Cable fault detection method and storage medium

By segmenting and fusing local and overall cable signals, and combining them with a fault detection model, the problem of insufficient accuracy in existing cable detection methods is solved, achieving efficient and accurate cable fault identification.

CN118981666APending Publication Date: 2024-11-19HANGZHOU JUQI INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411121915.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-15
Publication Date
2024-11-19

AI Technical Summary

Technical Problem

Existing cable testing methods cannot accurately obtain fault data, leading to frequent false fault reports and affecting the stability and reliability of the power system.

Method used

By acquiring local and overall cable signals, including images and sounds, over multiple time periods, segmentation, fusion, and processing are performed. A trained fault detection model is then used for comprehensive judgment to obtain cable fault data.

Benefits of technology

It improves the accuracy of cable fault detection, reduces the false alarm rate, reduces the time cost of fault diagnosis, and achieves efficient identification of cable faults.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118981666A_ABST
    Figure CN118981666A_ABST
Patent Text Reader

Abstract

The invention provides a cable fault detection method and a storage medium, and the method comprises the steps: obtaining a plurality of local signals and a plurality of overall signals of a cable in a plurality of time periods, and judging whether fault information exists in the plurality of local signals and the plurality of overall signals or not; if fault information exists in the local image or the overall image, the overall image is segmented and fused to obtain a first target image, and other local signals and / or overall signals without fault information are correspondingly integrated according to the sequence of the multiple time periods to obtain a second target image; if the local sound or the overall sound has the fault information, segmenting and fusing the overall sound to obtain a first target sound, and processing other local signals and overall signals without the fault information to obtain a second target sound; and inputting the first target image, the second target image, the first target sound and the second target sound into a trained fault detection model to obtain fault data of the current cable.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of cable technology, and in particular to a cable fault detection method and a storage medium. Background Art

[0002] Cables are an important part of the power system, and their safe operation is crucial to the stability and reliability of the entire power system. However, due to long-term use and the influence of environmental factors, cables may have various defects and faults, such as aging, wear, and rupture of the insulation layer. These problems will not only affect the efficiency and quality of power transmission, but may also cause serious safety accidents. Therefore, the timeliness of cable defect detection and identification accuracy are crucial to maintaining and improving the reliability of the cable system.

[0003] Existing methods often focus on monitoring and analyzing the abnormality of a certain power parameter of the cable, such as abnormal values ​​or abnormal fluctuations of voltage or current, in terms of cable fault detection. However, due to natural factors or other factors, voltage anomalies or current fluctuations are sometimes short-term and accidental, so this single parameter monitoring method may not accurately reflect whether there is a real fault in the cable, and may produce some false alarms. Summary of the invention

[0004] The problem solved by the present invention is that the existing cable detection method cannot accurately obtain fault data.

[0005] In order to solve the above problems, the present invention provides a method for detecting cable faults, the detection method comprising: Acquire multiple local signals and multiple overall signals of the cable in multiple time periods, each of the local signals includes multiple local images and multiple local sounds of the cable, and each of the overall signals includes an overall image and an overall sound of the cable, wherein the local signals and the overall signals acquired in the same time period correspond one to one; Determining whether there is fault information in the plurality of local signals and the plurality of overall signals; If at least one local image in any of the local signals or at least one overall image in any of the overall signals has the fault information, the overall image in the corresponding overall signal is segmented and fused according to the local images in the corresponding local signals to obtain a first target image, and the other local signals and / or the overall signals that do not have the fault information are integrated one by one according to the order of the multiple time periods to obtain a second target image; If at least one local sound in any of the local signals or at least one overall sound in any of the overall signals has the fault information, the overall sound in the corresponding overall signal is segmented and merged according to the local sounds in the corresponding local signals to obtain a first target sound, and the other local signals and the overall signal that do not have the fault information are processed to obtain a second target sound; The first target image, the second target image, the first target sound and the second target sound are input into a trained fault detection model to obtain fault data of the current cable.

[0006] Optionally, the step of segmenting and fusing the overall image in the corresponding overall signal according to the local image in the corresponding local signal to obtain the first target image includes: Acquire the number of local images containing the fault information in the local signal, and mark the local images containing the fault information as local images to be fused; Dividing the overall image in the overall signal into a plurality of sub-overall images according to the number, and obtaining similarities between the plurality of sub-overall images and each of the local images to be fused; Determining whether the overall image corresponding to the partial image contains the fault information; If the fault information exists in the overall image, the local image to be fused and the plurality of sub-overall images are fused according to the similarity and a first rule to obtain the first target image; If the fault information does not exist in the overall image, the local image to be fused and the plurality of sub-overall images are replaced and masked according to the similarity and a second rule to obtain the first target image.

[0007] Optionally, fusing the to-be-fused local image and the plurality of sub-whole images according to the similarity in accordance with a first rule to obtain the first target image includes: Marking the sub-whole images among the plurality of sub-whole images, the sub-whole images having a similarity with the partial image to be fused higher than a first preset value, as sub-whole images to be fused; The sub-image to be fused and the local image to be fused are sequentially spliced ​​in descending order of similarity to obtain the first target image.

[0008] Optionally, replacing and masking the to-be-fused local image and the plurality of sub-whole images according to the similarity in accordance with a second rule to obtain the first target image includes: replacing the sub-whole images among the plurality of sub-whole images, the similarity of which with the partial image to be fused is higher than a second preset value; Masking the sub-whole images whose similarity is lower than a third preset value to obtain a plurality of image blocks, wherein the second preset value is greater than the third preset value; The plurality of image blocks are spliced ​​together through an overlap-add operation to obtain the first target image.

[0009] Optionally, the step of segmenting and fusing the overall sound in the corresponding overall signal according to the local sound in the corresponding local signal to obtain the first target sound includes: Acquire the number of local sounds containing the fault information in the local signal, and mark the local sounds containing the fault information as local sounds to be fused; Classifying at least one of the local sounds to be fused according to frequency, pitch, and loudness to obtain M abnormal local data sets, where M is a natural number greater than 0; Dividing the overall sound in the overall signal into a plurality of sub-overall sounds according to the number, and dividing the plurality of sub-overall sounds into N abnormal overall data sets according to the frequency, the pitch and the loudness, where N is a natural number greater than 0; If M is equal to N, the sub-whole sound in the abnormal whole data set and the local sound to be fused in the abnormal local data set are fused according to a third rule to obtain the first target sound; If M is less than N, the sub-overall sound in the abnormal overall data set and the local sound to be fused in the abnormal local data set are fused according to a fourth rule to obtain the first target sound.

[0010] Optionally, fusing the sub-whole sound in the abnormal whole data set and the local sound to be fused in the abnormal local data set according to a third rule to obtain the first target sound includes: Obtaining a matching degree between the sub-whole sound and the local sound to be fused in the same time period among the multiple time periods; The corresponding sub-whole sound with a matching degree higher than a fourth preset value and the local sound to be fused are superimposed and the effect is increased to obtain a first sound to be fused; The corresponding sub-whole sound whose matching degree is lower than the fourth preset value and the local sound to be fused are superimposed and the effect is weakened to obtain a second sound to be fused; The plurality of first sounds to be fused and the plurality of second sounds to be fused are fused in the order of the plurality of time periods to obtain the first target sound.

[0011] Optionally, fusing the sub-whole sound in the abnormal whole data set and the local sound to be fused in the abnormal local data set according to a fourth rule to obtain the first target sound includes: Obtaining a matching degree between the sub-whole sound and the local sound to be fused in the same time period among the multiple time periods; The corresponding sub-whole sound with a matching degree higher than a fifth preset value and the local sound to be fused are superimposed and the effect is increased to obtain a third sound to be fused; Marking the corresponding sub-whole sound whose matching degree is lower than the fifth preset value as the fourth sound to be fused; At least one of the third sounds to be fused and at least one of the fourth sounds to be fused are fused in the order of the multiple time periods to obtain the first target sound.

[0012] Optionally, the step of inputting the first target image, the second target image, the first target sound, and the second target sound into a trained fault detection model to obtain fault data of the current cable includes: Preprocessing the first target image and the first target sound to obtain fault features; Preprocessing the second target image and the second target sound to obtain normal features; The fault detection model is controlled to correct and adjust the fault feature according to the normal feature to output fault data.

[0013] Optionally, the fault detection model is a Yogi-mLSTM network model, and the fault data includes the fault point location, fault type and fault cause corresponding to the cable.

[0014] An embodiment of the present application further provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, the steps of any of the above-mentioned monitoring methods are implemented.

[0015] The cable fault detection method provided in the embodiment of the present application processes the local images, local sounds, overall images and overall sounds obtained in multiple time periods to obtain the fault data of the current cable, and comprehensively determines whether the current cable has defects from the perspectives of the whole and the part, the sound and the image, avoiding the situation where a single parameter is abnormal and thus leading to misjudgment, which helps to reduce the false alarm rate of cable fault detection, thereby improving the comprehensive understanding of the cable line status. In addition, the present application obtains a new first target image by segmenting and fusing the overall image and / or the local image with fault information, and obtains a new first target sound by segmenting and fusing the overall sound and / or the local sound with fault information, so as to carry out detailed identification of potential faults in the cable, strengthen the data with fault information, weaken the data of other information, thereby improving the accuracy of the fault information, and enabling the fault detection model to quickly and accurately capture the fault information, greatly reducing the time cost and energy investment of troubleshooting, and realizing efficient identification of fault data. The overall signal and / or local signal without fault information is integrated in the order of time periods to obtain a second target image and a second target sound. The first target sound and the first target image with fault information are supplemented and adjusted by the second target image and the second target sound, thereby avoiding missed detection and false detection of fault information and further improving the accuracy of obtaining fault data. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 A schematic diagram of a flow chart of a cable fault detection method provided in an embodiment of the present application; Figure 2 for Figure 1 A schematic diagram of the process of segmenting and fusing the overall image according to the corresponding local images in the detection method shown; Figure 3 for Figure 1 A schematic diagram of the process of segmenting and fusing the overall sound according to the corresponding local sound in the detection method shown; Figure 4 for Figure 3 A schematic diagram of a process of obtaining a first target sound by fusing according to a third rule in the detection method shown; Figure 5 for Figure 3 A schematic diagram of a flow chart of obtaining a first target sound by fusing according to a fourth rule in the detection method shown. DETAILED DESCRIPTION

[0017] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, specific embodiments of the present invention will be described in detail below.

[0018] See also Figure 1 , Figure 1 A schematic diagram of a method for detecting a cable fault provided in an embodiment of the present application. The present application provides a method for detecting a cable fault, and the detection method includes the following process: 110. Acquire multiple local signals and multiple overall signals of the cable in multiple time periods, each local signal includes multiple local images and multiple local sounds of the cable, and each overall signal includes an overall image and an overall sound of the cable, wherein the local signals and overall signals acquired in the same time period correspond one to one.

[0019] The local signal and the overall signal are acquired in a time period at intervals, wherein the local signal and the overall signal acquired in the same time period correspond to each other. The local signal and the overall signal may be acquired multiple times in a time period, or may be acquired only once. The specific setting may be based on actual conditions, and no specific limitation is made here.

[0020] The local signal includes multiple local images and multiple local sounds. The local images can be provided by multiple local image acquisition units set along the cable distribution route, and the local sounds can be provided by multiple local sound acquisition units set along the cable distribution route, so as to obtain multiple local images and multiple local sounds at the same time when acquiring the local signal.

[0021] In some embodiments, the multiple local image acquisition units and the multiple local sound acquisition units can be a group of local acquisition units consisting of a local image acquisition unit and b local sound acquisition units, and the multiple local acquisition units are arranged in sequence along the cable distribution route, wherein the specific values ​​of a and b can be set according to actual conditions and are not specifically limited here.

[0022] In some embodiments, multiple local image acquisition units and multiple local sound acquisition units may also be independently and randomly set along the cable distribution route. No specific restrictions are made here. It only needs to ensure that the setting position is consistent with the cable distribution route.

[0023] The overall signal includes the overall image and overall sound of the cable, wherein the overall image can be provided by an overall image acquisition unit arranged along the cable distribution route, and the overall sound can be provided by an overall sound acquisition unit arranged along the cable distribution route. Similarly, the overall image acquisition unit and the overall sound acquisition unit can be arranged in combination or separately, and no specific limitation is made here.

[0024] It should be noted that since the cable is long and distributed over a wide area, the overall signal effect may be poor. Therefore, in some embodiments, the present application divides the cable into regions according to the actual situation of the cable, that is, the cable is divided into multiple regions, and local signals and overall signals are obtained in each region respectively. Therefore, the corresponding overall image obtained is the overall image corresponding to each region, thereby ensuring the effectiveness of the overall signal and the accuracy of the judgment.

[0025] It can be understood that the local signal can also include the local discharge position, discharge amplitude, discharge number, discharge time, fault position and fault time, and the overall signal can also include the corresponding overall discharge position, discharge amplitude, discharge number, discharge time, fault position and fault time.

[0026] 120. Determine whether there is fault information in multiple local signals and multiple overall signals.

[0027] Determine whether there is fault information based on abnormal sound or image. Abnormal sound may be caused by discontinuity, mutation, or ambiguity of the sound. Abnormal image may be caused by cable damage, corrosion, or breakage identified based on the image content.

[0028] 130. If at least one local image in any local signal or at least one overall image in any overall signal contains fault information, the overall image in the corresponding overall signal is segmented and fused according to the local images in the corresponding local signals to obtain a first target image, and other local signals and / or overall signals that do not have fault information are integrated one by one according to the order of multiple time periods to obtain a second target image.

[0029] It is understandable that as long as there is fault information in the local image or the overall image, the local signal and the overall signal are processed. If there is no fault information in both the local image and the overall image, in order to avoid missed detection or false detection, the present application integrates other local signals and / or overall signals without fault information in a one-to-one correspondence according to the order of multiple time periods to obtain a second target image. The first target image with fault information can be supplemented and adjusted by integrating in the order of multiple time periods to improve the accuracy of obtaining fault data.

[0030] The details of segmenting and fusing the overall image in the corresponding overall signal according to the local image in the corresponding local signal to obtain the first target image can be found in Figure 2 , Figure 2 for Figure 1 Schematic diagram of the process of segmenting and fusing the overall image according to the corresponding local images in the detection method shown.

[0031] 131. Obtain the number of local images with fault information in the local signal, and mark the local images with fault information as local images to be fused.

[0032] It is understandable that within the same time period, some of the multiple local images in a local signal may have faults, while other local images may not have fault information. Therefore, it is necessary to mark the local images with fault information to facilitate subsequent enhanced processing of the local images with fault information, thereby saving time in subsequent capture of fault information.

[0033] 132. Divide the overall image in the overall signal into a number of sub-overall images according to the number, and obtain the similarity between the sub-overall images and each local image to be fused.

[0034] Among them, the number of sub-whole images is positively correlated with the number of local images with fault information. That is, the more local images with fault information, the more sub-whole images are cut. It can be understood that when there is more fault information, the more the whole image is cut, the more refined the feature comparison can be, thereby ensuring the accuracy of the final fault data. On the contrary, when there is less fault information, the less the whole image is cut, the faster the part related to the fault information can be located, the interference of irrelevant information can be reduced, and the recognition efficiency can be improved.

[0035] 133. Determine whether the overall image corresponding to the partial image has fault information.

[0036] The overall image corresponding to the partial image can be understood as the partial image and the overall image acquired at the same time.

[0037] 134. If there is fault information in the overall image, the local image to be fused and the plurality of sub-overall images are fused according to the similarity and the first rule to obtain a first target image.

[0038] If the overall image contains fault information, it means that the shooting content of the overall image can assist the local image in obtaining fault data. Therefore, the sub-overall image and the local image to be fused are fused according to the similarity, which can further enhance the content related to the fault information and thus improve the speed of obtaining fault data.

[0039] Exemplarily, a sub-whole image among several sub-whole images whose similarity with the local image to be fused is higher than a first preset value is marked as a sub-whole image to be fused; the sub-whole image to be fused and the local image to be fused are sequentially spliced ​​in descending order of similarity to obtain a first target image. Among them, splicing in descending order of similarity can fuse the sub-whole images to be fused with the local images to be fused with high similarity, thereby ensuring the maximum enhancement of fault information, improving the accuracy of obtaining fault information, and avoiding the situation where the sub-whole image to be fused is spliced ​​with the local image to be fused with weaker similarity, resulting in unclear fault information. In addition, splicing in descending order can also place prominent fault information in the front, thereby enabling the fault detection model to quickly capture the fault information, thereby speeding up the fault detection speed.

[0040] The first preset value may be set and updated according to historical data, experimental data or empirical data, and is not specifically limited here.

[0041] 135. If there is no fault information in the overall image, the fused local image and several sub-overall images are replaced and masked according to the similarity and the second rule to obtain the first target image.

[0042] If there is no fault information in the overall image, it means that the field of view of the overall image is too large to display the fault information in the overall image, or the fault information is too small to display in the overall image. Therefore, the overall image needs to be adjusted to find content related to the fault information.

[0043] Exemplarily, a sub-whole image with a similarity higher than a second preset value to the local image to be fused among several sub-whole images is replaced; a sub-whole image with a similarity lower than a third preset value is masked to obtain a plurality of image blocks, wherein the second preset value is greater than the third preset value; and the plurality of image blocks are spliced ​​by overlap-addition operation to obtain a first target image. It can be understood that the pixels and the magnification ratio of the local image to be fused are both higher than those of the sub-whole image, so by replacing the sub-whole image with a similarity higher than the second preset value, the clarity and detail of the fused first target image can be improved, thereby facilitating the subsequent extraction of fault information. In addition, the sub-whole image with a similarity lower than the third preset value can be understood as having basically no fault information, and thus can be defined as an interference item for extracting fault information. Therefore, by masking the sub-whole image with a similarity lower than the third preset value, the interference information of the first target image can be reduced, thereby speeding up the speed at which the fault detection model acquires fault data according to the first target image, and also ensuring the accuracy of acquiring fault data.

[0044] The second preset value and the third preset value may be set and updated according to historical data, experimental data or empirical data, and no specific limitation is made here.

[0045] 140. If at least one local sound in any local signal or at least one overall sound in any overall signal contains fault information, the overall sound in the corresponding overall signal is segmented and fused according to the local sounds in the corresponding local signals to obtain a first target sound, and other local signals and overall signals that do not have fault information are processed to obtain a second target sound.

[0046] It is understandable that as long as there is fault information in the local sound or the overall sound, the local signal and the overall signal are processed; if there is no fault information in the local sound and the overall sound, in order to avoid missed detection or false detection, the present application integrates other local signals and / or overall signals without fault information in a one-to-one correspondence according to the order of multiple time periods to obtain the second target sound. The first target sound with fault information can be supplemented and adjusted by integrating in the order of multiple time periods to improve the accuracy of obtaining fault data.

[0047] The specific process of segmenting and fusing the overall sound in the corresponding overall signal according to the local sound in the corresponding local signal to obtain the first target sound can be referred to in Figure 3 , Figure 3 for Figure 1 The schematic diagram of the process of segmenting and fusing the overall sound according to the corresponding local sound in the detection method shown, the specific process is as follows: 141. Obtain the number of local sounds with fault information in the local signal, and mark the local sounds with fault information as local sounds to be fused.

[0048] 142. Classify at least one local sound to be fused according to frequency, pitch, and loudness to obtain M abnormal local data sets, where M is a natural number greater than 0.

[0049] Exemplarily, the frequency, pitch and loudness are weighted according to historical data, experimental data and empirical data to obtain a weighting function. If there are multiple local sounds to be fused, the frequency, pitch and loudness of each local sound to be fused are input into the weighting function to obtain multiple abnormal local data, and the multiple abnormal data are classified according to the preset division range to obtain M abnormal local data sets. By dividing the local sounds to be fused according to frequency, pitch and loudness, the local sounds to be fused with high correlation can be placed in one abnormal local data set, which is convenient for the subsequent screening and extraction of fault information and also improves the accuracy of fault information extraction.

[0050] 143. The overall sound in the overall signal is divided into a number of sub-overall sounds according to the number, and the sub-overall sounds are divided into N abnormal overall data sets according to the frequency, pitch and loudness, where N is a natural number greater than 0.

[0051] Among them, the number of sub-overall sounds is positively correlated with the number of local sounds with fault information. That is, the more local sounds with fault information, the more sub-overall sounds are cut. It can be understood that when there is more fault information, the more the overall sound is cut, the more refined the feature comparison can be, thereby ensuring the accuracy of the final fault data. On the contrary, when there is less fault information, the less the overall sound is cut, the faster the part related to the fault information can be located, and the interference of irrelevant information can be reduced, thereby improving the recognition efficiency.

[0052] Dividing several sub-whole sounds into N abnormal whole data sets according to frequency, pitch and loudness can be understood as inputting the frequency, pitch and loudness of each sub-whole sound into a weighting function to obtain several abnormal whole data, and classifying several abnormal whole data according to a preset division range to obtain N abnormal whole data sets. Dividing several sub-whole sounds according to frequency, pitch and loudness can put several sub-whole sounds with high correlation in one abnormal whole data set, which is convenient for subsequent screening and extraction of fault information, and further improves the accuracy of fault information extraction.

[0053] 144. If M is equal to N, the sub-overall sound in the abnormal overall data set and the local sound to be fused in the abnormal local data set are fused according to the third rule to obtain the first target sound.

[0054] If M is equal to N, that is, the number of abnormal local data sets is the same as the number of abnormal overall data sets, that is, the division and acquisition of abnormal local data sets are consistent with the division and acquisition of abnormal overall data sets, so they can be directly fused. Among them, the specific method of fusion according to the third rule can be referred to in Figure 4 , Figure 4 for Figure 3 The schematic diagram of the process of obtaining the first target sound by fusing according to the third rule in the detection method shown, the specific process is as follows: 1441. Obtain the matching degree between the sub-overall sound and the local sound to be fused in the same time period within multiple time periods.

[0055] The sub-overall sounds and the local sounds to be fused are matched one by one according to the same time period, and the matching degree between the corresponding sub-overall sounds and the local sounds to be fused is obtained.

[0056] 1442. Superimpose the corresponding sub-whole sound with a matching degree higher than a fourth preset value and the local sound to be fused and increase the effect to obtain a first sound to be fused.

[0057] Superimposing the sub-overall sound with a high degree of matching and the local sound to be fused to amplify the effect can make the fault information more obvious, thereby enabling the fault detection model to extract fault data faster and more accurately.

[0058] The fourth preset value may be set and updated according to historical data, experimental data or empirical data, and is not specifically limited here.

[0059] 1443. Superimpose the corresponding sub-whole sound with a matching degree lower than a fourth preset value and the local sound to be fused and weaken the effect to obtain a second sound to be fused.

[0060] Superimposing the sub-overall sound with low matching degree and the local sound to be fused and then weakening the effect can weaken the influence of interference information, thereby making the effective fault information display more obvious and further improving the accuracy of extracting fault information.

[0061] 1444. Fuse the plurality of first sounds to be fused and the plurality of second sounds to be fused in the order of a plurality of time periods to obtain a first target sound.

[0062] Fusion in the order of time periods can further clarify the timeline of fault occurrence, and thus know the time point, duration, and frequency of fault change, etc.

[0063] 145. If M is less than N, the sub-overall sound in the abnormal overall data set and the local sound to be fused in the abnormal local data set are fused according to the fourth rule to obtain the first target sound.

[0064] If M is less than N, that is, the number of abnormal local data sets is less than the number of abnormal overall data sets, that is to say, there is fault information in the overall sound, but the corresponding fault information is not obtained in the local sound. Therefore, it is necessary to perform fusion according to the fourth rule. For specific methods, see Figure 5 , Figure 5 for Figure 3 The schematic diagram of the process of obtaining the first target sound by fusing according to the fourth rule in the detection method shown, the specific process is as follows: 1451. Obtain the matching degree between the sub-overall sound and the local sound to be fused within the same time period within multiple time periods.

[0065] The specific content is shown in the above step 1441 and will not be repeated here.

[0066] 1452. Superimpose the corresponding sub-whole sound and the local sound to be fused whose matching degree is higher than the fifth preset value and increase the effect to obtain a third sound to be fused.

[0067] Superimposing the sub-overall sounds with high matching degree and the local sounds to be fused and increasing the effect can make the fault information clearer and more specific, thereby enabling the fault detection model to extract fault data faster and more accurately.

[0068] The fifth preset value may be set and updated according to historical data, experimental data or empirical data, and is not specifically limited here.

[0069] 1453. Mark the corresponding sub-whole sound whose matching degree is lower than the fifth preset value as the fourth sound to be fused.

[0070] In order to avoid missing the sub-whole sound containing fault information, all the sub-whole sounds with low matching degree are marked as the fourth sound to be fused to ensure the integrity of the fault information.

[0071] 1454. Fuse at least one third sound to be fused and at least one fourth sound to be fused in a sequence of multiple time periods to obtain a first target sound.

[0072] Fusion in the order of time periods can further clarify the timeline of fault occurrence, and thus know the time point, duration, and frequency of fault change, etc.

[0073] 150. Input the first target image, the second target image, the first target sound, and the second target sound into a trained fault detection model to obtain fault data of the current cable.

[0074] Exemplarily, the first target image and the first target sound are preprocessed to obtain fault features, the second target image and the second target sound are preprocessed to obtain normal features, and the fault detection model is controlled to correct and adjust the fault features according to the normal features to output fault data. The fault features are corrected and adjusted according to the normal features of the undetected fault information, thereby avoiding the situation where missed detection or misdetection is caused by errors in the fault information detected in the early stage, and further improving the accuracy and comprehensiveness of the fault data.

[0075] In addition, the fault detection model monitors the local and overall cable in real time. When the fault detection model predicts a fault risk, it automatically generates and issues a warning signal.

[0076] In some embodiments, the sample data used by the fault detection model is not fixed. The first target image, the second target image, the first target sound, and the second target sound in the sample data are input into the fault detection model for prediction, and the currently collected first target image, the second target image, the first target sound, and the second target sound are continuously written into the sample data to achieve self-update of the sample data, ensure the richness of the sample data, and gradually improve the accuracy of the prediction.

[0077] In some embodiments, the fault detection model is a Yogi-mLSTM network model, and the fault data includes the fault point location, fault category and fault cause corresponding to the cable.

[0078] Among them, the Yogi-mLSTM network model has the advantages of simple structure, few parameters, faster training and prediction speed, and higher recognition accuracy. The structure of mLSTM (multiplicative LSTM) increases the connection between the hidden state of mRNN and each threshold unit in LSTM. The goal is to use the hidden state output of mRNN at the previous moment as the weight of the input at this moment to update the hidden state, and combine the large delay and information flow control of LSTM. In order to speed up the convergence of the mLSTM model, Yogi optimization is introduced. Yogi is a variant of Adam, inheriting all the advantages of the Adam algorithm and solving the convergence problem when Adam estimates explode.

[0079] An embodiment of the present application further provides a readable storage medium, which stores computer executable instructions. When the computer executable instructions are read and executed by a processor, the readable storage medium is controlled to implement the cable fault detection method in the above embodiment.

[0080] In the several embodiments provided in the present application, it should be understood that the disclosed cable fault detection method and readable storage medium can also be implemented in other ways. The above described embodiments are merely illustrative.

[0081] If the function is implemented in the form of a software function module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the methods of each embodiment of the present invention. The aforementioned readable storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, and other media that can store program codes.

[0082] Although the present invention is disclosed as above, the present invention is not limited thereto. Any person skilled in the art can make various changes and modifications without departing from the spirit and scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the scope defined by the claims.

Claims

1. A method for detecting a cable fault, characterized in that: The detection method comprises: Acquire multiple local signals and multiple overall signals of the cable in multiple time periods, each of the local signals includes multiple local images and multiple local sounds of the cable, and each of the overall signals includes an overall image and an overall sound of the cable, wherein the local signals and the overall signals acquired in the same time period correspond one to one; Determining whether there is fault information in the plurality of local signals and the plurality of overall signals; If at least one local image in any of the local signals or at least one overall image in any of the overall signals has the fault information, the overall image in the corresponding overall signal is segmented and fused according to the local images in the corresponding local signals to obtain a first target image, and the other local signals and / or the overall signals that do not have the fault information are integrated one by one according to the order of the multiple time periods to obtain a second target image; If at least one local sound in any of the local signals or at least one overall sound in any of the overall signals has the fault information, the overall sound in the corresponding overall signal is segmented and merged according to the local sounds in the corresponding local signals to obtain a first target sound, and the other local signals and the overall signal that do not have the fault information are processed to obtain a second target sound; The first target image, the second target image, the first target sound and the second target sound are input into a trained fault detection model to obtain fault data of the current cable.

2. The detection method according to claim 1, characterized in that: The step of segmenting and fusing the overall image in the corresponding overall signal according to the local image in the corresponding local signal to obtain the first target image comprises: Acquire the number of local images containing the fault information in the local signal, and mark the local images containing the fault information as local images to be fused; Dividing the overall image in the overall signal into a plurality of sub-overall images according to the number, and obtaining similarities between the plurality of sub-overall images and each of the local images to be fused; Determining whether the overall image corresponding to the partial image contains the fault information; If the fault information exists in the overall image, the local image to be fused and the plurality of sub-overall images are fused according to the similarity and a first rule to obtain the first target image; If the fault information does not exist in the overall image, the local image to be fused and the plurality of sub-overall images are replaced and masked according to the similarity and a second rule to obtain the first target image.

3. The detection method according to claim 2, characterized in that: The fusing the partial image to be fused and the plurality of sub-whole images according to the similarity and the first rule to obtain the first target image comprises: Marking the sub-whole images among the plurality of sub-whole images, the sub-whole images having a similarity with the partial image to be fused higher than a first preset value, as sub-whole images to be fused; The sub-image to be fused and the local image to be fused are sequentially spliced ​​in descending order of similarity to obtain the first target image.

4. The detection method according to claim 2, characterized in that: The replacing and masking the partial image to be fused and the plurality of sub-whole images according to the similarity and the second rule to obtain the first target image comprises: replacing the sub-whole images among the plurality of sub-whole images, the similarity of which with the partial image to be fused is higher than a second preset value; Masking the sub-whole images whose similarity is lower than a third preset value to obtain a plurality of image blocks, wherein the second preset value is greater than the third preset value; The plurality of image blocks are spliced ​​together through an overlap-add operation to obtain the first target image.

5. The detection method according to any one of claims 1 to 4, characterized in that: The step of segmenting and fusing the overall sound in the corresponding overall signal according to the local sound in the corresponding local signal to obtain the first target sound comprises: Acquire the number of local sounds containing the fault information in the local signal, and mark the local sounds containing the fault information as local sounds to be fused; Classifying at least one of the local sounds to be fused according to frequency, pitch, and loudness to obtain M abnormal local data sets, where M is a natural number greater than 0; Dividing the overall sound in the overall signal into a plurality of sub-overall sounds according to the number, and dividing the plurality of sub-overall sounds into N abnormal overall data sets according to the frequency, the pitch and the loudness, where N is a natural number greater than 0; If M is equal to N, the sub-whole sound in the abnormal whole data set and the local sound to be fused in the abnormal local data set are fused according to a third rule to obtain the first target sound; If M is less than N, the sub-overall sound in the abnormal overall data set and the local sound to be fused in the abnormal local data set are fused according to a fourth rule to obtain the first target sound.

6. The detection method according to claim 5, characterized in that: The fusing the sub-whole sound in the abnormal whole data set and the local sound to be fused in the abnormal local data set according to a third rule to obtain the first target sound comprises: Obtaining a matching degree between the sub-whole sound and the local sound to be fused in the same time period among the multiple time periods; The corresponding sub-whole sound with a matching degree higher than a fourth preset value and the local sound to be fused are superimposed and the effect is increased to obtain a first sound to be fused; The corresponding sub-whole sound whose matching degree is lower than the fourth preset value and the local sound to be fused are superimposed and the effect is weakened to obtain a second sound to be fused; The plurality of first sounds to be fused and the plurality of second sounds to be fused are fused in the order of the plurality of time periods to obtain the first target sound.

7. The detection method according to claim 5, characterized in that: The fusing the sub-whole sound in the abnormal whole data set and the local sound to be fused in the abnormal local data set according to a fourth rule to obtain the first target sound comprises: Obtaining a matching degree between the sub-whole sound and the local sound to be fused in the same time period among the multiple time periods; The corresponding sub-whole sound with a matching degree higher than a fifth preset value and the local sound to be fused are superimposed and the effect is increased to obtain a third sound to be fused; Marking the corresponding sub-whole sound whose matching degree is lower than the fifth preset value as the fourth sound to be fused; At least one of the third sounds to be fused and at least one of the fourth sounds to be fused are fused in the order of the multiple time periods to obtain the first target sound.

8. The detection method according to any one of claims 1 to 4, characterized in that: The step of inputting the first target image, the second target image, the first target sound, and the second target sound into a trained fault detection model to obtain fault data of the current cable includes: Preprocessing the first target image and the first target sound to obtain fault features; Preprocessing the second target image and the second target sound to obtain normal features; The fault detection model is controlled to correct and adjust the fault feature according to the normal feature to output fault data.

9. The detection method according to any one of claims 1 to 4, characterized in that: The fault detection model is a Yogi-mLSTM network model, and the fault data includes the fault point location, fault type and fault cause corresponding to the cable.

10. A readable storage medium, characterized in that: The readable storage medium stores a program or instruction, and when the program or instruction is executed by a processor, the steps of the detection method according to any one of claims 1 to 9 are implemented.