Method for predicting external corrosion rate of buried pipeline based on IGWO-BP neural network

Abstract

The invention discloses a buried pipeline external corrosion rate prediction method based on an IGWO-BP neural network, and relates to the buried pipeline corrosion rate prediction field, and the method comprises the steps: obtaining experiment data, and carrying out the normalization processing to obtain sample data; constructing a BP neural network model, and determining the number of nodes of a network hidden layer according to the topological structure of the BP neural network; initializing parameters a, A and C by adopting an improved grey wolf algorithm in combination with a BP neural network topological structure, and randomly generating wolf pack heavy grey wolf position information; calculating fitness values based on the fitness function, sorting the fitness values in a descending order, dividing grey wolves into alpha wolves, beta wolves, delta wolves and omega wolves, and updating the positions and parameters of the grey wolves; iterating the improved grey wolf algorithm to obtain the position of a head wolf alpha wolf and a corresponding minimum error of a target result; and bringing the initial weight and the threshold value corresponding to the position of the result head wolf alpha wolf back to the neural network for judgment to obtain an output prediction result. The prediction precision of the model is improved.

Description

Technical Field

[0001] The invention relates to the field of buried pipeline corrosion rate prediction, in particular to a buried pipeline external corrosion rate prediction method based on IGWO-BP neural network. Background Art

[0002] The water supply network is an important lifeline of the city and an important infrastructure to ensure the normal operation of the city. As the operation time of buried pipelines increases, accidents such as perforation, leakage, and cracking also increase. In order to avoid burst pipes and leakage accidents that affect people's normal life and safety, people are paying more and more attention to the management and maintenance of buried pipelines.

[0003] There are many studies on pipeline corrosion, among which the prediction of pipeline corrosion rate has always been the focus of research.

[0004] So far, Chinese and foreign scholars have conducted a lot of research on pipeline corrosion prediction methods, which can be mainly summarized as probability statistics, reliability function analysis, grey system theory prediction, machine learning prediction, etc. Among them, machine learning prediction plays an important role in the prediction of pipeline corrosion rate.

[0005] The machine learning prediction method predicts unknown data by establishing relevant models based on learning a large amount of experimental data. It does not need to explore the specific mechanism that affects the pipeline corrosion rate. It has been widely used in the prediction of pipeline corrosion rate.

[0006] The Gray Wolf Optimization (GWO) algorithm and BP neural network are both important algorithms in the field of computer artificial intelligence. As a typical biomimetic intelligent optimization algorithm inspired by biology, the mathematical model of GWO is mainly derived from the simulation of the hierarchy and predation and hunting behavior of gray wolf packs in nature. Due to its simplicity, fewer control parameters and easy implementation, GWO has been widely used to solve different optimization problems.

[0007] BP neural network is a multi-layer feedforward network trained by the error back propagation algorithm. It is widely used in many fields because of its strong nonlinear mapping ability, strong generalization ability, high self-learning and self-adaptation ability, simple principle and high fault tolerance.

[0008] Taking into account the characteristic of the Grey Wolf Optimization Algorithm that it can seek the global optimal solution, the disadvantage that a single BP neural network is prone to fall into the local optimum is solved by optimizing the weights and thresholds of the BP neural network.

[0009] Therefore, there is an urgent need for a joint algorithm that uses the improved grey wolf algorithm to optimize the BP neural network to predict the pipeline corrosion rate. Summary of the invention

[0010] In order to solve the above problems, this application proposes a method for predicting the external corrosion rate of buried pipelines based on IGWO-BP neural network. The invention takes the external corrosion rate of buried pipelines as the objective function, and takes water content, salt content, C1 - Content, SO4 2- Eleven influencing factors such as content, pH value, soil resistivity, and redox potential are taken as independent variables. The independent variables and objective functions are analyzed and normalized, and then the mapping relationship between the independent variables and the objective function is established. The IGWO-BP neural network is continuously trained and learned, and finally the purpose of calculating the external corrosion rate prediction of buried pipelines is achieved.

[0011] The present application mentions a method for predicting the external corrosion rate of buried pipelines based on IGWO-BP neural network, and the specific steps are as follows: Step 1: Obtain the experimental data of the external corrosion rate of the buried pipeline and perform normalization processing to obtain sample data, and divide the sample data into groups to obtain training group data and test group data; Step 2: Construct a BP neural network model and determine the number of nodes in the hidden layer of the network according to the topological structure of the BP neural network itself; Step 3: Use the improved gray wolf algorithm combined with the BP neural network topology structure to initialize the parameters a, A, and C, and randomly generate the position information of the gray wolves in the wolf pack; Step 4: Determine the transfer function, training function and fitness function of the BP neural network model; Step 5: Calculate the fitness value based on the fitness function and sort it in descending order, divide the gray wolves into α wolf, β wolf, δ wolf, ω wolf, and update the position and parameters of the gray wolves; Step 6: Iterate the improved gray wolf algorithm to obtain the position of the target leader wolf α and the corresponding minimum error; Step 7: Bring the initial weight and threshold corresponding to the position of the leader wolf α back to the neural network, and make a judgment to obtain the output prediction result, that is, the predicted buried pipeline corrosion rate.

[0012] Preferably, in step 3, the improved gray wolf algorithm is combined with the BP neural network topology structure to initialize the parameters a, A, and C, and the specific content of randomly generating the position information of the heavy gray wolves in the wolf pack is: The gray wolf population is initialized, and the individual dimensions of the gray wolves, the population size, the maximum number of iterations, and the upper and lower bounds of the wolf search space are determined according to the number of weights and thresholds to be optimized.

[0013] Preferably, the improved grey wolf algorithm specifically includes: Optimizing population initialization to obtain an initial gray wolf population, wherein the process of optimizing population initialization is performed using a Circle chaos strategy; An improved nonlinear convergence method is used to optimize the convergence factor in the Grey Wolf Algorithm. Introducing weights W Optimize the global search capability of the Grey Wolf Algorithm.

[0014] Preferably, the process of optimizing population initialization is expressed as: ; In the formula, mod is the remainder function, C ( t )and C ( t +1) represent the t Second and t +1 Circle chaos iteration value, initial value C (0) is a random number between 0 and 1.

[0015] Preferably, the convergence factor a The calculation formula is as follows: ; in, e is a natural constant; is the current iteration number; M is the maximum number of iterations.

[0016] Preferably, the weight W The expression is as follows: ; Where: W 1. W 2. W 3 are the learning rate weights of ω wolf to α wolf, β wolf and δ wolf respectively; X 1. X 2. X 3 are the updated positions of wolf α, wolf β and wolf δ respectively.

[0017] Preferably, in step 4, the network transfer function is a logsig() type function, the network training function is tansig(), and the fitness function is the mean square error between the predicted output value and the measured value.

[0018] Preferably, in step 6, the specific contents of iterating the improved grey wolf algorithm to obtain the position of the target result leader wolf α and the corresponding minimum error are: Determine whether the set error or the maximum number of iterations is met. If not, repeat steps 5-6 until the conditions are met. Obtain the position of the target leader wolf α and the corresponding minimum error.

[0019] Preferably, the specific content of the output prediction result obtained by judging in step 7 is: Calculate the output error of the BP neural network; If the output error reaches the preset learning accuracy or the maximum number of training times, the prediction result can be output after denormalization; If the output error does not reach the preset learning accuracy or the maximum number of training times, first update the weights and thresholds, and then repeat step 7 until the output error reaches the preset learning accuracy or the maximum number of training times. After denormalization, output the prediction result. α wolf, β wolf, and δ wolf are the gray wolves corresponding to the first, second, and third in fitness value order, and ω wolf is the remaining gray wolves.

[0020] In summary, the present invention is a method for predicting the external corrosion rate of buried pipelines based on the IGWO-BP neural network. The present invention mainly takes three improvement measures on the gray wolf algorithm. Step 1: Optimize population initialization. The present invention introduces Circle chaotic mapping to initialize the gray wolf group. Compared with the traditional random search strategy to generate the initial population, the chaotic search strategy to generate the initial population can better cover the search space and ensure the distribution of the initial population in the search space. Step 2: Optimize the iteration coefficient. The present invention uses an improved nonlinear convergence method. Step 3: The present invention introduces weights W , optimize the global search ability of the algorithm. The improved gray wolf algorithm can better optimize the weights and thresholds of the prediction model. Finally, the improved gray wolf optimized BP neural network is used to predict the corrosion rate of buried pipelines. It has the advantages of high calculation accuracy, strong optimization ability, simple operation, etc. It can calculate the external corrosion rate of buried pipelines under different corrosion conditions.

[0021] Compared with traditional technologies, the present invention effectively improves the computational efficiency of the model, and uses the IGWO method to optimize model parameters, further improving the prediction accuracy of the model.

[0022] This application aims at the shortcomings of the traditional gray wolf algorithm, introduces a new nonlinear update mechanism and population update mechanism, and proposes an improved gray wolf optimization algorithm (IGWO) that integrates multiple strategies. Secondly, IGWO is applied to the weight and threshold search of the BP neural network, forming a BP neural network model based on IGWO (IGWO-BP).

[0023] In the specific operation, the improved gray wolf algorithm is used to optimize the weights and thresholds of the prediction model, and then the test samples are input into the corrosion prediction model of the trained BP neural network to obtain the predicted value. When the error between the predicted value and the measured value is less than or equal to the preset value, it means that the corrosion prediction model of the trained BP neural network is correct.

[0024] In actual prediction, the detection data of the external corrosion of the buried pipeline to be predicted can be input into the trained corrosion prediction model to obtain the external corrosion rate of the buried pipeline. The operation is simple, the optimization ability is strong, and the calculation accuracy is high.

[0025] The technical method of the present invention is further described in detail below through the drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 This is a flow chart of the external corrosion rate prediction method for buried pipelines based on IGWO-BP neural network; Figure 2 It is the fitness curve of the external corrosion rate prediction method of buried pipeline based on IGWO-BP neural network; Figure 3 Compare the predicted output values ​​of pipeline liquid holdup predicted by BP neural network, GWO-BP neural network and IGWO-BP neural network with the actual experimental data values; Figure 4 This is a comparison of the errors in predicting pipeline liquid holdup using BP neural network, GWO-BP neural network and IGWO-BP neural network. DETAILED DESCRIPTION

[0027] The technical method of the present invention is further described below by means of the accompanying drawings and embodiments. It should be noted that unless otherwise specifically stated, the relative arrangement of the components and steps, numerical expressions and numerical values ​​described in these embodiments do not limit the scope of the present application.

[0028] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the present application, its application, or uses.

[0029] Technologies, systems, and devices known to ordinary technicians in the relevant art may not be discussed in detail, but where appropriate, the technologies, systems, and devices should be considered part of the specification.

[0030] In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not limiting. Therefore, other examples of the exemplary embodiments may have different values.

[0031] Unless otherwise defined, technical or scientific terms used in the present invention shall have the common meanings understood by one having ordinary skills in the field to which the present invention belongs.

[0032] Embodiment 1 The present invention provides a method for predicting the external corrosion rate of buried pipelines based on IGWO-BP neural network. The specific calculation process is shown in Figure 1 , the steps include: Step 1: Obtain the experimental data of the external corrosion rate of the buried pipeline and perform normalization processing to obtain sample data, and divide the sample data into groups to obtain training group data and test group data; Step 2: Construct a BP neural network model and determine the number of nodes in the hidden layer of the network according to the topological structure of the BP neural network itself; Step 3: Use the improved gray wolf algorithm combined with the BP neural network topology structure to initialize the parameters a, A, and C, and randomly generate the position information of the gray wolves in the wolf pack; A is the convergence factor a. The convergence factor is a numerical factor that participates in regulating the rhythm of parameter updates during neural network training, prompting the model to move toward the minimum value of the loss function (converge) faster and more stably. It plays a role as a key component for adjusting the learning rate.

[0033] A is the activation function related parameter, activation function variant: Common activation functions of BP neural network include Sigmoid, ReLU, etc.

[0034] C is the regularization coefficient, Regularization Coefficient, Regularization principle: In order to avoid overfitting of BP neural network, regularization means are often introduced, and C appears in the regularization term.

[0035] Preferably, in step 3, the improved gray wolf algorithm is combined with the BP neural network topology structure to initialize the parameters a, A, and C, and the specific content of randomly generating the position information of the heavy gray wolves in the wolf pack is: The gray wolf population is initialized, and the individual dimensions of the gray wolves, the population size, the maximum number of iterations, and the upper and lower bounds of the wolf search space are determined according to the number of weights and thresholds to be optimized.

[0036] Preferably, the improved grey wolf algorithm specifically includes: Optimizing population initialization to obtain an initial gray wolf population, wherein the process of optimizing population initialization is performed using a Circle chaos strategy; The more uniform the distribution of the initial population in the solution space, the greater the probability that the algorithm will eventually escape the extreme value and obtain the global optimal value. When the distribution of the initial gray wolf group generated by the gray wolf optimization algorithm in the search space is more uniform, the greater the probability that the gray wolf searches for and captures prey, that is, the greater the probability of solving the global optimal value. Compared with the traditional random search strategy to generate the initial population, the chaotic search strategy to generate the initial population can better cover the search space and ensure the distribution of the initial population in the search space. Therefore, the present invention introduces a chaotic map to initialize the gray wolf group. However, the initial population generated by different chaotic maps has an impact on the performance of the algorithm.

[0037] Preferably, the process of optimizing population initialization is expressed as: ; In the formula, mod is the remainder function, C ( t )and C ( t +1) represent the t Second and t +1 Circle chaos iteration value, initial value C (0) is a random number between 0 and 1.

[0038] An improved nonlinear convergence method is used to optimize the convergence factor in the Grey Wolf Algorithm. In the Grey Wolf Algorithm, the original convergence factor a It decreases linearly until it reaches 0 during the iteration process. However, in the actual optimization process, the gray wolf algorithm does not converge linearly, so the original convergence factor a It cannot represent the optimization process of the algorithm well.

[0039] Preferably, the convergence factor a The calculation formula is as follows: ; in, e is a natural constant; l is the current iteration number; M is the maximum number of iterations.

[0040] Introducing weights W Optimize the global search capability of the Grey Wolf Algorithm.

[0041] In the optimization process of the grey wolf algorithm, α wolf is not necessarily always at the global optimal point. After continuous iterations, the algorithm is likely to fall into the local optimal point.

[0042] Preferably, the weight W The expression is as follows: ; Where: W 1. W 2. W 3 are the learning rate weights of ω wolf to α wolf, β wolf and δ wolf respectively; X 1. X 2. X 3 are the updated positions of wolf α, wolf β and wolf δ respectively.

[0043] Step 4: Determine the transfer function, training function and fitness function of the BP neural network model; Preferably, in step 4, the network transfer function is a logsig() type function, the network training function is tansig(), and the fitness function is the mean square error between the predicted output value and the measured value.

[0044] Step 5: Calculate the fitness value based on the fitness function and sort it in descending order, divide the gray wolves into α wolf, β wolf, δ wolf, ω wolf, and update the position and parameters of the gray wolves; Step 6: Iterate the improved gray wolf algorithm to obtain the position of the target leader wolf α and the corresponding minimum error; Preferably, in step 6, the specific contents of iterating the improved grey wolf algorithm to obtain the position of the target result leader wolf α and the corresponding minimum error are: Determine whether the set error or the maximum number of iterations is met. If not, repeat steps 5-6 until the conditions are met. Obtain the position of the target leader wolf α and the corresponding minimum error.

[0045] Step 7: Bring the initial weights and thresholds corresponding to the position of the leader wolf α back to the neural network for judgment and output prediction results.

[0046] Preferably, the specific content of the output prediction result obtained by judging in step 7 is: Calculate the output error of the BP neural network; If the output error reaches the preset learning accuracy or the maximum number of training times, the prediction result can be output after denormalization; If the output error does not reach the preset learning accuracy or the maximum number of training times, first update the weights and thresholds, and then repeat step 7 until the output error reaches the preset learning accuracy or the maximum number of training times. After denormalization, output the prediction result.

[0047] Embodiment 2 The external corrosion rate data of buried pipelines and the data affecting the corrosion rate are obtained as sample data.

[0048] In this embodiment, the corrosion rate data and other detection data affecting the corrosion rate include water content, salt content, C1 - Content, SO4 2- 11 influencing factors including content, pH value, soil resistivity, and redox potential.

[0049] The sample data is normalized to obtain normalized sample data. The normalization formula is: ; in, is the eigenvalue before normalization, is the normalized eigenvalue, are the maximum and minimum eigenvalues ​​before normalization, is the eigenvalue number, indicating the eigenvalues, ; It is the expected maximum and minimum value after normalization, and generally defaults to 1 and -1.

[0050] Determine the initial structure of the BP neural network. The BP neural network model with a single hidden layer can approximate any nonlinear mapping relationship. This application adopts a 3-layer network.

[0051] The normalized influencing factor data, i.e., the sample data, is used as the input of the BP neural network input layer, which has eleven neurons; the corrosion rate is used as the output of the output layer, which has only one neuron.

[0052] The number of nodes in the hidden layer is determined by the empirical formula for calculating the number of elements in the hidden layer. The empirical formula is as follows: ; in, is the number of nodes in the hidden layer of the BP neural network, is the number of nodes in the output layer, is the number of nodes in the input layer of the BP neural network, and c is a constant between 1 and 10. From the empirical formula, we know that when When , the optimal number of hidden layer nodes is calculated to be 14.

[0053] Divide the training set and the test set into a ratio of 3:1, randomly select 15 groups of data as the training set for training the model, and the remaining 5 groups of data as the test set for model prediction and error checking.

[0054] Determine the transfer function, training function and fitness function of the neural network. The network transfer function is a logsig() type function, the network training function is tansig(), and the fitness function is the mean square error between the predicted output value and the measured value.

[0055] Gray wolf population initialization: According to the weights and thresholds to be optimized, the individual dimensions of the gray wolf, the population size, the maximum number of iterations, the upper and lower bounds of the wolf search space are determined, and the wolf location information is randomly generated in combination with the initial parameters a, A, and C of the network topology structure. The present invention adopts the Circle chaos strategy to generate the initial gray wolf population. The Circle chaos strategy formula is as follows: ; In the formula, mod is the remainder function, C ( t )and C ( t +1) represent the t Second and t+1 Circle chaos iteration value, initial value C (0) is a random number between 0 and 1.

[0056] Iterative updates: The Gray Wolf Optimizer (GWO) is a typical bionic intelligent optimization algorithm inspired by biological factors. Its mathematical model is mainly derived from the simulation of the hierarchy and predation and hunting behavior of gray wolf packs in nature.

[0057] In the algorithm's search process, the three best gray wolves at each iteration are defined as α wolf, β wolf and δ wolf, which guide the remaining gray wolves ω to search for prey in the search space. The optimization process of the gray wolf algorithm is to surround the prey, hunt and track, and attack the prey.

[0058] The model of gray wolf hunting prey can be defined as follows: ; ; Where: D is the distance between the gray wolf and its prey; t is the number of iterations, is the maximum number of iterations; X P ( t ) is the location of the prey, X ( t +1) is the t +The position of the gray wolf at iteration 1; A and C are coefficient vectors respectively.

[0059] in, The present invention uses an improved nonlinear convergence method for the convergence factor. The calculation formula for the gray wolf's position update formula is as follows: ; Where: is a natural constant; l is the current iteration number; M is the maximum number of iterations.

[0060] Gray wolves have the ability to collaborate to surround prey, such as Figure 2 As shown in the fitness curve, the gray wolves with the best fitness, α wolf, β wolf and δ wolf, lead the rest of the gray wolves to update their positions.

[0061] The prey tracking model of an individual gray wolf is as follows: ; The above formula represents the distance between other wolves ω and α, β and δ, where: D α ,D β , D δ Respectively represent the distances between wolf α, wolf β and wolf δ and other gray wolf individuals ω; X α , X β , X δ Represent the current positions of α wolf, β wolf and δ wolf respectively; C 1 , C 2 , C 3 is a random vector.

[0062] ; The above indicates that the positions of other gray wolves are updated according to the positions of α wolf, β wolf and δ wolf. Where: X 1 , X 2 , X 3 They represent the step lengths of the gray wolf ω individual updating towards the wolf α, wolf β, and wolf δ respectively.

[0063] ; The final updated position of the gray wolf ω individual is determined by the above formula.

[0064] Among them, in the optimization process of the gray wolf algorithm, the α wolf is not necessarily always at the global optimal point. After continuous iterations, the algorithm is likely to fall into the local optimal point.

[0065] This application introduces weight W , optimize the global search capability of the algorithm. Weight W The calculation formula is: ; Where: W 1. W 2. W 3 are the learning rate weights of ω wolf to α wolf, β wolf and δ wolf respectively; X 1. X 2. X 3 are the updated positions of wolf α, wolf β and wolf δ respectively.

[0066] Wolf Pack Updated Locations The calculation formula is: ;

[0067] After continuous iterations and updates, the best position for the prey will be determined, which is the position of the alpha wolf.

[0068] Output the weights and thresholds determined by the position of the best individual.

[0069] The optimal weights and thresholds are reassigned to the BP neural network, and the test samples are substituted into the BP neural network to compare the calculated output results with the actual results. If the calculation error is satisfied, the model is successfully established. If it is not satisfied, the training and prediction are repeated until the calculation error meets the design requirements.

[0070] The BP neural network at this time is the optimal BP neural network optimized by the IGWO algorithm. The external corrosion rate data of the buried pipeline predicted at this time is the predicted value.

[0071] In order to verify the generalization performance of the improved model (the ability to learn different experimental samples), and to compare and analyze the prediction accuracy of the constructed IGWO-BP model with other models.

[0072] The present invention lists the comparison diagrams of the predicted values ​​and true values ​​of the external corrosion rate of buried pipelines respectively obtained by the IGWO-BP neural network model, the GWO-BP neural network model and the BP neural network model under the same data conditions of the external corrosion rate of buried pipelines, such as Figure 3 As shown; The relative error comparison diagram of IGWO-BP neural network model, GWO-BP neural network model and BP neural network model is shown in the figure. Figure 4 As shown, by comparison, it can be seen that the prediction effect of the improved model in this application is significantly improved.

[0073] Finally, taking a domestic buried pipeline as an example, the IGWO-BP buried pipeline external corrosion rate prediction model was established to verify the effectiveness of the prediction results of the IGWO-BP model. Table 1 shows 20 sets of buried pipeline external corrosion rate data.

[0074] Table 1 Data of external corrosion rates of 20 groups of buried pipelines ;

[0075] Finally, it should be noted that the above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical method of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical method to deviate from the spirit and scope of the technical method of the present invention.

Claims

1. A method for predicting the external corrosion rate of buried pipelines based on IGWO-BP neural network, characterized in that: The following steps are involved: Step 1: Obtain the experimental data of the external corrosion rate of the buried pipeline and perform normalization processing to obtain sample data, and divide the sample data into groups to obtain training group data and test group data; Step 2: Construct a BP neural network model and determine the number of nodes in the hidden layer of the network according to the topological structure of the BP neural network itself; Step 3: Use the improved gray wolf algorithm combined with the BP neural network topology structure to initialize the parameters a, A, and C, and randomly generate the position information of the gray wolves in the wolf pack; Step 4: Determine the transfer function, training function and fitness function of the BP neural network model; Step 5: Calculate the fitness value based on the fitness function and sort it in descending order, divide the gray wolf into α wolf, β wolf, δ wolf, ω wolf, and update the position and parameters of the gray wolf, among which α wolf, β wolf, δ wolf lead ω wolf; Step 6: Iterate the improved gray wolf algorithm to obtain the position of the target leader wolf α and the corresponding minimum error; Step 7: Bring the initial weight and threshold corresponding to the position of the leader wolf α back to the neural network, and make a judgment to obtain the output prediction result, that is, the predicted buried pipeline corrosion rate.

2. The method for predicting the external corrosion rate of buried pipelines based on IGWO-BP neural network according to claim 1 is characterized in that: In step 3, the improved gray wolf algorithm is combined with the BP neural network topology structure to initialize the parameters a, A, and C, and the specific content of the position information of the gray wolves in the wolf pack is randomly generated as follows: The gray wolf population is initialized, and the individual dimensions of the gray wolves, the population size, the maximum number of iterations, and the upper and lower bounds of the wolf search space are determined according to the number of weights and thresholds to be optimized.

3. The method for predicting the external corrosion rate of buried pipelines based on IGWO-BP neural network according to claim 2 is characterized in that: The specific contents of the improved grey wolf algorithm include: Optimizing population initialization to obtain an initial gray wolf population, wherein the process of optimizing population initialization is performed using a Circle chaos strategy; An improved nonlinear convergence method is used to optimize the convergence factor in the Grey Wolf Algorithm. Introducing weights W Optimize the global search capability of the Grey Wolf Algorithm.

4. The method for predicting the external corrosion rate of buried pipelines based on IGWO-BP neural network according to claim 3 is characterized in that: The process of optimizing population initialization is expressed as: ； In the formula, mod is the remainder function, C ( t )and C ( t +1) represent the t Second and t +1 Circle chaos iteration value, initial value C (0) is a random number between 0 and 1.

5. The method for predicting the external corrosion rate of buried pipelines based on IGWO-BP neural network according to claim 4 is characterized in that: Convergence Factor a The calculation formula is as follows: ； in, e is a natural constant; is the current iteration number; M is the maximum number of iterations.

6. The method for predicting the external corrosion rate of buried pipelines based on IGWO-BP neural network according to claim 5 is characterized in that: Weight W The expression is as follows: ； Where: W 1. W 2. W 3 are the learning rate weights of ω wolf to α wolf, β wolf and δ wolf respectively; X 1. X 2. X 3 are the updated positions of α wolf, β wolf and δ wolf respectively.

7. The method for predicting the external corrosion rate of buried pipelines based on IGWO-BP neural network according to claim 6 is characterized in that: In step 4, the network transfer function is a logsig() type function, the network training function is tansig(), and the fitness function is the mean square error between the predicted output value and the measured value.

8. The method for predicting the external corrosion rate of buried pipelines based on IGWO-BP neural network according to claim 7 is characterized in that: In step 6, the improved gray wolf algorithm is iterated to obtain the position of the target leader wolf α and the corresponding minimum error: Determine whether the set error or the maximum number of iterations is met. If not, repeat steps 5-6 until the conditions are met. Obtain the position of the target leader wolf α and the corresponding minimum error.

9. The method for predicting the external corrosion rate of buried pipelines based on IGWO-BP neural network according to claim 8, characterized in that: The specific content of the output prediction result obtained by judging in step 7 is: Calculate the output error of the BP neural network; If the output error reaches the preset learning accuracy or the maximum number of training times, the prediction result can be output after denormalization; If the output error does not reach the preset learning accuracy or the maximum number of training times, first update the weights and thresholds, and then repeat step 7 until the output error reaches the preset learning accuracy or the maximum number of training times. After denormalization, output the prediction result.

10. The method for predicting the external corrosion rate of buried pipelines based on IGWO-BP neural network according to claim 1, characterized in that: α wolf, β wolf, and δ wolf are the gray wolves corresponding to the first, second, and third in fitness value order, and ω wolf is the remaining gray wolves.

Images