Pipeline surrounding event pattern recognition method based on improved sparrow search algorithm optimization support vector machine

Abstract

The invention relates to a pipeline surrounding event pattern recognition method based on an improved sparrow search algorithm optimization support vector machine, and solves the problem that the working efficiency and prediction precision of pipeline monitoring are reduced due to the fact that an existing pipeline monitoring recognition model is easy to fall into local optimum and relatively slow in convergence speed. The method comprises the following steps: firstly, acquiring pipeline peripheral vibration data acquired by a distributed optical fiber vibration sensing system, preprocessing, performing feature extraction to obtain a feature vector, dividing the feature vector into a training set and a test set, and taking the training set as input to obtain a test set; and searching the penalty factor of the support vector machine model and the optimal value of the kernel parameter by using an improved sparrow search algorithm to obtain an ISSA-SVM identification and classification model, and testing the ISSA-SVM identification and classification model by using the test set. According to the method, the support vector machine is optimized based on the improved sparrow search algorithm, different types of vibration events can be accurately recognized, the real-time performance is high, the accuracy is high, and the pipeline monitoring efficiency is greatly improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of distributed optical fiber vibration sensing and artificial intelligence, in particular to a pipeline surrounding event pattern recognition method based on an improved sparrow search algorithm optimized support vector machine. BACKGROUND

[0002] In recent years, China's economic scale has been expanding, and people's living standards have improved significantly. This development has promoted the progress of various industries, and also driven the rapid increase in energy demand. With the continuous development of transportation, construction, electronics and other fields, especially the growth of energy-intensive industries, China's demand for various energy, especially oil, natural gas and other energy, has increased significantly. This increase in demand has placed higher demands on energy supply chains and transportation methods. Long-distance pipeline transportation methods have gradually become the main way of oil and gas transportation due to their efficiency, safety, and economy. Therefore, pipeline safety monitoring is an extremely important task. The emergence of distributed optical fiber vibration sensing technology provides a new way to achieve comprehensive monitoring of pipeline safety. By burying optical fiber sensors in the area where the pipeline is laid, real-time monitoring of pipeline vibrations can be achieved. Once an anomaly is detected, an alarm can be raised in a timely manner to reduce the loss of life and property caused by pipeline leaks and other abnormal events.

[0003] Traditional pipeline safety monitoring methods such as air sampling, in-pipe probe ball, and line observation can only cover a limited distance. In particular, in the monitoring of long-distance pipelines, multiple monitoring points often need to be installed, and rely on manual or periodic detection, with low monitoring frequency and efficiency, increasing the complexity and cost of monitoring. Pipeline safety monitoring based on distributed optical fiber vibration sensing technology can continuously monitor along the entire pipeline, and the optical fiber is not affected by corrosive substances, with lower installation and maintenance costs and 24-hour uninterrupted operation. The development of pattern recognition technology provides a basis for determining the type of pipeline surrounding events. Most optimization algorithms for support vector machines (SVM) use traditional intelligent optimization algorithms, which are prone to local optimization and have relatively slow convergence speed. The optimization of the SVM recognition model still needs to be improved. Using an improved optimization algorithm to adjust the parameters of the SVM model can improve the recognition accuracy of pipeline monitoring, thereby providing better assistance for subsequent pipeline monitoring work. SUMMARY

[0004] In order to solve the problems that the existing pipeline monitoring identification model is easy to fall into local optimum, the convergence speed is relatively slow, and the work efficiency and prediction accuracy of pipeline monitoring are reduced, the present application provides a pipeline surrounding event pattern recognition method based on improved sparrow search algorithm optimization support vector machine (SVM), which uses the improved sparrow search algorithm to find the global optimal value of the penalty factor and the kernel parameter to optimize the support vector machine, and obtains the ISSA-SVM identification classification model.

[0005] To solve the above problems, the present application adopts the following technical scheme:

[0006] The pipeline surrounding event pattern recognition method based on improved sparrow search algorithm optimization support vector machine comprises the following steps:

[0007] Step 1: Obtain the vibration data of the pipeline surrounding collected by the distributed optical fiber vibration sensing system, and pre-process the vibration data, then extract the features of the pre-processed data to obtain the feature vectors, and after the sample data set is formed by all the feature vectors, the sample data set is divided into a training set and a test set according to a preset ratio;

[0008] Step 2: Construct a support vector machine model;

[0009] Step 3: Take the training set as input, and use the improved sparrow search algorithm to iteratively find the optimal value of the penalty factor and the kernel parameter of the support vector machine model;

[0010] Step 4: Input the optimal value into the support vector machine model to obtain the ISSA-SVM identification classification model;

[0011] Step 5: Input the test set into the ISSA-SVM identification classification model for identification classification, and output the identification classification result.

[0012] The main advantages of the present application are as follows:

[0013] (1) The present application is based on a distributed optical fiber vibration sensing system, different types of vibration event data are collected, time domain features of events are extracted based on the difference of vibration events, and a sample data set is established as model input, which realizes more accurate classification of future vibration data, thereby improving the accuracy and reliability of pipeline vibration safety warning, improving the recognition degree when monitoring a large range of pipelines in complex environments, reducing the false alarm rate, and having great significance for reducing economic losses and protecting people's life and property safety;

[0014] (2) The improved sparrow search algorithm (ISSA) is used in the application to iteratively solve the penalty factor and the kernel parameter of the support vector machine (SVM) model, so that the recognition accuracy of the SVM model is improved, and the optimization ability of the optimization algorithm is also improved. The ISSA algorithm can significantly improve the problem that the sparrow search algorithm is prone to fall into a local optimal solution, and more quickly and efficiently find the optimal sparrow position and the optimal fitness value. Finally, the optimal parameters of the SVM, i.e., the optimal values of the penalty factor and the kernel parameter, are obtained through the optimal sparrow position, and an ISSA-SVM recognition classification model is obtained, so that the ISSA-SVM recognition classification model has higher recognition accuracy, and the ISSA-SVM recognition classification model has simple structure, high implementation efficiency and small error rate, and performs well in pipeline monitoring. BRIEF DESCRIPTION OF DRAWINGS

[0015] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0016] Figure 1 A flowchart of the pipeline surrounding event pattern recognition method based on the improved sparrow search algorithm optimization support vector machine is provided in the embodiments of the present application.

[0017] Figure 2 A principle diagram of the SVM is provided.

[0018] Figure 3 A flowchart of the improved sparrow search algorithm in the present application is provided. DETAILED DESCRIPTION

[0019] The technical solutions of the present application will be described in detail below with reference to the drawings and preferred embodiments.

[0020] In one embodiment, as shown in Figure 1 The present embodiment provides a pipeline surrounding event pattern recognition method based on an improved sparrow search algorithm optimization support vector machine, which specifically includes the following steps 1 to 5.

[0021] Step 1: Obtain the vibration data of the pipeline surrounding collected by the distributed optical fiber vibration sensing system, and pre-process the vibration data, then extract features from the pre-processed data, combine the extracted features to obtain a feature vector, and divide the sample data set into a training set and a test set according to a predetermined ratio, for example, a ratio of 6:4.

[0022] Firstly, the vibration data collected by the distributed optical fiber vibration sensing system is preprocessed, which includes data denoising processing and data normalization processing.

[0023] Data denoising processing: wavelet threshold denoising method is used for denoising processing of vibration data, which includes the following steps: (1) selecting appropriate wavelet basis function and decomposition level; (2) selecting a threshold and determining a threshold processing function for processing wavelet coefficients in the denoising process; (3) wavelet decomposition of vibration signal is carried out to obtain wavelet coefficients; (4) threshold processing is applied to wavelet coefficients to remove or reduce noise components; (5) through inverse wavelet transform, the signal is reconstructed to make the signal noise smaller.

[0024] Data normalization processing: in order to reduce the interference introduced by signal autocorrelation and alleviate the risk of overfitting, L2 norm normalization method is used to convert data to [0, 1] interval. After normalization processing, the data can significantly improve the feature extraction and recognition efficiency while retaining the original characteristics, thereby effectively improving the timeliness of recognition. The L2 norm normalization calculation formula is as follows:

[0025]

[0026] Where x i represents the i-th vector, x j represents the j-th vector, 1≤i,j≤n, n is the total number of vectors, the denominator represents L2 norm, that is, the square root of the sum of squares of each vector, x i ' represents the normalized i-th vector.

[0027] Feature extraction: this step extracts the characteristic indexes of each type of vibration signal through mathematical operation, the extracted time domain characteristic indexes include energy (E), amplitude (M), amplitude difference (MD), zero crossing rate (Z), pulse factor (I) and margin factor (L) of vibration signal, and the calculation formulas of each time domain characteristic index are as follows:

[0028]

[0029] In formula (2)-(7), s(i) represents the amplitude of the vibration signal sampling data point, and n represents the number of sampling data points of each sample. In formula (4), m represents the interval point number between the two points for calculating the amplitude difference, and m=1 is taken, that is, the amplitude difference between adjacent two points is calculated. In formula (5), sgn[·] represents the sign function, which is used to represent whether the vibration signal crosses the horizontal axis. In formula (6) and formula (7), max(s(i)) is the maximum amplitude of the sampling data point.

[0030] Step 2: construct a support vector machine model, such as Figure 2As shown, the circle and triangle symbols are used to represent the two types of training sample data respectively. The classification hyperplane H and the two parallel support hyperplanes H1 and H2 together constitute the classification model, where H1 and H2 pass through the sample points closest to H in their respective categories. These sample points are called support vectors, and the distance between H1 and H2 is called the classification interval. The optimal hyperplane needs to meet two conditions: one is to accurately distinguish the two types of samples, and the other is to maximize the classification interval. If the sample set is (X1, y1), (X2, y2),…, (X i ,y i ),X i ∈R n ,y i ∈{-1,1},R n represents n-dimensional Euclidean space, X i Represents a vector in n-dimensional Euclidean space. The goal of training is to construct a decision function that can correctly classify the training data as much as possible. Let the general form of the discriminant function of the hyperplane H be:

[0031] g(x)=W T X+b (8)

[0032] Then the classification surface equation is:

[0033] W T X+b=0 (9)

[0034] That is, when g(x) = 0, it means that the point is on the hyperplane.

[0035] If satisfied:

[0036] y i =+1,W T X i +b≥0 (10)

[0037] y i =-1,W T X i +b<0 (11)

[0038] A training set is said to be linearly separable.

[0039] Equations (10) and (11) can be combined into:

[0040] y i [(W T ·X i )+b / ≥0,i=1,2,...,l (12)

[0041] Wherein, l is an integer greater than 0.

[0042] Vector X to hyperplane W TThe distance of X+b=0 is To maximize the distance d, we want to minimize the value of ||W||.

[0043] |W T X+b| is normalized so that the support vectors X have |W T X+b| = 1, the distance of the support vectors to the classification hyperplane is The distance between H1 and H2 is

[0044] Support vector machines were originally designed as an algorithm for binary classification problems, but can be extended to multi-class classification tasks through structural improvements. A common implementation is to build a multi-class classification system by combining multiple binary classifiers, mainly using the "one-vs-rest" and "one-vs-one" strategies. In this step, a radial basis kernel function is used to build a multi-class SVM classification model based on the one-vs-one strategy. This strategy requires the construction of a binary classifier for each of the two classes in the set of m' class samples, a total of m'(m'-1) / 2 classifiers. In the classification stage, all binary classifiers vote on the test sample, and the final classification is the class with the most votes. In the experimental process, first, an initial feature set is constructed from the time domain features of each type of vibration signal, and then the selected feature vectors and corresponding class labels are input into the one-vs-one classifier for model training. In the classifier training process, the improved sparrow search algorithm is used to optimize the penalty factor and kernel parameter.

[0045] Step 3: Use the improved sparrow search algorithm to iteratively find the optimal values of the support vector machine model parameters (penalty factor and kernel parameter) with the training set as input.

[0046] Further, as shown in Figure 3 , the improved sparrow search algorithm specifically includes the following steps:

[0047] Step 3.1: Initialize relevant parameters, including the size of the sparrow population, the maximum number of iterations, the penalty factor C, and the kernel parameter σ;

[0048] Step 3.2: Use Chebyshev-LHS chaotic mapping to initialize the sparrow population and determine the initial values of the sparrow individual position vectors, i.e., the parameters C and σ;

[0049] Step 3.3: Calculate the initial fitness values of the sparrows;

[0050] Step 3.4: Sort the fitness values and select the optimal fitness value and the corresponding optimal sparrow position;

[0051] Step 3.5: Iteratively update the spatial positions of the discoverer, follower and alarmist, and update the current optimal fitness value and optimal sparrow position;

[0052] Step 3.6: Apply disturbance to the current optimal sparrow individual using the glowworm strategy, and then update the optimal sparrow position and optimal fitness value;

[0053] Step 3.7: Determine whether the current iteration number exceeds the maximum iteration number, if yes, end the iteration, and output the global optimal position of the sparrow individual, i.e. the optimal value of the parameters C and σ; otherwise, return to step 3.2 to re-initialize the sparrow population.

[0054] The specific steps of the traditional sparrow search algorithm are as follows:

[0055] In the process of sparrow searching for food, the sparrow population can be divided into three categories: discoverer, follower and alarmist. Let the population consisting of n sparrows be represented as:

[0056]

[0057] In equation (13), m is the dimension of the variables to be optimized.

[0058] The fitness value matrix of the sparrow population is as follows:

[0059]

[0060] In equation (14), F x The value of each row is the fitness value of the individual.

[0061] The position updating method of the discoverer is as follows:

[0062]

[0063] In equation (15), t is the current iteration number; is the position of the ith sparrow in the jth dimension in the tth generation, j = 1, 2,..., m; iter max is the maximum iteration number; α ∈ (0, 1], is a random number; R2 ∈ [0, 1], ST ∈ [0.5, 1.0] is the safety threshold; Q is a random number subject to [0, 1] normal distribution; L is a 1 × m matrix, and all elements are 1.

[0064] The position updating method of the follower is as follows:

[0065]

[0066] In equation (16): is the optimal position of the discoverer in the t+1th generation sparrow population; is the global worst position of the tth generation of sparrows; A is a 1xm matrix, each element of the matrix is randomly taken as 1 or -1, and the matrix A satisfies A + = A T (AA T ) -1 .

[0067] The scout position updating mode is as follows:

[0068]

[0069] In formula (17): is the global optimal position in the tth generation of the sparrow population; β ~ N(0, 1), k is a random number in [0, 1], and also represents the moving direction of the sparrow; β and k are used to adjust the step length; f i is the fitness value of the current sparrow individual, and f g and f w are the current global optimal fitness value and the current global worst fitness value respectively; ε is a constant not equal to zero, which is used to avoid the denominator equal to 0.

[0070] In this embodiment, a mapping method combining Chebyshev chaotic mapping and Latin hypercube sampling (LHS) (Chebyshev-LHS) is proposed to improve the uniformity of population space distribution, an adaptive inertia weight factor is introduced to improve the scout position updating formula, so that the search range can be expanded and the convergence speed can be improved, a golden sine algorithm is introduced to improve the follower position updating formula, in the case of fast convergence of the algorithm, the sparrow with low fitness can improve the tracking speed to obtain food, and the global exploration and local development capabilities are considered, in order to avoid the improved sparrow search algorithm falling into a local solution, a firefly disturbance is introduced after updating the optimal sparrow position, so as to improve the optimization precision. The improved sparrow search algorithm in this embodiment includes the following processes:

[0071] Because the traditional sparrow search algorithm uses a random initialization population method, aggregation and uneven distribution may occur, which is not conducive to the wide search of the algorithm. Chebyshev mapping is a typical representative of chaotic mapping, in order to meet the demand of uniform distribution of sequence in optimization theory, the present application proposes a mapping method combining Chebyshev chaotic mapping and LHS (Chebyshev-LHS) to improve the uniformity of population space distribution. The combination of the pseudo-random sequence of Chebyshev chaotic mapping and the layered uniformity of LHS introduces the sampling value of LHS into the result of Chebyshev chaotic mapping, realizes efficient population diversity and global coverage, and the mathematical expression is:

[0072] x t+1= cos[t arccos(x t )] lhs(i,j), -1≤x t ≤1 (18)

[0073] y t = x t+1 , t = 0, 1, 2, …, n (19)

[0074]

[0075] In formula (18)-(20), x t+1 is the value generated by the chaotic sequence of the t+1 generation, lhs(i,j) is the value of the i th individual, the j th dimension in the Latin hypercube sampling matrix, y t+1 is the sparrow individual after the t+1 mapping iteration, and M is an arbitrary constant.

[0076] The specific process of initializing sparrow population by using Chebyshev-LHS chaotic mapping is as follows:

[0077] Step 3.2.1: First, generate an m-dimensional vector representing the initial individual randomly in [-1, 1], and then substitute it into formula (18) to derive n-1 new individuals for each dimension;

[0078] Step 3.2.2: Then substitute the derived new individuals into formula (20) to map the variable value generated by the Chebyshev-LHS chaotic mapping to the sparrow individual.

[0079] An adaptive inertia weight factor is introduced, and its formula is as follows:

[0080]

[0081] In formula (21), f is the current sparrow fitness value, f is the average fitness value of all current sparrows, f is the minimum fitness value, ω min and ω max are the minimum inertia coefficient and the maximum inertia coefficient respectively, and ω min is generally taken as 0.45, and ω max is taken as 0.95. The improved follower position update formula of introducing the adaptive inertia weight factor is as follows:

[0082]

[0083] The improved follower position update formula of introducing the golden sine algorithm is as follows:

[0084]

[0085] In formula (23)-(24), r1∈(0, π)∪(π, 2π), r2∈(0, π); Q is a random number subject to [0, 1] normal distribution; L is a 1×m matrix, and all elements are 1; c1, c2 are golden sine segmentation coefficients; λ is a golden segmentation number, The initial values of a and b are set as π / 2 and -π / 2, and then the values of a and b change with the target value (i.e. the optimal value to be found), and c1 and c2 are updated accordingly.

[0086] In order to solve the problem that the original sparrow search algorithm is easy to fall into local optimum in the process of finding the optimal value, after updating the current optimal sparrow position and optimal fitness value, a firefly disturbance strategy is introduced, which uses the firefly strategy to disturb the current optimal sparrow individual. The strategy simulates the luminescent characteristics of fireflies, simulates the attraction and movement process of fireflies in the search and optimization process, disturbs and updates the position of the sparrow individual, compares the disturbed fitness value with the value before disturbance, and if the fitness value is better, the position is updated and the fitness value is updated. The attraction degree of firefly is described as follows:

[0087]

[0088] In formula (25), β0 is the maximum attraction degree, γ is the light intensity absorption coefficient, r ij is the distance between fireflies i and j.

[0089] The position update of firefly i attracted and moved to firefly j is described as follows:

[0090] x i = x i + β * (x j - x i ) + μ * (rand-1 / 2) (26)

[0091] In formula (26), x i and x j are the spatial positions of fireflies i and j, respectively, μ∈[0, 1] is a step factor, and rand is a random number subject to uniform distribution on [0, 1].

[0092] Step 4: input the optimal values of the penalty factor and the kernel parameter found into the SVM model to obtain an ISSA-SVM identification classification model, which can distinguish different event types acting on the pipeline with high accuracy.

[0093] Step 5: Model testing: input the test set into the trained ISSA-SVM identification classification model for identification classification, and output the identification classification result. The evaluation indexes adopted when testing the ISSA-SVM identification classification model include average identification accuracy rate (AIAR), intrusion detection rate (IDR), human-made event classification accuracy rate (HECR), and identification accuracy rate for each event category (IARE). After testing, the ISSA-SVM identification classification model can be directly used for pattern recognition of pipeline peripheral vibration events.

[0094] An experiment of monitoring pipeline leakage is given below to prove the effectiveness of the present application, and the specific steps are as follows:

[0095] Step 1: Data acquisition and preprocessing. Five events, pipeline leakage (recorded as event A), artificial oil theft (recorded as event B), vehicle passing (recorded as event C), human walking (recorded as event D), and quiet state (E), were manufactured in an experimental base. The distributed fiber vibration sensing system was used to collect vibration signal data, 300 samples for each event. After data denoising and data normalization processing, the time domain vibration signal feature vector was extracted, and the training set and data set were divided according to the ratio of 6:4.

[0096] Step 2: Constructing a support vector machine model.

[0097] Step 3: Taking the training set as input, the improved sparrow search algorithm is used to iteratively find the optimal value of the parameters (penalty factor and kernel parameter) of the support vector machine model.

[0098] Step 4: Input the optimal value found into the SVM model to obtain the ISSA-SVM identification classification model.

[0099] Step 5: Input the test set into the trained ISSA-SVM identification classification model for identification classification, and output the identification classification result. The evaluation indexes adopted when testing include average identification accuracy rate (AIAR), intrusion detection rate (IDR), human-made event classification accuracy rate (HECR), and identification accuracy rate for each event category (IARE).

[0100] The improved sparrow search algorithm and the sparrow search algorithm optimized support vector machine model are trained on the same data set, and the related evaluation indexes are compared, and the comparison results are shown in Tables 1 and 2.

[0101] Table 1 Improved sparrow search algorithm optimized support vector machine model

[0102]

[0103] Table 2 Sparrow search algorithm optimized support vector machine model

[0104]

[0105]

[0106] As can be seen from Tables 1 and 2, the pipeline peripheral event pattern recognition method based on the improved sparrow search algorithm optimized support vector machine proposed in the application increases the average recognition accuracy (AIAR) by 4.17 points, increases the intrusion detection rate (IDR) by 6.67 percent points, increases the human event classification accuracy (HECR) by 5 percent points, and increases the recognition accuracy of each type of event (IARE) by 5.83, 7.5, 2.5, 4.17 and 0.84 percent points, respectively.

[0107] The pipeline peripheral event pattern recognition method based on the improved sparrow search algorithm optimized support vector machine proposed in the application improves the traditional sparrow search algorithm by using an adaptive inertia weight factor and a golden sine algorithm, improves the richness of the population and the convergence speed, introduces firefly disturbance after updating the optimal sparrow position, further improves the optimization precision, optimizes the hyperparameters of SVM, and is more beneficial to the accurate prediction of SVM.

[0108] In summary, the improved sparrow search algorithm optimized SVM model proposed in the application can improve the accuracy of identification of different event types acting on the pipeline, and provides more reliable technical support for subsequent pipeline monitoring work.

[0109] The technical features of the above-described embodiments can be combined arbitrarily, and to make the description concise, all possible combinations of the technical features in the above-described embodiments are not described, however, as long as the combinations of the technical features do not exist contradictory, it should be considered that they are within the scope of the present application.

[0110] The above embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as a limitation on the scope of the patent. It should be noted that for ordinary skilled persons in the art, without departing from the concept of the present application, several modifications and improvements can be made, which are within the scope of protection of the present application. Therefore, the protection scope of the patent of the present application should be subject to the appended claims.

Claims

1. A pipeline peripheral event pattern recognition method based on an improved sparrow search algorithm and optimized support vector machine is characterized by: The following steps are involved: Step 1: Obtain vibration data around the pipeline collected by a distributed optical fiber vibration sensing system, preprocess the vibration data, and then perform feature extraction on the preprocessed data to obtain feature vectors. After all feature vectors are combined into a sample data set, the sample data set is divided into a training set and a test set according to a preset ratio; Step 2: Build a support vector machine model; Step 3: Using the training set as input, an improved sparrow search algorithm is used to iteratively search for optimal values ​​of the penalty factor and kernel parameters of the support vector machine model; Step 4: Input the optimal value into the support vector machine model to obtain the ISSA-SVM recognition and classification model; Step 5: Input the test set into the ISSA-SVM recognition and classification model for recognition and classification, and output the recognition and classification results.

2. The pipeline peripheral event pattern recognition method based on the improved sparrow search algorithm and support vector machine optimization according to claim 1 is characterized in that: The evaluation indicators used when testing the ISSA-SVM recognition and classification model include average recognition accuracy, intrusion detection rate, human-caused event classification accuracy, and recognition accuracy of each type of event.

3. The pipeline peripheral event pattern recognition method based on the improved sparrow search algorithm and support vector machine optimization according to claim 1 or 2 is characterized in that: Preprocessing the vibration data includes performing data denoising processing on the vibration data using a wavelet threshold denoising method and performing data normalization processing on the vibration data using an L2 norm normalization method.

4. The pipeline peripheral event pattern recognition method based on the improved sparrow search algorithm and support vector machine optimization according to claim 1 or 2 is characterized in that: In step 1, the extracted signal features include energy, amplitude, amplitude difference, zero-crossing rate, pulse factor and margin factor of the vibration signal.

5. The pipeline peripheral event pattern recognition method based on the improved sparrow search algorithm and support vector machine optimization according to claim 1 or 2 is characterized in that: Step 3 includes the following steps: Step 3.1: Initialize relevant parameters, including the sparrow population size, maximum number of iterations, penalty factor C, and kernel parameter σ; Step 3.2: Initialize the sparrow population using the Chebyshev-LHS chaotic map and determine the individual sparrow position vectors, i.e., the initial values ​​of the parameters C and σ; Step 3.3: Calculate the initial fitness value of the sparrow individual; Step 3.4: Sort the fitness values ​​and select the optimal fitness value and the corresponding optimal sparrow position; Step 3.5: Iteratively update the spatial positions of the discoverer, follower, and sentinel, and update the optimal fitness value and optimal sparrow position; Step 3.6: Use the firefly strategy to perturb the current optimal sparrow individual, and then update the optimal sparrow position and optimal fitness value; Step 3.7: Determine whether the current number of iterations exceeds the maximum number of iterations. If so, the training ends and the global optimal position of the sparrow individual is output, that is, the optimal values ​​of parameters C and σ; otherwise, return to step 3.

2.

6. The pipeline peripheral event pattern recognition method based on the improved sparrow search algorithm and support vector machine optimization according to claim 5 is characterized in that: The process of initializing the sparrow population using Chebyshev-LHS chaotic mapping includes: Step 3.2.1: Randomly generate an m-dimensional vector in [-1, 1] to represent the initial individual, substitute it into formula (18) and iterate each dimension to derive n-1 new individuals, where formula (18) is as follows: x t+1 =cos[t·arccos(x t )]lhs(i,j),-1≤x t ≤1 (18) Step 3.2.2: Substitute the derived new individual into formula (20) and map the variable values ​​generated by the Chebyshev-LHS chaotic map to the sparrow individual, where formula (20) is as follows: y t =x t+1 ,t=0,1,2,…,n (19) In formulas (18)-(20), x t+1 is the value generated by the t+1th generation chaotic sequence, lhs(i,j) is the value of the i-th individual and j-th dimension in the Latin hypercube sampling matrix, y t+1 is the sparrow individual after the t+1th mapping iteration, and M is an arbitrary constant.

7. The pipeline peripheral event pattern recognition method based on the improved sparrow search algorithm and support vector machine optimization according to claim 5 is characterized in that: An adaptive inertia weight factor is introduced to improve the finder position update formula, and the finder's spatial position is iteratively updated using the improved finder position update formula. The improved finder position update formula is as follows: in, is the adaptive inertia weight factor; is the current sparrow fitness value; is the average fitness value of all sparrows at present; is the minimum fitness value; ω min and ω max are the minimum and maximum inertia coefficients respectively; t is the current iteration number; is the position of the i-th sparrow in the t-th generation in the j-th dimension, j = 1, 2, ..., m; iter max is the maximum number of iterations; α∈(0,1); R2∈[0,1], ST∈[0.5,1.0] is the safety threshold; Q is a random number that obeys the [0,1] normal distribution; L is a 1×m matrix, and all elements are 1.

8. The pipeline periphery event pattern recognition method based on the improved sparrow search algorithm and support vector machine optimization according to claim 5 is characterized in that: The golden sine algorithm is introduced to improve the follower position update formula. The improved follower position update formula is used to iteratively update the spatial position of the follower. The improved follower position update formula is as follows: Among them, r1∈(0,π)∪(π,2π), r2∈(0,π); c1, c2 are the golden sine division coefficients; λ is the golden section number, The initial values ​​of a and b are set to π / 2 and -π / 2, and then the values ​​of a and b change with the target value, and c1 and c2 are updated accordingly; is the optimal position of the discoverer in the t+1 generation sparrow population; is the global worst position of the t-th generation sparrow; A is a 1×m matrix, each element in the matrix is ​​randomly 1 or -1, and the matrix A satisfies A + =A T (AA T ) -1 .

9. The pipeline periphery event pattern recognition method based on the improved sparrow search algorithm and support vector machine optimization according to claim 8 is characterized in that: Minimum inertia coefficient ω min Take 0.45, the maximum inertia coefficient ω max Take 0.95.

Images