currently a variety of cross-link density detection methods, among which the nuclear magnetic resonance method is widely used in cross-link density detection. When determining the cross-link density, the nuclear magnetic resonance method needs to first collect the nuclear magnetic resonance spectrum, and then calculate the cross-link density by analyzing the nuclear magnetic resonance spectrum. Therefore, the accuracy of the nuclear magnetic resonance spectrum will affect the detection accuracy of the cross-link density. In the process of collecting the nuclear magnetic resonance spectrum, it is disturbed by factors such as magnetic field and sensor noise, resulting in noise in the collected nuclear magnetic resonance spectrum, which will affect the subsequent calculation accuracy of the cross-link density. Therefore, denoising is required before using the nuclear magnetic resonance spectrum to calculate the cross-link density.

[0004] As a commonly used denoising algorithm, the wavelet filter algorithm has a good denoising effect. However, the accuracy of the wavelet filter algorithm denoising is affected by the parameter settings, such as the number of decomposition layers. When the number of decomposition layers is set too large, it will cause a large loss of useful information in the signal, thereby affecting the subsequent related analysis. When the number of decomposition layers is set too small, it will lead to the inability to extract the noise signal, thus affecting the denoising effect. Therefore, how to accurately set the relevant parameters in the wavelet filter algorithm has become the research focus of this invention.

[0005] The patent application document with publication number CN117647636A discloses a method and system for detecting the vulcanization uniformity of a composite insulator. The method in the patent application document mainly discloses how to use the crosslinking density to detect the vulcanization uniformity, and does not involve the denoising related content of the nuclear magnetic resonance spectrum. Therefore, the method in the patent application document cannot solve the technical problem in the present invention. Summary of the invention

[0006] In order to solve the problem of how to accurately set relevant parameters in a wavelet filtering algorithm, the present invention provides a method and device for detecting the vulcanization uniformity of a composite insulator.

[0007] In a first aspect, the present invention provides a method for detecting the vulcanization uniformity of a composite insulator, which adopts the following technical solution:

[0008] A method for detecting the vulcanization uniformity of a composite insulator comprises the following steps:

[0009] Obtaining NMR spectra of the insulator after vulcanization;

[0010] The first-layer decomposition of the NMR spectrum is performed using the wavelet filtering algorithm to obtain detail coefficients and approximate coefficients; the description vector of each detail coefficient constructed in advance is obtained; all detail coefficients are clustered into two categories according to the similarity of the wave vectors of every two detail coefficients; the regularity of the detail coefficients in each category is analyzed, and the detail coefficients in the category with small regularity and the detail coefficients in the category with large regularity are recorded as noise detail coefficients and signal detail coefficients respectively;

[0011] The approximate coefficients are decomposed in the second layer to obtain detail coefficients and approximate coefficients; the signal purity C is calculated. , , , Respectively represent the distribution interval length of the description vector of all signal detail coefficients, all noise detail coefficients and the detail coefficients obtained by the second layer decomposition processing in the direction of the i-th principal component, represents the number of principal component directions, represents a hyperbolic tangent function; in response to the signal purity being less than a preset purity threshold, filtering out the noise detail coefficients in the detail coefficients obtained by the second layer decomposition processing by analyzing the similarity with the distribution of the description vector of the signal detail coefficient; continuing to decompose and calculate the signal purity, and in response to the signal purity being not less than the preset purity threshold, stopping the decomposition;

[0012] After the noise detail coefficient is corrected and adjusted, the denoised NMR spectrum is reconstructed to achieve vulcanization uniformity detection.

[0013] The present invention controls the decomposition process of the NMR spectrum through a wavelet filtering algorithm to prevent the signal loss in the over-decomposition process and the problem of insufficient noise extraction caused by under-decomposition; further, when controlling the decomposition process of the NMR spectrum, the stop control is performed by analyzing the presence of noise information in the decomposed information, thereby preventing the over-decomposition caused by continuing the decomposition when there is less noise information; or preventing the under-decomposition caused by stopping the decomposition when there is still much noise, thereby improving the accuracy of the decomposition; further, when analyzing the presence of noise in the decomposed information, the distribution difference between the newly decomposed information and the noise information or the signal information in each main component direction is introduced to accurately determine the interference of the newly decomposed information with noise.

[0014] Preferably, the obtaining of the pre-constructed description vector of each detail coefficient includes:

[0015] A number of interception lengths are preset; based on each interception length, detail coefficients are equally intercepted on both sides of a detail coefficient with a detail coefficient as the center, and a data segment consisting of the intercepted detail coefficients is recorded as an analysis segment;

[0016] Fit the polynomial of each analysis segment and obtain the vector consisting of the coefficients of the polynomial, recorded as the coefficient vector;

[0017] The coefficient vectors obtained from all analysis segments of the detail coefficient are summed to obtain the description vector of the detail coefficient.

[0018] The present invention constructs a description vector of the detail coefficient by analyzing the fitting rules of analysis segments of different scales obtained around the detail coefficient, thereby more accurately describing the variation rule of the detail coefficient and providing a data basis for subsequent signal decomposition control.

[0019] Preferably, the method of clustering all detail coefficients into two categories according to the similarity of the fluctuation vectors of every two detail coefficients includes:

[0020] The cosine similarity of the fluctuation vectors of every two detail coefficients is calculated, and the cosine similarity is used as the clustering distance. The k-means algorithm is used to cluster all detail coefficients into two categories.

[0021] The present invention separates noise and signal by clustering, and the implementation method is simpler and more efficient.

[0022] Preferably, the analyzing the regularity of detail coefficients in each category includes:

[0023] For any detail coefficient in any category, a fitting value of the detail coefficient is obtained by using the polynomial of each analysis segment of the detail coefficient, and a fitting deviation of the detail coefficient in the analysis segment is calculated by using the fitting value;

[0024] The length of each analysis segment of the detail coefficient is used as a weight, and the fitting deviations obtained from all analysis segments of the detail coefficient are weighted summed up, and the reciprocal of the accumulated sum is taken to obtain the regularity of the detail coefficient.

[0025] The present invention utilizes the fitting deviation of detail coefficients at various scales to accurately reflect the regularity of the detail coefficients, thereby more accurately distinguishing noise information from signal information.

[0026] Preferably, the step of screening out the noise detail coefficients from the detail coefficients obtained by the second-layer decomposition processing by analyzing the similarity with the distribution of the description vector of the signal detail coefficient comprises:

[0027] Record the detail coefficient obtained by the second layer decomposition process as the analysis detail coefficient; obtain all principal component directions of the description vectors of all analysis detail coefficients within a preset range of the analysis detail coefficient;

[0028] Calculating distribution similarity , represents the length of the distribution interval of all analysis detail coefficients within a preset range of an analysis detail coefficient in the direction of the i-th principal component, and S represents the distribution similarity between the analysis detail coefficient and the signal detail coefficient;

[0029] According to the distribution similarity, the analysis detail coefficients are clustered into two categories, and the analysis detail coefficients in the category with a small distribution similarity mean are taken as noise detail coefficients.

[0030] The present invention screens out noise information by analyzing the distribution difference between the newly decomposed information and the signal information, providing a basis for subsequent accurate denoising.

[0031] Preferably, reconstructing the denoised NMR spectrum after correcting and adjusting the noise detail coefficient comprises:

[0032] Calculate the cosine similarity of the description vectors of the noise detail coefficient and each signal detail coefficient, and take the signal detail coefficient with the largest cosine similarity as the reference signal detail coefficient;

[0033] The product of the reference signal detail coefficient and the cosine similarity is added to the product of the noise detail coefficient and the difference between 1 and the cosine similarity to obtain a corrected noise detail coefficient;

[0034] Based on the corrected noise detail coefficient and signal detail coefficient, the denoised NMR spectrum was reconstructed using the wavelet filtering algorithm.

[0035] Preferably, the method for obtaining the distribution interval length includes:

[0036] Get the projection points of each description vector in the direction of each principal component, get the maximum and minimum values ​​of all projection points, and use the difference between the maximum and minimum values ​​of all projection points as the length of the distribution interval.

[0037] In a second aspect, the present invention provides a composite insulator vulcanization uniformity detection device, which adopts the following technical solution:

[0038] A composite insulator vulcanization uniformity detection device comprises: a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the composite insulator vulcanization uniformity detection method mentioned above is implemented.

[0039] By adopting the above technical solution, the above-mentioned composite insulator vulcanization uniformity detection method is generated into a computer program and stored in a memory so as to be loaded and executed by a processor, thereby manufacturing a terminal device based on the memory and the processor for easy use.

[0040] The present invention has the following technical effects:

[0041] The present invention controls the decomposition process of the NMR spectrum through a wavelet filtering algorithm to prevent the signal loss in the over-decomposition process and the problem of insufficient noise extraction caused by under-decomposition;

[0042] Furthermore, when controlling the decomposition process of the NMR spectrum, the stop control is performed by analyzing the presence of noise information in the decomposed information, thereby preventing the decomposition from continuing when there is less noise information, resulting in over-decomposition; or preventing the decomposition from stopping when there is still much noise, resulting in under-decomposition, thereby improving the accuracy of the decomposition;

[0043] Furthermore, when analyzing the presence of noise in the decomposed information, the distribution difference between the newly decomposed information and the noise information or the signal information in each principal component direction is introduced to accurately determine the interference of the newly decomposed information with the noise. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] By reading the detailed description below with reference to the accompanying drawings, the above and other purposes, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-restrictive manner, and the same or corresponding reference numerals represent the same or corresponding parts.

[0045] Figure 1 The present invention is a flowchart of a method for detecting the vulcanization uniformity of a composite insulator according to an embodiment of the present invention. DETAILED DESCRIPTION

[0046] 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 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 those skilled in the art without creative work are within the scope of protection of the present invention.

[0047] It should be understood that when the terms "first", "second", etc. are used in the claims, descriptions, and drawings of the present invention, they are only used to distinguish different objects, rather than to describe a specific order. The terms "include" and "comprise" used in the description and claims of the present invention indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their collections.

[0048] The embodiment of the present invention discloses a method for detecting the uniformity of vulcanization of a composite insulator, referring to Figure 1 , comprising steps S1 to S4:

[0049] S1: Obtain the NMR spectrum of the insulator after vulcanization.

[0050] Specifically, a portion of the insulators after vulcanization are extracted as measurement samples, and a nuclear magnetic resonance spectrum of the measurement samples is collected and recorded as an NMR spectrum.

[0051] S2: Use the wavelet filtering algorithm to perform the first-level decomposition processing on the NMR spectrum to obtain detail coefficients and approximate coefficients; obtain the pre-constructed description vector of each detail coefficient; cluster all detail coefficients into two categories according to the similarity of the wave vectors of every two detail coefficients; analyze the regularity of the detail coefficients in each category, and record the detail coefficients in the category with small regularity and the detail coefficients in the category with large regularity as noise detail coefficients and signal detail coefficients, respectively.

[0052] It should be noted that in order to prevent the phenomenon of over-decomposition and under-decomposition by directly setting the number of decomposition layers without analyzing the decomposition effect, it is necessary to analyze the information decomposed during the decomposition process to determine whether there is still noise in the decomposed information, so as to perform decomposition stop control according to the presence of noise in the latest decomposed information.

[0053] It should be further explained that the wavelet filtering algorithm adopts a recursive decomposition method. As the number of decomposition layers increases, the noise information will become less and less until the noise disappears. Since noise is inevitable, some information in the first decomposition layer will be noise information. Since the change characteristics of noise information are different from those of signal information, all information can be clustered by the change characteristics to separate noise information from signal information.

[0054] S20: Performing the first-level decomposition processing on the NMR spectrum using the wavelet filtering algorithm to obtain detail coefficients and approximate coefficients; obtaining a pre-constructed description vector for each detail coefficient; and clustering all detail coefficients into two categories according to the similarity of the wave vectors of every two detail coefficients.

[0055] It should be noted that since noise information is generally high-frequency information, detail coefficients mainly describe high-frequency information, and thus noise information mainly exists in detail coefficients. In order to use the variation characteristics to distinguish noise information from signal information in detail coefficients, it is necessary to first describe the variation characteristics of each detail coefficient, and then cluster them according to the variation characteristics to distinguish noise information from signal information.

[0056] Preferably, as an example, the first layer decomposition processing of the NMR spectrum is performed using a wavelet filtering algorithm to obtain detail coefficients and approximate coefficients; a description vector of each detail coefficient constructed in advance is obtained; and all detail coefficients are clustered into two categories according to the similarity of the wave vectors of every two detail coefficients, including:

[0057] The wavelet filtering algorithm is used to perform the first-level decomposition of the NMR spectrum to obtain detail coefficients and approximate coefficients;

[0058] Several truncation lengths are preset; based on each truncation length, with a detail coefficient as the center, detail coefficients are equally truncated on both sides of the detail coefficient, and the data segment composed of the truncated detail coefficients is recorded as an analysis segment; in this embodiment, five truncation lengths are preset, and these five truncation lengths are 5, 11, 17, 23, and 29 respectively. Other embodiments are adjusted according to actual conditions, and this embodiment does not make specific restrictions.

[0059] Fit the polynomial of each analysis segment and obtain the vector consisting of the coefficients of the polynomial, recorded as the coefficient vector;

[0060] The coefficient vectors obtained from all analysis segments of the detail coefficient are summed to obtain the description vector of the detail coefficient.

[0061] The cosine similarity of the fluctuation vectors of every two detail coefficients is calculated, and the cosine similarity is used as the clustering distance. The k-means algorithm is used to cluster all detail coefficients into two categories.

[0062] It should be noted that the use of a wavelet filtering algorithm to perform a first-level decomposition process on an NMR spectrum to obtain detail coefficients and approximate coefficients is a prior art and will not be described in detail here.

[0063] It can be understood that different analysis segments reflect the variation information of the detail coefficient at different scales. The coefficients of the polynomial reflect the variation characteristics of the analysis segment, and the description vector composed of the coefficients of the polynomials of all analysis segments reflects the variation characteristics of the detail coefficient at different scales.

[0064] S21: Analyze the regularity of the detail coefficients in each category, and record the detail coefficients in the category with small regularity and the detail coefficients in the category with large regularity as noise detail coefficients and signal detail coefficients, respectively.

[0065] It should be noted that although clustering separates noise information from signal information, it is impossible to distinguish which category is noise information. Noise information is highly random and has poor regularity, so regularity can be used to identify which category is noise information.

[0066] Preferably, as an example, the regularity of the detail coefficients in each category is analyzed, and the detail coefficients in the category with small regularity and the detail coefficients in the category with large regularity are recorded as noise detail coefficients and signal detail coefficients, respectively, including:

[0067] For any detail coefficient in any category, a fitting value of the detail coefficient is obtained by using the polynomial of each analysis segment of the detail coefficient, and the absolute value of the difference between the detail coefficient and the fitting value is taken as the fitting deviation of the detail coefficient in the analysis segment;

[0068] The length of each analysis segment of the detail coefficient is used as a weight, and the fitting deviations obtained from all analysis segments of the detail coefficient are weighted summed up, and the reciprocal of the accumulated sum is taken to obtain the regularity of the detail coefficient.

[0069] The regularity mean of all detail coefficients in each category is calculated, and the detail coefficients in the category with a small regularity mean are recorded as noise detail coefficients, and the detail coefficients in the category with a large regularity mean are recorded as signal detail coefficients.

[0070] It can be understood that the fitting deviation reflects the change regularity of the detail coefficient. The larger the value, the smaller the change regularity of the detail coefficient. At the same time, the longer the analysis segment is, the more information it contains, and the less the law fitted by the analysis segment is affected by noise. Therefore, the length of the analysis segment is used as the weight to perform weighted summation on the regularity.

[0071] S3: Perform a second-level decomposition process on the approximate coefficients to obtain detail coefficients and approximate coefficients; calculate signal purity; in response to the signal purity being less than a preset purity threshold, filter out the noise detail coefficients in the detail coefficients obtained by the second-level decomposition process by analyzing the similarity with the distribution of the description vector of the signal detail coefficients; continue to decompose and calculate the signal purity, and stop the decomposition in response to the signal purity being not less than the preset purity threshold.

[0072] It should be noted that in order to perform decomposition cutoff control, it is necessary to analyze the presence of noise in the decomposed information. Since the distribution similarity between signal information and signal information is large, and the distribution similarity between noise information and signal information is small, the presence of noise in the decomposed information can be judged by analyzing the distribution similarity between the decomposed information and the noise information.

[0073] S30: performing a second-level decomposition process on the approximate coefficients to obtain detail coefficients and approximate coefficients; and calculating signal purity.

[0074] Preferably, as an example, performing a second-layer decomposition process on the approximate coefficient to obtain a detail coefficient and an approximate coefficient; calculating the signal purity includes:

[0075] The approximate coefficients are decomposed into detail coefficients and approximate coefficients using the wavelet filtering algorithm.

[0076] Calculate signal purity:

[0077]

[0078] in, Represents the distribution interval length of the description vector of all signal detail coefficients in the direction of the i-th principal component, Represents the distribution interval length of the description vector of all noise detail coefficients in the direction of the i-th principal component, Represents the distribution interval length of the description vector of the detail coefficient obtained by the second-level decomposition process in the direction of the i-th principal component, represents the number of principal component directions, represents the hyperbolic tangent function, which is used for normalization processing, and C represents the signal purity of the detail coefficient obtained by the second-layer decomposition processing.

[0079] Understandably, It reflects the similarity between the description vector of the detail coefficients obtained by the second-level decomposition and the description vector of the signal detail coefficients. The larger the value, the more similar the information obtained by the second-level decomposition is to the signal information distribution. Therefore, the information obtained by the second-level decomposition contains more signal information and less noise information. It reflects the similarity between the description vector of the detail coefficient obtained by the second-level decomposition processing and the description vector of the noise detail coefficient. The larger the value is, the more similar the distribution of the information obtained by the second-level decomposition processing is to that of the noise information. Therefore, there is less signal information and more noise information in the information obtained by the second-level decomposition processing.

[0080] It should be added that the method for obtaining the length of the distribution interval includes:

[0081] The principal component analysis algorithm is used to perform principal component analysis on all description vectors to obtain several principal component directions, obtain the projection points of each description vector in each principal component direction, obtain the maximum and minimum values ​​of all projection points, and take the difference between the maximum and minimum values ​​of all projection points as the distribution interval length.

[0082] S31: In response to the signal purity being less than a preset purity threshold, the noise detail coefficients in the detail coefficients obtained by the second layer decomposition processing are screened out by analyzing the similarity with the distribution of the description vector of the signal detail coefficients; the decomposition is continued and the signal purity is calculated, and in response to the signal purity being not less than the preset purity threshold, the decomposition is stopped.

[0083] It should be noted that, during the decomposition process, if the decomposed information does not contain noise information or has less noise information, there is no need to decompose it again to prevent unnecessary information loss caused by excessive decomposition. Signal purity reflects the presence of noise information in the decomposed signal, so the decomposition stop control can be performed through signal purity.

[0084] It should be further explained that, in order to facilitate the subsequent denoising process, the noise information needs to be screened out after each decomposition.

[0085] Preferably, as an example, in response to the signal purity being less than a preset purity threshold, filtering out the noise detail coefficients in the detail coefficients obtained by the second-layer decomposition processing by analyzing the similarity with the distribution of the description vector of the signal detail coefficient; continuing to decompose and calculate the signal purity, and in response to the signal purity being not less than the preset purity threshold, stopping the decomposition, including:

[0086] In response to the signal purity being less than a preset purity threshold, the detail coefficient obtained by the second layer decomposition processing is recorded as the analysis detail coefficient; all analysis detail coefficients within a preset range of the analysis detail coefficient are obtained, and the principal component analysis algorithm is used to perform principal component analysis on the description vectors of all analysis detail coefficients within the preset range of the analysis detail coefficient to obtain a number of principal component directions; this embodiment is described using the preset range of 21 as an example, and other embodiments may take other values, and this embodiment does not impose any specific restrictions.

[0087] Calculating distribution similarity , represents the distribution interval length of all analysis detail coefficients within a preset range of an analysis detail coefficient in the direction of the i-th principal component, and S represents the distribution similarity between the analysis detail coefficient and the signal detail coefficient;

[0088] Understandably, It reflects the distribution similarity between the variation characteristics of the analysis detail coefficient and the variation characteristics of the signal detail coefficient. The larger the value is, the greater the distribution similarity between the variation characteristics of the analysis detail coefficient and the variation characteristics of the signal detail coefficient is, and therefore the possibility that the analysis detail coefficient is a noise detail coefficient is smaller.

[0089] According to the distribution similarity, the k-means clustering algorithm is used to cluster the analysis detail coefficients into two categories, and the analysis detail coefficients in the category with a small distribution similarity mean are taken as noise detail coefficients.

[0090] It should be supplemented that the method for obtaining all analysis detail coefficients within a preset range of the analysis detail coefficient includes: taking an analysis detail coefficient as the center, and equally obtaining analysis detail coefficients within a preset range on both sides of the analysis detail coefficient as all analysis detail coefficients within the preset range of the analysis detail coefficient.

[0091] The signal is further decomposed according to the method in step S30 and the signal purity is calculated. In response to the signal purity being not less than a preset purity threshold, the decomposition ends.

[0092] It should be noted that the above steps complete the decomposition control and screen out the noise detail coefficients. Next, denoising processing needs to be performed based on the screened noise detail coefficients.

[0093] S4: After correcting and adjusting the noise detail coefficient, reconstruct the denoised NMR spectrum to achieve vulcanization uniformity detection.

[0094] It should be noted that the screened noise detail coefficients may contain not only noise information but also signal information. By removing the noise detail coefficients by setting them to zero, excessive denoising may easily occur, resulting in signal information loss.

[0095] S40: After correcting and adjusting the noise detail coefficient, a denoised NMR spectrum is reconstructed.

[0096] Preferably, as an example, reconstructing the denoised NMR spectrum after correcting and adjusting the noise detail coefficient includes:

[0097] Calculate the cosine similarity of the description vectors of the noise detail coefficient and each signal detail coefficient, and take the signal detail coefficient with the largest cosine similarity as the reference signal detail coefficient;

[0098] The product of the reference signal detail coefficient and the cosine similarity is added to the product of the noise detail coefficient and the difference between 1 and the cosine similarity to obtain a corrected noise detail coefficient;

[0099] Based on the corrected noise detail coefficient and signal detail coefficient, the denoised NMR spectrum was reconstructed using the wavelet filtering algorithm.

[0100] It can be understood that the similarity of the description vector is used to select the reference signal detail coefficient that is similar to the noise detail coefficient, and then the reference detail coefficient is used to correct the value of the noise detail coefficient to remove the noise. This method can effectively prevent the loss of useful information caused by excessive denoising.

[0101] S41: To achieve vulcanization uniformity detection.

[0102] Preferably, as an example, to realize vulcanization uniformity detection, the method includes:

[0103] The cross-link density of the measured sample was calculated based on the NMR spectrum using the method introduced in the article "Comparison of Three Nuclear Magnetic Resonance Methods for Measuring Rubber Cross-link Density" published in the fourth issue of Spectroscopy in 2017;

[0104] Then, the cross-linking density of the insulator is calculated based on the cross-linking density using the method in the patent application document with publication number CN117647636A and titled “A method and system for detecting the vulcanization uniformity of a composite insulator”.

[0105] An embodiment of the present invention further discloses a composite insulator vulcanization uniformity detection device, comprising a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a composite insulator vulcanization uniformity detection method according to the present invention is implemented.

[0106] The above-mentioned device also includes other components well known to those skilled in the art, such as a communication bus and a communication interface, whose configuration and functions are known in the art and thus will not be described in detail here.

[0107] In the present invention, the aforementioned memory may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. For example, a computer-readable storage medium may be any suitable magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory, a dynamic random access memory, a static random access memory, an enhanced dynamic random access memory, a high bandwidth memory, a hybrid storage cube, etc., or any other medium that can be used to store the required information and can be accessed by an application, a module, or both. Any such computer storage medium may be part of the device or accessible or connectable to the device.

[0108] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

[0109] The above are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for detecting the vulcanization uniformity of a composite insulator, characterized in that: Includes steps: Obtaining NMR spectra of the insulator after vulcanization; The first-layer decomposition of the NMR spectrum is performed using the wavelet filtering algorithm to obtain detail coefficients and approximate coefficients; the description vector of each detail coefficient constructed in advance is obtained; all detail coefficients are clustered into two categories according to the similarity of the wave vectors of every two detail coefficients; the regularity of the detail coefficients in each category is analyzed, and the detail coefficients in the category with small regularity and the detail coefficients in the category with large regularity are recorded as noise detail coefficients and signal detail coefficients respectively; The approximate coefficients are decomposed in the second layer to obtain detail coefficients and approximate coefficients; the signal purity C is calculated. , , , Respectively represent the distribution interval length of the description vector of all signal detail coefficients, all noise detail coefficients and the detail coefficients obtained by the second layer decomposition processing in the direction of the i-th principal component, represents the number of principal component directions, represents the hyperbolic tangent function; In response to the signal purity being less than a preset purity threshold, the noise detail coefficients in the detail coefficients obtained by the second-layer decomposition processing are screened out by analyzing the similarity with the distribution of the description vector of the signal detail coefficients, and the decomposition is continued and the signal purity is calculated. The subsequent decomposition is the same as the second-layer decomposition method; In response to the signal purity being not less than a preset purity threshold, stopping the decomposition; After correcting and adjusting the noise detail coefficients of each layer, the denoised NMR spectrum is reconstructed to achieve vulcanization uniformity detection.

2. A composite insulator vulcanization uniformity detection method according to claim 1, characterized in that: The step of obtaining a pre-constructed description vector of each detail coefficient includes: A number of interception lengths are preset; based on each interception length, detail coefficients are equally intercepted on both sides of a detail coefficient with a detail coefficient as the center, and a data segment consisting of the intercepted detail coefficients is recorded as an analysis segment; Fit the polynomial of each analysis segment and obtain the vector consisting of the coefficients of the polynomial, recorded as the coefficient vector; The coefficient vectors obtained from all analysis segments of the detail coefficient are summed to obtain the description vector of the detail coefficient.

3. A composite insulator vulcanization uniformity detection method according to claim 1, characterized in that: According to the similarity of the fluctuation vectors of every two detail coefficients, all detail coefficients are clustered into two categories, including: The cosine similarity of the fluctuation vectors of every two detail coefficients is calculated, and the cosine similarity is used as the clustering distance. The k-means algorithm is used to cluster all detail coefficients into two categories.

4. A composite insulator vulcanization uniformity detection method according to claim 2, characterized in that: The analysis of the regularity of detail coefficients in each category includes: For any detail coefficient in any category, a fitting value of the detail coefficient is obtained by using the polynomial of each analysis segment of the detail coefficient, and a fitting deviation of the detail coefficient in the analysis segment is calculated by using the fitting value; The length of each analysis segment of the detail coefficient is used as a weight, and the fitting deviations obtained from all analysis segments of the detail coefficient are weighted summed up, and the reciprocal of the accumulated sum is taken to obtain the regularity of the detail coefficient.

5. The method for detecting vulcanization uniformity of composite insulators according to claim 1, characterized in that: The step of screening out the noise detail coefficients from the detail coefficients obtained by the second-layer decomposition process by analyzing the similarity with the distribution of the description vector of the signal detail coefficients includes: Record the detail coefficient obtained by the second layer decomposition process as the analysis detail coefficient; obtain all principal component directions of the description vectors of all analysis detail coefficients within a preset range of the analysis detail coefficient; Calculating distribution similarity , represents the length of the distribution interval of all analysis detail coefficients within a preset range of an analysis detail coefficient in the direction of the i-th principal component, and S represents the distribution similarity between the analysis detail coefficient and the signal detail coefficient; According to the distribution similarity, the analysis detail coefficients are clustered into two categories, and the analysis detail coefficients in the category with a small distribution similarity mean are taken as noise detail coefficients.

6. A composite insulator vulcanization uniformity detection method according to claim 1, characterized in that: The reconstructing the denoised NMR spectrum after correcting and adjusting the noise detail coefficient comprises: Calculate the cosine similarity of the description vectors of the noise detail coefficient and each signal detail coefficient, and take the signal detail coefficient with the largest cosine similarity as the reference signal detail coefficient; The product of the reference signal detail coefficient and the cosine similarity is added to the product of the noise detail coefficient and the difference between 1 and the cosine similarity to obtain a corrected noise detail coefficient; Based on the approximation coefficient, signal detail coefficient and corrected noise detail coefficient, the denoised NMR spectrum was reconstructed using the wavelet filtering algorithm.

7. A composite insulator vulcanization uniformity detection method according to claim 1, characterized in that: The method for obtaining the distribution interval length includes: Get the projection points of each description vector in the direction of each principal component, get the maximum and minimum values ​​of all projection points, and use the difference between the maximum and minimum values ​​of all projection points as the length of the distribution interval.

8. A composite insulator vulcanization uniformity detection device, characterized in that: include: A processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, a composite insulator vulcanization uniformity detection method according to any one of claims 1 to 7 is implemented.

Images