Pipeline residual strength prediction method and system based on artificial intelligence algorithm

Abstract

The invention discloses a pipeline residual strength prediction method and system based on an artificial intelligence algorithm, and the method comprises the steps: constructing a CNN-BiLSTM-Adaboost model; the CNN-BiLSTM-Adaboost model comprises a CNN feature extraction module, a BiLSTM time sequence module and an Adaboost optimization module, wherein the CNN feature extraction module is used for extracting features of the corroded pipeline according to parameter data of the corroded pipeline, the BiLSTM time sequence module is used for obtaining a feature time sequence of the corroded pipeline by utilizing a bidirectional LSTM propagation mechanism, and the Adaboost optimization module is used for adjusting sample weights in an iteration process to generate a strong predictor; and during prediction, the residual strength of the pipeline is predicted through the trained CNN-BiLSTM-Adaboost model. The method is high in prediction precision and wide in adaptability.

Description

Technical Field

[0001] The present invention relates to the field of pipeline engineering, and in particular to a method and system for predicting the residual strength of a corroded pipeline by using an artificial intelligence algorithm. Background Art

[0002] Pipeline transportation is an important mode of transportation in China. However, due to the huge amount of detection data and the reliance on manual interpretation for defect identification, traditional methods are often time-consuming and labor-intensive, and prone to errors such as false positives and missed positives. Therefore, there is an urgent need to develop intelligent defect identification methods to improve the efficiency of defect detection. One of the key points in defect detection is the prediction of pipeline residual strength. Pipeline residual strength refers to the maximum pressure or load that the pipeline can withstand after corrosion, wear or other forms of damage. Specifically, it involves the pipeline under actual use conditions. Due to factors such as corrosion and erosion, the wall thickness of the pipeline is thinned, but a certain strength and safety must be maintained to ensure the normal operation of the pipeline and avoid accidents. If the residual strength of the pipeline can be accurately predicted and effectively maintained, the occurrence of major accidents (such as oil and gas leaks) will be significantly reduced, thereby reducing economic losses and protecting the ecological environment.

[0003] In recent years, researchers have increasingly adopted data-driven intelligent prediction methods. For example, Oh et al. (Oh, D., J. Race, S. Oterkus, and B. Koo. 2020. “Burst pressure prediction of API5L X-grade dented pipelines using deep neural network.” J. Mar. Sci. Eng. 8(10): 766) explored the application of deep neural networks in predicting the burst pressure of API 5L X-grade pipelines and conducted experimental verification. Lu et al. (Lu, H., T. Iseley, J. Matthews, W. Liao, and M. Azimi. 2021a. “Anensemble model based on relevance vector machine and multi-objective SALPswarm algorithm for predicting burst pressure of corroded pipelines.” J. Pet. Sci. Eng. 203 (Jun): 108585) proposed an ensemble model based on relevance vector machine and combined with a multi-objective optimization algorithm to predict the burst pressure of corroded pipelines, and evaluated the accuracy and stability of the model through a case study. In addition, Lu et al. (Lu, H., ZD Xu, T. Iseley, and JC Matthews. 2021b. “Novel data-drivenframework for predicting residual strength of corroded pipelines.” J. Pipeline Syst. Eng. Pract. 12 (4): 04021045) used a hybrid data-driven model to predict the residual strength of single-point corroded pipelines and optimized the input data using principal component analysis.Ma et al. (Ma, Y., J. Zheng, Y. Liang, JJ Klemeš, J. Du, Q. Liao, H. Lu, and B. Wang. 2022. “Deep pipe: Theory-guided neural network method for predicting burst pressure of corrodedpipelines.” Process Saf. Environ. Prot. 162 (Jun): 595–609) introduced a neural network model based on theoretical principles to predict the burst pressure of corroded pipelines. They combined traditional burst pressure prediction formulas with empirical formula knowledge and verified the superior prediction accuracy of the model through experiments and finite element simulations using open data sets.

[0004] Although some progress has been made in corrosion depth prediction, most of the current research on pipeline residual strength prediction is still focused on improving and applying pipeline strength assessment formulas and experimental testing using finite element analysis, while algorithm exploration is still limited. Specifically, CNN (convolutional neural network), as a typical deep learning model, has been introduced into pipeline leak diagnosis due to its unique convolution structure, weight sharing, and powerful autonomous feature extraction capabilities. Studies have shown that CNN has good classification prediction performance in pipeline leak detection. For example, Chuang et al. (Chuang WY, Tsai YL, Wang L H. Leak detection in water distribution pipes based on CNNwith mel frequency cepstral coefficients [C] Proceedings of the 3 rd International Conference on Innovation in Artificial Intelligence.2019) proposed a CNN-based groundwater pipeline leakage detection method, and the classification accuracy of its model reached 98%.

[0005] However, existing CNN models have limitations when processing time series data and cannot effectively capture long-term dependencies. In addition, although LSTM (Long Short-Term Memory Network) can alleviate the gradient vanishing problem of traditional recurrent neural networks when processing long time series data, it is still insufficient when processing high-dimensional spatial features. BiLSTM (Bidirectional Long Short-Term Memory Network) further improves the performance of LSTM, but still faces challenges when processing high-dimensional spatial data. Summary of the invention

[0006] Purpose of the invention: The present invention aims to solve the shortcomings of existing pipeline residual strength prediction methods, especially to improve the accuracy and robustness of the prediction, and proposes a pipeline residual strength prediction method based on artificial intelligence algorithm. By combining CNN, BiLSTM and Adaboost algorithms, a CNN-BiLSTM-Adaboost hybrid model is provided, which can effectively process complex corroded pipeline data and can be applied to the safety assessment and maintenance management of energy pipelines such as oil and natural gas, thereby improving the prediction accuracy and adaptability.

[0007] Technical solution: To achieve the above purpose, the technical solution adopted by the present invention is: A pipeline residual strength prediction method based on artificial intelligence algorithm includes the following steps: Step 1: Collect and preprocess the parameter data of the corroded pipeline to obtain a training set.

[0008] Step 2, construct a CNN-BiLSTM-Adaboost model, the CNN-BiLSTM-Adaboost model includes a CNN feature extraction module, a BiLSTM time series module, and an Adaboost optimization module, wherein the CNN feature extraction module is used to extract corrosion pipeline features according to the parameter data of the corrosion pipeline. The BiLSTM time series module uses a bidirectional LSTM propagation mechanism to capture past and future information in the corrosion pipeline features and obtain a corrosion pipeline feature time series. The Adaboost optimization module is used to adjust sample weights during the iteration process of the BiLSTM time series module to generate a strong predictor.

[0009] Step 3: Train the CNN-BiLSTM-Adaboost model according to the training set to obtain a trained CNN-BiLSTM-Adaboost model.

[0010] Step 4: During prediction, the parameter data of the corroded pipeline to be predicted is collected and preprocessed, and the preprocessed parameter data of the corroded pipeline is input into the trained CNN-BiLSTM-Adaboost model to obtain the prediction of the residual strength of the corroded pipeline to be predicted.

[0011] Preferably: the Adaboost optimization module is used in each iteration , given a weight distribution , train a weak classifier , weighted error The calculation formula is as follows: ; in, represents the weighted error, Indicates In the round of iteration, The weight of the samples, is the indicator function, Indicates In the round of iteration, the weak classifier Samples The prediction results, Indicates The input features of samples, is the true label of the sample, For the samples.

[0012] Weights of weak classifiers The calculation formula is as follows: ; in, represents the weight of the weak classifier, is the learning rate that is adjusted with the number of iterations according to the current error level.

[0013] if , the weight distribution of training samples is updated as: ; in, Indicates In the round of iteration, The weight of the samples, Indicates The weight of the samples.

[0014] if , the weight distribution of training samples is updated as: ; Finally, the output of Adaboost is the weighted result of all weak classifiers.

[0015] ; in, represents the weighted result of all weak classifiers, represents the weight of the weak classifier, Indicates The output function of the round weak classifier, Represents the total number of weak classifiers, that is, the number of iterations in the Adaboost algorithm.

[0016] Preferably: the BiLSTM time series module is calculated by the following formula: Input Gate: ; Forget Gate: ; Candidate memory cells: ; Memory unit update: ; Output Gate: ; Hide status updates: ; in, represents the output of the input gate, is the Sigmoid activation function, represents the weight matrix of the input gate, Represents the output at the previous moment, Represents the input at the current moment, represents the bias term of the input gate, represents the output of the forget gate, represents the forget gate adjustment coefficient, represents the weight matrix of the forget gate, represents the bias term of the forget gate, Represents candidate memory cells, showing new information at the current moment, represents the weight matrix of the candidate memory unit, represents the bias term of the candidate memory unit, Indicates the current state of the memory unit, represents the output of the output gate, represents the output gate adjustment coefficient, represents the weight matrix of the output gate, represents the bias term of the output gate, Represents the hidden state at the current moment and displays the output of the network at the current moment.

[0017] Preferably: the CNN feature extraction module includes a first convolutional layer 1, a first maximum pooling layer 1, a first convolutional layer 2, a first maximum pooling layer 2, and a first Flatten layer connected in sequence.

[0018] Preferably: the first convolutional layer 1 and the first convolutional layer 2 use a convolution kernel of size 3x3, and introduce nonlinearity through a ReLU activation function.

[0019] Preferably: the collected parameter data of the corroded pipeline is normalized, and all features are scaled to the interval [0, 1] to obtain a training set.

[0020] Preferred: Optimize model parameters by evaluating the performance of the CNN-BiLSTM-Adaboost model.

[0021] Another object of the present invention is to provide a pipeline residual strength prediction system based on an artificial intelligence algorithm, which is used to implement the pipeline residual strength prediction method based on an artificial intelligence algorithm, including an input unit, a preprocessing unit, a CNN-BiLSTM-Adaboost model unit, and an output unit, wherein: The input unit is used to input parameter data of the corroded pipeline.

[0022] The preprocessing unit is used to preprocess the parameter data input into the corrosion pipeline.

[0023] The CNN-BiLSTM-Adaboost model unit is used to obtain a prediction of the pipeline residual strength of the corroded pipeline to be predicted based on the preprocessed parameter data of the corroded pipeline.

[0024] Compared with the prior art, the present invention has the following beneficial effects: The present invention introduces the Adaboost algorithm, combines the advantages of CNN and BiLSTM, and further improves the prediction performance. The core of Adaboost is to iteratively adjust the sample weights according to the learning results of the previous prediction, focus on the classified samples, and train the next predictor. This process continues until all basic predictors are trained. Adaboost performs well in processing noisy and unbalanced data, and can effectively improve the stability and accuracy of the model. By combining Adaboost with CNN, the learning ability and generalization performance of the model are optimized, thereby improving the accuracy of pipeline residual strength prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] Figure 1 This is a schematic diagram of the CNN-BiLSTM-Adaboost model structure.

[0026] Figure 2 It is the data of pipe inner diameter and defect length.

[0027] Figure 3 Pipeline wall thickness and defect depth data.

[0028] Figure 4 Pipeline bursting pressure data.

[0029] Figure 5 This is the test set prediction result diagram of the CNN-BiLSTM-Adaboost model.

[0030] Figure 6 This is the training set prediction result diagram of the CNN-BiLSTM-Adaboost model.

[0031] Figure 7 This is the test set prediction result diagram of the CNN model.

[0032] Figure 8 This is the prediction result diagram of the training set of the CNN model.

[0033] Fig. 9 This is the test set prediction result diagram of the CNN-BiLSTM model.

[0034] Fig.10 This is the training set prediction result diagram of the CNN-BiLSTM model.

[0035] Fig.11 This is the test set prediction result diagram of the CNN-LSTM model.

[0036] Fig.12 This is the training set prediction result diagram of the CNN-LSTM model.

[0037] Fig.13 This is the test set prediction result graph of the BiLSTM-Adaboost model.

[0038] Fig.14 This is the training set prediction result diagram of the BiLSTM-Adaboost model. DETAILED DESCRIPTION

[0039] The present invention is further explained below in conjunction with the accompanying drawings and specific embodiments. It should be understood that these examples are only used to illustrate the present invention and are not used to limit the scope of the present invention. After reading the present invention, various equivalent forms of modifications to the present invention by those skilled in the art all fall within the scope defined by the claims attached to this application.

[0040] A pipeline residual strength prediction method based on artificial intelligence algorithm includes the following steps: Step 1: Collect and preprocess the parameter data of the corroded pipeline to obtain a training set.

[0041] Collect and organize relevant data of pipelines, and perform necessary preprocessing to ensure the uniformity and accuracy of data. Mainly collect parameters such as pipeline wall thickness, diameter, corrosion depth, defect length, bursting pressure, etc.

[0042] The collected parameter data of the corroded pipeline are normalized, and all features are scaled to the interval [0, 1] to obtain the training set to improve the efficiency and stability of model training. The data set is randomly divided into training set and test set to evaluate the generalization ability of the model.

[0043] In this embodiment, if Figures 2 to 4As shown in the figure, the main parameters collected are pipeline wall thickness (mm), pipeline inner diameter (mm), corrosion depth (mm), defect length (mm), burst pressure (mm), etc. The pipeline parameter data is collected and normalized using MATLAB to ensure that all data are in the same range. 80% of the data is randomly selected as the training set and 20% as the test set to ensure the fairness of the model evaluation.

[0044] Step 2: Build a CNN-BiLSTM-Adaboost model, such as Figure 1 As shown, the CNN-BiLSTM-Adaboost model includes a CNN feature extraction module, a BiLSTM time series module, and an Adaboost optimization module.

[0045] The CNN feature extraction module is used to extract corrosion pipeline features according to parameter data of the corrosion pipeline. The CNN feature extraction module extracts spatial features of input data through a convolutional neural network, processes the input data through a convolution layer and a pooling layer, and generates a feature map.

[0046] The CNN feature extraction module includes the first convolution layer 1, the first maximum pooling layer 1, the first convolution layer 2, the first maximum pooling layer 2, and the first Flatten flattening layer, which are connected in sequence. The convolution kernel size of the first convolution layer 1 is 3×3, the number is 64, and the activation function is ReLU. The window size of the first maximum pooling layer 1 is 2×2, and the step size is 2. The convolution kernel size of the first convolution layer 2 is 3×3, the number is 128, and the activation function is ReLU. The window size of the first maximum pooling layer 2 is 2×2, and the step size is 2. The first Flatten flattening layer flattens the output through the Flatten layer for BiLSTM processing, uses convolution operations to extract local features, reduces data dimensions to reduce computational complexity, uses convolution kernels of size 3x3, introduces nonlinearity through the ReLU activation function, and uses maximum pooling operations in the pooling layer to reduce the dimension of the feature map.

[0047] The input data passes through the convolution layer and uses the convolution kernel to extract features. The convolution operation can be expressed as: ; in, It is The output feature map of the layer, is the convolution weight matrix, represents the convolution operation, is the input, is the bias, is the activation function (ReLU). The pooling layer reduces the dimension of the feature map and reduces the computational complexity.

[0048] The BiLSTM time series module uses a bidirectional LSTM propagation mechanism to capture past and future information in the corrosion pipeline characteristics and obtain the corrosion pipeline characteristic time series. The BiLSTM time series module uses a bidirectional LSTM to capture the long-term dependency in the time series data, inputs the features output by the CNN into the BiLSTM time series module, processes the data bidirectionally, and captures the past and future information.

[0049] The BiLSTM time series module processes the input features, forward and backward propagation respectively to capture the long-term dependencies of the time series. The BiLSTM time series module uses 25 LSTM units and a bidirectional structure to process time series data. The output mode is "the last time step" and the final hidden state is obtained as a feature.

[0050] The BiLSTM time series module is calculated by the following formula: Input Gate: ; Forget Gate: ; Candidate memory cells: ; Memory unit update: ; Output Gate: ; Hide status updates: ; in, represents the output of the input gate, is the Sigmoid activation function, represents the weight matrix of the input gate, Represents the output at the previous moment, Represents the input at the current moment, represents the bias term of the input gate, represents the output of the forget gate, represents the forget gate adjustment coefficient, represents the weight matrix of the forget gate, represents the bias term of the forget gate, Represents candidate memory cells, showing new information at the current moment, represents the weight matrix of the candidate memory unit, represents the bias term of the candidate memory unit, Indicates the current state of the memory unit, represents the output of the output gate, represents the output gate adjustment coefficient, represents the weight matrix of the output gate, represents the bias term of the output gate, Represents the hidden state at the current moment and displays the output of the network at the current moment.

[0051] BiLSTM captures past and future information simultaneously through forward and backward propagation. The forget gate adds an adjustment coefficient Control the output size of the forget gate, thereby adjusting the proportion of memory unit information retained at the previous moment. Keep more historical information, smaller It is easier to forget old information. The output gate introduces the adjustment coefficient Control the output amplitude of the output gate, thereby affecting the hidden state The output strength of the larger Will increase the network's output to the current memory unit, smaller The output is suppressed.

[0052] The Adaboost optimization module is used to adjust the sample weights during the iteration of the BiLSTM time series module to generate a strong predictor. The Adaboost optimization module trains multiple weak regressors, iteratively adjusts the sample weights, focuses on the classified samples, combines the results of all weak regressors, generates a strong predictor, and improves the prediction accuracy of the model through ensemble learning.

[0053] The Adaboost optimization module iteratively trains multiple weak regressors and adjusts the weight distribution. In each iteration, a weak classifier is trained and its weighted error rate is calculated. The weight distribution is updated according to the error rate to ensure that the classified samples get higher weights in the next training. The final output is the weighted result of all weak regressors.

[0054] The Adaboost optimization module performs , given a weight distribution , train a weak classifier , weighted error The calculation formula is as follows: ; in, represents the weighted error, Indicates In the round of iteration, The weight of the samples, is the indicator function, Indicates In the round of iteration, the weak classifier Samples The prediction results, Indicates The input features of samples, is the true label of the sample, For the samples.

[0055] Weights of weak classifiers The calculation formula is as follows: ; in, represents the weight of the weak classifier, It is the learning rate that is adjusted with the number of iterations according to the current error level, so that it converges faster when the error is large and remains stable when the error is small.

[0056] if , the weight distribution of training samples is updated as: ; in, Indicates In the round of iteration, The weight of the samples, Indicates The weight of the samples.

[0057] if , the weight distribution of training samples is updated as: ; Finally, the output of Adaboost is the weighted result of all weak classifiers.

[0058] ; in, represents the weighted result of all weak classifiers, represents the weight of the weak classifier, Indicates The output function of the round weak classifier, Represents the total number of weak classifiers, that is, the number of iterations in the Adaboost algorithm.

[0059] In this embodiment, 10 weak regressors (K=10) are used. In each iteration, the weak regressors are trained according to the current weight distribution of the training set, and the weighted error rate is calculated. The sample weights are updated so that the classified samples can be focused on in the next iteration.

[0060] Step 3: Train the CNN-BiLSTM-Adaboost model according to the training set to obtain a trained CNN-BiLSTM-Adaboost model.

[0061] In this embodiment, Figure 5 Figure 6As shown in the figure, the model is trained using the training set, and the cross-validation method is used to optimize the parameters and select the best hyperparameters (such as the number of convolution kernels, the number of LSTM units, etc.). The loss value and accuracy during the training process are recorded to monitor the learning progress of the model.

[0062] Step 4: During prediction, the parameter data of the corroded pipeline to be predicted is collected and preprocessed, and the preprocessed parameter data of the corroded pipeline is input into the trained CNN-BiLSTM-Adaboost model to obtain the prediction of the residual strength of the corroded pipeline to be predicted.

[0063] Step 5, model evaluation and optimization, evaluate the performance of the model and make necessary optimizations. Specifically, the model is evaluated by calculating indicators such as mean relative error (MAE) and R² value. Based on the evaluation results, the model parameters are optimized to improve the prediction accuracy.

[0064] like Figures 7 to 14 As shown, the required data (pipeline inner diameter, pipe wall thickness, defect length, defect depth) is first input into it, and the input data is normalized, and then 80% of the data is randomly divided into a training set, and the remaining 20% ​​is divided into a test set. The processed data is trained by the CNN-BiLSTM-Adaboost algorithm, and a prediction model for the residual strength of the pipeline is obtained after training. The model is evaluated using the test set, the mean relative error (MAE) and the coefficient of determination (R² value) are calculated, the prediction performance of the model is judged, and the parameters of CNN and BiLSTM are iteratively adjusted to ensure the best prediction effect. The prediction effects of different models are compared, including the method of the present invention, CNN, LSTM, BiLSTM, BiLSTM-Adaboost, CNN-LSTM, CNN-BiLSTM and CNN-BiLSTM-Adaboost, and the advantages and disadvantages of each model are analyzed.

[0065] The average relative percentage error (MAPE) of the model of the present invention on the test set is 4.6944%, while the average relative errors of traditional evaluation methods such as ASME B31G, DNV RP-F101 and PCORRC are 33.595%, 48.085% and 45.4% respectively. This shows that the model of the present invention has obvious advantages in prediction accuracy and can provide more reliable pipeline residual strength evaluation.

[0066] Through ensemble learning, Adaboost can effectively handle the noise and imbalance problems in the training data, making the model more stable when facing complex corrosion data. Experiments show that the model can maintain high prediction accuracy under various corrosion conditions, especially under high pressure conditions, with an accuracy improvement of more than 15%.

[0067] The present invention demonstrates good real-time prediction capabilities when processing large-scale data, can respond quickly in pipeline maintenance management, and provide timely support for decision-making. Compared with traditional finite element analysis, the method's calculation speed is increased by about 30%, making on-site assessment and monitoring more efficient.

[0068] By accurately predicting the remaining strength of the pipeline, potential corrosion problems can be discovered in time, reducing economic losses and environmental pollution caused by pipeline leakage. It is expected that in actual application, a large amount of maintenance costs can be saved every year.

[0069] The advantages of the present invention are due to the powerful ability of CNN in spatial feature extraction, the advantages of BiLSTM in time series modeling, and the effect of Adaboost in improving model stability. These advantages have been verified by actual experimental data. The experiments were conducted under different pipeline conditions, covering a variety of corrosion depths and defect types, ensuring the wide applicability of the results.

[0070] The present invention combines convolutional neural network (CNN), bidirectional long short-term memory network (BiLSTM) and adaptive boosting algorithm (AdaBoost) to form a new pipeline residual strength prediction model. Multiple weak regressors are integrated through the Adaboost algorithm, which significantly improves the robustness and generalization ability of the model. The model effectively captures the spatial characteristics and time dependencies in pipeline data, thereby improving the accuracy of prediction, especially when processing complex corrosion data. The model can flexibly respond to input data of different types and sizes, and is suitable for prediction tasks of various pipeline parameters and corrosion conditions. Through efficient data processing and model training, it provides real-time prediction capabilities, which helps to reduce pipeline maintenance costs and potential economic losses.

[0071] In another embodiment, a pipeline remaining strength prediction system based on an artificial intelligence algorithm is provided, which is used to implement the pipeline remaining strength prediction method based on an artificial intelligence algorithm, including an input unit, a preprocessing unit, a CNN-BiLSTM-Adaboost model unit, and an output unit, wherein: The input unit is used to input parameter data of the corroded pipeline.

[0072] The preprocessing unit is used to preprocess the parameter data input into the corrosion pipeline.

[0073] The CNN-BiLSTM-Adaboost model unit is used to obtain a prediction of the pipeline residual strength of the corroded pipeline to be predicted based on the preprocessed parameter data of the corroded pipeline.

[0074] 1. The present invention combines the advantages of CNN, BiLSTM and AdaBoost, making full use of CNN's ability in spatial feature extraction, BiLSTM's advantage in capturing time dependencies, and AdaBoost's effectiveness in ensemble learning. This combination significantly improves the overall prediction ability of the model, reducing its average relative error to 4.6944% when processing complex corrosion data, which is much lower than the 33.595% to 48.085% of traditional evaluation methods. Therefore, the model can provide more accurate predictions of pipeline remaining strength and help decision makers make more reasonable maintenance plans.

[0075] 2. Through the introduction of the Adaboost algorithm, the present invention can effectively deal with the noise and imbalance problems in the data. During the training process, Adaboost continuously adjusts the sample weights and focuses on the classified samples, thereby enhancing the model's adaptability to abnormal data. This allows the model to maintain high stability and accuracy when facing complex environments in practical applications.

[0076] 3. The model structure design and training algorithm of the present invention enable it to quickly process large-scale pipeline data. Compared with the traditional finite element analysis method, this model not only has higher prediction accuracy, but also has significantly improved processing speed, and can achieve real-time prediction. This provides timely decision support for pipeline maintenance and helps reduce economic losses and environmental risks caused by corrosion.

[0077] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A pipeline residual strength prediction method based on artificial intelligence algorithm, characterized in that: The following steps are involved: Step 1, collect and preprocess parameter data of the corroded pipeline to obtain a training set; Step 2, constructing a CNN-BiLSTM-Adaboost model, wherein the CNN-BiLSTM-Adaboost model includes a CNN feature extraction module, a BiLSTM time series module, and an Adaboost optimization module, wherein the CNN feature extraction module is used to extract corrosion pipeline features according to parameter data of the corrosion pipeline; the BiLSTM time series module uses a bidirectional LSTM propagation mechanism to capture past and future information in the corrosion pipeline features to obtain a corrosion pipeline feature time series; the Adaboost optimization module is used to adjust sample weights during the iteration process of the BiLSTM time series module to generate a strong predictor; Step 3, train the CNN-BiLSTM-Adaboost model according to the training set to obtain a trained CNN-BiLSTM-Adaboost model; Step 4: During prediction, the parameter data of the corroded pipeline to be predicted is collected and preprocessed, and the preprocessed parameter data of the corroded pipeline is input into the trained CNN-BiLSTM-Adaboost model to obtain the prediction of the residual strength of the corroded pipeline to be predicted.

2. The pipeline residual strength prediction method based on artificial intelligence algorithm according to claim 1 is characterized by: The Adaboost optimization module performs , given a weight distribution , train a weak classifier , weighted error The calculation formula is as follows: ； in, represents the weighted error, Indicates In the round of iteration, The weight of the samples, is the indicator function, Indicates In the round of iteration, the weak classifier Samples The prediction results, Indicates The input features of samples, is the true label of the sample, For the samples; Weights of weak classifiers The calculation formula is as follows: ； in, represents the weight of the weak classifier, is the learning rate adjusted according to the current error level with the number of iterations; if , the weight distribution of training samples is updated as: ； in, Indicates In the round of iteration, The weight of the samples, Indicates The weight of the samples; if , the weight distribution of training samples is updated as: ； Finally, the output of Adaboost is the weighted result of all weak classifiers; ； in, represents the weighted result of all weak classifiers, represents the weight of the weak classifier, Indicates The output function of the round weak classifier, Represents the total number of weak classifiers, that is, the number of iterations in the Adaboost algorithm.

3. The pipeline residual strength prediction method based on artificial intelligence algorithm according to claim 2 is characterized in that: The BiLSTM time series module is calculated by the following formula: Input Gate: ; Forget Gate: ; Candidate memory cells: ; Memory unit update: ; Output Gate: ; Hide status updates: ; in, represents the output of the input gate, is the Sigmoid activation function, represents the weight matrix of the input gate, Represents the output at the previous moment, Represents the input at the current moment, represents the bias term of the input gate, represents the output of the forget gate, represents the forget gate adjustment coefficient, represents the weight matrix of the forget gate, represents the bias term of the forget gate, Represents candidate memory cells, showing new information at the current moment, represents the weight matrix of the candidate memory unit, represents the bias term of the candidate memory unit, Indicates the current state of the memory unit, represents the output of the output gate, represents the output gate adjustment coefficient, represents the weight matrix of the output gate, represents the bias term of the output gate, Represents the hidden state at the current moment and displays the output of the network at the current moment.

4. The pipeline residual strength prediction method based on artificial intelligence algorithm according to claim 3 is characterized by: The CNN feature extraction module includes a first convolutional layer 1, a first maximum pooling layer 1, a first convolutional layer 2, a first maximum pooling layer 2, and a first Flatten layer which are connected in sequence.

5. The pipeline residual strength prediction method based on artificial intelligence algorithm according to claim 4 is characterized in that: The first convolutional layer 1 and the first convolutional layer 2 use a convolution kernel of size 3x3 and introduce nonlinearity through the ReLU activation function.

6. The pipeline residual strength prediction method based on artificial intelligence algorithm according to claim 5 is characterized by: The collected parameter data of the corroded pipeline are normalized, and all features are scaled to the interval [0, 1] to obtain the training set.

7. The pipeline residual strength prediction method based on artificial intelligence algorithm according to claim 6 is characterized by: By evaluating the performance of the CNN-BiLSTM-Adaboost model, the model parameters are optimized.

8. A pipeline residual strength prediction system based on artificial intelligence algorithm, characterized in that: The method for predicting the remaining strength of a pipeline based on an artificial intelligence algorithm according to any one of claims 1 to 7 comprises an input unit, a preprocessing unit, a CNN-BiLSTM-Adaboost model unit, and an output unit, wherein: The input unit is used to input parameter data of the corroded pipeline; The preprocessing unit is used to preprocess the parameter data input into the corrosion pipeline; The CNN-BiLSTM-Adaboost model unit is used to obtain a prediction of the pipeline residual strength of the corroded pipeline to be predicted based on the preprocessed parameter data of the corroded pipeline.

9. The pipeline residual strength prediction system based on artificial intelligence algorithm according to claim 8 is characterized in that: The CNN-BiLSTM-Adaboost model includes a CNN feature extraction module, a BiLSTM time series module, and an Adaboost optimization module, wherein the CNN feature extraction module is used to extract corrosion pipeline features according to parameter data of the corrosion pipeline; the BiLSTM time series module uses a bidirectional LSTM propagation mechanism to capture past and future information in the corrosion pipeline features to obtain a corrosion pipeline feature time series; and the Adaboost optimization module is used to adjust sample weights during the iteration process of the BiLSTM time series module to generate a strong predictor.

Images