Cable deflection test system based on multi-source data

Abstract

A cable deflection test system based on multi-source data relates to the technical field of cable deflection test, and sets cable test conditions, collects various test data of a cable in the test process and after the test is finished, and performs threshold analysis on a surface image and a numerical time sequence of various test indexes; key test indexes of cables with different functional characteristics in different application scenes are obtained; performing correlation analysis and trend analysis on key test indexes in the historical deflection test result, obtaining correlation among the key test indexes and a fluctuation change coefficient of each key test index, comparing the correlation and the fluctuation change coefficient with a pre-estimation result output by the deflection test pre-estimation model, and constructing an abnormal feature library according to a comparison result; and then fusion analysis is performed on the cables in the to-be-fused analysis state, and the cables in the to-be-fused analysis state are divided into qualified cables or unqualified cables according to the fusion analysis result, so that the accuracy and comprehensiveness of cable quality judgment are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of cable flexure testing, and specifically to a cable flexure testing system based on multi-source data. Background Art

[0002] In the production, quality inspection, and practical application of cables, the cable flexure performance is one of the key indicators for evaluating the quality and reliability of cables. Traditional cable flexure testing methods usually rely only on a single data source. For example, only by monitoring the resistance change of the cable to judge its flexure performance, this method has great limitations. Single data cannot comprehensively reflect the actual situation of the cable during the flexure process, and it is easy to lead to misjudgment of the cable performance. For example, some subtle damages to the internal structure may not immediately cause an obvious change in resistance, but will have a serious impact on the long-term use of the cable. In addition, traditional testing methods often cannot obtain various performance parameters of the cable in complex flexure working conditions in real time and accurately, and it is difficult to meet the requirements of modern cable manufacturing for high-quality and high-precision testing. Summary of the Invention

[0003] In order to solve the above technical problems, the purpose of the present invention is to provide a cable flexure testing system based on multi-source data, including a cloud. The cloud is communicatively connected to a multi-source data acquisition module, a data detection module, a data detection module, a data processing module, a historical data analysis module, a model construction module, and a fusion analysis module;

[0004] The multi-source data acquisition module is used to set cable test conditions and collect various test data of the cable during and after the test. The various test data include surface images and numerical time series of various types of test indicators;

[0005] The data detection module is used to perform threshold analysis on the surface images and numerical time series of various types of test indicators, and mark the cable as a non-conforming cable or a state to be fusion analyzed;

[0006] The data processing module is used to obtain the key test indicators of cables with different functional characteristics in different application scenarios;

[0007] The historical data analysis module is used to perform correlation analysis and trend analysis on the key test indicators in the historical flexure test results, obtain the correlations between the key test indicators and the fluctuation change coefficients of the key test indicators, compare the correlations between the key test indicators and the fluctuation change coefficients of the key test indicators with the estimated results output by the flexure test prediction model, and construct an abnormal feature library according to the comparison results;

[0008] The model construction module is used to construct a flexure test prediction model;

[0009] The fusion analysis module is used to perform fusion analysis on the cables in the state to be fused and analyzed, and classify the cables in the state to be fused and analyzed into qualified cables or unqualified cables according to the fusion analysis results.

[0010] Furthermore, the multi-source data acquisition module sets the cable test conditions, and the process of acquiring various test data of the cable during and after the test includes:

[0011] Obtain the functional characteristics and application scenarios of the cable, construct a test condition comparison table, where the test condition comparison table includes the test conditions of cables with different functional characteristics under different application scenario conditions, match the functional characteristics and application scenarios of the cable with the test condition comparison table to obtain the flexure test conditions of the cable (including bending radius, flexure frequency, flexure stroke, and ambient temperature, etc.), set the flexure test content of the cable according to the flexure test conditions, and test the cable according to the flexure test content to acquire various test data during and after the test.

[0012] Furthermore, the process of the data detection module performing threshold analysis on the surface image includes:

[0013] Perform image preprocessing operations on the surface image of the cable. Improving the image quality through image preprocessing operations is a well-known image processing method for those skilled in the art. In the embodiments of the invention, noise in the image is removed by the Gaussian filtering method, and then the image data is equalized to improve the contrast within the image. Subsequently, the surface image is grayscale processed to convert the surface image into a grayscale image. The grayscale process is usually achieved by averaging the weights of the color channels. Among them, denoising equalization processing, semantic segmentation network, and grayscale processing are well-known technical means for those skilled in the art, and the specific steps will not be elaborated. Use the gray-level co-occurrence matrix technology to obtain the gray-level co-occurrence matrix within the neighborhood of each pixel point in the grayscale image. Specific details are as follows: Quantize the gray-level of the surface image data into discrete gray-levels, divide the gray-value range into several levels. For example, an 8-bit image can be divided into 16, 32, or 64 levels, and define the parameters required for the gray-level co-occurrence matrix, including distance (d) and direction (θ), which are determined according to actual application requirements, and the specific steps will not be elaborated. Obtain the texture feature values within the neighborhood of each pixel point according to the gray-level co-occurrence matrix. The texture feature values include contrast, energy, and entropy, etc. Preset the texture feature threshold interval of the cable, compare the texture feature value within the neighborhood of each pixel point with the texture feature threshold interval. If there is a texture feature value within the neighborhood of a pixel point that is not within the texture feature threshold interval, then mark the cable as an unqualified cable;

[0014] If the texture feature values within the neighborhood of each pixel point are all within the texture feature threshold interval, then perform threshold analysis on the numerical time series of each type of test index.

[0015] Further, the process of the data detection module performing threshold analysis on the numerical time series of various types of test indicators includes:

[0016] Preset the standard threshold intervals corresponding to various types of test indicators. The various types of test indicators include electrical parameters such as the insulation resistance, withstand voltage strength, and conductor resistance of the cable, and dimensional parameters such as the outer diameter, insulation thickness, and sheath thickness of the cable before and after flexure. Extract the numerical time series corresponding to various types of test indicators from the multi-source test data, and compare the numerical time series corresponding to various types of test indicators with the corresponding standard threshold intervals. If there is a numerical time series corresponding to a type of test indicator that is not within the corresponding standard threshold interval, mark the type of test indicator as a non-conforming test indicator and mark the cable as a non-conforming cable;

[0017] If the numerical time series corresponding to all types of test indicators are within the corresponding standard threshold intervals, mark the cable as in the state of pending fusion analysis.

[0018] Further, the process of the data processing module obtaining the key test indicators of cables with different functional characteristics in different application scenarios includes:

[0019] Obtain the historical flexure test results of several cables with different functional characteristics in different application scenarios, perform statistical analysis on the historical flexure test results, and obtain the probabilities that various types of test indicators are marked as non-conforming test indicators for cables with different functional characteristics in different application scenarios;

[0020] Preset a probability threshold, compare the probabilities that various types of test indicators are marked as non-conforming test indicators in different functional characteristics and application scenarios with the probability threshold, obtain the types of test indicators with probabilities greater than or equal to the probability threshold in different functional characteristics and application scenarios, and mark the types of test indicators as key test indicators.

[0021] Further, the process of the historical data analysis module performing correlation analysis and trend analysis on the key test indicators in the historical flexure test results to obtain the correlations between the key test indicators and the fluctuation change coefficients of the key test indicators includes:

[0022] Screen out the test data of non-conforming cables during and after the test when the cable is marked as a non-conforming cable from the historical flexure test results, extract the numerical time series of each key test indicator from the test data, perform correlation analysis on the numerical time series of each key test indicator to obtain the correlations between the key test indicators, and at the same time perform trend analysis on the numerical time series of each key test indicator to obtain the fluctuation change coefficients of the key test indicators, and associate the functional characteristics and application scenarios of the non-conforming cable with the correlations between the key test indicators and the fluctuation change coefficients of the key test indicators;

[0023] Among them, the calculation process for obtaining the correlation between each key test index is as follows:

[0024]

[0025] Among them, q(ij) represents the correlation between key test index i and key test index j, D it represents the value of key test index i at the t-th moment, D jt represents the value of key test index j at the t-th moment, DiA represents the average value of the numerical values of key test index i, DjA represents the average value of the numerical values of key test index j, and n represents the total number of moments;

[0026] The calculation process for obtaining the fluctuation change coefficient of each key test index is as follows:

[0027]

[0028] Among them, z i represents the average fluctuation coefficient of key test index i, x it represents the value of key test index i at the t-th moment. The above formulas are all calculated by removing the dimension and taking their numerical values. The formula is obtained by collecting a large amount of data for software simulation to get a formula closest to the real situation. The preset parameters and preset thresholds in the formula are set by those skilled in the art according to the actual situation or obtained by simulating a large amount of data.

[0029] Furthermore, the historical data analysis module compares the correlation between each key test index and the fluctuation change coefficient of each key test index with the estimated results output by the flexure test prediction model. The process of constructing the abnormal feature library according to the comparison results includes:

[0030] Obtain the flexure test prediction model, and according to the flexure test prediction model, output the estimated correlation between each key test index of the qualified cable and the estimated fluctuation change coefficient of the key test index under the functional characteristics and application scenario conditions. Perform a significance analysis of the difference between the correlation between each key test index and the estimated correlation, and the fluctuation change coefficient between each key test index and the estimated fluctuation change coefficient, to obtain the significance level of the correlation between each key test index and the estimated correlation, and the significance level of the fluctuation change coefficient between each key test index and the estimated fluctuation change coefficient;

[0031] Preset a significance level threshold. If the significance level of the correlation between key test indicators and the estimated correlation is greater than the preset significance level threshold, it is considered that the correlation between the key test indicators deviates from the expected dynamic trajectory and has a significant difference from the estimated correlation. The change in the correlation between the key test indicators may affect the state of the cable. Then, mark the correlation between the key test indicators as the first key feature. If the significance level of the fluctuation change coefficient of the key test indicator and the estimated fluctuation change coefficient is greater than the preset significance level threshold, it is considered that the fluctuation change coefficient of the key test indicator deviates from the expected dynamic trajectory and has a significant difference from the estimated fluctuation change coefficient. The change in the fluctuation change coefficient of the key test indicator may affect the state of the cable. Mark the fluctuation change coefficient of the key test indicator as the second key feature. Construct a key feature set of unqualified cables based on the first key feature and the second key feature, and associate the functional characteristics and application scenarios of the unqualified cables with the key feature set;

[0032] Construct an abnormal feature library and store the key feature sets of several unqualified cables in the abnormal feature library.

[0033] Furthermore, the process of the model construction module constructing the flexure test prediction model includes:

[0034] Construct a flexure test prediction model based on deep learning. Obtain the correlations between various key test indicators and the fluctuation change coefficients of various key test indicators of cables with different functional characteristics in different application scenarios according to historical flexure test results. Use the correlations between the various key test indicators and the fluctuation change coefficients of the various key test indicators as the training set and the test set. Input the training set into the flexure test prediction model for training until the loss function is trained stably, and save the model parameters. Test the flexure test prediction model through the test set until it meets the preset requirements, and output the flexure test prediction model;

[0035] Constructing a flexure test prediction model based on deep learning is a complex process that involves multiple steps such as model selection, training, validation, and testing. The following is a detailed supplementary description of this process:

[0036] In this embodiment, a convolutional neural network (CNN) suitable for time series analysis is selected as the deep learning architecture. After determining the model architecture, the mean squared error loss function (MES) is selected as the optimization objective. Subsequently, the prepared training set is input into the selected deep learning model to start training. During the training process, the weights are continuously updated through the backpropagation algorithm, causing the loss function to gradually decrease until a stable state is reached. During this period, techniques such as early stopping are used to avoid overfitting. In addition to the basic training process, various parameters of the model are tuned through grid search. The parameters include the learning rate, batch size, regularization coefficient, etc.

[0037] When the model training is completed and the parameters are adjusted, the final evaluation is carried out through the test set to obtain the evaluation results of the model. The evaluation results include classification metrics such as accuracy, recall rate, F1 score, etc. According to the evaluation results on the test set, it is judged whether the model meets the expected standards. If the requirements are met, the model parameters are saved and preparation for deployment is made; if not, it is necessary to return to a previous stage to re-examine issues such as data quality, model structure, or training strategy.

[0038] Furthermore, the fusion analysis module performs fusion analysis on the cables in the to-be-fused analysis state. The process of classifying the cables in the to-be-fused analysis state into qualified cables or unqualified cables according to the fusion analysis results includes:

[0039] Obtain the correlation between each key test index of the cable in the to-be-fused analysis state and the fluctuation change coefficient of each key test index. Retrieve and compare in the abnormal feature library according to the correlation between each key test index and the fluctuation change coefficient of each key test index, and obtain the similarity between the correlation between each key test index and the fluctuation change coefficient of each key test index and each feature data set in the abnormal feature library. If there is a corresponding feature data set in the abnormal feature library with a similarity greater than the preset standard, mark the cable as an unqualified cable; if there is no corresponding feature data set in the abnormal feature library with a similarity greater than the preset standard, mark the cable as a qualified cable.

[0040] Compared with the prior art, the beneficial effects of the present invention are:

[0041] 1. Comprehensive and accurate data collection and analysis: The multi-source data collection module can construct a test condition comparison table according to the functional characteristics and application scenarios of the cable, accurately match and set appropriate flexure test conditions, and then comprehensively collect various data during and after the test, including surface images and numerical time series of various types of test indicators. Compared with traditional test systems, it breaks through the singularity and limitations of test condition settings, making the collected data more in line with the actual usage conditions of the cable, providing a more comprehensive and targeted basis for subsequent analysis, and greatly improving the accuracy and reliability of test results.

[0042] 2. Efficient preliminary screening mechanism: The data detection module can quickly screen out obviously unqualified cables by performing image preprocessing, grayscale processing on the surface image, extracting texture feature values based on the gray-level co-occurrence matrix technology for threshold analysis, and at the same time performing threshold comparison on the numerical time series of various types of test indicators. This mechanism effectively reduces the workload of subsequent complex analysis, avoids unnecessary in-depth analysis of a large number of unqualified cables, improves the operating efficiency of the entire test system, and reduces the test cost.

[0043] 3. Precise extraction of key indicators: The data processing module statistically analyzes a large number of historical flexure test results and screens out key test indicators based on the probability that various types of test indicators are marked as unqualified. This method enables subsequent analysis to focus on the key factors that truly affect the quality and performance of the cable, avoids blind analysis among numerous test indicators, improves the efficiency and pertinence of analysis, and can more quickly and accurately discover potential problems existing in the cable.

[0044] 4. Intelligent construction of abnormal feature library: The historical data analysis module performs correlation analysis and trend analysis on the key test indicators of unqualified cables, obtains the correlation and fluctuation change coefficients among key indicators, and compares them with the estimated results output by the flexure test prediction model to construct an abnormal feature library. The establishment of the abnormal feature library provides a powerful reference basis for subsequent cable testing. By comparing the data of the cable to be tested with the features in the library, potential abnormal situations can be quickly identified, greatly improving the detection ability and early warning level for cable quality problems.

[0045] 5. Precise fusion analysis and judgment: The fusion analysis module accurately judges whether the cable is qualified by comparing the similarity between the key test indicator data of the cable to be fused and analyzed and the data in the abnormal feature library. This fusion analysis method comprehensively considers various factors, can discover some subtle defects and potential problems that are difficult to detect by relying on single data alone, improves the accuracy and comprehensiveness of cable quality judgment, effectively avoids misjudgment and missed judgment, and ensures the quality of cables put into use. Description of the Drawings

[0046] Figure 1 This is the schematic diagram of a cable flexure test system based on multi-source data according to an embodiment of the present application. Specific implementation manners

[0047] Next, in combination with the accompanying drawings in the embodiments of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described. Obviously, the described embodiments are part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts shall fall within the protection scope of the present application.

[0048] As Figure 1 shown, a cable flexure test system based on multi-source data includes a cloud, and the cloud is communicatively connected to a multi-source data acquisition module, a data detection module, a data detection module, a data processing module, a historical data analysis module, a model construction module, and a fusion analysis module;

[0049] The multi-source data acquisition module is used to set cable test conditions and acquire various test data during and after the cable test process. The various test data include surface images and numerical time series of various types of test indicators;

[0050] The data detection module is used to perform threshold analysis on the surface images and numerical time series of various types of test indicators, and mark the cable as a non-conforming cable or a cable to be fusion-analyzed;

[0051] The data processing module is used to obtain key test indicators of cables with different functional characteristics in different application scenarios;

[0052] The historical data analysis module is used to perform correlation analysis and trend analysis on the key test indicators in the historical flexure test results, obtain the correlations between the key test indicators and the fluctuation change coefficients of the key test indicators, compare the correlations between the key test indicators and the fluctuation change coefficients of the key test indicators with the estimated results output by the flexure test prediction model, and construct an abnormal feature library according to the comparison results;

[0053] The model construction module is used to construct a flexure test prediction model;

[0054] The fusion analysis module is used to perform fusion analysis on the cables to be fusion-analyzed, and classify the cables to be fusion-analyzed as conforming cables or non-conforming cables according to the fusion analysis results;

[0055] It should be further noted that in the specific implementation process, the process of the multi-source data acquisition module setting cable test conditions and acquiring various test data during and after the cable test process includes:

[0056] Obtain the functional characteristics and application scenarios of the cable, construct a test condition comparison table, where the test condition comparison table includes the test conditions for cables with different functional characteristics under different application scenario conditions. Match the functional characteristics and application scenarios of the cable with the test condition comparison table to obtain the flexure test conditions of the cable (including bending radius, flexure frequency, flexure stroke, ambient temperature, etc.). Set the flexure test content of the cable according to the flexure test conditions, and test the cable according to the flexure test content, and collect various test data during and after the test.

[0057] It should be further noted that, in the specific implementation process, the process of the data detection module performing threshold analysis on the surface image includes:

[0058] Perform image preprocessing operations on the surface image of the cable. Improving the image quality through image preprocessing operations is a well-known image processing method for those skilled in the art. In the embodiment of the invention, Gaussian filtering method is used to remove the noise in the image, and then the image data is equalized to improve the contrast in the image. Subsequently, the surface image is grayscale processed to convert the surface image into a grayscale image. The grayscale process is usually achieved by averaging the weights of the color channels. Among them, denoising equalization processing, semantic segmentation network, and grayscale processing are well-known technical means for those skilled in the art, and the specific steps will not be elaborated. Use the gray-level co-occurrence matrix technology to obtain the gray-level co-occurrence matrix in the neighborhood of each pixel point in the grayscale image. Specific details are as follows: Quantize the gray-level of the surface image data into discrete gray-levels, divide the gray-value range into several levels, for example, an 8-bit image can be divided into 16, 32, 64 levels, and define the parameters required for the gray-level co-occurrence matrix, including distance (d) and direction (θ), which are determined according to actual application requirements, and the specific steps will not be elaborated. Obtain the texture feature values in the neighborhood of each pixel point according to the gray-level co-occurrence matrix. The texture feature values include contrast, energy, entropy, etc. Preset the texture feature threshold interval of the cable, and compare the texture feature values in the neighborhood of each pixel point with the texture feature threshold interval. If there are texture feature values in the neighborhood of a pixel point that are not within the texture feature threshold interval, then mark the cable as a non-conforming cable;

[0059] If the texture feature values in the neighborhood of each pixel point are all within the texture feature threshold interval, then perform threshold analysis on the numerical time series of each type of test index.

[0060] It should be further noted that, in the specific implementation process, the process of the data detection module performing threshold analysis on the numerical time series of each type of test index includes:

[0061] Preset the standard threshold intervals corresponding to various types of test indicators. The various types of test indicators include electrical parameters such as the insulation resistance, withstand voltage strength, and conductor resistance of the cable, and dimensional parameters such as the outer diameter, insulation thickness, and sheath thickness of the cable before and after flexure. Extract the numerical time series corresponding to each type of test indicator from the multi-source test data, and compare the numerical time series corresponding to each type of test indicator with the corresponding standard threshold interval. If there is a numerical time series corresponding to a type of test indicator that is not within the corresponding standard threshold interval, mark the type of test indicator as a non-conforming test indicator and mark the cable as a non-conforming cable;

[0062] If the numerical time series corresponding to each type of test indicator are all within the corresponding standard threshold intervals, mark the cable as in a state of pending fusion analysis.

[0063] It should be further noted that in the specific implementation process, the process by which the data processing module obtains the key test indicators of cables with different functional characteristics in different application scenarios includes:

[0064] Obtain the historical flexure test results of several cables with different functional characteristics in different application scenarios, conduct statistical analysis on the historical flexure test results, and obtain the probabilities that various types of test indicators of cables with different functional characteristics in different application scenarios are marked as non-conforming test indicators;

[0065] Preset a probability threshold, compare the probabilities that various types of test indicators in different functional characteristics and application scenarios are marked as non-conforming test indicators with the probability threshold, obtain the types of test indicators with probabilities greater than or equal to the probability threshold in different functional characteristics and application scenarios, and mark the types of test indicators as key test indicators.

[0066] It should be further noted that in the specific implementation process, the process by which the historical data analysis module conducts correlation analysis and trend analysis on the key test indicators in the historical flexure test results to obtain the correlations between the key test indicators and the fluctuation change coefficients of the key test indicators includes:

[0067] Screen out the test data of non-conforming cables during and after the test when the cable is marked as a non-conforming cable from the historical flexure test results, extract the numerical time series of each key test indicator from the test data, conduct correlation analysis on the numerical time series of each key test indicator to obtain the correlations between the key test indicators, and at the same time conduct trend analysis on the numerical time series of each key test indicator to obtain the fluctuation change coefficients of the key test indicators, and associate the functional characteristics and application scenarios of the non-conforming cable with the correlations between the key test indicators and the fluctuation change coefficients of the key test indicators;

[0068] Among them, the calculation process for obtaining the correlation between each key test index is as follows:

[0069]

[0070] Among them, q(ij) represents the correlation between key test index i and key test index j, D it represents the value of key test index i at the t-th moment, D jt represents the value of key test index j at the t-th moment, DiA represents the average value of the numerical values of key test index i, DjA represents the average value of the numerical values of key test index j, and n represents the total number of moments;

[0071] The calculation process for obtaining the fluctuation change coefficient of each key test index is as follows:

[0072]

[0073] Among them, z i represents the average fluctuation coefficient of key test index i, x it represents the value of key test index i at the t-th moment. The above formulas are all calculated by removing the dimension and taking their numerical values. The formulas are obtained by collecting a large amount of data for software simulation to get a formula closest to the actual situation. The preset parameters and preset thresholds in the formulas are set by those skilled in the art according to the actual situation or obtained by simulating a large amount of data.

[0074] It should be further noted that in the specific implementation process, the historical data analysis module compares the correlation between each key test index and the fluctuation change coefficient of each key test index with the estimated results output by the flexure test prediction model. The process of constructing the abnormal feature library according to the comparison results includes:

[0075] Obtain the flexure test prediction model, and according to the flexure test prediction model, output the estimated correlation between each key test index of a qualified cable and the estimated fluctuation change coefficient of the key test index under the functional characteristics and application scenario conditions. Perform a significance analysis of the difference between the correlation between each key test index and the estimated correlation, and the fluctuation change coefficient of each key test index and the estimated fluctuation change coefficient, and obtain the significance level of the correlation between each key test index and the estimated correlation, and the significance level of the fluctuation change coefficient of each key test index and the estimated fluctuation change coefficient;

[0076] Preset a significance level threshold. If the significance level of the correlation between key test indicators and the estimated correlation is greater than the preset significance level threshold, it is considered that the correlation between the key test indicators deviates from the expected dynamic trajectory and has a significant difference from the estimated correlation. The change in the correlation between the key test indicators may affect the state of the cable. Then, mark the correlation between the key test indicators as the first key feature. If the significance level of the fluctuation change coefficient of the key test indicator and the estimated fluctuation change coefficient is greater than the preset significance level threshold, it is considered that the fluctuation change coefficient of the key test indicator deviates from the expected dynamic trajectory and has a significant difference from the estimated fluctuation change coefficient. The change in the fluctuation change coefficient of the key test indicator may affect the state of the cable. Mark the fluctuation change coefficient of the key test indicator as the second key feature. Construct a key feature set of unqualified cables based on the first key feature and the second key feature, and associate the functional characteristics and application scenarios of the unqualified cables with the key feature set;

[0077] Construct an abnormal feature library and store the key feature sets of several unqualified cables in the abnormal feature library.

[0078] It should be further noted that in the specific implementation process, the process of the model construction module constructing the flexure test prediction model includes:

[0079] Construct a flexure test prediction model based on deep learning. Obtain the correlation between each key test indicator and the fluctuation change coefficient of each key test indicator of cables with different functional characteristics in different application scenarios according to historical flexure test results. Use the correlation between each key test indicator and the fluctuation change coefficient of each key test indicator as the training set and the test set. Input the training set into the flexure test prediction model for training until the loss function is trained stably, and save the model parameters. Test the flexure test prediction model through the test set until it meets the preset requirements, and output the flexure test prediction model;

[0080] Constructing a flexure test prediction model based on deep learning is a complex process, which involves multiple steps such as model selection, training, validation, and testing. The following is a detailed supplementary description of this process:

[0081] In this embodiment, a convolutional neural network (CNN) suitable for time series analysis is selected as the deep learning architecture. After determining the model architecture, the mean squared error loss function (MES) is selected as the optimization objective. Subsequently, the prepared training set is input into the selected deep learning model to start training. During the training process, the weights are continuously updated through the backpropagation algorithm, causing the loss function to gradually decrease until a stable state is reached. During this period, techniques such as early stopping are used to avoid overfitting. In addition to the basic training process, various parameters of the model are tuned through grid search. The parameters include the learning rate, batch size, regularization coefficient, etc.

[0082] When the model training is completed and the parameters are adjusted, the final evaluation is carried out through the test set to obtain the evaluation results of the model. The evaluation results include classification metrics such as accuracy, recall rate, F1 score, etc. According to the evaluation results on the test set, it is judged whether the model meets the expected standards. If the requirements are met, the model parameters are saved and prepared for deployment; if not, it is necessary to return to a previous stage to re-examine issues such as data quality, model structure, or training strategy.

[0083] It should be further noted that in the specific implementation process, the process of the fusion analysis module performing fusion analysis on the cables in the to-be-fused analysis state and classifying the cables in the to-be-fused analysis state into qualified cables or unqualified cables according to the fusion analysis results includes:

[0084] Obtain the correlation between each key test index of the cable in the to-be-fused analysis state and the fluctuation change coefficient of each key test index. Retrieve and compare in the abnormal feature library according to the correlation between each key test index and the fluctuation change coefficient of each key test index, and obtain the similarity between the correlation between each key test index and the fluctuation change coefficient of each key test index and each feature data set in the abnormal feature library. If there is a feature data set in the abnormal feature library with a corresponding similarity greater than the preset standard, mark the cable as an unqualified cable; if there is no feature data set in the abnormal feature library with a corresponding similarity greater than the preset standard, mark the cable as a qualified cable.

[0085] The above embodiments are only used to illustrate the technical method of the present invention and not to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical method of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A cable bending test system based on multi-source data, characterized in that: It includes a cloud, wherein the cloud is communicatively connected to a multi-source data acquisition module, a data detection module, a data processing module, a historical data analysis module, a model building module and a fusion analysis module; The multi-source data acquisition module is used to set the cable test conditions and collect various test data of the cable during and after the test, wherein the various test data include surface images and numerical time series of various types of test indicators; The data detection module is used to perform threshold analysis on the surface image and the numerical time series of each type of test index, marking the cable as an unqualified cable or a state to be fused and analyzed; The data processing module is used to obtain key test indicators of cables with different functional characteristics in different application scenarios; The historical data analysis module is used to perform correlation analysis and trend analysis on key test indicators in historical flexural test results, obtain the correlation between key test indicators and the fluctuation coefficient of each key test indicator, compare the correlation between key test indicators and the fluctuation coefficient of each key test indicator with the estimated results output by the flexural test prediction model, and build an abnormal feature library based on the comparison results; The model building module is used to build a flexural test prediction model; The fusion analysis module is used to perform fusion analysis on the cables in the fusion analysis state, and classify the cables in the fusion analysis state into qualified cables or unqualified cables according to the fusion analysis results.

2. A cable bending test system based on multi-source data according to claim 1, characterized in that: The multi-source data acquisition module sets the cable test conditions and collects various test data of the cable during and after the test, including: Acquire the functional characteristics and application scenarios of the cable, and construct a test condition comparison table, wherein the test condition comparison table includes the test conditions of cables with different functional characteristics under different application scenarios, match the functional characteristics and application scenarios of the cable with the test condition comparison table for test conditions, acquire the flexural test conditions of the cable, set the flexural test content of the cable according to the flexural test conditions, test the cable according to the flexural test content, and collect various test data during and after the test.

3. A cable bending test system based on multi-source data according to claim 2, characterized in that: The process of threshold analysis of surface images by the data detection module includes: Perform image preprocessing on the surface image of the cable, then perform grayscale processing on the surface image, convert the surface image into a grayscale image, use the grayscale co-occurrence matrix technology to obtain the grayscale co-occurrence matrix in the neighborhood of each pixel in the grayscale image, obtain the texture feature value in the neighborhood of each pixel according to the grayscale co-occurrence matrix, preset the texture feature threshold interval of the cable, compare the texture feature value in the neighborhood of each pixel with the texture feature threshold interval, and if there is a texture feature value in the neighborhood of a pixel that is not within the texture feature threshold interval, the cable is marked as an unqualified cable; If the texture feature values ​​in the neighborhood of each pixel point are all within the texture feature threshold interval, threshold analysis is performed on the numerical time series of each type of test index.

4. A cable bending test system based on multi-source data according to claim 3, characterized in that: The process of the data detection module performing threshold analysis on the numerical time series of each type of test indicator includes: Preset the standard threshold interval corresponding to each type of test indicator, extract the numerical time series sequence corresponding to each type of test indicator in the multi-source test data, compare the numerical time series sequence corresponding to each type of test indicator with the corresponding standard threshold interval, if there is a type of test indicator The numerical time series sequence is not located in the corresponding standard threshold interval, then the type of test indicator is marked as an unqualified test indicator, and the cable is marked as an unqualified cable; If the numerical timing sequences corresponding to each type of test indicator are all within the corresponding standard threshold range, the cable is marked as a state to be fused and analyzed.

5. A cable bending test system based on multi-source data according to claim 4, characterized in that: The process of the data processing module obtaining key test indicators of cables with different functional characteristics in different application scenarios includes: Obtain historical flex test results of cables with different functional characteristics in different application scenarios, perform statistical analysis on the historical flex test results, and obtain the probability of each type of test indicator being marked as an unqualified test indicator for cables with different functional characteristics in different application scenarios; Preset a probability threshold, compare the probability of each type of test indicator under different functional characteristics and application scenarios being marked as an unqualified test indicator with the probability threshold, obtain the type of test indicators under different functional characteristics and application scenarios whose probability is greater than or equal to the probability threshold, and mark the type of test indicators as key test indicators.

6. A cable bending test system based on multi-source data according to claim 5, characterized in that: The historical data analysis module performs correlation analysis and trend analysis on key test indicators in historical flexure test results. The process of obtaining the correlation between key test indicators and the fluctuation coefficient of each key test indicator includes: From the historical flex test results, various test data of the unqualified cable during and after the test when the cable is marked as an unqualified cable are screened out, the numerical time series sequence of each key test indicator is extracted from the various test data, and the numerical time series sequence of each key test indicator is correlated, and the correlation between each key test indicator is obtained. At the same time, trend analysis is performed on the numerical time series sequence of each key test indicator to obtain the fluctuation variation coefficient of each key test indicator, and the functional characteristics and application scenarios of the unqualified cable are associated with the correlation between each key test indicator and the fluctuation variation coefficient of each key test indicator.

7. A cable bending test system based on multi-source data according to claim 6, characterized in that: The historical data analysis module compares the correlation between each key test indicator and the fluctuation coefficient of each key test indicator with the estimated result output by the flexural test prediction model. The process of constructing an abnormal feature library based on the comparison results includes: Obtain a flexural test prediction model, output the estimated correlations between the key test indicators of qualified cables and the estimated fluctuation coefficients of the key test indicators under the functional characteristics and application scenarios of the flexural test prediction model, perform a significance analysis on the correlations between the key test indicators and the estimated correlations, and the fluctuation coefficients of the key test indicators and the estimated fluctuation coefficients, and obtain the significance levels of the correlations between the key test indicators and the estimated correlations, and the fluctuation coefficients of the key test indicators and the estimated fluctuation coefficients; A significance level threshold is preset. If the significance level of the correlation between the key test indicators and the estimated correlation is greater than the preset significance level threshold, the correlation between the key test indicators is marked as the first key feature. If the significance level of the fluctuation variation coefficient of the key test indicators and the estimated fluctuation variation coefficient is greater than the preset significance level threshold, the fluctuation variation coefficient of the key test indicators is marked as the second key feature. A key feature set of unqualified cables is constructed based on the first key feature and the second key feature, and the functional characteristics and application scenarios of the unqualified cables are associated with the key feature set. An abnormal feature library is constructed, and key feature sets of several unqualified cables are stored in the abnormal feature library.

8. A cable bending test system based on multi-source data according to claim 7, characterized in that: The process of building a flexural test prediction model in the model building module includes: A flexural test prediction model is constructed based on deep learning. The correlation between key test indicators of cables with different functional characteristics in different application scenarios and the fluctuation coefficient of each key test indicator are obtained according to historical flexural test results. The correlation between the key test indicators and the fluctuation coefficient of each key test indicator are used as training sets and test sets. The training set is input into the flexural test prediction model for training until the loss function training is stable, and the model parameters are saved. The flexural test prediction model is tested by the test set until it meets the preset requirements, and the flexural test prediction model is output.

9. The cable bending test system based on multi-source data according to claim 8, characterized in that: The fusion analysis module performs fusion analysis on the cables in the fusion analysis state, and classifies the cables in the fusion analysis state into qualified cables or unqualified cables according to the fusion analysis results, including: Obtain the correlation between each key test indicator of the cable to be integrated and analyzed and the fluctuation variation coefficient of each key test indicator, perform search and comparison in the abnormal feature library according to the correlation between each key test indicator and the fluctuation variation coefficient of each key test indicator, obtain the similarity between the correlation between each key test indicator and the fluctuation variation coefficient of each key test indicator and each feature data set in the abnormal feature library, if there is a corresponding feature data set in the abnormal feature library with a similarity greater than the preset standard, the cable is marked as an unqualified cable, if there is no corresponding feature data set in the abnormal feature library with a similarity greater than the preset standard, the cable is marked as a qualified cable.

Images