Polyethylene gas pipeline mechanical response prediction method and system

Abstract

The invention discloses a polyethylene gas pipeline mechanical response prediction method and system, and belongs to the technical field of pipeline burying prediction. The polyethylene gas pipeline mechanical response prediction method comprises the following steps: constructing a PE gas pipeline finite element model according to self characteristics of a PE gas pipeline and multi-factor influence data of laying requirements of the PE gas pipeline; constructing a three-dimensional geometric model based on the finite element model, and obtaining a PE gas pipeline data set sample; performing normalization processing on the multi-factor influence data and the PE gas pipeline data set sample to obtain overall sample data, and dividing the overall sample data into training sample data and test sample data; according to the PE gas pipeline mechanical response calculation based on the multi-factor influence, the finite element model under the multi-factor influence is constructed, the back propagation neural network is combined for training optimization, and the bearing capacity and the deformation calculation speed of the PE gas pipeline under the multi-factor effect are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of pipeline burial prediction, in particular to a method and system for predicting the mechanical response of a polyethylene gas pipeline. Background Art

[0002] As a form of urban infrastructure, buried pipelines date back to the ancient Roman period, initially used for urban drainage. With the expansion of urban areas and socioeconomic development, the application of buried pipelines in modern cities has become extensive and complex. As the lifeline of cities, buried pipelines play a vital role in municipal administration, transportation, water conservancy projects, and other projects, directly impacting the daily lives of urban residents. Therefore, they are of great significance to urban development.

[0003] The performance and service life of buried polyethylene (PE) gas pipelines in urban areas are closely related to the complex load environments they encounter during actual use. These complex load environments include operating pressure within the pipeline, ground subsidence, ground loads, and temperature fluctuations. The combined effects of these multiple loads can damage and cause failure of pipeline materials and structures, seriously threatening the safe operation of the pipeline system. Therefore, the design and maintenance of buried PE gas pipelines in urban areas require full consideration of these complex load environments to ensure the long-term stability and reliability of the pipelines.

[0004] Current research has largely focused on the impact of a single factor on buried PE gas pipelines, neglecting the study of damage caused by the synergistic effects of multiple factors. Furthermore, the complex nature of these factors places high demands on modeling parameters, making it difficult to adapt to the demands of large-scale simulations. With the rapid development of artificial intelligence (AI) technology, its application to the calculation of damage responses of PE gas pipelines under the influence of multiple factors not only improves prediction accuracy but also provides valuable insights for practical engineering applications. Summary of the Invention

[0005] In view of the above problems existing in the existing technology, the present invention is proposed.

[0006] Therefore, the technical problem to be solved by the present invention is a problem.

[0007] To achieve the above objectives, the present invention provides the following technical solutions: a method and system for predicting the mechanical response of a polyethylene gas pipeline, comprising:

[0008] Based on the characteristics of PE gas pipeline and the multi-factor influencing data of its laying requirements, a PE gas pipeline finite element model is constructed;

[0009] A 3D geometric model was constructed based on the finite element model, and a sample PE gas pipeline dataset was obtained;

[0010] Normalize the multi-factor impact data and the PE gas pipeline dataset samples to obtain the overall sample data, and divide the overall sample data into training sample data and test sample data;

[0011] Establish a back-propagation neural network model, determine the parameters of the back-propagation neural network model, and then initialize the parameters;

[0012] The training sample data is input into the back propagation neural network model for training, and the model is evaluated by the test sample data to construct a back propagation neural network model for PE gas pipeline under multiple factors;

[0013] Determine the initial weights and biases of the back-propagation neural network model, and apply them to the back-propagation neural network model for optimization iteration to obtain an optimized back-propagation neural network model.

[0014] As a further solution of the present invention: according to the inherent characteristics of the PE gas pipeline including mechanical parameters and geometric parameters; the laying requirements of the PE gas pipeline include ambient temperature, occupation load, subsidence depth, pipeline internal pressure and pipeline burial depth.

[0015] As a further solution of the present invention: the steps of constructing a three-dimensional geometric model based on the finite element model and obtaining a PE gas pipeline data set sample include:

[0016] Generate a cylindrical pipe model based on the PE gas pipeline of the finite element model;

[0017] Add soil layers around the pipeline to simulate buried conditions;

[0018] Process the interface between the pipeline and the soil and construct a three-dimensional geometric model of the pipeline;

[0019] Divide the three-dimensional geometric model into several finite element units;

[0020] Different mechanical parameters are assigned to the pipeline material according to the different temperatures of the PE gas pipeline, and a PE gas pipeline dataset sample is obtained.

[0021] As a further solution of the present invention: the PE gas pipeline data set sample includes the displacement component of the PE gas pipeline and the Mises stress of the PE gas pipeline.

[0022] As a further solution of the present invention: the normalization processing formula is:

[0023]

[0024] Among them, f(x) is the normalized data; x is the data of the dataset; x min is the minimum value of the data set; x maxis the maximum value of the data set.

[0025] As a further solution of the present invention: the back propagation neural network model includes an input layer, a hidden layer and an output layer;

[0026] Among them, the number of units in the hidden layer is N hid The calculation formula is:

[0027]

[0028] Among them, N hid is the number of hidden layer neurons, N input is the number of neurons in the input layer, N output is the output layer neuron, c is an integer between 2 and 10, and s is the number of samples.

[0029] As a further solution of the present invention: inputting the training sample data into the back propagation neural network model for training, and evaluating the model through the test sample data, the steps of constructing the back propagation neural network model of the PE gas pipeline under multiple factors include:

[0030] Initialize the weights and biases of the backpropagation neural network model, import the input features and output features, and then obtain the predicted value through forward propagation calculation;

[0031] Calculate the loss function based on the predicted value and the true label, and use the backpropagation algorithm to update the weights and biases to reduce the value of the loss function;

[0032] Adjust the parameters of the neural network based on the gradient descent method to minimize the loss function;

[0033] Repeat the above steps until the loss function converges or reaches the set stopping condition, completing the training;

[0034] The trained back propagation neural network model is evaluated using test sample data, and the hyperparameters of the back propagation neural network model are optimized based on the evaluation results to obtain a back propagation neural network model of the mechanical response of PE gas pipelines under the influence of multiple factors.

[0035] As a further solution of the present invention: the formula of the loss function is,

[0036]

[0037] Among them, t i is the finite element numerical simulation value of the displacement component and Mises stress of the i-th sample, o i is the neural network prediction value of the displacement component and Mises stress of the i-th sample, and n is the total number of samples.

[0038] As a further solution of the present invention: the back propagation neural network model is optimized using the minnow optimization algorithm.

[0039] As a further solution of the present invention: a system applied to the above-mentioned polyethylene gas pipeline mechanical response prediction method comprises:

[0040] Model building module, used to build the finite element model of PE gas pipeline and obtain multi-factor influencing data;

[0041] The geometric modeling and meshing module is connected to the model building module to build a three-dimensional geometric model and obtain a sample of the PE gas pipeline dataset;

[0042] The data processing and preprocessing module is used to normalize the multi-factor influence data and the PE gas pipeline data set samples to obtain the overall sample data, and divide the overall sample data into training sample data and test sample data; it is connected to the geometric modeling and meshing module;

[0043] Neural network modeling module, used to establish a back-propagation neural network model, determine its architecture and related parameters, and initialize the model parameters;

[0044] The model training module is connected to the neural network modeling module and the data processing and preprocessing module, which inputs the training sample data into the neural network model for training and constructs a back propagation neural network model of the PE gas pipeline under the influence of multiple factors;

[0045] The model optimization module determines the initial weights and biases of the back-propagation neural network model and improves the model performance through an iterative optimization process, ultimately obtaining an optimized back-propagation neural network model. This module is connected to the model training module.

[0046] The prediction and evaluation module feeds the new input feature data into the optimized back-propagation neural network model, calculates and outputs the predicted value of the mechanical response of the PE gas pipeline, including displacement components and Mises stress, analyzes the model output results, and combines the engineering background and standards to determine whether the pipeline is within the safe range.

[0047] Compared with the existing technology, the beneficial effects of the present invention are: based on the calculation of the mechanical response of PE gas pipelines under the influence of multiple factors, the present invention constructs a finite element model under the influence of multiple factors, and combines it with back-propagation neural network for training and optimization. Compared with traditional mechanical modeling methods, the speed of calculating the bearing capacity and deformation of PE gas pipelines under the influence of multiple factors is significantly improved; this method ensures the accuracy and reliability of the calculation results, can handle the mechanical response of pipelines under complex conditions, and provide faster and more accurate technical support. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be derived from these drawings without inventive effort. Among them:

[0049] Figure 1 This is a schematic diagram of the overall process of an embodiment provided by the present invention.

[0050] Figure 2 Schematic diagram of the back propagation neural network model of the embodiment provided by the present invention.

[0051] Figure 3 This is a flow chart of constructing a finite element model according to an embodiment of the present invention.

[0052] Figure 4 Schematic diagram of comparative experiments of embodiments provided by the present invention. DETAILED DESCRIPTION

[0053] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0054] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0055] Next, the present invention is described in detail with reference to schematic diagrams. For ease of illustration, when describing embodiments of the present invention, cross-sectional views illustrating device structures may be partially enlarged and not to scale. Furthermore, the schematic diagrams are merely illustrative and should not limit the scope of protection of the present invention. Furthermore, in actual production, three-dimensional dimensions, including length, width, and depth, should be included.

[0056] Furthermore, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it necessarily refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0057] Example 1

[0058] like Figures 1-2 As shown, the present invention provides a technical solution: a method for predicting the mechanical response of a polyethylene gas pipeline, comprising:

[0059] S1: A PE gas pipeline finite element model is constructed based on the characteristics of the PE gas pipeline and the multi-factor influencing data of its laying requirements;

[0060] It should be noted that the inherent characteristics of PE gas pipelines include mechanical parameters and geometric parameters. Among them, mechanical parameters specifically include elastic modulus, Poisson's ratio, etc. The mechanical parameters can refer to ISO 4427 (Polyethylene Pipe System Specification GB15558.1-2015 (Buried Polyethylene Pipe System for Gas) and EN 1555 (European Standard for PE Pipe for Gas); geometric parameters specifically include pipe diameter, wall thickness, etc. The geometric parameters can refer to SDR (Standard Dimension Ratio): such as SDR11 and SDR17, which are specified by ISO4427 and represent the ratio of pipe diameter to wall thickness. Design pressure corresponds to wall thickness: selected according to the pressure level in GB15558.1 (such as 0.4MPa, 0.8MPa); PE gas pipeline laying requirements include ambient temperature (obtained through meteorological data or design specifications (such as minimum / maximum operating temperature, usually -20°C to 40°C), which can define the temperature field and correct the material elastic modulus), and load (which can be obtained by referring to the "Highway Engineering Technical Standard" (JTG B01) in the vehicle load (e.g. equivalent uniformly distributed load 700kN / m 2 ), which can be reflected by applying uniformly distributed pressure or concentrated force on the top of the pipeline), settlement depth (obtained through geological exploration reports or soil mechanical parameters (such as settlement, soil elastic modulus), simulating uneven foundation settlement through displacement boundary conditions, or combining with soil constitutive models), pipeline internal pressure (design pressure (such as 0.4 MPa) or actual operating pressure monitoring data, which can reflect the uniform pressure load applied to the inner wall), and pipeline burial depth (design specifications (such as GB 50028 stipulates a minimum cover depth of 0.8-1.2 m, which can be used to calculate the vertical soil pressure (yx burial depth) and apply it as an external load);

[0061] S2: Construct a 3D geometric model based on the finite element model and obtain a PE gas pipeline dataset sample;

[0062] It should be noted that the PE gas pipeline dataset sample includes the displacement component and the Mises stress of the PE gas pipeline.

[0063] S3: Normalize the multi-factor impact data and the PE gas pipeline dataset samples to obtain the overall sample data, and divide the overall sample data into training sample data and test sample data;

[0064] Specifically, the pipeline characteristics, ambient temperature, occupation load, subsidence depth, pipeline internal pressure, and pipeline burial depth in the multi-factor influencing data are used as input features of the neural network (the neural network used here is the Back Propagation Neural Network (BPNN)), and the displacement component and Mises stress of the PE gas pipeline are used as output features; then the pipeline characteristics, ambient temperature, occupation load, subsidence depth, pipeline internal pressure, pipeline burial depth, displacement component and Mises stress are normalized to obtain the overall sample data, and the overall sample data is randomly shuffled and divided into training sample data and test sample data according to a ratio of 7:3. The training sample data and test sample data are used for model training and evaluation of the final model performance, respectively.

[0065] Furthermore, in step S3, the normalization formula is:

[0066]

[0067] Where f(x) is the normalized data; x is the data set (specifically referring to the following two types of data: 1. The input feature data set, which contains 7 features: pipeline diameter, pipeline length, ambient temperature, occupied load, subsidence depth, pipeline internal pressure, and pipeline burial depth; 2. The output feature data set, which contains 4732 output units: the displacement component (X / Y / Z direction) and Mises stress of each finite element unit); x min is the minimum value of the data set; x max is the maximum value of the data set.

[0068] S4: Establish a back-propagation neural network model, determine the parameters of the back-propagation neural network model, and then initialize the parameters;

[0069] It should be noted that the architecture refers to the number of main neural network layers and the number of input and output neurons. The related parameters are the number of hidden layer neurons. The specific confirmation method will be provided later. The model parameter initialization is to initialize the weights and biases. These weights and biases are the initial weights and biases, which are randomly generated and are usually random numbers between [-1,1].

[0070] In this step, the back propagation neural network model includes an input layer, a hidden layer, and an output layer. The parameters of the back propagation neural network model include weights, biases, and transfer functions. The number of neurons in the input layer is 7, representing the pipeline diameter, pipeline length, ambient temperature, occupation load, subsidence depth, pipeline internal pressure, and pipeline burial depth. The number of neurons in the output layer is 4732, representing the Mises stress and displacement of all units in the PE gas pipeline.

[0071] Specifically, the parameter weights of a backpropagation neural network model are determined by random initialization (such as Xavier or He initialization) before training and self-adjusted during training using the backpropagation algorithm. Similarly, the bias (incorrectly called "bias") is determined by initialization to a small random number and dynamically adjusted during training. The beneficial effects of parameter setting in a backpropagation neural network model include accelerated convergence (properly initializing weights and biases can prevent vanishing / exploding gradients) and improved accuracy (appropriate activation functions can enhance the model's expressiveness).

[0072] Furthermore, the number of hidden layer units is N hid The calculation formula is:

[0073]

[0074] Among them, N hid is the number of hidden layer neurons, N input is the number of neurons in the input layer, N output is the output layer neuron, c is an integer between 2 and 10, and s is the number of samples.

[0075] Furthermore, the expression of the transfer function (activation function) is as follows:

[0076]

[0077] Among them, n is the hidden layer input.

[0078] S5: Input the training sample data into the back propagation neural network model for training, and evaluate the model through the test sample data to build a back propagation neural network model for PE gas pipeline under multiple factors;

[0079] S6: Determine the initial weights and biases of the back-propagation neural network model, and apply them to the back-propagation neural network model for optimization iteration to obtain an optimized back-propagation neural network model.

[0080] Furthermore, step S1 specifically includes:

[0081] S11: The complex ground loads on buried PE pipelines are transmitted through the soil. Therefore, the properties of the soil have a significant impact on the structural mechanical behavior of the pipeline, and the soil properties are reflected through its constitutive model. Considering the common problems and conditions in the modeling and analysis process, this paper assumes that environmental factors only consider the impact of ambient temperature on the mechanical properties of the pipeline, and does not consider the impact of ambient temperature on the mechanical properties of the soil. Considering the type and properties of the pipeline fill, the soil model adopts the Drucker-Prager model commonly used in engineering.

[0082] S12: The material properties of PE pipes (such as elastic modulus, yield stress, etc.) will change with temperature. In order to accurately simulate this effect in the finite element model, the material parameters of PE pipes at different temperatures need to be clearly defined, as shown in the table below:

[0083] Table 1 PE pipe parameters at different temperatures

[0084]

[0085] S13. Simulate and calculate the PE gas pipeline affected by multiple factors to obtain its stress results.

[0086] Simulation calculation: After considering various influencing factors (such as ambient temperature, fill type and properties, soil model, etc.), a complete finite element model is constructed to simulate the PE gas pipeline.

[0087] Stress results: Through simulation calculation, the stress results of the pipeline under different conditions are obtained (the stress results are displayed in the following four ways: 1. Contour Plot: 1.1 Mises stress contour plot: The stress distribution on the pipeline surface is displayed in a color gradient, and red indicates high stress areas (such as those near the pressure load). 1.2 Displacement contour plot: Displays the deformation of the pipeline, such as the maximum vertical displacement of 3.2mm. 2. Vector Plot: 2.1 Displacement vector: The arrow indicates the direction and magnitude of the displacement, which is used to analyze the overall deformation trend of the pipeline. 3. Data tables and curves: 3.1 Key point data: Extract the stress-time curve of high-risk areas (such as elbows and interfaces). 3.2 Statistical values: Maximum stress, average stress, safety factor (yield strength / actual stress). 4. Animation demonstration: 4.1 Dynamic deformation process: Displays the progressive deformation and stress evolution of the pipeline during load application), including key mechanical response indicators such as Mises stress; these results can be used to further analyze the safety and reliability of the pipeline in actual applications.

[0088] It should be noted that the finite element model (see Figure 3 Specific construction steps shown):

[0089] 1. Geometric modeling

[0090] Tool selection: Use CAD software (AutoCAD) and finite element pre-processing software (ABAQUS) to create three-dimensional geometric models.

[0091] Pipeline model: Generate cylindrical pipes based on design parameters (pipe diameter, wall thickness, length).

[0092] Soil model: Construct a rectangular or cylindrical soil area around the pipeline to simulate the buried environment. The soil size needs to cover at least five times the diameter of the pipeline to avoid boundary effects.

[0093] 2. Material parameter definition

[0094] PE pipe:

[0095] Elastic modulus: Look up the table according to the temperature (for example, the elastic modulus of PE100 at -15°C is 595.82 MPa, see Table 1).

[0096] Poisson's ratio: fixed value 0.45 (refer to GB15558.1).

[0097] Yield strength: adjusted according to temperature (e.g. 23.08 MPa at -15°C).

[0098] soil:

[0099] Constitutive model: Drucker-Prager model (Document S12)

[0100] Parameter input: including density, internal friction angle, cohesion, etc.

[0101] 3. Grid division

[0102] Pipeline mesh: Use hexahedral elements (e.g., C3D8R) for structured meshing, with 3-5 layers in the wall thickness direction. Soil meshes use tetrahedral elements (e.g., C3D4) or reduced integration elements (C3D8R), with mesh size increasing gradually (getting denser near the pipeline).

[0103] Contact surface treatment: define the contact pair between the outer wall of the pipe and the inner wall of the soil, and set the friction coefficient to 0.4.

[0104] 4. Boundary conditions and load application

[0105] Pipeline constraint: Axial free constraints are imposed on both ends (simulating the real displacement limit under the soil surface).

[0106] Soil boundary: The bottom is fixed (simulating bedrock) and the sides restrict horizontal displacement.

[0107] Load Type:

[0108] Internal pressure: Apply uniform pressure (such as 0.4MPa) on the inner wall of the pipe.

[0109] Occupancy load: A uniformly distributed pressure (e.g. 70 kN / m2, simulating vehicle load) is applied to the soil surface at the top of the pipeline.

[0110] Temperature field: define the ambient temperature gradient and correct the material elastic modulus.

[0111] Settlement displacement: Simulate foundation settlement by applying non-uniform displacement to the bottom of the soil.

[0112] It should be noted that the finite element simulation calculation process is as follows:

[0113] 1. Solver Settings

[0114] Solution type: Static structural analysis (for long-term loads) or transient analysis (for dynamic loads)

[0115] Nonlinear options: Enable large deformation effect (PE material may undergo plastic deformation)

[0116] 2. Multiphysics coupling

[0117] Temperature-mechanical coupling: Map temperature field data to structural fields and dynamically adjust material parameters.

[0118] Soil-Structure Interaction: Calculates the interaction forces between soil and pipelines using a contact algorithm.

[0119] 3. Calculation and convergence control

[0120] Step size selection: Adaptive step size control ensures convergence stability. Convergence criteria: displacement residual <1e-5, force residual <1e-3.

[0121] 4. Post-processing and result extraction

[0122] Stress Results: Extract VonMisesStress distribution

[0123] Displacement results: Extract the displacement components of the pipeline nodes (X / / Z directions).

[0124] Furthermore, the steps of constructing a three-dimensional geometric model based on the finite element model and obtaining a PE gas pipeline data set sample include:

[0125] S21: Generate a cylindrical pipe model based on the design parameters (pipe diameter, wall thickness, length) of the PE gas pipe of the finite element model;

[0126] S22: Add soil layers around the pipeline to simulate buried conditions;

[0127] It should be noted that this step is based on environmental modeling, which can more realistically simulate the environment in which the pipeline is laid and buried in the ground, and the soil layer is a rectangular or cylindrical area.

[0128] S23: Process the contact surface between the pipeline and the soil (interface processing to ensure the accuracy of load transfer) and construct a three-dimensional geometric model of the pipeline;

[0129] It should be noted that three-dimensional geometric models are usually constructed through CAD software or finite element pre-processing software (such as ABAQUS).

[0130] S24: Mesh the three-dimensional geometric model and discretize it into several finite element units;

[0131] Among them, the grid division process:

[0132] 1. Automatic meshing: Use the automatic mesh generation function of the finite element software to generate tetrahedral (Tetra) or hexahedral (Hexa) units according to the geometric shape;

[0133] 2. Local encryption: Manually encrypt the grid in stress concentration areas (such as the interface between the pipe and the soil, and elbows) to improve calculation accuracy.

[0134] 3. Element type selection: Shell elements (such as S4R, 4-node shell elements) are used to simplify wall pipe analysis.

[0135] The unit display format is as follows:

[0136] Wireframe mode: displays element boundaries and nodes (useful for checking mesh quality).

[0137] Solid rendering mode: Display stress, strain and other results through color mapping (such as Mises stress cloud map).

[0138] S25: Assign different mechanical parameters to the pipe material according to the different temperatures of the PE gas pipe to obtain a PE gas pipe dataset sample.

[0139] Among them, different mechanical parameters are assigned to pipeline materials based on reference to industry standards, such as elastic modulus and Poisson's ratio: refer to GB15558.1-2015 or IS04427 (for example, the elastic modulus of PE100 at room temperature is 0.8-1.2GPa); yield strength: based on the long-term hydrostatic strength specified in the standard (for example, the yield stress of PE100 is 23MPa).

[0140] Furthermore, the training sample data is input into the back propagation neural network model for training, and the model is evaluated by the test sample data. The steps of constructing the back propagation neural network model of PE gas pipeline under multiple factors include:

[0141] S51: Initialize the weights and biases of the backpropagation neural network model, import the input features and output features, and then obtain the predicted value through forward propagation calculation;

[0142] Among them, the expression of the loss function is as follows:

[0143]

[0144] Where: t i is the finite element numerical simulation value of the displacement component and Mises stress of the i-th sample, o iis the neural network prediction value of the displacement component and Mises stress of the i-th sample, and n is the total number of samples.

[0145] Among them, the input features are 7 parameters of pipeline mechanical response (pipeline diameter, length, ambient temperature, occupation load, subsidence depth, pipeline internal pressure, and pipeline burial depth). Each data in the training sample contains these 7 input features; the output features are the displacement components (XZ directions) and Mises stress (a total of 4732 output values) of all units in the finite element model: the output feature of each training sample is the real mechanical response value obtained by finite element simulation calculation; the training sample data is each sample as an (input feature, output feature) pair, for example: sample = ([D, L, T, P, S, p, h], [σ1, u x1 ,u y1 ,u z1 ,.....,σ n ])

[0146] Where D is the pipe diameter, σ is the Mises stress, and u is the displacement.

[0147] It should be noted that forward propagation is the layer-by-layer calculation process of input data from the input layer through the hidden layer to the output layer. The steps are as follows:

[0148] 1. Input layer → hidden layer:

[0149] Hidden layer input = Σ(weight 输入→隐藏 × input features) + bias 隐藏

[0150] 2. Activation via transfer function:

[0151] Hidden layer output = f(hidden layer input)

[0152] 3. Hidden layer → output layer:

[0153] Output layer input = Σ(weight 隐藏→输出 × hidden layer output) + bias 输出

[0154] S52: Calculate the loss function based on the predicted value and the true label, and use the backpropagation algorithm to update the weights and biases to reduce the value of the loss function;

[0155] It should be noted that the displacement components and Mises stresses of the real labels are obtained by finite element simulation (data set samples generated in step S2); the acquisition steps are:

[0156] 1. Steps S1-S2: Build a finite element model and simulate it to generate a data set containing input features and true outputs.

[0157] 2. Step S3: Use the true output as the label and pair it with the input features to form a training sample.

[0158] It should be noted that the backpropagation algorithm calculates the gradient of the loss function with respect to the weights and adjusts the parameters backward from the output layer to the input layer to minimize the error. The specific steps are as follows:

[0159] 1. Calculate the loss function: mean square error (MSE)

[0160]

[0161] Among them, t i is the true label, o i is the predicted value.

[0162] 2. Back propagation gradient, output layer gradient:

[0163]

[0164] 3. Update weights and biases, using gradient descent to calculate the gradient:

[0165]

[0166] Update parameters:

[0167]

[0168] Where η is the learning rate (hyperparameter).

[0169] S53: Adjust the parameters of the neural network based on the gradient descent method to minimize the loss function;

[0170] It should be noted that the gradient descent method calculates the gradient of the loss function with respect to the neural network parameters (weights, biases), and adjusts the parameters in the opposite direction of the gradient to minimize the loss function. The steps are as follows:

[0171] 1 Calculate the gradient:

[0172]

[0173] 2 Update parameters:

[0174]

[0175] Where η is the learning rate (hyperparameter).

[0176] Among them, the gradient direction indicates the direction in which the loss function grows fastest, and the reverse adjustment parameters move the loss function toward the local minimum.

[0177] S54: Repeat the above steps S51 to S53 until the loss function converges or reaches the set stopping condition, and the training is completed;

[0178] S55: Use the test sample data to evaluate the trained back propagation neural network model, and optimize the hyperparameters of the back propagation neural network model based on the evaluation results to obtain the back propagation neural network model of the mechanical response of the PE gas pipeline under the influence of multiple factors.

[0179] It should be noted that the evaluation criteria in this step are as follows:

[0180] 1. Mean Squared Error (MSE): Measures the average deviation between the predicted value and the true value.

[0181] 2. Coefficient of determination (R2): reflects the model's ability to explain data changes.

[0182] 3. Maximum absolute error (MAE): directly reflects the magnitude of the prediction error.

[0183] Specific evaluation steps:

[0184] 1. Use the test set data to input the trained model to obtain the predicted value;

[0185] 2. Calculate the error index (such as MSE) between the predicted value and the true value;

[0186] 3. Judge the model generalization ability based on the error index.

[0187] It should be noted that the hyperparameters in this step are the learning rate (η), the number of hidden layers, and the number of neurons, which need to be set manually and affect the model structure and training process; the parameters of the neural network refer to the weights and biases, which are automatically adjusted through backpropagation during training.

[0188] Attachment Figure 1 The input data is multi-factor characteristic data related to the mechanical response of polyethylene (PE) gas pipelines, including: 1. Mechanical parameters: elastic modulus, Poisson's ratio, yield strength (temperature-dependent); 2. Geometric parameters: pipe diameter, wall thickness, and pipe length; 3. Environmental and load parameters: ambient temperature, occupancy load (e.g., vehicle load), subsidence depth, internal pipeline pressure, and pipeline burial depth; 4. Output target values: displacement components (X / Z directions) and Mises stress obtained through finite element simulation. Preprocessing: 1. Data cleaning (handling missing values: deleting or interpolating (e.g., mean or median filling) missing data; outlier removal: identifying and removing anomalous samples using boxplots or the 3σ principle); 2. Normalization; 3. Feature engineering (feature selection: selecting important features through correlation analysis; feature construction: generating new features based on domain knowledge, such as stress concentration factors); 4. Data partitioning (dividing the dataset into training and test sets in a 7:3 ratio).

[0189] S7: Input the new input feature data into the optimized back-propagation neural network model.

[0190] It should be noted that the newly input feature data is collected through sensors (dynamic data such as ambient temperature and internal pipe pressure are acquired in real time through sensors embedded around the pipeline (such as pressure sensors and temperature sensors); design parameters are simultaneously input (geometric parameters such as pipe diameter, wall thickness, and burial depth are directly entered through design documents or a database); and then external loads are calculated (occupancy loads (such as vehicle loads) are estimated through traffic flow statistics or on-site monitoring data). The new input feature data is presented in a structured data table (presented in tabular form, containing the values of seven input features (for example: pipe diameter 200mm, ambient temperature 25°C, occupancy load 50kN / m2)).

[0191] The back-propagation neural network model calculates based on the input features and outputs the predicted values of the mechanical response of the PE gas pipeline, including displacement components and Mises stress. The displacement components and Mises stress data in this step come from the optimized neural network that directly calculates the output based on the new input features. It is used to predict the mechanical response of the pipeline under actual working conditions and is used for engineering decision-making.

[0192] The displacement components and Mises stress output by analytical models (models that directly describe the mechanical behavior of pipelines through mathematical or physical equations, such as the equations of elasticity) are combined with engineering context and standards to determine whether the pipeline is within the safe range under the predicted conditions. (This is based on Mises stress (predicted value ≤ material yield strength (e.g., 23 MPa for PE100)) and displacement components (displacement ≤ design allowable deformation (e.g., 5% of the pipe diameter)). For example, if the predicted stress is 20 MPa (<23 MPa), the pipeline is safe; if it is 25 MPa (>23 MPa), a warning and standard are required.) For example, the predicted Mises stress exceeds the material yield strength to assess the pipeline's structural integrity.

[0193] It should be noted that the engineering background includes pipeline material properties (such as the yield strength of PE100), buried environment (soil type), operating conditions (design pressure), etc.; relevant standards: GB15558.1-2015 (specifies the long-term hydrostatic strength of PE gas pipelines (such as the yield stress of PE100 is 23MPa)) and ISO4427 (defines the relationship between pipeline design pressure and wall thickness (such as SDR11 corresponding to wall thickness calculation)).

[0194] Combination of engineering background and standards: Compare the predicted Mises stress with the allowable stress in the standard. If the predicted stress is ≤ allowable stress, it is considered safe.

[0195] Compared with the prior art, the advantages of the present invention are:

[0196] 1. Improved Computational Efficiency and Accuracy: This method, based on the multi-factor PE gas pipeline mechanical response calculation method, constructs a finite element model under these conditions and combines it with a back-propagation neural network for training and optimization. Compared to traditional mechanical modeling methods, this method significantly improves the speed of calculating the bearing capacity and deformation of PE gas pipelines under these conditions. This method ensures the accuracy and reliability of the calculation results, is capable of handling pipeline mechanical responses under complex conditions, and provides faster and more accurate technical support.

[0197] 2. Efficient Big Data Processing: By combining mechanical simulation and machine learning, this invention can more effectively process big data and accelerate the analysis and evaluation of pipeline mechanical responses under complex conditions. This data-driven algorithm replaces traditional modeling methods, optimizes the analysis process for pipeline design and operation, and improves the overall efficiency and accuracy of the system.

[0198] 3. Strong Adaptability: This method is suitable for situations where a physical model of a PE gas pipeline is difficult or non-existent. Even when a complete physical model cannot be established, the method can calculate response results for a variety of practical scenarios using a limited pipeline model. This feature facilitates early warning and maintenance work when pipelines are subject to multiple complex environmental factors, providing personnel with rapid and accurate pipeline structural assessment results and maintenance decisions.

[0199] Example 2

[0200] Unlike the previous embodiment, since the initial weights and thresholds of the neural network are randomly generated, the results of each neural network training will be different and it is easy to fall into the local optimal solution. Therefore, an optimization algorithm is used to optimize its initial weights and thresholds. Specifically, the backpropagation neural network model is optimized using the Griffin Optimization (GFO) algorithm. The Griffin Optimization (GFO) algorithm replaces the traditional gradient descent method and optimizes the neural network weights and thresholds by simulating the foraging behavior of fish to avoid falling into the local optimal solution. The mechanical response of PE pipelines is affected by multiple factors (temperature, load, etc.) nonlinearly. The global search capability of GFO is more suitable for handling such complex problems.

[0201] Specifically, the optimization steps of the back propagation neural network model using the Ginger Fish Optimization (GFO) algorithm are as follows:

[0202] S61. Randomly generate a certain number of minnow individuals or minnow eggs as an initial population;

[0203] S611. Each minnow or minnow egg represents a solution in the current generation and can be encoded in a D-dimensional vector by the following expression:

[0204]

[0205] Among them: the value of D is the dimension of each solution or the number of decision variables (corresponding to weights and thresholds), and the value of i is also the solution to the problem. represents the j-th dimension of the i-th solution.

[0206] S612. Consider n randomly generated fish populations, the expression is as follows:

[0207]

[0208] Where: The value of F is a prime matrix of n.

[0209] S613. Create an initial population in the interval [1,u] using the following formula:

[0210]

[0211] Where: r is a random number between [0,1].

[0212] S62, determining the individual fitness function;

[0213] In the minnow optimization algorithm, each solution to a problem is a fish. To evaluate each solution, we can focus on its oyster selection. The quality of each fish depends on whether its chosen oyster is suitable and attractive to other female fish. An objective function (e.g., f) is used to evaluate each fish or solution. The capability of each solution is defined in the capability matrix, as shown below:

[0214]

[0215] S63. Search and seizure of oysters

[0216] Any solution or fish can search the problem space and find a suitable oyster to mate with. In the proposed method, each fish is eligible to reside in an area with more optimal shells. A fish can roam in search of oysters, in which case it will target a specific oyster and move toward it. In this case, the target oyster will not be discovered by another fish and will be captured by that fish. The following equation is used to represent the oyster possession state:

[0217]

[0218] in: and are the current and new positions of the i-th solution in iterations t and t+1, F * is the optimal solution, F + is the local optimal solution, δ and r are random numbers between 0 and 1, J is the number of steps or movement speed of the fish to escape from or approach the oyster, and this coefficient is reduced by algorithm iteration.

[0219] The reason why parameter J decreases is that after a period of time, the male fish's activity decreases because they have successfully mated. Over time, the decrease in J causes the global search to shift to a local search. The parameter J is reduced using the following formula:

[0220]

[0221] Where: J(1) is the step length and jump value of each fish in the initial iteration, t and Maxt are the current iteration number and the maximum iteration number respectively, and J(t) is the step length and jump value of each fish in searching for oysters in the tth iteration. The random function U generates a random sequence using the following formula:

[0222] U(t+1)=cos(t×cos -1 (U(t)))

[0223] Among them: U(1) can be considered equal to 1;

[0224] S64. Escape from the oyster, don't catch the oyster

[0225] In the "escape" and "no oyster capture" behaviors, another fish will take care of the oyster, causing the fish to no longer approach the oyster and choose another location. The following formula is used to represent a fish that escapes or accidentally searches for an oyster without successfully capturing it:

[0226]

[0227] The value of M is equal to the average position of the school. In this respect, due to escape, minnows can search the space between the average and the optimal value, or search the random space problem. The following formula is used to calculate the population gravity point:

[0228]

[0229] S65, Production

[0230] After finding an oyster, the male fish can attract the female fish and lay eggs in the oyster. The male fish can fertilize the eggs to produce new fish around the current fish, as shown in the following formula:

[0231]

[0232] Where: R is the radius of the fish's distribution around the shell. This radius has an initial value in the range [0, 2]. Typically, this parameter is initially assumed to be equal to 2 and is decreased over time according to the iterations of the BFO algorithm.

[0233] S66, Fishing

[0234] A single oyster can hold several juveniles, and some of the weaker juveniles may be killed by the oyster or predators. The probability of losing a solution can be considered to be inversely proportional to its merit. Assume that the problem is of the minimization type, and the probability of losing a fish is calculated as follows:

[0235]

[0236] Among them: Delete solution The probability of is the objective function value of the solution.

[0237] It should be emphasized that the specific way in which each CNC of the polyethylene (PE) gas pipe is embedded or used in S61 to S66 is as follows:

[0238] Each "crab mackerel individual" represents an encoding of a set of neural network weights and biases (step S611).

[0239] D-dimensional vector: The dimension D is equal to the total number of neural network parameters (such as the number of weights + the number of biases).

[0240] Fitness evaluation: Use the current parameters to calculate the neural network prediction results and evaluate the fitness through the loss function value (step S62); the smaller the loss function value (the more accurate the prediction), the higher the individual fitness.

[0241] Example 3

[0242] The difference from the previous embodiment is that: a system for predicting the mechanical response of a polyethylene gas pipeline is provided, including:

[0243] Model building module, used to build the finite element model of PE gas pipeline and obtain multi-factor influencing data;

[0244] The geometric modeling and meshing module is connected to the model building module to build a three-dimensional geometric model and obtain a sample of the PE gas pipeline dataset;

[0245] The data processing and preprocessing module is used to normalize the multi-factor influence data and the PE gas pipeline data set samples to obtain the overall sample data, and divide the overall sample data into training sample data and test sample data; it is connected to the geometric modeling and meshing module;

[0246] Neural network modeling module, used to establish a back-propagation neural network model, determine its architecture and related parameters, and initialize the model parameters;

[0247] The model training module is connected to the neural network modeling module and the data processing and preprocessing module, which inputs the training sample data into the neural network model for training and constructs a back propagation neural network model of the PE gas pipeline under the influence of multiple factors;

[0248] The model optimization module determines the initial weights and biases of the back-propagation neural network model and improves the model performance through an iterative optimization process, ultimately obtaining an optimized back-propagation neural network model. This module is connected to the model training module.

[0249] The prediction and evaluation module feeds the new input feature data into the optimized back-propagation neural network model, calculates and outputs the predicted value of the mechanical response of the PE gas pipeline, including displacement components and Mises stress, analyzes the model output results, and combines the engineering background and standards to determine whether the pipeline is within the safe range.

[0250] The specific corresponding products / components and functions of each module are as follows:

[0251]

[0252] Example 4

[0253] The difference from the previous embodiment is that this embodiment provides a comparative experiment, specifically,

[0254] 1. This comparative experiment provides three existing technical solutions for comparison with the method of the present invention. The three existing methods are as follows:

[0255] Option 1: Traditional finite element analysis

[0256] Method: Completely rely on finite element simulation (such as ANSYS or ABAQUS), and each time a new working condition is required, new modeling, meshing, and stress / displacement calculation are required.

[0257] Disadvantages: The calculation takes a long time (a single simulation takes several hours to several days) and cannot quickly respond to dynamically changing engineering needs.

[0258] Solution 2: Single machine learning model (linear regression) Method: Use a linear regression model to directly predict Mises stress based on input features (such as pipe diameter and temperature).

[0259] Disadvantages: The model has limited expressive power, cannot capture the nonlinear relationship between multiple factors, and has large prediction errors.

[0260] Solution 3: Unoptimized Shallow Neural Network

[0261] Method: A three-layer BP neural network was used, and the input features and output features were the same as those of the present invention, but no optimization algorithm (such as GFO) was used.

[0262] Disadvantages: easy to fall into local optimality, long training time, and weak generalization ability.

[0263] 2. In order to ensure the fairness of the experiment, a comparative experiment was conducted under the same conditions. The equivalent conditions are as follows:

[0264] Dataset: The same set of PE gas pipeline operating data (1000 samples), including 7 input features and 4732 output features (displacement and stress)

[0265] Hardware environment: GPU, 32GB memory;

[0266] Test scenario:

[0267] Scenario 1: Static load (constant occupied and internal pressure);

[0268] Scenario 2: Dynamic temperature change (-20°C to 40°C);

[0269] Scenario 3: Soil subsidence (non-uniform foundation settlement).

[0270] 3. Comparative Experimental Results

[0271] The performance of the present invention is compared with that of the prior art by the following indicators:

[0272] Computational efficiency (time required for a single prediction):

[0273]

[0274] Prediction accuracy (mean square error, MSE):

[0275] plan Static load Dynamic temperature Soil subsidence Traditional finite element method (Scheme 1) 0 (baseline) 0 (baseline) 0 (baseline) Linear regression (Scheme 2) 12.5 18.7 25.3 Shallow Neural Network (Solution 3) 3.2 5.8 8.1 The present invention (GFO-BPNN) 0.9 1.5 2.3

[0276] As attached Figure 4 As shown in the figure, under complex working conditions, the prediction error of the present invention is between 0 and 5%, the prediction error of the shallow neural network (Scheme 3) is between 10 and 20%, and the prediction error of the linear regression (Scheme 2) is between 30 and 50%. Through this comparison, it is not difficult to find that the prediction error of the present invention is much lower than that of the shallow neural network (Scheme 3) and linear regression (Scheme 2), which further demonstrates that the generalization ability of the method of the present invention is strong;

[0277] The results in the above table and accompanying experimental figures clearly demonstrate that the present invention is more efficient, accurate, and adaptable than existing methods in calculating the mechanical response of PE gas pipelines. It can handle pipeline mechanical responses under complex conditions, providing faster and more accurate technical support. It also more effectively processes big data, accelerating the analysis and evaluation of pipeline mechanical responses under complex conditions. This data-driven algorithm model replaces traditional modeling methods, optimizes the analysis process for pipeline design and operation, and improves the overall efficiency and accuracy of the system. It also facilitates early warning and maintenance work when multiple complex environmental factors are present in the pipeline, providing relevant personnel with rapid and accurate pipeline structural assessment results and maintenance decisions.

[0278] Example 5

[0279] Unlike the previous embodiment, the present invention provides a terminal device comprising a processor and a memory, wherein the memory is used to store a computer program, the computer program including program instructions, and the processor is used to execute the program instructions stored in the computer storage medium. The processor may be a central processing unit (CPU), or may be another general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal and is suitable for implementing one or more instructions, specifically for loading and executing one or more instructions to implement the corresponding method flow or corresponding function. The processor described in the embodiment of the present invention can be used to operate the polyethylene gas pipeline mechanical response prediction method.

[0280] In another embodiment of the present invention, a storage medium is provided, specifically a computer-readable storage medium (Memory), which is a memory device in a terminal device for storing programs and data. It can be understood that the computer-readable storage medium here can include both the built-in storage medium in the terminal device and, of course, the extended storage medium supported by the terminal device. The computer-readable storage medium provides a storage space, which stores the operating system of the terminal. In addition, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space. These instructions can be one or more computer programs (including program codes). It should be noted that the computer-readable storage medium here can be a high-speed RAM memory or a non-volatile memory, such as at least one disk memory.

[0281] One or more instructions stored in a computer-readable storage medium can be loaded and executed by a processor to implement the corresponding steps of the polyethylene gas pipeline mechanical response prediction method in the above embodiment; one or more instructions in the computer-readable storage medium are loaded and executed by the processor.

[0282] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0283] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0284] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0285] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0286] Those skilled in the art will appreciate that the embodiments described herein are intended to help readers understand the implementation methods of the present invention, and it should be understood that the scope of protection of the present invention is not limited to such specific descriptions and embodiments. Those skilled in the art can make various other specific variations and combinations based on the technical teachings disclosed in the present invention without departing from the essence of the present invention, and such variations and combinations are still within the scope of protection of the present invention.

Claims

1. A method for predicting the mechanical response of a polyethylene gas pipeline, characterized by: include, Based on the characteristics of PE gas pipeline and the multi-factor influencing data of its laying requirements, a PE gas pipeline finite element model is constructed; A 3D geometric model was constructed based on the finite element model, and a sample PE gas pipeline dataset was obtained; Normalize the multi-factor impact data and the PE gas pipeline dataset samples to obtain the overall sample data, and divide the overall sample data into training sample data and test sample data; Establish a back-propagation neural network model, determine the parameters of the back-propagation neural network model, and then initialize the parameters; The training sample data is input for training, and the model is evaluated through the test sample data to build a back propagation neural network model for PE gas pipelines under multiple factors; Determine the initial weights and biases of the back-propagation neural network model, and apply them to the back-propagation neural network model for optimization iteration to obtain an optimized back-propagation neural network model.

2. The method for predicting the mechanical response of a polyethylene gas pipeline according to claim 1, wherein: According to the inherent characteristics of PE gas pipelines, including mechanical parameters and geometric parameters; the laying requirements of PE gas pipelines include ambient temperature, pressure load, subsidence depth, pipeline internal pressure and pipeline burial depth.

3. The method for predicting the mechanical response of a polyethylene gas pipeline according to claim 1 or 2, wherein: The steps of building a three-dimensional geometric model based on the finite element model and obtaining a PE gas pipeline data set sample include: Generate a cylindrical pipe model based on the PE gas pipeline of the finite element model; Add soil layers around the pipeline to simulate buried conditions; Process the interface between the pipeline and the soil and construct a three-dimensional geometric model of the pipeline; Divide the three-dimensional geometric model into several finite element units; Different mechanical parameters are assigned to the pipeline material according to the different temperatures of the PE gas pipeline, and a PE gas pipeline dataset sample is obtained.

4. The method for predicting the mechanical response of a polyethylene gas pipeline according to claim 3, wherein: The PE gas pipeline dataset sample includes the displacement components and Mises stress of the PE gas pipeline.

5. The method for predicting the mechanical response of a polyethylene gas pipeline according to claim 4, wherein: The normalization formula is: Among them, f(x) is the normalized data; x is the data of the dataset; x min is the minimum value of the data set; x max is the maximum value of the data set.

6. The method for predicting the mechanical response of a polyethylene gas pipeline according to claim 5, wherein: The back propagation neural network model includes an input layer, a hidden layer and an output layer; Among them, the number of units in the hidden layer is N hid The calculation formula is: Among them, N hid is the number of hidden layer neurons, N input is the number of neurons in the input layer, N output is the output layer neuron, c is an integer between 2 and 10, and s is the number of samples.

7. The method for predicting the mechanical response of a polyethylene gas pipeline according to claim 6, wherein: The training sample data is input into the back propagation neural network model for training, and the model is evaluated by the test sample data. The steps of constructing the back propagation neural network model of PE gas pipeline under multiple factors include: Initialize the weights and biases of the backpropagation neural network model, import the input features and output features, and then obtain the predicted value through forward propagation calculation; Calculate the loss function based on the predicted value and the true label, and use the backpropagation algorithm to update the weights and biases to reduce the value of the loss function; Adjust the parameters of the neural network based on the gradient descent method to minimize the loss function; Repeat the above steps until the loss function converges or the set stopping condition is reached, and the training is completed; The trained back propagation neural network model is evaluated using test sample data, and the hyperparameters of the back propagation neural network model are optimized based on the evaluation results to obtain a back propagation neural network model of the mechanical response of PE gas pipelines under the influence of multiple factors.

8. The method for predicting the mechanical response of a polyethylene gas pipeline according to claim 7, wherein: The formula of the loss function is, Among them, t i is the finite element numerical simulation value of the displacement component and Mises stress of the i-th sample, o i is the neural network prediction value of the displacement component and Mises stress of the i-th sample, and n is the total number of samples.

9. The method for predicting the mechanical response of a polyethylene gas pipeline according to claim 8, wherein: The back propagation neural network model is optimized using the minnow optimization algorithm.

10. A system for use in the method for predicting mechanical response of a polyethylene gas pipeline according to any one of claims 1 to 9, characterized in that: include, Model building module, used to build the finite element model of PE gas pipeline and obtain multi-factor influencing data; The geometric modeling and meshing module is connected to the model building module to build a three-dimensional geometric model and obtain a sample of the PE gas pipeline dataset; The data processing and preprocessing module is used to normalize the multi-factor influence data and the PE gas pipeline data set samples to obtain the overall sample data, and divide the overall sample data into training sample data and test sample data; It is connected with the geometry modeling and meshing modules; Neural network modeling module, used to establish a back-propagation neural network model, determine its architecture and related parameters, and initialize the model parameters; The model training module is connected to the neural network modeling module and the data processing and preprocessing module, which inputs the training sample data into the neural network model for training and constructs a back propagation neural network model of the PE gas pipeline under the influence of multiple factors; The model optimization module determines the initial weights and biases of the back-propagation neural network model and improves the model performance through an iterative optimization process, ultimately obtaining an optimized back-propagation neural network model. This module is connected to the model training module. The prediction and evaluation module feeds the new input feature data into the optimized back-propagation neural network model, calculates and outputs the predicted value of the mechanical response of the PE gas pipeline, including displacement components and Mises stress, analyzes the model output results, and combines the engineering background and standards to determine whether the pipeline is within the safe range.

Images