Carbon fiber composite core wire real-time health monitoring method based on deep learning

Abstract

The invention discloses a carbon fiber composite core wire real-time health monitoring method based on deep learning, and the method comprises the steps: collecting real-time monitoring data of a first carbon fiber composite core wire, and carrying out the preprocessing of the real-time monitoring data, and obtaining first health data; constructing an initial health monitoring model, and training and testing the real-time health monitoring model by using a training set and a test set to obtain a real-time health monitoring model; deploying the real-time health monitoring model into an embedded system to obtain real-time sampling data, and performing health rating on the carbon fiber composite core wire according to the real-time sampling data to obtain a health grade corresponding to the carbon fiber composite core wire; and maintaining the second carbon fiber composite core wire. According to the method, accurate real-time monitoring and detection of the carbon fiber composite core wire are achieved, normal work of the wire is guaranteed, and the power supply requirement is met.

Description

Technical Field

[0001] The present invention relates to the technical field of carbon fiber composite core conductor performance research, and in particular to a real-time health monitoring method for carbon fiber composite core conductors based on deep learning. Background Art

[0002] Carbon fiber composite core conductors play an important role in power transmission systems, but long-term operation and external environmental factors may cause damage or wear to the conductors, so a real-time health monitoring system is needed to detect potential problems and provide early warnings. Deep learning technology has advantages in processing complex data and feature extraction, and can be applied to real-time health monitoring of carbon fiber composite core conductors.

[0003] The Chinese patent with publication number CN111445445A discloses a carbon fiber composite core wire damage detection method based on semi-supervised anomaly detection, which only uses non-destructive carbon fiber composite core wire X-ray image training model to detect wire damage. However, it is not comprehensive enough to detect wire damage only from image data. It is also necessary to comprehensively evaluate the health status of the carbon fiber composite core wire from the surface physical data of the wire and the internal electrical data of the wire to avoid omissions caused by evaluation only from image data. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a real-time health monitoring method for carbon fiber composite core conductors based on deep learning. A real-time health monitoring model is constructed through deep learning neural networks and long short-term memory networks. The multi-dimensional data of the carbon fiber composite core conductors are identified and predicted respectively, and the health status of the carbon fiber composite core conductors is comprehensively evaluated. Different maintenance measures are adopted according to different health levels, thereby realizing accurate real-time monitoring and detection of the carbon fiber composite core conductors, ensuring the normal operation of the conductors, and meeting power supply requirements.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions, a real-time health monitoring method of a carbon fiber composite core conductor based on deep learning, comprising:

[0007] Collecting real-time monitoring data of a first carbon fiber composite core conductor, and preprocessing the real-time monitoring data to obtain first health data, wherein the first carbon fiber composite core conductor is a carbon fiber composite core conductor with normal real-time health detection;

[0008] Constructing an initial health monitoring model, dividing the first health data into a training set and a test set, and using the training set and the test set to train and test the real-time health monitoring model to obtain a real-time health monitoring model;

[0009] The real-time health monitoring model is deployed in an embedded system, a second carbon fiber composite core conductor is sampled through the embedded system to obtain real-time sampling data, a health rating is performed on the carbon fiber composite core conductor according to the real-time sampling data to obtain a health grade corresponding to the carbon fiber composite core conductor, wherein the second carbon fiber composite core conductor is a carbon fiber composite core conductor to be tested in a working state;

[0010] The second carbon fiber composite core conductor is maintained according to the maintenance measure table of the carbon fiber composite core conductor and the health level.

[0011] As a preferred solution of the real-time health monitoring method of a carbon fiber composite core conductor based on deep learning described in the present invention, wherein: the preprocessing includes collecting real-time monitoring data of a first carbon fiber composite core conductor, and the real-time monitoring data includes conductor physical data, conductor electrical data, and conductor image data;

[0012] The real-time monitoring data is preprocessed, and the conductor physical data, conductor electrical data, and conductor image data are cleaned, standardized, and fused to obtain first health data; the conductor physical data includes conductor temperature, conductor strain, and conductor vibration frequency; the conductor electrical data includes conductor current value and conductor voltage value; the conductor image data includes conductor ultrasonic image and conductor magnetic particle detection image.

[0013] As a preferred solution of the real-time health monitoring method of carbon fiber composite core conductor based on deep learning described in the present invention, wherein: the preprocessing also includes performing outlier processing and missing value processing on the conductor physical data and the conductor electrical data to obtain the first conductor physical data and the first conductor electrical data, performing image denoising and image enhancement processing on the conductor image data to obtain the first conductor image data;

[0014] Performing feature extraction and Z-Score normalization processing on the first conductor physical data and the first conductor electrical data respectively to obtain conductor physical features and conductor electrical features, and merging the conductor physical features and the conductor electrical features to obtain a conductor feature matrix, wherein the conductor feature matrix includes numerical values ​​corresponding to conductor temperature, conductor strain, conductor vibration frequency, conductor current value, and conductor voltage value, and timestamps corresponding to acquisition of the numerical values;

[0015] The first health data includes first guide wire image data and a guide wire feature matrix.

[0016] As a preferred solution of the real-time health monitoring method of carbon fiber composite core conductor based on deep learning described in the present invention, wherein: the construction of the initial health monitoring model includes: constructing the initial health monitoring model based on a deep learning neural network and a long short-term memory network, dividing the first health data into a training set and a test set according to a ratio of 7:3; iteratively training the initial health monitoring model through the training set, configuring the optimizer, loss function and hyperparameters until the iteration stop condition is reached, and then obtaining the trained health monitoring model; using the test set to test the trained health monitoring model, and evaluating the model through the mAP value. If the model evaluation meets the requirements, the training and verification are completed, that is, the real-time health monitoring model is obtained. If the model evaluation does not meet the requirements, the hyperparameters are adjusted and the model is iteratively trained again;

[0017] The initial health monitoring model includes an input layer, a network combination layer, a mixing layer, an attention mechanism, and an output layer; the input layer inputs the first health data; the network combination layer includes a DNN and an LSTM, the DNN inputs the first wire image data in the first health data, and outputs the wire defect type and the probability of the detection result; the LSTM inputs the wire feature matrix in the first health data, and outputs the time-varying function of each type of data in the first health data; the mixing layer adopts a Softmax function; the attention mechanism adopts a Multi-head Attention function; and the output layer adopts a Quantileregression function.

[0018] As a preferred solution of the real-time health monitoring method of carbon fiber composite core conductor based on deep learning described in the present invention, wherein: the construction of the initial health monitoring model also includes the calculation formula of the loss function is:

[0019]

[0020] Among them, N is the number of data categories in the conductor feature matrix, T is the total time stamp step corresponding to the collection of various data values ​​in the conductor feature matrix, and y i,t is the probability of the actual value corresponding to the t-th timestamp of the i-th type of data, is the predicted probability of the model for the tth timestamp of the i-th category data;

[0021] The iteration stopping condition is that the loss function converges or reaches the total number of iterations.

[0022] As a preferred solution of the real-time health monitoring method of carbon fiber composite core conductor based on deep learning described in the present invention, wherein: the construction of the initial health monitoring model also includes evaluating the model by mAP value, and the calculation method of mAP value is:

[0023]

[0024] Where b is the number of data categories in the wire feature matrix, R(t) is the recall rate of the current timestamp, R(t+1) is the recall rate of the next timestamp, and precision(t,t+1) is the accuracy of the current timestamp and the next timestamp;

[0025] When the mAP value is not less than the preset threshold, the iterative training and testing are terminated and the real-time health monitoring model is output.

[0026] As a preferred solution of the real-time health monitoring method of carbon fiber composite core conductor based on deep learning described in the present invention, wherein: the real-time sampling data includes time-varying functions of various types of data, conductor defect types and probabilities corresponding to the conductor defect types;

[0027] A plurality of time-varying coefficient thresholds are preset to correspond to various types of data one by one, and the calculation formula for performing health rating on the carbon fiber composite core conductor according to the real-time sampling data is:

[0028]

[0029] Wherein, W is the health grade of the second carbon fiber composite core conductor, W 1 1 The first health level, W 1 2 The second health level, The third health level, The fourth health level, q a is the time-varying coefficient weight of the time-varying function of the a-th type of data, n is the total number of data categories of the second carbon fiber composite core conductor, V is the conductor image defect type of the second carbon fiber composite core conductor, when V is 0, the conductor image defect type is normal, when V is 1, the conductor image defect type is abnormal, α a is the time-varying coefficient of the time-varying function of the a-th type of data, σ a is the time-varying coefficient threshold of the time-varying function of the a-th type of data.

[0030] As a preferred solution of the real-time health monitoring method of carbon fiber composite core conductor based on deep learning described in the present invention, wherein: the maintenance measure table includes, when the second carbon fiber composite core conductor is at the first health level, regularly performing a conductor surface inspection on a one-year cycle to detect defects or damage on the surface of the second carbon fiber composite core conductor;

[0031] When the second carbon fiber composite core conductor is at the second health level, regularly inspecting the conductor surface in a quarterly cycle to detect defects or damage on the surface of the second carbon fiber composite core conductor;

[0032] When the second carbon fiber composite core conductor is at the third health level, ultrasonic testing and magnetic particle testing are performed regularly on a monthly basis to detect damage inside the conductor;

[0033] When the second carbon fiber composite core conductor is at the fourth health level, the second carbon fiber composite core conductor is replaced immediately.

[0034] A computer device comprises a memory and a processor, wherein the memory stores a computer program, and is characterized in that when the processor executes the computer program, steps of a real-time health monitoring method for a carbon fiber composite core conductor based on deep learning are implemented.

[0035] A computer-readable storage medium having a computer program stored thereon, characterized in that when the computer program is executed by a processor, the steps of a real-time health monitoring method for a carbon fiber composite core conductor based on deep learning are implemented.

[0036] Beneficial effects of the present invention: The present invention constructs a real-time health monitoring model through deep learning neural networks and long short-term memory networks, identifies and predicts the multi-dimensional data of carbon fiber composite core conductors, comprehensively evaluates the health status of carbon fiber composite core conductors, and adopts different maintenance measures according to different health levels, thereby realizing accurate real-time monitoring and detection of carbon fiber composite core conductors, ensuring the normal operation of the conductors and meeting power supply requirements;

[0037] The present invention collects conductor physical data, electrical data and image data to comprehensively monitor various indicators and state changes of the conductor, and helps to understand the surface and internal health status of the conductor in real time. Different types of data provide multi-dimensional feature information. The conductor physical data reflects the physical condition of the conductor, the electrical data reflects the electrical performance of the conductor, and the image data provides visual information. The comprehensive use of these data can more comprehensively describe the state of the conductor.

[0038] The present invention combines a deep learning neural network with a long short-term memory network to construct an initial health monitoring model, and divides the first health data into a training set and a test set in proportion for iterative training and optimization of the model. Ultimately, a trained and verified real-time health monitoring model is obtained, which can accurately monitor the health status of the wire and make real-time predictions. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0040] Figure 1A schematic flow chart of a method for real-time health monitoring of a carbon fiber composite core conductor based on deep learning provided in accordance with an embodiment of the present invention.

[0041] Figure 2 A schematic diagram of the structure of an initial health monitoring model of a real-time health monitoring method for a carbon fiber composite core conductor based on deep learning provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0042] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0043] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0044] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0045] The present invention is described in detail with reference to schematic diagrams. When describing the embodiments of the present invention, for the sake of convenience, the cross-sectional diagrams showing the device structure will not be partially enlarged according to the general scale, and the schematic diagrams are only examples, which should not limit the scope of protection of the present invention. In addition, in actual production, the three-dimensional dimensions of length, width and depth should be included.

[0046] At the same time, in the description of the present invention, it should be noted that the directions or positional relationships indicated by the terms "upper, lower, inner and outer" are based on the directions or positional relationships shown in the drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore cannot be understood as limiting the present invention. In addition, the terms "first, second or third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0047] In the present invention, unless otherwise clearly specified and limited, the terms "install, connect, connect" should be understood in a broad sense, for example: it can be a fixed connection, a detachable connection or an integral connection; it can also be a mechanical connection, an electrical connection or a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two components. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0048] Example 1

[0049] Reference Figure 1-Figure 2 , which is the first embodiment of the present invention, and provides a real-time health monitoring method for a carbon fiber composite core conductor based on deep learning, comprising:

[0050] S1, collecting real-time monitoring data of a first carbon fiber composite core conductor, and preprocessing the real-time monitoring data to obtain first health data, wherein the first carbon fiber composite core conductor is a carbon fiber composite core conductor with normal real-time health detection;

[0051] S2, constructing an initial health monitoring model, dividing the first health data into a training set and a test set, using the training set and the test set to train and test the real-time health monitoring model to obtain a real-time health monitoring model;

[0052] S3, deploying the real-time health monitoring model into an embedded system, sampling a second carbon fiber composite core conductor through the embedded system to obtain real-time sampling data, and performing a health rating on the carbon fiber composite core conductor according to the real-time sampling data to obtain a health grade corresponding to the carbon fiber composite core conductor, wherein the second carbon fiber composite core conductor is a carbon fiber composite core conductor to be tested in a working state;

[0053] S4, maintaining the second carbon fiber composite core conductor according to the maintenance measure table of carbon fiber composite core conductors and the health level.

[0054] It should be noted that:

[0055] Traditional carbon fiber composite core conductor monitoring methods rely on manual inspection or rule-based algorithms, including semi-supervised anomaly detection carbon fiber composite core conductor damage detection methods, which train models through image data to achieve the identification of carbon fiber composite core conductors.

[0056] However, although image data can provide certain visual information, it cannot provide detailed structural and status information inside the wire. It is impossible to accurately assess problems such as fine cracks and fatigue areas that may exist inside the wire. Although ultrasonic images can provide internal structural information, they still cannot capture small or complex problems, such as small cracks or local fatigue. Further research is needed to ensure accurate assessment of all internal problems. Ultrasonic images require professional equipment and technicians to acquire and interpret, which may increase costs and operational complexity, limiting convenience in practical applications.

[0057] Therefore, in this embodiment, combining multiple data types can comprehensively evaluate the health status of the carbon fiber composite core conductor from different angles, including external visual information, internal structure, and electrical performance, etc., thereby improving the comprehensiveness and accuracy of the evaluation.

[0058] The conductor physical data and conductor electrical data of this embodiment can provide more detailed quantitative information, which is deeper than simple image data, and can achieve more accurate analysis and evaluation of the health status of the carbon fiber composite core conductor.

[0059] The comprehensive use of multiple data types can make up for the limitations of each data, verify each other, improve the reliability of the assessment results, and reduce the risk of misjudgment. Combining multiple data types can also better establish a health assessment model, use deep learning and other technologies to learn and predict the health status of the conductor from the comprehensive data, and further improve the monitoring and early warning capabilities.

[0060] Step S1 includes:

[0061] S11, collecting real-time monitoring data of the first carbon fiber composite core conductor, wherein the real-time monitoring data includes conductor physical data, conductor electrical data, and conductor image data;

[0062] S12, preprocessing the real-time monitoring data, cleaning, standardizing and fusing the conductor physical data, conductor electrical data and conductor image data to obtain first health data.

[0063] It should be noted that:

[0064] The first health data is obtained by using the real-time monitoring data collection and preprocessing method, combining the wire physical data, wire electrical data and wire image data. Such health data integrates various information, including the physical state, electrical performance and appearance characteristics of the wire. Through data cleaning, standardization and data fusion, it can achieve the effect of comprehensive evaluation, data optimization and data fusion, and provide a reliable data source for subsequent model training and testing.

[0065] Combining the conductor's physical, electrical and image (visual information) data can comprehensively assess the conductor's health status, including its internal structure, electrical performance and appearance characteristics, thus improving the comprehensiveness and accuracy of the assessment.

[0066] Through preprocessing and data cleaning, the quality of real-time monitoring data can be optimized, noise and errors can be removed, and the accuracy and reliability of the data can be ensured, which is conducive to subsequent analysis and application.

[0067] Fusing different types of data together can provide richer information and more comprehensive features, helping to discover patterns and regularities hidden in the data and better understand the health of the wires.

[0068] Step S11 includes:

[0069] The conductor physical data includes conductor temperature, conductor strain, and conductor vibration frequency;

[0070] The conductor electrical data includes the conductor current and the conductor voltage;

[0071] The conductor image data includes a conductor ultrasonic image and a conductor magnetic particle detection image.

[0072] It should be noted that:

[0073] This embodiment involves the collection of conductor physical data, conductor electrical data and conductor image data, specifically including conductor temperature, conductor strain, conductor vibration frequency, conductor current, conductor voltage, conductor ultrasonic image and conductor magnetic particle detection image.

[0074] By collecting the conductor's physical data, electrical data, and image data, it is used to comprehensively monitor the conductor's various indicators and status changes, helping to understand the conductor's surface and internal health status in real time.

[0075] Different types of data provide multi-dimensional feature information. The physical data of the conductor reflects the physical condition of the conductor, the electrical data reflects the electrical performance of the conductor, and the image data provides visual information. The comprehensive use of these data can more comprehensively describe the status of the conductor.

[0076] Step S12 includes:

[0077] Performing outlier processing and missing value processing on the conductor physical data and the conductor electrical data to obtain first conductor physical data and first conductor electrical data, and performing image denoising and image enhancement processing on the conductor image data to obtain first conductor image data;

[0078] Performing feature extraction and Z-Score normalization processing on the first conductor physical data and the first conductor electrical data respectively to obtain conductor physical features and conductor electrical features, and merging the conductor physical features and the conductor electrical features to obtain a conductor feature matrix, wherein the conductor feature matrix includes numerical values ​​corresponding to conductor temperature, conductor strain, conductor vibration frequency, conductor current, and conductor voltage, and timestamps corresponding to the acquisition of the numerical values;

[0079] The first health data includes first guide wire image data and a guide wire feature matrix.

[0080] It should be noted that:

[0081] In this embodiment, outlier processing and missing value processing are performed on the conductor physical data and the conductor electrical data by using a regression interpolation method, image denoising is performed on the conductor image data by using a Gaussian blur or median filtering method, and image enhancement is performed on the conductor image data by using histogram equalization to obtain first conductor image data.

[0082] Then, the principal component analysis method is used to reduce the dimension of the first conductor physical data and the first conductor electrical data, respectively, and the main features of the first conductor physical data and the first conductor electrical data are extracted, including the conductor temperature, conductor strain, conductor vibration frequency, conductor current, and conductor voltage. Then, through Z-Score standardization processing, the conductor temperature, conductor strain, conductor vibration frequency, conductor current, and conductor voltage are converted to the same scale according to the normal distribution, and the values ​​are all between 0 and 1.

[0083] It should be noted that, in this embodiment, the first conductor physical data and the first conductor electrical data include data of the entire time period of an acquisition cycle. In this embodiment, all data are evenly divided into T segments at a fixed time interval, and the time step of each segment is equal. The timestamp of each time period is marked, and the number of samples 1 is used as the first dimension, the number of data categories 5 is used as the second dimension, and the time step T is used as the third dimension to form a three-dimensional tensor conductor feature matrix. Specifically, the three-dimensional tensor conductor feature matrix includes a matrix, which is 5 rows and T columns; the number of matrices is the number of carbon fiber composite core conductors monitored, the number of rows is the number of categories of conductor physical data and conductor electrical data collected by the carbon fiber composite core conductor, and the number of columns is the number of timestamp divisions. For example:

[0084]

[0085] The above three-dimensional tensor represents 1 sample (wire), 3 rows (categories), and 3 columns (timestamps). 10 means that the value of the first category of data of the wire at the first timestamp is 10, 13 means that the value of the first category of data of the wire at the second timestamp is 13, and 15 means that the value of the first category of data of the wire at the third timestamp is 15.

[0086] The wire feature matrix is ​​used as the input of LSTM (Long Short-Term Memory), and the first wire image data is used as the input of DNN (Deep Neural Network).

[0087] Step S2 includes:

[0088] S21, constructing an initial health monitoring model based on a deep learning neural network and a long short-term memory network, and dividing the first health data into a training set and a test set in a ratio of 7:3;

[0089] S22, iteratively training the initial health monitoring model through the training set, configuring the optimizer, loss function and hyperparameters until the iteration stop condition is reached, and obtaining the trained health monitoring model;

[0090] S23, use the test set to test the trained health monitoring model, and evaluate the model through the mAP value. If the model evaluation meets the requirements, the training and verification are completed, and the real-time health monitoring model is obtained. If the model evaluation does not meet the requirements, return to step S22 to adjust the hyperparameters and re-train the model iteratively.

[0091] It should be noted that:

[0092] This embodiment combines a deep learning neural network with a long short-term memory network to construct an initial health monitoring model, and divides the first health data into a training set and a test set in a ratio of 7:3 for iterative training and optimization of the model. Finally, a trained and verified real-time health monitoring model is obtained, which can accurately monitor the health status of the wire and make real-time predictions.

[0093] Step S21 includes:

[0094] The initial health monitoring model includes an input layer, a network combination layer, a mixing layer, an attention mechanism, and an output layer;

[0095] The input layer inputs first health data;

[0096] The network combination layer includes a DNN and an LSTM, wherein the DNN inputs the first wire image data in the first healthy data and outputs the wire defect type and the detection result probability; the LSTM inputs the wire feature matrix in the first healthy data and outputs the time-varying function of each type of data in the first healthy data;

[0097] The mixing layer adopts the Softmax function;

[0098] The attention mechanism adopts the Multi-head Attention function;

[0099] The output layer adopts the Quantile regression function.

[0100] It should be noted that:

[0101] like Figure 2 As shown, the initial health monitoring model includes an input layer, a network combination layer, a mixing layer, an attention mechanism, and an output layer;

[0102] The input layer inputs the first health data to provide an original data source for subsequent models;

[0103] The network combination layer includes two parts: DNN and LSTM:

[0104] The DNN part is responsible for processing the first wire image data in the first healthy data, and outputting the wire defect type and the probability of the detection result.

[0105] The LSTM part is responsible for processing the wire feature matrix and outputting the time-varying functions of various types of data, that is, modeling and predicting the changes of data over time.

[0106] The hybrid layer uses a Softmax function to process the multi-classification problem of the network output and convert the output into a probability distribution.

[0107] The Softmax function in this embodiment is:

[0108]

[0109] Where Z = (z 1 ,z 2 ) is the vector composed of the two outputs z1 and z2 of DNN and LSTM, σ(z) l is the output of the Softmax function, K is the dimension of vector Z, and l is the element type of vector Z.

[0110] The attention mechanism adopts the Multi-headAttention function to introduce the attention mechanism into the model, which helps the model focus on important input information and improves the learning and generalization capabilities of the model.

[0111] The output layer uses the Quantile regression function to perform quantile regression, which can better predict the output and provide a confidence interval, which is suitable for handling uncertainties.

[0112] The initial health monitoring model combines the recognition of image data and the time series prediction capabilities of feature sequences, and can comprehensively learn and predict the health information of the wires. The introduction of attention mechanism and quantile regression helps to improve the model effect and output accuracy.

[0113] Step S22 includes:

[0114] The calculation formula of the loss function is:

[0115]

[0116] Among them, N is the number of data categories in the conductor feature matrix, T is the total time stamp step corresponding to the collection of various data values ​​in the conductor feature matrix, and y i,t is the probability of the actual value corresponding to the t-th timestamp of the i-th type of data, is the predicted probability of the model for the tth timestamp of the i-th category data;

[0117] The iteration stopping condition is that the loss function converges or reaches the total number of iterations.

[0118] The optimizer in this embodiment adopts the Adam optimizer, and the number of iterations, learning rate, and total timestamp step size are set as hyperparameters.

[0119] It should be noted that:

[0120] In this embodiment, by defining a reasonable loss function, the model can more accurately predict the probability of various types of data in the wire feature matrix at different time stamps.

[0121] By setting the iteration stop condition, we can effectively prevent the model from overtraining and maintain the generalization ability of the model.

[0122] The convergence of the loss function is used as the iteration stopping condition, which can reduce unnecessary calculations and improve training efficiency.

[0123] Set the total number of iterations to 100. When the number of iterations reaches 100 or the above loss function converges, stop iterative training.

[0124] Step S23 includes:

[0125] The model is evaluated by the mAP value, which is calculated as follows:

[0126]

[0127] Where b is the number of data categories in the wire feature matrix, R(t) is the recall rate of the current timestamp, R(t+1) is the recall rate of the next timestamp, and precision(t,t+1) is the accuracy of the current timestamp and the next timestamp;

[0128] When the mAP value is not less than the preset threshold, the iterative training and testing are terminated and the real-time health monitoring model is output.

[0129] It should be noted that:

[0130] In this embodiment, the following parameters are specified:

[0131] TP: The number of samples correctly classified as positive samples; they are actually positive samples and are also classified as positive samples by the model;

[0132] FP: The number of samples that are misclassified as positive samples; they are actually negative samples, but are classified as positive samples by the model;

[0133] TN: The number of samples correctly classified as negative samples; actually negative samples, but also classified as negative samples by the model;

[0134] FN: The number of samples that are misclassified as negative samples; they are actually positive samples, but are classified as negative samples by the model;

[0135] Precision = TP / (TP+FP), recall = TP / (TP+FN);

[0136] Using the Precision and Recall of each type as the horizontal and vertical axes, finding the AP value is actually equivalent to finding the area enclosed by Precision and Recall. Since the curve area is very jittery, interpolation is used here to smooth the curve area. If the current recall value is R(t), the interpolation value is the maximum Precision value between the next position recall R(t+1). Where t represents the current position. The mAP value is the average of all types of AP values, where b represents the number of data categories in the wire feature matrix.

[0137] A mAP threshold is preset. When the required mAP value is not less than the mAP threshold, the iterative training and optimization are terminated to obtain a real-time health monitoring model.

[0138] Step S3 includes:

[0139] The real-time sampling data includes time-varying functions of various types of data, wire defect types, and probabilities corresponding to the wire defect types;

[0140] A plurality of time-varying coefficient thresholds are preset to correspond to various types of data one by one, and the calculation formula for performing health rating on the carbon fiber composite core conductor according to the real-time sampling data is:

[0141]

[0142] Wherein, W is the health grade of the second carbon fiber composite core conductor, W 1 1 The first health level, W 1 2 The second health level, The third health level, The fourth health level, q a is the time-varying coefficient weight of the time-varying function of the a-th type of data, n is the total number of data categories of the second carbon fiber composite core conductor, V is the conductor image defect type of the second carbon fiber composite core conductor, when V is 0, the conductor image defect type is normal, when V is 1, the conductor image defect type is abnormal, α a is the time-varying coefficient of the time-varying function of the a-th type of data, σ a is the time-varying coefficient threshold of the time-varying function of the a-th type of data.

[0143] It should be noted that:

[0144] In this embodiment, the real-time health monitoring model is deployed in an embedded system, and the second carbon fiber composite core conductor is sampled and monitored in real time through the system, thereby achieving health rating and grading of the carbon fiber composite core conductor.

[0145] First, the trained and verified real-time health monitoring model is deployed to a specially designed embedded system to ensure that the model can run correctly on the hardware and process real-time sampled data;

[0146] Secondly, the embedded system is used to sample the second carbon fiber composite core conductor in real time to obtain instant sensor data. The sampled data are processed and analyzed through the deployed real-time health monitoring model to achieve real-time monitoring of the status of the second carbon fiber composite core conductor and obtain real-time sampled data.

[0147] Finally, based on the real-time sampling data, the health status of the second carbon fiber composite core conductor is rated to obtain the health grade of the corresponding carbon fiber composite core conductor, such as good, average, problematic, etc.

[0148] In this embodiment, the physical and electrical health status of the second carbon fiber composite core conductor is evaluated based on the relationship between the time-varying coefficient of the time-varying function of various types of data of the second carbon fiber composite core conductor and the preset time-varying coefficient threshold of the time-varying function of various types of data, and the conductor image defect type of the second carbon fiber composite core conductor is evaluated by sampling the conductor image defect type of the second carbon fiber composite core conductor in the real-time data. Finally, the health rating of the second carbon fiber composite core conductor is performed based on the physical and electrical health status and the conductor image defect type.

[0149] It should be noted that the time-varying coefficient threshold of the time-varying function of the various types of data is determined according to the time-varying coefficient of the various types of data of the first carbon fiber composite core conductor, the time-varying coefficient threshold of the time-varying function of the various types of data of the second carbon fiber composite core conductor deviates from the time-varying coefficient of the various types of data of the first carbon fiber composite core conductor by ±5%, the physical and electrical health status of the second carbon fiber composite core conductor is specifically the comprehensive sum of the size relationship of the time-varying coefficient threshold of the time-varying function of the various types of data, and the conductor image defect type of the second carbon fiber composite core conductor is directly output by the real-time monitoring model.

[0150] In this embodiment, the health level of the second carbon fiber composite core conductor is described by a first health level, a second health level, a third health level and a fourth health level, wherein:

[0151] The first health level is when the time-varying coefficient of the time-varying function of less than half of the categories of data exceeds its time-varying coefficient threshold, the conductor image defect type is a normal carbon fiber composite core conductor;

[0152] The second health level is when the time-varying coefficient of the time-varying function of less than half of the categories of data exceeds its time-varying coefficient threshold, the wire image defect type is an abnormal carbon fiber composite core wire;

[0153] The third health level is when the time-varying coefficient of the time-varying function of at least half of the categories of data exceeds its time-varying coefficient threshold, the conductor image defect type is a normal carbon fiber composite core conductor;

[0154] The fourth health level is when the time-varying coefficient of the time-varying function of data of no less than half of the categories exceeds its time-varying coefficient threshold, and the wire image defect type is an abnormal carbon fiber composite core wire.

[0155] Step S4 includes:

[0156] The maintenance measures table includes:

[0157] When the second carbon fiber composite core conductor is at the first health level, regularly inspecting the conductor surface on a one-year basis to detect defects or damage on the surface of the second carbon fiber composite core conductor;

[0158] When the second carbon fiber composite core conductor is at the second health level, regularly inspecting the conductor surface in a quarterly cycle to detect defects or damage on the surface of the second carbon fiber composite core conductor;

[0159] When the second carbon fiber composite core conductor is at the third health level, ultrasonic testing and magnetic particle testing are performed regularly on a monthly basis to detect damage inside the conductor;

[0160] When the second carbon fiber composite core conductor is at the fourth health level, the second carbon fiber composite core conductor is replaced immediately.

[0161] It should be noted that:

[0162] According to the health level and maintenance measures table, corresponding maintenance measures are performed on the second carbon fiber composite core conductor to ensure its safety and stability.

[0163] When the second carbon fiber composite core conductor is assessed as having different health levels, corresponding periodic maintenance measures are performed according to the guidance in the maintenance measures table.

[0164] Since the carbon fiber composite core conductors of the first and second health levels are normal in terms of main physical and electrical data, that is, they can work normally, no special measures are needed, only attention needs to be paid to the frequency of inspection and maintenance; while the carbon fiber composite core conductors of the third and fourth health levels have problems not only in the type of conductor defects, but also in physical and electrical data, which have affected normal work, so they need short-term (one month) high-frequency inspection and maintenance, or even direct replacement.

[0165] For wires of the first health level, surface inspections are performed regularly (once a year) to promptly detect and address surface defects or damage to ensure that the wires work in good condition.

[0166] For conductors in the second health class, the inspection frequency increases to once a quarter to ensure regular inspection of any damage to the conductor surface.

[0167] For conductors of the third health level, ultrasonic testing and magnetic particle testing are performed regularly (once a month) to detect any damage or abnormality inside the conductor so that repair measures can be taken in time.

[0168] If the second carbon fiber composite core conductor reaches the fourth health level, the conductor is immediately replaced to avoid any serious damage to the conductor affecting the system.

[0169] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

[0170] Example 2

[0171] The second embodiment of the present invention is different from the previous embodiment in that:

[0172] If the functions are implemented in the form of software functional units and sold or used as independent products, they 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, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned 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, etc., which can store program codes.

[0173] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0174] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0175] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0176] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0177] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A real-time health monitoring method for carbon fiber composite core conductors based on deep learning, characterized in that: include, Collecting real-time monitoring data of a first carbon fiber composite core conductor, and preprocessing the real-time monitoring data to obtain first health data, wherein the first carbon fiber composite core conductor is a carbon fiber composite core conductor with normal real-time health detection; Constructing an initial health monitoring model, dividing the first health data into a training set and a test set, and using the training set and the test set to train and test the real-time health monitoring model to obtain a real-time health monitoring model; The real-time health monitoring model is deployed in an embedded system, a second carbon fiber composite core conductor is sampled through the embedded system to obtain real-time sampling data, a health rating is performed on the carbon fiber composite core conductor according to the real-time sampling data to obtain a health grade corresponding to the carbon fiber composite core conductor, wherein the second carbon fiber composite core conductor is a carbon fiber composite core conductor to be tested in a working state; The second carbon fiber composite core conductor is maintained according to the maintenance measure table of the carbon fiber composite core conductor and the health level.

2. The real-time health monitoring method of a carbon fiber composite core conductor based on deep learning according to claim 1, characterized in that: The preprocessing includes collecting real-time monitoring data of the first carbon fiber composite core conductor, wherein the real-time monitoring data includes conductor physical data, conductor electrical data, and conductor image data; Preprocessing the real-time monitoring data, cleaning, standardizing and fusing the wire physical data, wire electrical data and wire image data to obtain first health data; The conductor physical data includes conductor temperature, conductor strain, and conductor vibration frequency; the conductor electrical data includes conductor current value and conductor voltage value; The conductor image data includes a conductor ultrasonic image and a conductor magnetic particle detection image.

3. The real-time health monitoring method of a carbon fiber composite core conductor based on deep learning according to claim 2, characterized in that: The preprocessing further includes performing outlier processing and missing value processing on the conductor physical data and the conductor electrical data to obtain first conductor physical data and first conductor electrical data, and performing image denoising and image enhancement processing on the conductor image data to obtain first conductor image data; Performing feature extraction and Z-Score normalization processing on the first conductor physical data and the first conductor electrical data respectively to obtain conductor physical features and conductor electrical features, and merging the conductor physical features and the conductor electrical features to obtain a conductor feature matrix, wherein the conductor feature matrix includes numerical values ​​corresponding to conductor temperature, conductor strain, conductor vibration frequency, conductor current value, and conductor voltage value, and timestamps corresponding to acquisition of the numerical values; The first health data includes first guide wire image data and a guide wire feature matrix.

4. The real-time health monitoring method of a carbon fiber composite core conductor based on deep learning according to claim 3, characterized in that: The constructing of the initial health monitoring model includes constructing the initial health monitoring model based on a deep learning neural network and a long short-term memory network, dividing the first health data into a training set and a test set in a ratio of 7:3; iteratively training the initial health monitoring model through the training set, configuring the optimizer, the loss function and the hyperparameters until the iteration stop condition is reached, thereby obtaining the trained health monitoring model; testing the trained health monitoring model with the test set, and evaluating the model through the mAP value. If the model evaluation meets the requirements, the training and verification are completed, and the real-time health monitoring model is obtained. If the model evaluation does not meet the requirements, the hyperparameters are adjusted and the model is iteratively trained again; The initial health monitoring model includes an input layer, a network combination layer, a mixing layer, an attention mechanism, and an output layer; the input layer inputs the first health data; the network combination layer includes a DNN and an LSTM, the DNN inputs the first wire image data in the first health data, and outputs the wire defect type and the detection result probability; The LSTM inputs the wire feature matrix in the first health data and outputs the time-varying function of each type of data in the first health data; the mixed layer adopts the Softmax function; the attention mechanism adopts the Multi-head Attention function; and the output layer adopts the Quantileregression function.

5. The real-time health monitoring method for carbon fiber composite core conductor based on deep learning according to claim 4, characterized in that: The construction of the initial health monitoring model also includes that the calculation formula of the loss function is: Among them, N is the number of data categories in the conductor feature matrix, T is the total time stamp step corresponding to the collection of various data values ​​in the conductor feature matrix, and y i,t is the probability of the actual value corresponding to the t-th timestamp of the i-th type of data, is the predicted probability of the model for the tth timestamp of the i-th category data; The iteration stopping condition is that the loss function converges or reaches the total number of iterations.

6. The real-time health monitoring method for carbon fiber composite core conductor based on deep learning according to claim 5, characterized in that: The constructing of the initial health monitoring model further includes evaluating the model through the mAP value, and the mAP value is calculated as follows: Where b is the number of data categories in the wire feature matrix, R(t) is the recall rate of the current timestamp, R(t+1) is the recall rate of the next timestamp, and precision(t,t+1) is the accuracy of the current timestamp and the next timestamp; When the mAP value is not less than the preset threshold, the iterative training and testing are terminated and the real-time health monitoring model is output.

7. The real-time health monitoring method for carbon fiber composite core conductor based on deep learning according to claim 6, characterized in that: The real-time sampling data includes time-varying functions of various types of data, wire defect types, and probabilities corresponding to the wire defect types; A plurality of time-varying coefficient thresholds are preset to correspond to various types of data one by one, and the calculation formula for performing health rating on the carbon fiber composite core conductor according to the real-time sampling data is: Wherein, W is the health level of the second carbon fiber composite core conductor, The first health level, The second health level, The third health level, The fourth health level, q a is the time-varying coefficient weight of the time-varying function of the a-th type of data, n is the total number of data categories of the second carbon fiber composite core conductor, V is the conductor image defect type of the second carbon fiber composite core conductor, when V is 0, the conductor image defect type is normal, when V is 1, the conductor image defect type is abnormal, α a is the time-varying coefficient of the time-varying function of the a-th type of data, σ a is the time-varying coefficient threshold of the time-varying function of the a-th type of data.

8. The real-time health monitoring method for carbon fiber composite core conductor based on deep learning according to claim 7, characterized in that: The maintenance measures table includes: when the second carbon fiber composite core conductor is at the first health level, regularly inspecting the conductor surface with one year as a cycle to detect defects or damage on the surface of the second carbon fiber composite core conductor; When the second carbon fiber composite core conductor is at the second health level, regularly inspecting the conductor surface in a quarterly cycle to detect defects or damage on the surface of the second carbon fiber composite core conductor; When the second carbon fiber composite core conductor is at the third health level, ultrasonic testing and magnetic particle testing are performed regularly on a monthly basis to detect damage inside the conductor; When the second carbon fiber composite core conductor is at the fourth health level, the second carbon fiber composite core conductor is replaced immediately.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 8 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 8 are implemented.

Images