Cable defect identification method, device, electronic equipment and storage medium

Abstract

The present invention discloses a method, device, electronic device and storage medium for cable defect identification, which belongs to the technical field of distribution cable defect identification. The method comprises the following steps: collecting voiceprint data of a target cable; optimizing the filter length and period of a maximum correlation kurtosis filter method by a triangular topology aggregation optimization method, and performing denoising and enhancement processing on the voiceprint data by using the optimized maximum correlation kurtosis filter method; optimizing the input weights and hidden layer bias of an online sequence extreme learning machine by a triangular topology aggregation optimization method, and constructing a cable defect identification model based on the online sequence extreme learning machine after parameter optimization; training the model based on voiceprint data, and updating the current output weights of the cable defect identification model in real time by using the chain update rule of the output weights, and outputting the cable defect identification result. The present invention significantly improves the efficiency and effect of voiceprint enhancement and extreme learning machine parameter optimization, and greatly reduces hardware resources and computing costs.

Description

Technical Field

[0001] The present invention relates to a method, a device, an electronic device and a storage medium for cable defect identification, and belongs to the technical field of distribution cable defect identification. Background Art

[0002] As the power network continues to expand and become more complex, it is particularly important to ensure the stable operation of the cable system. Cables are key components for power transmission, and their internal defects may cause major accidents, such as power outages and fires, affecting public safety and economic operations. Therefore, the development of efficient and accurate cable defect detection technology has become an urgent need in the power industry.

[0003] Traditional cable detection methods such as infrared thermal imaging and withstand voltage testing may not be able to effectively detect tiny defects or early faults inside the cable in some cases, especially for hidden problems such as insulation aging and partial discharge. In recent years, with the rapid development of sensor technology, signal processing algorithms (especially digital signal processing and pattern recognition), and artificial intelligence technology (such as machine learning and deep learning), the application of voiceprint recognition technology in cable defect detection has become more accurate and efficient. These technological advances enable the system to separate useful signals from complex background noise and realize rapid identification and classification of cable defects.

[0004] In order to improve the defect perception ability of the distribution cable voiceprint, it is necessary to optimize the parameters of the corresponding algorithm and process them in real time. Therefore, the present invention proposes an ELM cable defect recognition method based on triangle topology aggregation optimization. Summary of the invention

[0005] In order to solve the above problems, the present invention proposes a method, device, electronic device and storage medium for cable defect identification, which can effectively identify cable defects and greatly reduce hardware resources and computing costs.

[0006] The technical solution adopted by the present invention to solve the technical problem is:

[0007] In a first aspect, an embodiment of the present invention provides a method for identifying cable defects, comprising the following steps:

[0008] Step S1, collecting the voiceprint data of the target cable;

[0009] Step S2, optimizing the filter length and period of the maximum correlation kurtosis filter method by using the triangle topology aggregation optimization method, and using the optimized maximum correlation kurtosis filter method to perform denoising and enhancement processing on the voiceprint data;

[0010] Step S3, optimizing the input weights and hidden layer biases of an online sequential extreme learning machine (OS-ELM) by a triangular topology aggregation optimization method, and constructing a cable defect recognition model based on the online sequential extreme learning machine after parameter optimization;

[0011] Step S4, performing model training based on voiceprint data, using the chain update rule of output weights to update the current output weights of the cable defect recognition model in real time, and outputting the cable defect recognition results.

[0012] As a possible implementation of this embodiment, the triangle topology aggregation optimization method includes the following steps:

[0013] (1) Initialize the restricted range of optimization parameters and the number of iterations T, construct a triangular topological space model, randomly select initial vertices, and then construct an initial triangular unit. All subsets within the triangular unit are configuration options of the optimization parameters. The triangular topological space model is:

[0014] ,

[0015] in, and are the maximum and minimum values ​​of the optimization parameters, respectively. , and are the initial vertices, r0 is a random number in [0, 1], represents the size of a triangular topological unit, t is the current iteration number, and represent Guided by and The direction vector of is a random angle between [0, π];

[0016] (2) Determine the fitness function and verify the parameter configuration options of the three vertices and subsets;

[0017] (3) Determine the optimal position of the subset based on the fitness function and the local optimal constraint function :

[0018] ;

[0019] (4) Select the optimal location As the new initial vertex, use the method in substep (1) to obtain and Create a new triangle topology cell:

[0020] ;

[0021] (5) Determine whether the current number of iterations t is greater than T. If so, output the optimal position of the final subset and determine the optimal solution of the optimization parameters. If so, repeat sub-steps (3) and (4).

[0022] As a possible implementation manner of this embodiment, the fitness function is a root mean square error.

[0023] As a possible implementation of this embodiment, the formula of the maximum correlation kurtosis filtering method is:

[0024] ,

[0025] Among them, T is the filtering period, M is the displacement number, and N is the output discrete data. is the regularization term of the cable voiceprint.

[0026] As a possible implementation of this embodiment, the filter length of the maximum correlation kurtosis filter method is [400, 600], the period is [100s, 300s], and both the filter length and the period are initialized and randomly selected to satisfy the Gaussian distribution.

[0027] As a possible implementation of this embodiment, the input weights and hidden layer biases of the online sequence extreme learning machine satisfy Gaussian random distribution.

[0028] As a possible implementation of this embodiment, the cable defect recognition model is:

[0029] ,

[0030] in, is a voiceprint data sample, is the input weight, is the bias, L is the number of hidden layer nodes, β0 is the output weight of the hidden layer nodes, N0 is the number of voiceprint data samples, is the network output of the extreme learning machine.

[0031] As a possible implementation of this embodiment, the chain update rule is:

[0032] Let H be the hidden layer output matrix, ,but , use the following formula to chain update output weights:

[0033] ,

[0034] in is the output weight of the n+1th time, Y n+1 is the network output of the voiceprint sample arriving at the n+1th time, is the Gaussian constraint term for the output weights.

[0035] As a possible implementation of this embodiment, the hidden layer output matrix is:

[0036] .

[0037] As a possible implementation of this embodiment, step S4 includes:

[0038] The process of optimizing network structure using triangular topology aggregation;

[0039] The process of online learning on cable voiceprint data.

[0040] In a second aspect, an embodiment of the present invention provides a device for identifying cable defects, including:

[0041] A data acquisition module, used to collect the voiceprint data of the target cable;

[0042] The data processing module is used to optimize the filter length and period of the maximum correlation kurtosis filter method through the triangle topology aggregation optimization method, and use the optimized maximum correlation kurtosis filter method to perform denoising and enhancement processing on the voiceprint data;

[0043] A model building module is used to optimize the input weights and hidden layer biases of the online sequence extreme learning machine through a triangular topology aggregation optimization method, and to build a cable defect recognition model based on the online sequence extreme learning machine after parameter optimization;

[0044] The defect recognition module is used to perform model training based on voiceprint data, adopt the chain update rule of output weights to update the current output weights of the cable defect recognition model in real time, and output the cable defect recognition results.

[0045] As a possible implementation of this embodiment, the triangle topology aggregation optimization method includes the following steps:

[0046] (1) Initialize the restricted range of optimization parameters and the number of iterations T, construct a triangular topological space model, randomly select initial vertices, and then construct an initial triangular unit. All subsets within the triangular unit are configuration options of the optimization parameters. The triangular topological space model is:

[0047] ,

[0048] in, and are the maximum and minimum values ​​of the optimization parameters, respectively. , and are the initial vertices, r0 is a random number in [0, 1], represents the size of a triangular topological unit, t is the current iteration number, and represent Guided by and The direction vector of is a random angle between [0, π];

[0049] (2) Determine the fitness function and verify the parameter configuration options of the three vertices and subsets;

[0050] (3) Determine the optimal position of the subset based on the fitness function and the local optimal constraint function :

[0051] ;

[0052] (4) Select the optimal location As the new initial vertex, use the method in substep (1) to obtain and Create a new triangle topology cell:

[0053] ;

[0054] (5) Determine whether the current number of iterations t is greater than T. If so, output the optimal position of the final subset and determine the optimal solution of the optimization parameters. If so, repeat sub-steps (3) and (4).

[0055] As a possible implementation manner of this embodiment, the fitness function is a root mean square error.

[0056] As a possible implementation of this embodiment, the formula of the maximum correlation kurtosis filtering method is:

[0057] ,

[0058] Among them, T is the filtering period, M is the displacement number, and N is the output discrete data. is the regularization term of the cable voiceprint.

[0059] As a possible implementation of this embodiment, the filter length of the maximum correlation kurtosis filter method is [400, 600], the period is [100s, 300s], and both the filter length and the period are initialized and randomly selected to satisfy the Gaussian distribution.

[0060] As a possible implementation of this embodiment, the input weights and hidden layer biases of the online sequence extreme learning machine satisfy Gaussian random distribution.

[0061] As a possible implementation of this embodiment, the cable defect recognition model is:

[0062] ,

[0063] in, is a voiceprint data sample, is the input weight, is the bias, L is the number of hidden layer nodes, β0 is the output weight of the hidden layer nodes, N0 is the number of voiceprint data samples, is the network output of the extreme learning machine.

[0064] As a possible implementation of this embodiment, the chain update rule is:

[0065] Let H be the hidden layer output matrix, ,but , use the following formula to chain update output weights:

[0066] ,

[0067] in is the output weight of the n+1th time, Y n+1 is the network output of the voiceprint sample arriving at the n+1th time, is the Gaussian constraint term for the output weights.

[0068] As a possible implementation of this embodiment, the hidden layer output matrix is:

[0069] .

[0070] As a possible implementation of this embodiment, the process in which the defect recognition module outputs the cable defect recognition result includes:

[0071] The process of optimizing network structure using triangular topology aggregation;

[0072] The process of online learning on cable voiceprint data.

[0073] In a third aspect, an embodiment of the present invention provides an electronic device, comprising a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the memory through the bus, and the processor executes the machine-readable instructions to perform the steps of any cable defect identification method as described above.

[0074] In a fourth aspect, an embodiment of the present invention provides a storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above-mentioned cable defect identification methods are executed.

[0075] The beneficial effects of the technical solution of the embodiment of the present invention are as follows:

[0076] A method for cable defect identification according to the technical solution of an embodiment of the present invention comprises the following steps: step S1, collecting voiceprint data of a target cable; step S2, optimizing the filtering length and period of a maximum correlation kurtosis filtering method by a triangular topology aggregation optimization method, and using the optimized maximum correlation kurtosis filtering method to perform denoising and enhancement processing on the voiceprint data; step S3, optimizing the input weights and hidden layer biases of an online sequence extreme learning machine by a triangular topology aggregation optimization method, and constructing a cable defect identification model based on the online sequence extreme learning machine after parameter optimization; step S4, training a model based on the voiceprint data, using the chain update rule of the output weights to update the current output weights of the cable defect identification model in real time, and outputting the cable defect identification result. According to the structural characteristics of cable voiceprint enhancement and online sequence extreme learning machine, the present invention selects the filtering length and period of the maximum correlation kurtosis filtering method and the input weight and hidden layer bias of the online sequence extreme learning machine as optimization parameters, and iteratively constructs multiple similar triangle topology units based on the triangle topology structure, generates new vertices through forward information exchange between different triangle topology units, and constructs new optimal positions according to the local optimal vertex of each triangle to balance the global and local search of the optimal parameters, thereby obtaining more competitive cable defect classification performance and realizing effective identification of cable defects. The present invention not only significantly improves the efficiency and effect of voiceprint enhancement and extreme learning machine parameter optimization, but also greatly reduces hardware resources and computing costs.

[0077] The present invention constructs a triangular topology structure according to the limited range of optimization parameters, and generates the optimal triangular stability architecture through iteration to achieve global and local optimal search. Thus, a more competitive cable defect classification performance is obtained. The efficiency and cost of voiceprint enhancement and extreme learning machine parameter optimization are significantly improved, the cable defect classification potential of the machine learning architecture is maximized, the hardware resources and computing costs of cable voiceprint enhancement technology and OS-ELM for defect identification are saved, and it has certain application value in engineering practice. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 is a flow chart of a method for cable defect identification according to an exemplary embodiment;

[0079] Figure 2 is a schematic structural diagram of a device for identifying cable defects according to an exemplary embodiment;

[0080] Figure 3 It is a specific implementation flow chart of cable defect identification using the technical solution of the present invention according to an exemplary embodiment;

[0081] Figure 4 The figure is a curve diagram showing the relationship between the initial sample size and the accuracy rate of the present invention according to an exemplary embodiment. DETAILED DESCRIPTION

[0082] In order to more clearly illustrate the technical features of the solution of the present invention, the present invention is described in detail below through specific implementation methods and in conjunction with the accompanying drawings.

[0083] like Figure 1 As shown, a method for identifying cable defects provided by an embodiment of the present invention includes the following steps:

[0084] Step S1, collecting the voiceprint data of the target cable;

[0085] Step S2, optimizing the filter length and period of the maximum correlation kurtosis filter method by using the triangle topology aggregation optimization method, and using the optimized maximum correlation kurtosis filter method to perform denoising and enhancement processing on the voiceprint data;

[0086] Step S3, optimizing the input weights and hidden layer biases of the online sequence extreme learning machine by using a triangular topology aggregation optimization method, and building a cable defect recognition model based on the online sequence extreme learning machine after parameter optimization;

[0087] Step S4, performing model training based on voiceprint data, using the chain update rule of output weights to update the current output weights of the cable defect recognition model in real time, and outputting the cable defect recognition results.

[0088] As a possible implementation of this embodiment, the triangle topology aggregation optimization method includes the following steps:

[0089] (1) Initialize the restricted range of optimization parameters and the number of iterations T, construct a triangular topological space model, randomly select initial vertices, and then construct an initial triangular unit. All subsets within the triangular unit are configuration options of the optimization parameters. The triangular topological space model is:

[0090] ,

[0091] in, and are the maximum and minimum values ​​of the optimization parameters, respectively. , and are the initial vertices, r0 is a random number in [0, 1], represents the size of a triangular topological unit, t is the current iteration number, and represent Guided by and The direction vector of is a random angle between [0, π];

[0092] (2) Determine the fitness function and verify the parameter configuration options of the three vertices and subsets;

[0093] (3) Determine the optimal position of the subset based on the fitness function and the local optimal constraint function :

[0094] ;

[0095] (4) Select the optimal location As the new initial vertex, use the method in substep (1) to obtain and Create a new triangle topology cell:

[0096] ;

[0097] (5) Determine whether the current number of iterations t is greater than T. If so, output the optimal position of the final subset and determine the optimal solution of the optimization parameters. If so, repeat sub-steps (3) and (4).

[0098] As a possible implementation manner of this embodiment, the fitness function is a root mean square error.

[0099] As a possible implementation of this embodiment, the formula of the maximum correlation kurtosis filtering method is:

[0100] ,

[0101] Among them, T is the filtering period, M is the displacement number, and N is the output discrete data. is the regularization term of the cable voiceprint.

[0102] As a possible implementation of this embodiment, the filter length of the maximum correlation kurtosis filter method is [400, 600], the period is [100s, 300s], and both the filter length and the period are initialized and randomly selected to satisfy the Gaussian distribution.

[0103] As a possible implementation of this embodiment, the input weights and hidden layer biases of the online sequence extreme learning machine satisfy Gaussian random distribution.

[0104] As a possible implementation of this embodiment, the cable defect recognition model is:

[0105] ,

[0106] in, is a voiceprint data sample, is the input weight, is the bias, L is the number of hidden layer nodes, β0 is the output weight of the hidden layer nodes, N0 is the number of voiceprint data samples, is the network output of the extreme learning machine.

[0107] As a possible implementation of this embodiment, the chain update rule is:

[0108] Let H be the hidden layer output matrix, ,but , use the following formula to chain update output weights:

[0109]

[0110] in is the output weight of the n+1th time, Y n+1 is the network output of the voiceprint sample arriving at the n+1th time, is the Gaussian constraint term for the output weights.

[0111] As a possible implementation of this embodiment, the hidden layer output matrix is:

[0112] .

[0113] As a possible implementation of this embodiment, step S4 includes:

[0114] The process of optimizing network structure using triangular topology aggregation;

[0115] The process of online learning on cable voiceprint data.

[0116] like Figure 2 As shown, an embodiment of the present invention provides a device for identifying cable defects, including:

[0117] A data acquisition module, used to collect the voiceprint data of the target cable;

[0118] The data processing module is used to optimize the filter length and period of the maximum correlation kurtosis filter method through the triangle topology aggregation optimization method, and use the optimized maximum correlation kurtosis filter method to perform denoising and enhancement processing on the voiceprint data;

[0119] A model building module is used to optimize the input weights and hidden layer biases of the online sequence extreme learning machine through a triangular topology aggregation optimization method, and to build a cable defect recognition model based on the online sequence extreme learning machine after parameter optimization;

[0120] The defect recognition module is used to perform model training based on voiceprint data, adopt the chain update rule of output weights to update the current output weights of the cable defect recognition model in real time, and output the cable defect recognition results.

[0121] As a possible implementation of this embodiment, the triangle topology aggregation optimization method includes the following steps:

[0122] (1) Initialize the restricted range of optimization parameters and the number of iterations T, construct a triangular topological space model, randomly select initial vertices, and then construct an initial triangular unit. All subsets within the triangular unit are configuration options of the optimization parameters. The triangular topological space model is:

[0123] ,

[0124] in, and are the maximum and minimum values ​​of the optimization parameters, respectively. , and are the initial vertices, r0 is a random number in [0, 1], represents the size of a triangular topological unit, t is the current iteration number, and represent Guided by and The direction vector of is a random angle between [0, π];

[0125] (2) Determine the fitness function and verify the parameter configuration options of the three vertices and subsets;

[0126] (3) Determine the optimal position of the subset based on the fitness function and the local optimal constraint function :

[0127] ;

[0128] (4) Select the optimal location As the new initial vertex, use the method in substep (1) to obtain and Create a new triangle topology cell:

[0129] ;

[0130] (5) Determine whether the current number of iterations t is greater than T. If so, output the optimal position of the final subset and determine the optimal solution of the optimization parameters. If so, repeat sub-steps (3) and (4).

[0131] As a possible implementation manner of this embodiment, the fitness function is a root mean square error.

[0132] As a possible implementation of this embodiment, the formula of the maximum correlation kurtosis filtering method is:

[0133] ,

[0134] Among them, T is the filtering period, M is the displacement number, and N is the output discrete data. is the regularization term of the cable voiceprint.

[0135] As a possible implementation of this embodiment, the filter length of the maximum correlation kurtosis filter method is [400, 600], the period is [100s, 300s], and both the filter length and the period are initialized and randomly selected to satisfy the Gaussian distribution.

[0136] As a possible implementation of this embodiment, the input weights and hidden layer biases of the online sequence extreme learning machine satisfy Gaussian random distribution.

[0137] As a possible implementation of this embodiment, the cable defect recognition model is:

[0138] ,

[0139] in, is a voiceprint data sample, is the input weight, is the bias, L is the number of hidden layer nodes, β0 is the output weight of the hidden layer nodes, N0 is the number of voiceprint data samples, is the network output of the extreme learning machine.

[0140] As a possible implementation of this embodiment, the chain update rule is:

[0141] Let H be the hidden layer output matrix, ,but , use the following formula to chain update output weights:

[0142] ,

[0143] in is the output weight of the n+1th time, Y n+1 is the network output of the voiceprint sample arriving at the n+1th time, is the Gaussian constraint term for the output weights.

[0144] As a possible implementation of this embodiment, the hidden layer output matrix is:

[0145] .

[0146] As a possible implementation of this embodiment, the process in which the defect recognition module outputs the cable defect recognition result includes:

[0147] The process of optimizing network structure using triangular topology aggregation;

[0148] The process of online learning on cable voiceprint data.

[0149] This paper proposes an ELM cable defect recognition method based on triangular topology aggregation optimization. Through the envelope feature analysis and observation of the collected cable voiceprint data, it is found that there are obvious distribution features such as envelope mutation in the defective cable voiceprint data. The core of the maximum correlation kurtosis filtering method MCKD is to use kurtosis as the optimization target. Kurtosis is a statistic that measures the sharpness of signal distribution. The sharper the peak, the higher the kurtosis of the signal. By maximizing the kurtosis, MCKD can highlight the characteristics of defects or fault events.

[0150] The purpose of MCKD is to find an inverse FIR filter to restore the original input signal from the actual output fault information and maximize the correlation kurtosis of the original impulse component, thereby extracting the periodic impulse component in the voiceprint and removing noise interference.

[0151] like Figure 3 As shown, the specific process of cable defect identification using the technical solution of the present invention is as follows.

[0152] Step 1: First, collect the voiceprint data of normal or defective cables through the hardware API. Here, the hardware can be the circuit sensor management unit of the distribution network. Define the collected voiceprint data as x(n) and the mixed noise data as y(n). The following definitions can be used:

[0153] ,

[0154] Among them, h(n) is the impulse response during the hardware API transmission process, and e(n) is the background noise.

[0155] In order to effectively remove background noise and highlight the voiceprint features, the period P and filter length L of MCKD need to be optimized before the collected cable voiceprint is enhanced.

[0156] Step 2: The cable soundprint denoising enhancement process is performed by optimizing the filter length and period of the maximum correlation kurtosis filter method through the triangle topology aggregation optimization method.

[0157] The result of the MCKD filter is expressed in matrix form as follows:

[0158] ,

[0159] The filtering steps are: 1) set the period P, filter length L and displacement M; 2) calculate the covariance matrix of the input signal x(n) and ; 3) Obtain the filtered output signal y(n); 4) Calculate the intermediate variable based on y(n) and B; 5) Update the filter result w; 6) If the correlation kurtosis of the voiceprint data before and after filtering is less than the given threshold, stop the iteration, otherwise, return to step 3).

[0160] The triangle topology aggregation optimization method specifically includes:

[0161] (1) Initialize the optimization parameter period P, the filter length L, the number of iterations T, build a triangular topological space model and randomly select initial vertices to build an initial triangular unit. All subsets inside the triangular unit are configuration options of the optimization parameters.

[0162] ,

[0163] in, and are the maximum and minimum values ​​of the optimization parameters, corresponding to the maximum and minimum values ​​of the optimization parameter filter length and period in step 2, , and are the initial vertices, r0 is a random number in [0, 1], represents the size of a triangular topological unit, t is the current iteration number, and represent Guided by and The direction vector of is a random angle between [0, π];

[0164] (2) Determine the fitness function G(x) and verify the parameter configuration options of the three vertices and subsets;

[0165] (3) Determine the optimal position of the subset based on the fitness function and the local optimal constraint function :

[0166] ;

[0167] (4) Update triangle topology unit: Select As the new initial vertex, use the method in substep (1) to obtain and Create a new triangle topology cell:

[0168] ;

[0169] (5) Determine whether the current number of iterations t is greater than T. If so, output the optimal position of the final subset and determine the optimal solution of the optimization parameters. If so, repeat sub-steps (3) and (4).

[0170] In step 2, the optimization range of the filter period is [100s, 300s], and the optimization range of the length is [400, 600], with the maximum values ​​being 300s and 600, and the minimum values ​​being 100s and 400, respectively. The initialization of the two is randomly selected to satisfy the Gaussian distribution.

[0171] In step 2, the parameter optimization of the maximum correlation kurtosis filter method is set to the adaptive function G(x):

[0172] ,

[0173] Among them, P is the filtering period, M is the displacement number, and N is the output discrete data. is the regularization term of the cable voiceprint.

[0174] After iterative optimization, the optimal parameter solution is obtained and the voiceprint data is enhanced, which improves the availability and rationality of subsequent cable defect identification samples.

[0175] Step 3: Optimize the input weights and hidden layer biases of the online sequence extreme learning machine through the triangle topology aggregation optimization method to build the initialization model.

[0176] The specific contents of initialization learning are:

[0177] First, select N0 sample voiceprint data information sets as initial samples for initial learning, then the network output of the extreme learning machine for:

[0178] ,

[0179] in, is a voiceprint data sample, is the input weight, is the bias, L is the number of hidden layer nodes, and the initial learning problem is to solve the output weight β0 connecting the output vector and the hidden layer nodes, which is equivalent to satisfying min‖Hβ- ‖ of β. The output matrix H of the hidden layer is expressed as:

[0180] ,

[0181] Calculate the network hidden layer output matrix H0 to obtain the output weight β0, store the recognition results, and complete the initial training.

[0182] Extreme learning machine is a type of algorithm that uses a random mechanism to set and select parameters for the hidden layer of a single hidden layer neural network, thereby greatly improving the learning speed and ensuring generalization ability. Therefore, the factors affecting the performance of OS-ELM are mainly the number of nodes in the network hidden layer, the network connection weights and the threshold. Therefore, appropriate methods are needed to optimize the network connection weights and hidden layer bias when identifying cable defects.

[0183] The triangle topology aggregation optimization method used in step 3 is the same as the triangle topology aggregation optimization method used in step 2.

[0184] The fitness function for parameter optimization of the line sequence extreme learning machine is the root mean square error and the optimization limit range of the input weights and hidden layer bias in step 3 satisfies the Gaussian random distribution.

[0185] After building the initialization model, subsequent voiceprint data needs to be learned online to output the final cable defect type.

[0186] Step 4: For the voiceprint data information that arrives successively, use the chain update rule of the output weight to update the current output weight in real time and output the defect recognition result.

[0187] In the process of defect recognition based on cable voiceprint, it is generally impossible to obtain all samples at one time. The standard extreme learning machine adopts a batch learning mode. When there are new samples, all samples need to be relearned, which is time-consuming. As the amount of data increases, the demand for memory space will also increase significantly. Using the online learning method, only the new samples at the current moment need to be learned, and all samples do not need to be learned. In addition, the online learning algorithm has relatively low requirements on time and space.

[0188] The online sequence extreme learning machine is a type of algorithm proposed to address the problem that ordinary extreme learning machines cannot effectively learn real-time data sets, that is, new data is constantly added to the network during the learning process. The core idea is to introduce online learning into the ELM algorithm.

[0189] The chain update rule in step 4 is:

[0190] Let H be the hidden layer output matrix, ,but , use the following formula to chain update output weights:

[0191] ,

[0192] in is the output weight of the n+1th time in step 4, Y n+1 is the network output of the voiceprint sample arriving at the n+1th time, is the Gaussian constraint term for the output weight, is the input layer output matrix, is the first output matrix of the hidden layer, is the nth output matrix of the hidden layer, is the initial coefficient of chain update, is the first chain update coefficient, is the n+1th chain update coefficient;

[0193] When new voiceprint data is used as input, the hidden layer output matrix is ​​calculated according to the activation function, and the output weights are updated using the above chain rule. The update is repeated until there is no new voiceprint data and the final defect category is output.

[0194] The overall steps of outputting defect recognition results are:

[0195] It is mainly divided into two parts: one is the process of optimizing the network structure using triangular topology aggregation, and the other is the online learning of cable voiceprint data.

[0196] In order to better illustrate the advantages of the optimization algorithm of the present invention, the algorithm is evaluated from the two aspects of accuracy and training time, and the triangle topology optimization OSELM algorithm of the present invention is compared with the general classification algorithms (SVM, ELM, OS-ELM) to illustrate the speed and generalization performance of the triangle topology optimization OSELM algorithm.

[0197] Since there is no public cable voiceprint database, this simulation uses the database of a distribution management system as the experimental database. 4181 data were obtained from the cable defect information and log table in the database, of which 3000 were used as training data sets and 1181 were used as test data sets. In order to avoid the accidental nature of the experiment, each algorithm was run 20 times, and the average time and average accuracy were taken as the experimental results.

[0198] In addition to the number of hidden nodes in the network, the size of the initial sample also has a significant impact on the accuracy of the recognition results. In order to obtain the best recognition results, the accuracy of the OS-ELM and triangular topology optimization OSELM algorithms was compared under different initial training set sizes. The results are shown in Figure 2. Figure 4 As shown in the figure, the algorithm accuracy rate starts to change rapidly with the increase of the number of initial training samples, and gradually stabilizes after 310, with a small increase. Therefore, 310 is selected as the number of samples in the initial learning stage of the training data set.

[0199] The maximum number of iterations in the triangular topology optimization OSELM algorithm of the present invention is set to 120, and r0 and . We conducted 20 experiments on each of the four algorithms on the data set, and the average results are shown in the following table.

[0200]

[0201] It can be seen that the OSELM algorithm of triangular topology optimization of the present invention has a similar accuracy to SVM, but the H1-OSELM training time is greatly reduced, especially when the number is large, the time advantage will be more obvious. Compared with the other two extreme learning machine algorithms ELM and OS-ELM, due to the addition of the online learning part and the optimization parameter selection process, the time is obviously longer than these two algorithms, but both the training accuracy and the test accuracy are significantly improved, and it can be said that the overall performance is better.

[0202] An embodiment of the present invention provides an electronic device, including a processor, a memory and a bus, wherein the memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory through the bus, and the processor executes the machine-readable instructions to perform the steps of any cable defect identification method as described above.

[0203] Specifically, the above-mentioned memory and processor can be general-purpose memory and processor, which are not specifically limited here. When the processor runs the computer program stored in the memory, the above-mentioned cable defect identification method can be executed.

[0204] Those skilled in the art will appreciate that the structure of the electronic device does not limit the electronic device and may include more or fewer components than shown in the figure, or combine or split certain components, or arrange the components differently.

[0205] In some embodiments, the electronic device may also include a touch screen that can be used to display a graphical user interface (e.g., a startup interface of an application) and receive user operations on the graphical user interface (e.g., startup operations on an application). The specific touch screen may include a display panel and a touch panel. The display panel may be configured in the form of LCD (Liquid Crystal Display), OLED (Organic Light-Emitting Diode), etc. The touch panel may collect the user's contact or non-contact operations on or near it, and generate pre-set operation instructions, for example, the user uses any suitable object such as a finger, stylus, or accessories on or near the touch panel. In addition, the touch panel may include two parts: a touch detection device and a touch controller. The touch detection device detects the user's touch position and posture, detects the signal brought by the touch operation, and transmits the signal to the touch controller; the touch controller receives the touch information from the touch detection device, converts it into information that the processor can process, and then sends it to the processor, and can receive and execute commands from the processor. In addition, the touch panel can be implemented by various types such as resistive, capacitive, infrared and surface acoustic wave, and any technology developed in the future can also be used to implement the touch panel. Further, the touch panel can cover the display panel, and the user can operate on or near the touch panel covered on the display panel according to the graphical user interface displayed on the display panel. After the touch panel detects the operation on or near it, it is transmitted to the processor to determine the user input, and then the processor provides corresponding visual output on the display panel in response to the user input. In addition, the touch panel and the display panel can be implemented as two independent components or integrated.

[0206] Corresponding to the method for starting the above application, an embodiment of the present invention further provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of any of the above cable defect identification methods are executed.

[0207] The application startup device provided in the embodiment of the present application can be specific hardware on the device or software or firmware installed on the device. The device provided in the embodiment of the present application, its implementation principle and the technical effect produced are the same as those in the aforementioned method embodiment. For the sake of brief description, the parts not mentioned in the device embodiment can refer to the corresponding contents in the aforementioned method embodiment. Those skilled in the art can clearly understand that for the convenience and simplicity of description, the specific working processes of the systems, devices and units described above can all refer to the corresponding processes in the aforementioned method embodiment, and will not be repeated here.

[0208] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0209] In the embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely schematic. For example, the division of modules is only a logical function division. There may be other division methods in actual implementation. For example, multiple modules or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some communication interfaces, indirect coupling or communication connection of devices or modules, which can be electrical, mechanical or other forms.

[0210] The modules described as separate components may or may not be physically separated, and the components shown as modules may or may not be physical modules, that is, they may be located in one place or distributed on multiple network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0211] In addition, each functional module in the embodiments provided in the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0212] 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.

[0213] 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.

[0214] 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.

[0215] Finally, 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 above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for identifying cable defects, characterized in that: The steps include: Step S1, collecting the voiceprint data of the target cable; Step S2, optimizing the filter length and period of the maximum correlation kurtosis filter method by using the triangle topology aggregation optimization method, and using the optimized maximum correlation kurtosis filter method to perform denoising and enhancement processing on the voiceprint data; Step S3, optimizing the input weights and hidden layer biases of the online sequence extreme learning machine by using a triangular topology aggregation optimization method, and building a cable defect recognition model based on the online sequence extreme learning machine after parameter optimization; Step S4, performing model training based on voiceprint data, using the chain update rule of output weights to update the current output weights of the cable defect recognition model in real time, and outputting the cable defect recognition results; The triangle topology aggregation optimization method comprises the following steps: (1) Initialize the restricted range of optimization parameters and the number of iterations T, construct a triangular topological space model, randomly select initial vertices, and then construct an initial triangular unit. All subsets within the triangular unit are configuration options of the optimization parameters. The triangular topological space model is: ， in, and are the maximum and minimum values ​​of the optimization parameters, respectively. , and are the initial vertices, r 0 is a random number in [0, 1], represents the size of a triangular topological unit, t is the current iteration number, and represent Guided by and The direction vector of is a random angle between [0, π]; (2) Determine the fitness function and verify the parameter configuration options of the three vertices and subsets; (3) Determine the optimal position of the subset based on the fitness function and the local optimal constraint function : ； (4) Select the optimal location As the new initial vertex, use the method in substep (1) to obtain and Create a new triangle topology cell: ； (5) Determine whether the current number of iterations t is greater than T. If so, output the optimal position of the final subset and determine the optimal solution of the optimization parameters. If so, repeat sub-steps (3) and (4). The cable defect recognition model is: ， in, is a voiceprint data sample, is the input weight, is the bias, L is the number of hidden layer nodes, β 0 is the output weight of the hidden layer node, N 0 is the number of voiceprint data samples, is the network output of the extreme learning machine.

2. The method for cable defect identification according to claim 1, characterized in that: The fitness function is the root mean square error.

3. The method for cable defect identification according to claim 1, characterized in that: The formula of the maximum correlation kurtosis filtering method is: ， Among them, T is the filtering period, M is the displacement number, and N is the output discrete data. is the regularization term of the cable voiceprint.

4. The method for cable defect identification according to claim 1, characterized in that: The filter length of the maximum correlation kurtosis filter method is [400, 600], the period is [100s, 300s], and both the filter length and the period are initialized and randomly selected to satisfy the Gaussian distribution.

5. The method for cable defect identification according to claim 1, characterized in that: The input weights and hidden layer biases of the online sequence extreme learning machine satisfy Gaussian random distribution.

6. The method for cable defect identification according to claim 1, characterized in that: The chain update rule is: Let H be the hidden layer output matrix, ,but , use the following formula to chain update output weights: ， in is the output weight of the n+1th time, Y n+1 is the network output of the voiceprint sample arriving at the n+1th time, is the Gaussian constraint term for the output weights.

7. The cable defect identification method according to claim 6, characterized in that: The hidden layer output matrix is: 。 8. The method for identifying cable defects according to any one of claims 1 to 7, characterized in that: The step S4 comprises: The process of optimizing network structure using triangular topology aggregation; The process of online learning on cable voiceprint data.

9. A device for identifying cable defects, characterized in that: include: A data acquisition module, used to collect the voiceprint data of the target cable; The data processing module is used to optimize the filter length and period of the maximum correlation kurtosis filter method through the triangle topology aggregation optimization method, and use the optimized maximum correlation kurtosis filter method to perform denoising and enhancement processing on the voiceprint data; A model building module is used to optimize the input weights and hidden layer biases of the online sequence extreme learning machine through a triangular topology aggregation optimization method, and to build a cable defect recognition model based on the online sequence extreme learning machine after parameter optimization; The defect recognition module is used to train the model based on the voiceprint data, update the current output weight of the cable defect recognition model in real time using the chain update rule of the output weight, and output the cable defect recognition result; The triangle topology aggregation optimization method comprises the following steps: (1) Initialize the restricted range of optimization parameters and the number of iterations T, construct a triangular topological space model, randomly select initial vertices, and then construct an initial triangular unit. All subsets within the triangular unit are configuration options of the optimization parameters. The triangular topological space model is: ， in, and are the maximum and minimum values ​​of the optimization parameters, respectively. , and are the initial vertices, r 0 is a random number in [0, 1], represents the size of a triangular topological unit, t is the current iteration number, and represent Guided by and The direction vector of is a random angle between [0, π]; (2) Determine the fitness function and verify the parameter configuration options of the three vertices and subsets; (3) Determine the optimal position of the subset based on the fitness function and the local optimal constraint function : ； (4) Select the optimal location As the new initial vertex, use the method in substep (1) to obtain and Create a new triangle topology cell: ； (5) Determine whether the current number of iterations t is greater than T. If so, output the optimal position of the final subset and determine the optimal solution of the optimization parameters. If so, repeat sub-steps (3) and (4). The cable defect recognition model is: ， in, is a voiceprint data sample, is the input weight, is the bias, L is the number of hidden layer nodes, β 0 is the output weight of the hidden layer node, N 0 is the number of voiceprint data samples, is the network output of the extreme learning machine.

10. The cable defect identification device according to claim 9, characterized in that: The formula of the maximum correlation kurtosis filtering method is: ， Among them, T is the filtering period, M is the displacement number, and N is the output discrete data. is the regularization term of the cable voiceprint.

11. The cable defect identification device according to claim 9, characterized in that: The input weights and hidden layer biases of the online sequence extreme learning machine satisfy Gaussian random distribution.

12. An electronic device, characterized in that: It includes a processor, a memory and a bus, the memory stores machine-readable instructions executable by the processor, when the electronic device is running, the processor and the memory communicate through the bus, and the processor executes the machine-readable instructions to perform the steps of the method for cable defect identification as described in any one of claims 1-8.

13. A storage medium, characterized in that: The storage medium stores a computer program, which, when executed by a processor, executes the steps of the method for identifying cable defects as described in any one of claims 1 to 8.

Images