Multi-objective intelligent optimization method for stability control of pipeline sinking process

Abstract

The invention discloses a multi-objective intelligent optimization method for stability control of a pipeline sinking process, and the method comprises the steps: S1, building a pipeline sinking database, and determining design variable characteristic parameters and performance index parameters; s2, establishing machine learning prediction models, and selecting the machine learning prediction model with the highest prediction precision to be applied to subsequent constraint conditions by comparing prediction results, so as to establish a pipeline sinking most stable optimization model; s3, establishing a multi-objective function with minimum cable force and minimum pipeline position deviation as targets; s4, setting constraint conditions; s5, proposing an intelligent optimization design method for the pipeline sinking stability under multi-target requirements; and S6, evaluating and screening out a matching scheme which is optimal in the aspects of economy and safety. According to the method, the stability and safety of pipeline sinking construction can be effectively improved, the construction cost and risk can be effectively reduced, and the engineering construction efficiency and economic benefits can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of immersed tube tunnel construction, and in particular to a multi-objective intelligent optimization method for stability control during a pipeline sinking process. Background Art

[0002] The study on the stability of the laying process of long and irregular pipelines is of great significance for improving the safety of pipeline laying projects, ensuring project quality, reducing maintenance costs, adapting to complex environments, promoting technological innovation, improving economic benefits, coping with extreme conditions, and promoting regional development.

[0003] In the existing immersed tube tunnel construction process, the construction technology is that engineers comprehensively calculate the theoretical value of ballast water, the tide level of the construction water area, the location of the immersed tube tunnel, the manually measured offset and settlement of the immersed tube tunnel and other data, and then use the engineers' technical experience to direct underwater divers and workers in the immersed tube to operate the water tank valves to achieve the purpose of controlling the underwater posture of the immersed tube tunnel.

[0004] The above-mentioned immersed tube tunnel construction process is highly dependent on the engineer's experience level and manual operation level, which will limit the construction accuracy, stability and progress, and also increase the construction risk and cost. Therefore, there is an urgent need for automated and intelligent solutions to improve construction efficiency and quality.

[0005] Since the pipeline is long and may be bent, the traditional construction method based on engineer experience and manual operation is difficult to ensure the stability of the pipeline during the laying process and the accuracy of the posture adjustment, which may cause the pipeline to deviate, tilt or even be damaged during the laying process, seriously affecting the construction quality and safety. In fact, the control factors of pipeline laying stability include water injection time, wave flow, wind, etc. The pipeline laying process is to optimize multiple variables (such as water injection time, waves, water flow, wind), multiple performance (pipe displacement, pipe body angle, pipe deviation angle) and multiple objectives (minimizing cable force, minimizing pipe position deviation) at the same time. When dealing with this multi-objective optimization problem, traditional methods often consume a lot of manpower and material resources, and the effect is limited. Summary of the invention

[0006] The purpose of the present invention is to provide a multi-objective intelligent optimization method for stability control of the pipeline sinking process in view of the deficiencies in the prior art. The multi-objective intelligent optimization method for stability control of the pipeline sinking process can effectively improve the stability and safety of pipeline sinking construction, effectively reduce construction costs and risks, and effectively improve engineering construction efficiency and economic benefits.

[0007] To achieve the above object, the present invention is implemented through the following technical solutions.

[0008] A multi-objective intelligent optimization method for stability control of a pipeline sinking process includes the following steps:

[0009] Step S1, establishing a pipeline laying database, collecting design variable characteristic parameters and performance index parameters that affect pipeline laying stability, and taking water injection time, waves, water flow, and wind as design variable characteristic parameters, and taking pipeline displacement, pipe body rotation angle, and pipeline deviation angle as performance index parameters;

[0010] Step S2, establishing a machine learning prediction model for pipeline displacement, pipe body rotation angle, and pipeline deviation angle, wherein the established machine learning prediction model includes an adaptive enhancement machine learning prediction model, a multi-layer perceptron machine learning prediction model, and a gradient boosting tree machine learning prediction model;

[0011] In the above machine learning prediction model, the characteristic parameters of each design variable of pipeline laying are used as input, and the pipeline position, pipe body posture, and pipeline deviation angle are used as output. The characteristic parameters of each design variable are: water injection time x1, wave x2, water flow x3, wind x4;

[0012] The Bayesian optimization based on the probability model is used to determine the optimal hyperparameter combination of each machine learning prediction model. By comparing the prediction results of the three machine learning prediction models, namely the adaptive enhancement machine learning prediction model, the multi-layer perceptron machine learning prediction model, and the gradient boosting tree machine learning prediction model, the machine learning prediction model with the highest prediction accuracy is selected and applied to the subsequent constraints, so as to establish the most stable optimization model for pipeline laying.

[0013] Step S3, constructing the objective function of multi-objective optimization of pipeline laying. Based on the machine learning prediction model selected in step S2, a multi-objective function is constructed with minimizing cable force and minimizing pipeline position deviation as the objectives. The specific objective function is:

[0014] X = (x1, x2, x3, x4);

[0015] f(X)=c1N min +c2Deviation min ;

[0016] f(X) represents the objective function; c1 and c2 are the weights corresponding to each objective value, and the weights are determined according to the importance of the objective; N min Represents the minimization of cable force, Deviation min represents minimizing pipeline position deviation;

[0017] Among them, N min =a i N i -N mean ; a iTo affect the cable force coefficient;

[0018] The position affected by the characteristic parameters of the design variables during pipeline placement is P i (x i ,y i , z i ), the target position of the pipeline is P0(x0, y0, z0), and the minimum pipeline position deviation affected by the characteristic parameters of the design variables is:

[0019]

[0020] b i The influence coefficient of the design variables on the degree of deviation of the pipeline position;

[0021] Step S4, setting constraints, which include pipe body laying performance constraints, objective function value constraints, and other constraints. The pipe body laying performance constraints belong to equality constraints, and the objective function value constraints belong to inequality constraints;

[0022] The equality constraints are:

[0023] h i (x) = 0, h i (x) is the i-th equality constraint function;

[0024] U t =U;

[0025] V t =V;

[0026] ω t =ω;

[0027] ω′ t =ω;

[0028] U t 、V t ,ω t ,ω′ t represents the target values ​​of horizontal displacement, vertical displacement, pipe body angle, and pipe deviation angle in turn; U, V, ω, and ω′ represent the prediction results of comparing the three machine learning prediction models in step S2 in turn;

[0029] The inequality constraints are:

[0030] g j (X)≤0,g j (X) is the jth inequality constraint function;

[0031] g1(X)=ω′(X)-ω′ max ≤0;

[0032] g2(X)=ω(X)-ωma x≤0;

[0033] g3(X)=U(X)-U max ≤0;

[0034] g4(X)=V(X)-V max ≤0;

[0035] g5(X)=N i (X)-N imax ≤0;

[0036] g6(X)=T i (X)-T imax ≤0;

[0037] X l ≤X≤X u ;

[0038] X l , X u They represent the upper and lower limits of the characteristic parameters of the design variables that affect the stability of pipeline sinking;

[0039] Other constraints include: pipe buoyancy, gravity, residual buoyancy, and average cable force;

[0040]

[0041] G 管 =πr 2 lρ 管 g;

[0042] G 水 =πr 2 lρ 水 g;

[0043]

[0044] The hybrid penalty function method is used to transform the constrained optimization problem into an unconstrained optimization problem, and the optimization algorithm is used to solve the multi-objective and multi-constraint optimization problem. Specifically:

[0045]

[0046] Among them, μ is the penalty factor, 0<μ<1;

[0047] Step S5, solving the above multi-objective and multi-constraint optimization problem through an optimization algorithm, based on the construction of a pipeline sinking design variable mix ratio test database and a multi-objective joint optimization mathematical model, the multi-objective joint optimization mathematical model includes an objective function and a constraint condition system, and proposes an intelligent optimization design method for pipeline sinking stability under multi-objective requirements;

[0048] Step S6: Determine the weight coefficient of each optimization target according to the specific needs and importance of the construction project to form a weight vector c; use the ideal point method to comprehensively evaluate the pipeline design variables and select the best ratio scheme in terms of economy and safety.

[0049] Among them, in the step S5, the specific implementation process of the intelligent optimization design method for pipeline laying stability under multi-objective requirements includes the following steps:

[0050] Step a: According to the multi-objective and multi-constraint requirements for the stability of pipeline sinking during construction, the design variables and corresponding constraints are determined, and the objective functions of minimizing the cable force and minimizing the pipeline position deviation are established;

[0051] Step b, using three machine learning prediction models, namely adaptive enhancement, multi-layer perceptron and gradient boosting tree, to use the training set to train and build prediction models for pipeline displacement, pipe body angle and pipeline deviation angle, and use the test set to perform error analysis on the established performance model to determine whether the model is sufficiently accurate and robust; if it does not meet the accuracy requirements, it is necessary to increase the amount of test data and retrain the model; if it meets the requirements, it is directly used as a mapping relationship constraint for the performance requirements;

[0052] Step c, screening out all mix proportions that meet the constraint conditions from the current test database;

[0053] Step d: perform initial setting of the pipe body position according to the pipeline placement mix ratio that meets the constraint conditions selected from the database, and then use the optimization algorithm to iteratively calculate the pipeline position deviation to minimize the pipeline placement.

[0054] Wherein, in step d, the local steps of iterative calculation of pipeline position deviation minimization during pipeline laying include:

[0055] Step d1: Define the target environment as an N×D-dimensional space, where N represents the population size and D represents the spatial dimension; consider the horizontal displacement, vertical displacement and deviation angle of the pipeline, where D is a six-dimensional space; the initial position of each search agent is defined as Where rand(0,1) represents a random number in the range [0,1];

[0056] Step d2: By observing the subject's position, the agent t is the current iteration number, the maximum iteration number is K, Levy (D) is used to autonomously move to the main position, and the average position of the entire population is observed To further determine the subject's position and direction, the updated position is

[0057] Step d3: Randomly select the search agent location to update, denoted as

[0058] Step d4: Based on the fitness information, update the position according to the information sharing between agents, denoted as

[0059] Step d5: Avoid bad solutions and move closer to better solutions. The updated position is recorded as

[0060] Step d6, evaluating the target values ​​of all agents in the current population N, i.e., fitness values;

[0061] Step d7, checking whether the location of the search agent is within the allowed boundaries;

[0062] Step d8, iterate and repeat the above steps d1 to d5 continuously until the number of iterations reaches the maximum and the loop stops. The algorithm will calculate the position deviation of the pipeline in the fluid with wave flow based on the current pipeline position and design variables, and minimize the pipeline position deviation through iteration.

[0063] Wherein, in said step S1, a data preprocessing step is also included;

[0064] Data preprocessing is to preprocess the data through data cleaning and normalization, eliminate abnormal data and eliminate the dimensional influence between data, and finally determine the characteristic parameters of the design variables and performance index parameters.

[0065] Compared with the prior art, the present invention has the following beneficial effects, specifically:

[0066] 1. The present invention can effectively improve the stability and safety of pipeline laying construction: The present invention overcomes the unstable laying of the pipe body caused by the uneven laying speed of the cable force under the influence of the external environment in the traditional pipeline engineering construction; by introducing machine learning means, the inherent laws between the multivariate characteristic parameters affecting the stability of pipeline laying and the pipeline design variables are deeply explored, and a high-precision prediction model is established. The prediction model can effectively predict the position, posture and deviation angle of the pipeline in a complex environment (such as waves, water flow, wind, etc.), thereby significantly improving the stability and safety of pipeline laying;

[0067] 2. The present invention can effectively reduce construction costs and risks: the present invention not only takes into account the multiple performance requirements of pipeline laying, but also takes the minimum cable force and pipeline position deviation as optimization targets; by optimizing the design and construction process, the present invention successfully achieves the minimum cable force and pipeline position deviation while meeting the performance requirements, thereby significantly reducing construction costs and risks;

[0068] 3. The present invention can effectively improve engineering efficiency and economic benefits: Traditional empirical methods have the problems of large workload, low efficiency, high cost and difficulty in ensuring the optimal solution when dealing with multi-variable, multi-performance and multi-objective optimization problems; the present invention greatly improves engineering efficiency, reduces the number of experiments and the amount of calculation of numerical simulation, and thus significantly improves the economic benefits of the project through the introduction of machine learning prediction models and the innovation of multi-objective optimization methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] The present invention is further described below with the aid of the accompanying drawings. However, the embodiments in the accompanying drawings do not constitute any limitation to the present invention.

[0070] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0071] The present invention is described below in conjunction with specific implementation methods.

[0072] A multi-objective intelligent optimization method for stability control of a pipeline sinking process includes the following steps:

[0073] Step S1, establish a pipeline laying database, collect design variable characteristic parameters and performance index parameters that affect pipeline laying stability, and use water injection time, waves, water flow, and wind as design variable characteristic parameters, and use pipeline displacement, pipe body angle, and pipeline deviation angle as performance index parameters, as shown in Table 1:

[0074] Table 1 Database information

[0075]

[0076] Step S2, establishing a machine learning prediction model for pipeline displacement, pipe body rotation angle, and pipeline deviation angle, wherein the established machine learning prediction model includes an adaptive enhancement machine learning prediction model, a multi-layer perceptron machine learning prediction model, and a gradient boosting tree machine learning prediction model;

[0077] In the above machine learning prediction model, the characteristic parameters of each design variable of pipeline laying are used as input, and the pipeline position, pipe body posture, and pipeline deviation angle are used as output. The characteristic parameters of each design variable are: water injection time x1, wave x2, water flow x3, wind x4;

[0078] The Bayesian optimization based on the probability model is used to determine the optimal hyperparameter combination of each machine learning prediction model. By comparing the prediction results of the three machine learning prediction models, namely the adaptive enhancement machine learning prediction model, the multi-layer perceptron machine learning prediction model, and the gradient boosting tree machine learning prediction model, the machine learning prediction model with the highest prediction accuracy is selected and applied to the subsequent constraints, so as to establish the most stable optimization model for pipeline laying.

[0079] Step S3, constructing the objective function of multi-objective optimization of pipeline laying. Based on the machine learning prediction model selected in step S2, a multi-objective function is constructed with minimizing cable force and minimizing pipeline position deviation as the objectives. The specific objective function is:

[0080] X = (x1, x2, x3, x4);

[0081] f(X)=c1N min +c2Deviation min ;

[0082] f(X) represents the objective function; c1 and c2 are the weights corresponding to each objective value, and the weights are determined according to the importance of the objective; N min Represents the minimization of cable force, Deviation min represents minimizing pipeline position deviation;

[0083] Among them, N min =a i N i -N mean ; a i To affect the cable force coefficient;

[0084] The position affected by the characteristic parameters of the design variables during pipeline placement is P i (x i ,y i , z i ), the target position of the pipeline is P0(x0, y0, z0), and the minimum pipeline position deviation affected by the characteristic parameters of the design variables is:

[0085]

[0086] b i The influence coefficient of the design variables on the degree of deviation of the pipeline position;

[0087] Step S4, setting constraints, which include pipe body laying performance constraints, objective function value constraints, and other constraints. The pipe body laying performance constraints belong to equality constraints, and the objective function value constraints belong to inequality constraints;

[0088] The equality constraints are:

[0089] h i (x) = 0, h i (x) is the i-th equality constraint function;

[0090] U t =U;

[0091] V t =V;

[0092] ω t =ω;

[0093] ω′ t =ω;

[0094] U t 、V t ,ω t ,ω′ t represents the target values ​​of horizontal displacement, vertical displacement, pipe body angle, and pipe deviation angle in turn; U, V, ω, and ω′ represent the prediction results of comparing the three machine learning prediction models in step S2 in turn;

[0095] The inequality constraints are:

[0096] g j (X)≤0,g j (X) is the jth inequality constraint function;

[0097] g1(X)=ω′(X)-ω′ max ≤0;

[0098] g2(X)=ω(X)-ω max ≤0;

[0099] g3(X)=U(X)-U max ≤0;

[0100] g4(X)=V(X)-V max ≤0;

[0101] g5(X)=Ni(X)-N imax ≤0;

[0102] g6(X)=T i (X)-T imax ≤0;

[0103] X l ≤X≤X u ;

[0104] X l , X u They represent the upper and lower limits of the characteristic parameters of the design variables that affect the stability of pipeline sinking;

[0105] Other constraints include: pipe buoyancy, gravity, residual buoyancy, and average cable force;

[0106]

[0107] G 管 =πr 2 lρ 管 g;

[0108] G 水 =πr 2 lρ 水 g;

[0109]

[0110] The hybrid penalty function method is used to transform the constrained optimization problem into an unconstrained optimization problem, and the optimization algorithm is used to solve the multi-objective and multi-constraint optimization problem. Specifically:

[0111]

[0112] Among them, μ is the penalty factor, 0<μ<1;

[0113] Step S5, solving the above multi-objective and multi-constraint optimization problem through an optimization algorithm, based on the construction of a pipeline laying design variable mix ratio test database and a multi-objective joint optimization mathematical model, the multi-objective joint optimization mathematical model includes an objective function and a constraint condition system, and proposes an intelligent optimization design method for pipeline laying stability under multi-objective requirements;

[0114] Step S6: Determine the weight coefficient of each optimization target according to the specific needs and importance of the construction project to form a weight vector c; use the ideal point method to comprehensively evaluate the pipeline design variables and select the best ratio scheme in terms of economy and safety.

[0115] Specifically, in step S5, the specific implementation process of the intelligent optimization design method for pipeline laying stability under multi-objective requirements includes the following steps:

[0116] Step a: According to the multi-objective and multi-constraint requirements for the stability of pipeline sinking during construction, the design variables and corresponding constraints are determined, and the objective functions of minimizing the cable force and minimizing the pipeline position deviation are established;

[0117] Step b, using three machine learning prediction models, namely adaptive enhancement, multi-layer perceptron and gradient boosting tree, to use the training set to train and build prediction models for pipeline displacement, pipe body angle and pipeline deviation angle, and use the test set to perform error analysis on the established performance model to determine whether the model is sufficiently accurate and robust; if it does not meet the accuracy requirements, it is necessary to increase the amount of test data and retrain the model; if it meets the requirements, it is directly used as a mapping relationship constraint for the performance requirements;

[0118] Step c, screening out all mix proportions that meet the constraint conditions from the current test database;

[0119] Step d: perform initial setting of the pipe body position according to the pipeline placement mix ratio that meets the constraint conditions selected from the database, and then use the optimization algorithm to iteratively calculate the pipeline position deviation to minimize the pipeline placement.

[0120] Furthermore, in step d, the local steps of iterative calculation of pipeline position deviation minimization for pipeline laying include:

[0121] Step d1: Define the target environment as an N×D-dimensional space, where N represents the population size and D represents the spatial dimension; consider the horizontal displacement, vertical displacement and deviation angle of the pipeline, where D is a six-dimensional space; the initial position of each search agent is defined as Where rand(0,1) represents a random number in the range [0,1];

[0122] Step d2: By observing the subject's position, the agent t is the current iteration number, the maximum iteration number is K, Levy (D) is used to autonomously move to the main position, and the average position of the entire population is observed To further determine the subject's position and direction, the updated position is

[0123] Step d3: Randomly select the search agent location to update, denoted as

[0124] Step d4: Based on the fitness information, update the position according to the information sharing between agents, denoted as

[0125] Step d5: Avoid bad solutions and move closer to better solutions. The updated position is recorded as

[0126] Step d6, evaluating the target values ​​of all agents in the current population N, i.e., fitness values;

[0127] Step d7, checking whether the location of the search agent is within the allowed boundaries;

[0128] Step d8, iterate and repeat the above steps d1 to d5 continuously until the number of iterations reaches the maximum and the loop stops. The algorithm will calculate the position deviation of the pipeline in the fluid with wave flow based on the current pipeline position and design variables, and minimize the pipeline position deviation through iteration.

[0129] It should be pointed out that step S1 also includes a data preprocessing step; data preprocessing is to preprocess the data through data cleaning and normalization, eliminate abnormal data and eliminate the dimensional influence between data, and finally determine the design variable characteristic parameters and performance index parameters.

[0130] Through the above steps, the multi-objective intelligent optimization method for stability control of pipeline sinking process of the present invention has the following advantages, specifically:

[0131] 1. The present invention can effectively improve the stability and safety of pipeline laying construction: The present invention overcomes the unstable laying of the pipe body caused by the uneven laying speed of the cable force under the influence of the external environment in the traditional pipeline engineering construction; by introducing machine learning means, the inherent laws between the multivariate characteristic parameters affecting the stability of pipeline laying and the pipeline design variables are deeply explored, and a high-precision prediction model is established. The prediction model can effectively predict the position, posture and deviation angle of the pipeline in a complex environment (such as waves, water flow, wind, etc.), thereby significantly improving the stability and safety of pipeline laying;

[0132] 2. The present invention can effectively reduce construction costs and risks: the present invention not only takes into account the multiple performance requirements of pipeline laying, but also takes the minimum cable force and pipeline position deviation as optimization targets; by optimizing the design and construction process, the present invention successfully achieves the minimum cable force and pipeline position deviation while meeting the performance requirements, thereby significantly reducing construction costs and risks;

[0133] 3. The present invention can effectively improve engineering efficiency and economic benefits: Traditional empirical methods have the problems of large workload, low efficiency, high cost and difficulty in ensuring the optimal solution when dealing with multi-variable, multi-performance and multi-objective optimization problems; the present invention greatly improves engineering efficiency, reduces the number of experiments and the amount of calculation of numerical simulation, and thus significantly improves the economic benefits of the project through the introduction of machine learning prediction models and the innovation of multi-objective optimization methods.

[0134] 4. Enhance the flexibility and adaptability of the optimization method: The present invention adopts a hybrid penalty function method and an advanced optimization algorithm to achieve a good balance between exploration (global search) and development (local search), effectively avoiding falling into the local optimal solution; in addition, the parameters in the algorithm can be adaptively adjusted according to the iterative process to adapt to different search stages and solution space characteristics, thereby enhancing the flexibility and adaptability of the optimization method.

[0135] The above contents are only preferred embodiments of the present invention. For ordinary technicians in this field, according to the concept of the present invention, there will be changes in the specific implementation methods and application scopes. The content of this specification should not be understood as limiting the present invention.

Claims

1. A multi-objective intelligent optimization method for stability control of pipeline laying process, characterized in that: The following steps are included: Step S1, establishing a pipeline laying database, collecting design variable characteristic parameters and performance index parameters that affect pipeline laying stability, and taking water injection time, waves, water flow, and wind as design variable characteristic parameters, and taking pipeline displacement, pipe body rotation angle, and pipeline deviation angle as performance index parameters; Step S2, establishing a machine learning prediction model for pipeline displacement, pipe body rotation angle, and pipeline deviation angle, wherein the established machine learning prediction model includes an adaptive enhancement machine learning prediction model, a multi-layer perceptron machine learning prediction model, and a gradient boosting tree machine learning prediction model; In the above machine learning prediction model, the characteristic parameters of each design variable of pipeline laying are used as input, and the pipeline position, pipe body posture, and pipeline deviation angle are used as output. The characteristic parameters of each design variable are: water injection time x1, wave x2, water flow x3, wind x4; The Bayesian optimization based on the probability model is used to determine the optimal hyperparameter combination of each machine learning prediction model. By comparing the prediction results of the three machine learning prediction models, namely the adaptive enhancement machine learning prediction model, the multi-layer perceptron machine learning prediction model, and the gradient boosting tree machine learning prediction model, the machine learning prediction model with the highest prediction accuracy is selected and applied to the subsequent constraints, so as to establish the most stable optimization model for pipeline laying. Step S3, constructing the objective function of multi-objective optimization of pipeline laying. Based on the machine learning prediction model selected in step S2, a multi-objective function is constructed with minimizing cable force and minimizing pipeline position deviation as the objectives. The specific objective function is: X = (x1, x2, x3, x4); f(X)=c1N min +c2Deviation min ; f(X) represents the objective function; c1 and c2 are the weights corresponding to each objective value, and the weights are determined according to the importance of the objective; N min Represents the minimization of cable force, Deviation min represents minimizing pipeline position deviation; Among them, N min =a i N i -N mean ; a i To affect the cable force coefficient; The position affected by the characteristic parameters of the design variables during pipeline placement is P i (x i ,y i , z i ), the target position of the pipeline is P0(x0, y0, z0), and the minimum pipeline position deviation affected by the characteristic parameters of the design variables is: b i The influence coefficient of the design variables on the degree of deviation of the pipeline position; Step S4, setting constraints, which include pipe body laying performance constraints, objective function value constraints, and other constraints. The pipe body laying performance constraints belong to equality constraints, and the objective function value constraints belong to inequality constraints; The equality constraints are: h i (x) = 0, h i (x) is the i-th equality constraint function; IN t =U; V t =V; oh t =ω; oh' t =ω; U t 、V t ,ω t ,ω′ t represents the target values ​​of horizontal displacement, vertical displacement, pipe body angle, and pipe deviation angle in turn; U, V, ω, and ω′ represent the prediction results of comparing the three machine learning prediction models in step S2 in turn; The inequality constraints are: g j (X)≤0,g j (X) is the jth inequality constraint function; g1(X)=ω′(X)-ω′ max ≤0; g2(X)=ω(X)-ω max ≤0; g3(X)=U(X)-U max ≤0; g4(X)=V(X)-V max ≤0; g5(X)=N i (X)-N imax ≤0; g6(X)=T i (X)-T imax ≤0; X l ≤X≤X u ; X l , X u They represent the upper and lower limits of the characteristic parameters of the design variables that affect the stability of pipeline sinking; Other constraints include: pipe buoyancy, gravity, residual buoyancy, and average cable force; G 管 =πr 2 lρ 管 g; G 水 =πr 2 lρ 水 g; The hybrid penalty function method is used to transform the constrained optimization problem into an unconstrained optimization problem, and the optimization algorithm is used to solve the multi-objective and multi-constraint optimization problem. Specifically: Among them, μ is the penalty factor, 0<μ<1; Step S5, solving the above multi-objective and multi-constraint optimization problem through an optimization algorithm, based on the construction of a pipeline sinking design variable mix ratio test database and a multi-objective joint optimization mathematical model, the multi-objective joint optimization mathematical model includes an objective function and a constraint condition system, and proposes an intelligent optimization design method for pipeline sinking stability under multi-objective requirements; Step S6: Determine the weight coefficient of each optimization target according to the specific needs and importance of the construction project to form a weight vector c; use the ideal point method to comprehensively evaluate the pipeline design variables and select the best ratio scheme in terms of economy and safety.

2. A multi-objective intelligent optimization method for stability control of pipeline laying process according to claim 1, characterized in that: In step S5, the specific implementation process of the intelligent optimization design method for pipeline laying stability under multi-objective requirements includes the following steps: Step a: According to the multi-objective and multi-constraint requirements for the stability of pipeline sinking during construction, the design variables and corresponding constraints are determined, and the objective functions of minimizing the cable force and minimizing the pipeline position deviation are established; Step b, using three machine learning prediction models, namely adaptive enhancement, multi-layer perceptron and gradient boosting tree, to use the training set to train and build prediction models for pipeline displacement, pipe body angle and pipeline deviation angle, and use the test set to perform error analysis on the established performance model to determine whether the model is sufficiently accurate and robust; if it does not meet the accuracy requirements, it is necessary to increase the amount of test data and retrain the model; if it meets the requirements, it is directly used as a mapping relationship constraint for the performance requirements; Step c, screening out all mix proportions that meet the constraint conditions from the current test database; Step d: perform initial setting of the pipe body position according to the pipeline placement mix ratio that meets the constraint conditions selected from the database, and then use the optimization algorithm to iteratively calculate the pipeline position deviation to minimize the pipeline placement.

3. A multi-objective intelligent optimization method for stability control of pipeline laying process according to claim 2, characterized in that: In step d, the local steps of iterative calculation of pipeline position deviation minimization during pipeline laying include: Step d1: Define the target environment as an N×D-dimensional space, where N represents the population size and D represents the spatial dimension; consider the horizontal displacement, vertical displacement and deviation angle of the pipeline, where D is a six-dimensional space; the initial position of each search agent is defined as Where rand(0,1) represents a random number in the range [0,1]; Step d2: By observing the subject's position, the agent t is the current iteration number, the maximum iteration number is K, Levy (D) is used to autonomously move to the main position, and the average position of the entire population is observed To further determine the subject's position and direction, the updated position is Step d3: Randomly select the search agent location to update, denoted as Step d4: Based on the fitness information, update the position according to the information sharing between agents, denoted as Step d5: Avoid bad solutions and move closer to better solutions. The updated position is recorded as Step d6, evaluating the target values ​​of all agents in the current population N, i.e., fitness values; Step d7, checking whether the location of the search agent is within the allowed boundaries; Step d8, iterate and repeat the above steps d1 to d5 continuously until the number of iterations reaches the maximum and the loop stops. The algorithm will calculate the position deviation of the pipeline in the fluid with wave flow based on the current pipeline position and design variables, and minimize the pipeline position deviation through iteration.

4. The multi-objective intelligent optimization method for stability control of pipeline laying process according to claim 1 is characterized in that: In the step S1, a data preprocessing step is also included; Data preprocessing is to preprocess the data through data cleaning and normalization, eliminate abnormal data and eliminate the dimensional influence between data, and finally determine the characteristic parameters of the design variables and performance index parameters.

Images