Flexible pipe fatigue sea condition prediction method and system based on BP neural network

Abstract

The present invention relates to a method and system for predicting sea conditions of fatigue of flexible pipes based on BP neural network. The method comprises: collecting sea condition data through sensors to generate a wave scatter diagram; extracting sample points to screen target sea condition data; inputting experimental sea condition data into a basic hydrodynamic analysis model to obtain the tension and bending radius of the flexible pipe as the standard tensile and bending performance of the flexible pipe; using the target sea condition data as input and the standard tensile and bending performance of the flexible pipe as output, using BP neural network to perform model construction training to obtain a proxy model; establishing an optimization model, using a multi-objective genetic algorithm to perform global optimization on the proxy model to obtain a global optimal solution, and completing the fatigue prediction of the flexible pipe according to the global optimal solution. By training the proxy model and optimizing the model, there is no need to perform calculation and analysis on the flexible pipe containing each layer structure, and the fatigue prediction of the flexible pipe can be completed according to the global optimal solution, thereby improving the accuracy and efficiency of the flexible pipe fatigue detection.

Description

Technical Field

[0001] The invention relates to the technical field of marine flexible pipe fatigue analysis, and in particular to a method and system for predicting flexible pipe fatigue sea conditions based on a BP neural network. Background Art

[0002] Marine flexible pipes are connecting pipelines from the seabed oil and gas wellheads to the oil and gas gathering and transportation stations, and play a key role in the development of marine oil and gas resources. Flexible pipes are made of metal and polymer composites, and are easy to bend, easy to lay, recyclable, more economical, and more suitable for the marine environment. The use of flexible pipes in the development of marginal oil fields can significantly reduce engineering construction and operating costs. Existing flexible pipes are susceptible to fatigue damage during long-term use, especially under high temperature and high pressure conditions. These fatigue damages may cause pipeline leakage and breakage, endangering the safety of personnel and the environment. Fatigue life, as a key indicator to measure the fatigue performance of flexible risers, directly determines the reliability of risers during long-term service. In the field of offshore transportation and oil and gas extraction, fatigue failure of unbonded flexible pipes is a serious problem. The traditional way to detect whether a flexible pipe is fatigued is often to directly calculate and analyze the flexible pipe containing various layers of structure after obtaining the load condition.

[0003] However, the traditional flexible pipe fatigue detection method has the problems of complex detection and low accuracy. Summary of the invention

[0004] Based on this, in order to solve the above technical problems, a flexible pipe fatigue sea condition prediction method and system based on BP neural network are provided, which can predict flexible pipe fatigue and improve the accuracy of flexible pipe fatigue detection.

[0005] A method for predicting fatigue sea conditions of flexible pipes based on BP neural network, the method comprising:

[0006] Collecting sea condition data of the target sea area through sensors, and generating a wave scatter diagram according to the sea condition data;

[0007] According to the wave scatter diagram, sample points are extracted using a statistical method to screen out target sea condition data;

[0008] The experimental sea condition data is input into the basic hydrodynamic analysis model, and the tension and bending radius of the flexible pipe under the experimental sea condition are analyzed and obtained as the standard tensile bending performance of the flexible pipe;

[0009] Taking the target sea condition data as input and the standard tensile bending performance of the flexible pipe as output, a BP neural network is used for model building training to obtain a proxy model;

[0010] An optimization model is established, and a multi-objective genetic algorithm is used to perform global optimization on the proxy model to obtain a global optimal solution, and the fatigue prediction of the flexible pipe is completed according to the global optimal solution.

[0011] In one embodiment, collecting sea condition data of the target sea area through a sensor and generating a wave scatter diagram according to the sea condition data includes:

[0012] Installing a buoy in the target sea area, placing the sensor on the buoy, and collecting sea condition data of the target sea area;

[0013] The sensor is provided with a collection unit, through which the sea condition data is regularly transmitted to a signal receiver of a sea condition observation station, and the wave scatter diagram is obtained through processing by a processor of the sea condition observation station.

[0014] In one embodiment, the method of extracting sample points by using a statistical method according to the wave scatter diagram to screen out target sea condition data includes:

[0015] According to the wave scatter diagram, in the sea condition data, a sample interval range is set using statistics to determine a number of sample points;

[0016] The marine environment information is obtained, each of the sample points is screened according to the marine environment information, and the screened sample points are used as target sea condition data.

[0017] In one embodiment, the inputting of experimental sea condition data into the basic hydrodynamic analysis model, analyzing and obtaining the tension and bending radius of the flexible pipe under the experimental sea condition as the standard tensile bending performance of the flexible pipe, includes:

[0018] The Latin hypercube method is used to sample and select the experimental sea state data from a number of data;

[0019] By using the basic hydrodynamic analysis model and combining the marine engineering analysis software, the tension and bending radius of the flexible pipe under the experimental sea conditions are obtained according to the experimental sea condition data analysis;

[0020] The tension and bending radius are used as standard bending performance of the flexible pipe, and the ultimate load-bearing bending capacity curve of the flexible pipe is obtained.

[0021] In one embodiment, the target sea state data is used as input, the standard bending performance of the flexible pipe is used as output, and a BP neural network is used for model building training to obtain a proxy model, including:

[0022] Inputting the target sea state data into the input layer of the BP neural network, and adjusting the connection weight between the hidden layer and the output layer in the BP neural network under the stimulation of the target sea state data;

[0023] Obtain output data through the output layer, and compare the output data with the standard bending performance of the flexible pipe to obtain error data;

[0024] The weight and bias coefficient of the BP neural network are updated according to the error data until the error data between the output data and the standard bending performance of the flexible pipe is minimized.

[0025] In one embodiment, the method further comprises:

[0026] The consistency and accuracy of the proxy model are evaluated by using the average error, upper error limit, and correlation coefficient method to obtain a first evaluation result;

[0027] The Pearson product-moment correlation coefficient is introduced as a quality evaluation criterion to perform quality evaluation on the proxy model to obtain a second evaluation result.

[0028] In one embodiment, the multi-objective genetic algorithm is used to perform global optimization on the proxy model to obtain a global optimal solution, including:

[0029] Encoding the parameters to be optimized in the proxy model in binary format and establishing an initial population;

[0030] The initial population undergoes non-dominated sorting, selection, crossover and mutation to generate a progeny population;

[0031] The crowding degree of individuals in each non-dominated layer is calculated, and the parent population is determined based on the non-dominated relationship and the crowding degree of the individuals;

[0032] Based on the multi-objective genetic algorithm, the offspring population and the parent population are merged to form a new population until a global optimization is achieved.

[0033] In one embodiment, the method of performing the flexible pipe fatigue prediction according to the global optimal solution includes:

[0034] Decoding the global optimal solution to obtain current sea condition data;

[0035] The current sea condition data is used as the flexible pipe fatigue data to complete the flexible pipe fatigue prediction.

[0036] A flexible pipe fatigue sea condition prediction system based on BP neural network, the system comprising:

[0037] A data acquisition module, used to collect sea condition data of a target sea area through sensors, and generate a wave scatter diagram according to the sea condition data;

[0038] A data screening module, used to extract sample points using a statistical method according to the wave scatter diagram, and screen out target sea condition data;

[0039] Flexible pipe performance calculation module, used to input experimental sea condition data into the basic hydrodynamic analysis model, analyze and obtain the tension and bending radius of the flexible pipe under the experimental sea condition as the standard tensile and bending performance of the flexible pipe;

[0040] A model training module is used to use the target sea state data as input, the standard tensile bending performance of the flexible pipe as output, and use a BP neural network to perform model construction training to obtain a proxy model;

[0041] The model optimization module is used to establish an optimization model, use a multi-objective genetic algorithm to perform global optimization on the proxy model, obtain a global optimal solution, and complete the fatigue prediction of the flexible pipe according to the global optimal solution.

[0042] The above-mentioned flexible pipe fatigue sea condition prediction method and system based on BP neural network, by training the proxy model and optimizing the model, does not need to perform calculation and analysis on the flexible pipe containing various layer structures. The fatigue prediction of the flexible pipe can be completed according to the global optimal solution, thereby improving the accuracy and efficiency of flexible pipe fatigue detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 This is an application environment diagram of a method for predicting fatigue sea conditions of flexible pipes based on a BP neural network in one embodiment;

[0044] Figure 2 It is a schematic flow chart of a method for predicting fatigue sea conditions of flexible pipes based on BP neural network in one embodiment;

[0045] Figure 3 A schematic diagram of a process for obtaining sea state data and a wave scatter diagram in one embodiment;

[0046] Figure 4 A schematic diagram of a sea state sampling design process in one embodiment;

[0047] Figure 5 A schematic diagram of the structure of a BP neural network in one embodiment;

[0048] Figure 6 A schematic diagram of the operation process of a BP neural network in one embodiment;

[0049] Figure 7 A schematic diagram of an embodiment of using a proxy model construction process and a model validation and evaluation process as an extreme value response model establishment and implementation process;

[0050] Figure 8A schematic diagram of using a multi-objective genetic algorithm to perform global optimization on an agent model in one embodiment;

[0051] Fig. 9 A schematic diagram of Pareto level division in one embodiment;

[0052] Fig.10 A schematic diagram of congestion calculation in one embodiment;

[0053] Fig.11 It is a structural block diagram of a flexible pipe fatigue sea condition prediction system based on BP neural network in one embodiment;

[0054] Fig.12 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0056] The flexible pipe fatigue sea condition prediction method based on BP neural network provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Figure 1 As shown, the application environment includes a computer device 110. The computer device 110 can collect sea condition data of the target sea area through sensors, and generate a wave scatter diagram according to the sea condition data; the computer device 110 can extract sample points according to the wave scatter diagram using statistical methods to filter out the target sea condition data; the computer device 110 can input the experimental sea condition data into the basic hydrodynamic analysis model, analyze the tension and bending radius under the experimental sea condition, and calculate the standard tensile bending performance of the flexible pipe; the computer device 110 can use the target sea condition data as input and the standard tensile bending performance of the flexible pipe as output, use the BP neural network to build and train the model, and obtain the proxy model; the computer device 110 can establish an optimization model, use a multi-objective genetic algorithm to globally optimize the proxy model, obtain the global optimal solution, and complete the fatigue prediction of the flexible pipe according to the global optimal solution. Among them, the computer device 110 can be, but not limited to, various personal computers, laptops, robots, tablet computers and other devices.

[0057] In one embodiment, Figure 2 As shown, a method for predicting the fatigue sea state of a flexible pipe based on a BP neural network is provided, comprising the following steps:

[0058] Step 202, collecting sea condition data of the target sea area through sensors, and generating a wave scatter diagram based on the sea condition data.

[0059] The sensor may be a wave sensor, a temperature sensor, etc., which is used to collect sea condition data such as wave height, wave direction, period, amplitude, frequency, phase parameters, etc. in the target sea area, and obtain a wave spectrum. Among them, the model of the wave sensor may be SVS-603HR, and after the sensor collects the sea condition data, the computer device may generate a wave scatter diagram.

[0060] Step 204, extracting sample points using statistical methods according to the wave scatter diagram to filter out target sea condition data.

[0061] From the real sea condition data collected by sensors, statistical methods can be used to reasonably extract sample points to form a sample set and complete the sea condition sampling design. Then, computer equipment can screen it according to the actual ocean environment to obtain the final wave height, period and wave direction as the target sea condition data.

[0062] Step 206, inputting the experimental sea condition data into the basic hydrodynamic analysis model, analyzing and obtaining the tension and bending radius of the flexible pipe under the experimental sea condition as the standard tension and bending performance of the flexible pipe.

[0063] The experimental sea condition data can be used to represent the sea conditions designed for the experiment. By inputting the experimental sea condition data into the basic hydrodynamic analysis model for analysis, the tension and bending radius under the experimental sea conditions can be obtained as the standard tensile and bending performance of the flexible pipe.

[0064] Step 208, using the target sea state data as input and the standard bending performance of the flexible pipe as output, a BP neural network is used to perform model construction training to obtain a proxy model.

[0065] Specifically, the computer device can use target sea condition data such as wave height, period, and wave direction as input, tension and bending radius as output, and use BP neural network for training to complete the construction of a proxy model for predicting the maximum value of flexible pipe tension and the minimum value of bending radius. In this embodiment, since the sampling design and the establishment of the proxy model need to be based on a certain number of samples and corresponding response values, a proxy model is used to simulate the relationship between variables and extreme value responses; then an accurate proxy model is obtained through four steps of experimental design, acquisition of design sample response values, construction of the proxy model, and model verification.

[0066] Step 210, establishing an optimization model, using a multi-objective genetic algorithm to perform global optimization on the proxy model, obtaining a global optimal solution, and completing the flexible pipe fatigue prediction based on the global optimal solution.

[0067] Computer equipment can use the maximum tension and the minimum bending radius as optimization targets. The design needs to meet the limit bending line capacity curve, determine the three basic elements of design variables, objective function and constraint conditions, and establish an optimization model.

[0068] Among them, the ultimate tensile bending capacity curve of the flexible pipe is generally obtained by analyzing the key components that bear the load. For the flexible pipe, the compressive armor layer is mainly subjected to internal and external pressures, and the tensile armor layer is a load-bearing component that bears tensile and bending loads. Therefore, the tensile-bending load bearing capacity of the flexible pipe can be obtained by stress analysis of the armor layer unit. The compressive armor layer and the tensile armor layer can be analyzed as a whole. The armor layer unit will be subjected to internal pressure, tension, bending and other loads, and the generated circumferential, axial and radial stress components can be calculated by formulas.

[0069] Specifically, the hoop stress is calculated as follows: the compression layer is the layer that bears the internal and external radial pressures and is under stress for a long time. The hoop stress calculation of the compression layer is mainly obtained by the following formula: Among them, σ h is the hoop stress of the pressure-resistant layer, p is the design internal pressure of the pressure-resistant layer or the vector sum of the internal pressure and the external pressure, L t and l t are the outer diameter and inner diameter of the pressure-resistant layer respectively.

[0070] The calculation method of radial stress is as follows: The radial stress of the compression layer is generated by internal pressure and external pressure. Considering the maximum pressure, the calculation formula is: σ r =-p, where σ r is the radial stress of the compression layer.

[0071] The calculation method of axial stress is as follows: the axial stress of the tensile armor layer of the flexible pipe is generated by the tensile load, bending load and the internal pressure of the end closed part. The axial pressure caused by each part is as follows: The axial stress directly generated by the tensile load is expressed as: σ t =S t ·F T , where S t is the tension-stress coefficient of the armor layer, F T is the tensile force on the flexible pipe; the axial stress caused by bending is expressed as: c =S c c, where S c is the curvature-stress coefficient of the armor layer in the full sliding stage, c is the curvature response value of the flexible pipe; the internal pressure acts on the connection flange between the flexible pipe and SEMI, causing the axial stress on the pipe body to be: Among them, A t is the cross-sectional area of ​​the armor layer ring.

[0072] The total axial stress of the armor layer is the vector sum of each axial component, that is:

[0073] σ a =σ r +σt +σ c +σ f .

[0074] The ultimate bearing capacity of the entire armor layer is determined by its ultimate stress. The Von Mises Stress yield criterion is usually used to check whether the stress meets the material yield strength. Therefore, the equivalent stress obtained from the above circumferential, radial and axial stresses is as shown in the formula:

[0075]

[0076] The combined load of internal pressure, tension and bending when the VonMises stress of the armor layer reaches the yield strength is the bending capacity curve of the flexible pipe under a specific internal pressure. The stress limit state of the armor layer is expressed as: γσ y =σ e , where γ is the engineering safety factor, σ y is the yield strength of the armor layer steel.

[0077] Expanding the above formula, we can regard it as the equation about the axial stress σ a A quadratic equation of one variable can be solved to obtain the combination of tensile and bending loads corresponding to the ultimate bearing capacity of the flexible pipe, which together constitute the tensile-bending capacity curve.

[0078] The actual use of deep-sea flexible pipes needs to meet the ultimate bending capacity curve. The search for representative sea conditions also needs to comprehensively consider the effects of tension and bending in order to quickly find representative sea conditions that may be encountered. For an optimization problem, design variables, objective functions, and constraints are basic elements. Therefore, the maximum tension and the minimum bending radius can be used as optimization targets, and the established optimization model is:

[0079] In order to find: (H, T, D)

[0080] minimum: (-δ max , ρ min )

[0081]

[0082] Among them, H, T, D are wave height, period and wave direction respectively, δ max , ρ min To determine the maximum tension response and minimum bending radius of the flexible pipe under sea conditions.

[0083] After the optimization model is established, it can be determined to use a multi-objective genetic algorithm (NSGA-Ⅱ, Non-dominated Sorting Genetic Algorithm-Ⅱ) to globally optimize the model and obtain the global optimal solution.

[0084] Among them, the global optimal solution can represent the final representative sea condition. Through comprehensive analysis, it can be confirmed that the flexible pipe is most likely to fail due to fatigue under this sea condition, and the fatigue prediction of the flexible pipe can be completed.

[0085] In this embodiment, by training the proxy model and optimizing the model, there is no need to perform calculation and analysis on the flexible pipe containing each layer structure, and the fatigue prediction of the flexible pipe can be completed according to the global optimal solution, thereby improving the accuracy and efficiency of flexible pipe fatigue detection.

[0086] In one embodiment, a flexible pipe fatigue sea condition prediction method based on a BP neural network is provided, which may also include a process of generating a wave scatter diagram, and the specific process includes: installing a buoy in the target sea area, setting a sensor on the buoy, and collecting sea condition data of the target sea area; a collection unit is provided on the sensor, and the sea condition data is transmitted to a signal receiver of a sea condition observation station at a regular interval through the collection unit, and the wave scatter diagram is obtained by processing the processor of the sea condition observation station.

[0087] like Figure 3 As shown, in the process of generating a wave scatter diagram, the target sea area needs to be located first, and a buoy is installed after the target sea area is determined. The sensor is installed on the buoy to obtain data, and the obtained data is transmitted to the sea condition observation station. The wave scatter diagram is obtained through data encoding and other processing.

[0088] The sensor can be connected to the data acquisition unit with a wire or wireless connection, and the sensor can be selected according to the model of the data acquisition unit. As a data acquisition device, the sensor can be installed and fixed on the buoy. The sensor reads the sea conditions and hydrology in the target sea area with the buoy, and records the reading results every time an equal time difference passes, and stores them in the medium carried by the buoy. Through multi-sensor fusion technology, the obtained sea condition data is integrated and processed; according to the data collected by the acquisition unit carried on the sensor, it is transmitted to the signal receiver of the sea condition observation station on the shore to obtain basic sea condition data, which is further processed to obtain the wave scatter diagram.

[0089] In one embodiment, a flexible pipe fatigue sea condition prediction method based on a BP neural network is provided, which may also include a process of screening target sea condition data, the specific process including: according to a wave scatter diagram, in the sea condition data, using statistics to set a sample interval range and determine a number of sample points; obtaining marine environmental information, screening each sample point according to the marine environmental information, and using the screened sample points as target sea condition data.

[0090] The design process of sea state sampling is as follows: Figure 4 As shown, Figure 4 It includes the integration and processing of experimental sea condition data and real sea condition data. In the process of processing experimental sea condition data, the tension and bending radius are calculated by establishing a hydrodynamic model to obtain the tensile bending performance; in the process of processing real sea condition data, the distribution of sea condition parameters is obtained by setting up sample intervals, selecting sample points, and filtering and screening the sample points. The experimental sea condition data and the real sea condition data are integrated to complete the sea condition sampling design.

[0091] Specifically, in the acquired sea condition data, statistics are used to set the sample interval range, reasonable sample points are selected, and a sample data model is established; then, computer equipment is used to screen according to the actual ocean environment, and the wave direction is limited between 0° and 360°, and the wave amplitude, frequency, and phase parameters are counted at the same time; the selected data sample matrix is ​​composed of the target sea condition data wave height, period, and wave direction, which are organized into a table for subsequent model training.

[0092] In one embodiment, a flexible pipe fatigue sea condition prediction method based on a BP neural network is provided, which may also include a process of obtaining standard tensile and bending performance of the flexible pipe, the specific process including: using the Latin hypercube method to sample and select experimental sea condition data from a number of data; using a basic hydrodynamic analysis model and combining marine engineering analysis software to analyze the experimental sea condition data to obtain the tension and bending radius of the flexible pipe under the experimental sea condition; using the tension and bending radius as the standard tensile and bending performance of the flexible pipe, and obtaining the ultimate load-bearing tensile and bending capacity curve of the flexible pipe.

[0093] Before calculating the standard tensile and bending performance of flexible pipes, experimental design is required. A selection plan for variable sample space can be drawn up first. Specifically, the Latin hypercube method in stratified sampling can be selected for sampling, which can meet the requirements of randomness and uniformity. After obtaining the experimental design samples, hydrodynamic analysis is required in nonlinear time domain theory. Through the basic hydrodynamic analysis model, combined with professional marine analysis software Sesam / Sima, the corresponding tension and bending radius are calculated and derived.

[0094] In one embodiment, a flexible pipe fatigue sea condition prediction method based on a BP neural network is provided, which may also include a process of training to obtain a proxy model, and the specific process includes: inputting target sea condition data into the input layer of the BP neural network, and adjusting the connection weights between the hidden layer and the output layer in the BP neural network under the stimulation of the target sea condition data; obtaining output data through the output layer, and comparing the output data with the standard tensile and bending performance of the flexible pipe to obtain error data; and updating the weight and bias coefficient of the BP neural network according to the error data until the error data between the output data and the standard tensile and bending performance of the flexible pipe is minimized.

[0095] After obtaining the tension and bending radius, you can choose to build a model, the common ones are response surface model, Kriging model, radial basis model and neural network model. There are many kinds of agent model building methods. The neural network model after training has fast response, strong nonlinear adaptability, powerful mapping function, and good performance in system modeling, performance optimization, etc. In this embodiment, BP neural network modeling is selected.

[0096] The main feature of the BP neural network is to adjust the neuron weights by back propagating the error. In this embodiment, a multi-layer feedforward neural network trained by the back propagation algorithm is selected. A simple schematic diagram thereof can be shown as follows: Figure 5 shown.

[0097] like Figure 5 As shown in the figure, the BP neural network actually consists of three parts: input layer, hidden layer and output layer. The neurons in each layer influence each other layer by layer. Its basic principle is that under the stimulation of external input sample X, the neural network adjusts the connection weight R between the hidden layer and the output layer. ij , in order to keep the network output O close to the expected response value Y, the output of the hidden layer can be expressed as: H = f(R H ·X+b H ), where b H is defined as the bias coefficient, f is defined as the activation function, and the activation function in the neural network used in this embodiment is: In the final output layer, the hidden layer output is multiplied by the weight and combined with the bias coefficient b O The output of the output layer can be obtained as: O = R H ·H+b O ; Then, the output O of the neural network can be further compared with the expected response Y to obtain the error, and then the weights and bias coefficients of each layer of the neural network can be updated according to the error until the error between the output value and the response is small, thereby completing the training process.

[0098] In this embodiment, the operation of the BP neural network is as follows: Figure 6 As shown, by setting the number of BP network layers, training function, learning function, and performance function; setting the number of neurons in each layer and the transfer function; setting the number of training and training targets; setting the training sample input and target; starting training after determining that the set data is correct; setting the prediction sample input and determining whether it is correct, thereby achieving prediction; saving the prediction results and training results, and ending the operation of the BP neural network.

[0099] In one embodiment, a flexible pipe fatigue sea condition prediction method based on a BP neural network is provided, which may also include a process of evaluating a proxy model, and the specific process includes: evaluating the consistency and accuracy of the proxy model using the average error, upper error limit, and correlation coefficient method to obtain a first evaluation result; introducing the Pearson product-moment correlation coefficient as a quality evaluation criterion to perform a quality evaluation on the proxy model to obtain a second evaluation result.

[0100] In the process of evaluating the proxy model, the average error, upper error limit and correlation coefficient methods can be used to evaluate the model's fit and accuracy quality, and then the Parson product-moment correlation coefficient is introduced as a model quality evaluation criterion.

[0101] Specifically, the average error formula is expressed as: Where n represents the number of test sets. represents the predicted value of the response, Indicates the actual value of the response. In this embodiment, the Parson product-moment correlation coefficient is used as the quality evaluation criterion of this model, and its expression is: in, and represent the average of the actual response and the predicted response, respectively.

[0102] like Figure 7 As shown in the figure, the process of constructing the proxy model and the process of model verification and evaluation can be used as the process of establishing and implementing the extreme response model. The process of constructing the proxy model can include: constructing the proxy model, selecting the BP neural network, expressing the hidden layer output, updating the error, and completing the training; the process of model verification and evaluation can include: model verification, determining the average error expression, introducing the model quality evaluation criteria, obtaining the actual response value and the predicted response value, and obtaining the error expression.

[0103] In one embodiment, a flexible pipe fatigue sea condition prediction method based on a BP neural network is provided, which may also include a process of globally optimizing an agent model using a multi-objective genetic algorithm, and the specific process includes: encoding the parameters to be optimized in the agent model in binary and establishing an initial population; the initial population undergoes non-dominated sorting, selection, crossover and mutation to generate a child population; the crowding degree of individuals in each non-dominated layer is calculated, and the parent population is determined according to the non-dominated relationship and the crowding degree of the individual; based on the multi-objective genetic algorithm, the child population and the parent population are merged to form a new population until global optimization is obtained.

[0104] Since there are many uncontrollable factors in the ocean, the relationship between wave direction, wave height, period and tension, and bending radius in the sea conditions shows strong nonlinearity, so it is necessary to use an optimization algorithm with global optimization ability for optimization. According to the established optimization model, it can be determined to use a multi-objective genetic algorithm (NSGA-Ⅱ, Non-dominated Sorting GeneticAlgorithm-Ⅱ, non-dominated sorting genetic algorithm) to perform global optimization on the model. Among them, the multi-objective genetic algorithm is an intelligent bionic multi-objective optimization algorithm based on the basic genetic algorithm. It has strong adaptability to optimization problems under various circumstances and can effectively avoid falling into the dilemma of local optimal solutions. The multi-objective optimization problem in this embodiment needs to be compared and discussed in multiple dimensions. For all mutually non-dominated vectors, it is called a Pareto solution set (ParetoSolution). Among all solution sets, the solution set with the highest fitness is called the Pareto frontier, which is considered to be the optimal solution.

[0105] In this embodiment, if Figure 8 As shown, the process of using a multi-objective genetic algorithm to globally optimize the agent model can include: establishing the optimization goal to establish the optimization model, encoding and population initialization, fast non-dominated sorting, crossover and mutation, crowding calculation, elite retention strategy, and looping to obtain the global optimal solution, thereby decoding to obtain representative sea conditions.

[0106] Specifically, the entire optimization process of the multi-objective genetic algorithm includes the following specific steps:

[0107] The first is encoding and population initialization. Using 0 and 1 in binary numbers to express the solution set of the optimization problem, imitating the existence of chromosomes in the biological field and establishing a population, it is concise and clear, and very conducive to the analysis of the problem. The parameters in the optimization problem are encoded in binary, and the binary digital combination formed after encoding is called a chromosome. The 0 and 1 in it become genes, and each chromosome can be converted into a solution to the optimization problem through decoding. Organisms with different chromosomes are called populations. Genetic algorithms operate on populations, so multiple starting search points are required. The starting search point is called the initial population.

[0108] Then, we will perform a fast non-dominated sort. We need to search the entire population and find the number of dominated and the dominating set corresponding to each individual. For example, if the number of dominated individuals is 0, we will take it as a set with a Pareto level of 1, and then exclude it from the dominating set of the remaining individuals. After that, the level of the individual whose number of dominated individuals becomes 0 is set to a Pareto level of 2, and the rest are divided by analogy to complete the classification of the population level. The Pareto level classification is shown in the figure below: Fig. 9 shown.

[0109] Further selection operations are performed. A preliminary fitness judgment calculation is performed for each individual in the initial population. The degree of conformity of the individual is assessed according to its fitness, and the quality of the individual is judged accordingly, and then a decision is made on whether to retain its chromosome. It is generally believed that individuals with high fitness have a higher probability of being inherited by the next generation.

[0110] Then, crossover and mutation are performed. Both crossover and mutation are the operation processes of generating new individuals. Crossover is to exchange part of chromosomes between two individuals with a certain probability; mutation is to change a gene in a single individual, usually exchanging 0 and 1. For the individuals with chromosomes in the obtained population, random crossover and mutation occur in the population.

[0111] Then the crowding degree is calculated. The concept of crowding degree is introduced, and the individuals in each Pareto level solution set are overall adjusted and sorted according to the size of the objective function parameter. The difference in distance between the adjacent points on both sides of each point is called the crowding degree. Fig.10 As shown, the congestion at both ends can be considered infinite. The larger the congestion coefficient, the easier it is to be selected in the selection stage, which also means that the calculation results are more uniformly distributed in space and more meet the requirements.

[0112] Then, the elite retention strategy is implemented. The offspring and the parent generation are merged into a new group, the new group is sorted non-dominatedly and the crowding degree is calculated, and the Pareto levels in the new group are added to the new parent generation in descending order, so that the new parent generation group reaches the group size, and then the selection operation is performed to generate a new offspring population, and the local optimal solution set is obtained in a cycle.

[0113] The local optimal solution is cycled sequentially to obtain the global optimal solution.

[0114] In one embodiment, a flexible pipe fatigue sea condition prediction method based on a BP neural network is provided, which may also include a process for completing the flexible pipe fatigue prediction, the specific process including: decoding the global optimal solution to obtain current sea condition data; using the current sea condition data as the flexible pipe fatigue data to complete the flexible pipe fatigue prediction.

[0115] After obtaining the global optimal solution, the computer equipment can perform decoding operations, decode it into the corresponding wave height, period, and wave direction, and obtain the final representative sea condition. Among them, the representative sea condition generally obtained has only one solution, that is, the dangerous sea condition that is most likely to cause fatigue failure in production. All processes are closely linked to provide support for production safety, so as to achieve the prediction of sea condition fatigue failure of marine flexible pipes.

[0116] It should be understood that, although the steps in the above-mentioned flowcharts are shown in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the above-mentioned flowcharts may include a plurality of sub-steps or a plurality of stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these sub-steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0117] In one embodiment, Fig.11 As shown, a flexible pipe fatigue sea condition prediction system based on BP neural network is provided, including: a data acquisition module 1110, a data screening module 1120, a flexible pipe performance calculation module 1130, a model training module 1140 and a model optimization module 1150, wherein:

[0118] The data acquisition module 1110 is used to collect sea condition data of the target sea area through sensors and generate a wave scatter diagram according to the sea condition data;

[0119] The data screening module 1120 is used to extract sample points using statistical methods according to the wave scatter diagram to screen out target sea condition data;

[0120] The flexible pipe performance calculation module 1130 is used to input the experimental sea condition data into the basic hydrodynamic analysis model, analyze and obtain the tension and bending radius of the flexible pipe under the experimental sea condition as the standard tensile bending performance of the flexible pipe;

[0121] The model training module 1140 is used to use the target sea state data as input and the standard bending performance of the flexible pipe as output, and use the BP neural network to perform model construction training to obtain a proxy model;

[0122] The model optimization module 1150 is used to establish an optimization model, use a multi-objective genetic algorithm to perform global optimization on the proxy model, obtain a global optimal solution, and complete the flexible pipe fatigue prediction according to the global optimal solution.

[0123] In one embodiment, the data acquisition module 1110 is also used to install a buoy in the target sea area, set a sensor on the buoy, and collect sea condition data of the target sea area; a collection unit is set on the sensor, and the sea condition data is transmitted to a signal receiver of a sea condition observation station at a regular interval through the collection unit, and the wave distribution diagram is obtained by processing through a processor of the sea condition observation station.

[0124] In one embodiment, the data screening module 1120 is also used to set the sample interval range in the sea condition data based on the wave scatter diagram, and determine a number of sample points using statistics; obtain the marine environment information, screen each sample point based on the marine environment information, and use the screened sample points as the target sea condition data.

[0125] In one embodiment, the flexible pipe performance calculation module 1130 is also used to sample and select experimental sea condition data from a number of data using the Latin hypercube method; obtain the tension and bending radius of the flexible pipe under the experimental sea condition based on the experimental sea condition data analysis through a basic hydrodynamic analysis model combined with marine engineering analysis software; use the tension and bending radius as the standard tensile and bending performance of the flexible pipe, and obtain the ultimate load-bearing tensile and bending capacity curve of the flexible pipe.

[0126] In one embodiment, the model training module 1140 is also used to input the target sea condition data into the input layer of the BP neural network, and adjust the connection weights between the hidden layer and the output layer in the BP neural network under the stimulation of the target sea condition data; obtain output data through the output layer, compare the output data with the standard bending performance of the flexible pipe, and obtain error data; update the weight and bias coefficient of the BP neural network according to the error data until the error data between the output data and the standard bending performance of the flexible pipe is minimized.

[0127] In one embodiment, the model optimization module 1150 is also used to evaluate the fit and accuracy of the proxy model using the methods of average error, upper error limit, and correlation coefficient to obtain a first evaluation result; and introduce the Pearson product-moment correlation coefficient as a quality evaluation criterion to perform quality evaluation on the proxy model to obtain a second evaluation result.

[0128] In one embodiment, the model optimization module 1150 is also used to encode the parameters to be optimized in the proxy model in binary and establish an initial population; the initial population undergoes non-dominated sorting, selection, crossover and mutation to generate a child population; the crowding degree of individuals in each non-dominated layer is calculated, and the parent population is determined based on the non-dominated relationship and the crowding degree of the individuals; based on a multi-objective genetic algorithm, the child population and the parent population are merged to form a new population until global optimization is achieved.

[0129] In one embodiment, the model optimization module 1150 is further used to decode the global optimal solution to obtain current sea state data; and use the current sea state data as flexible pipe fatigue data to complete flexible pipe fatigue prediction.

[0130] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Fig.11As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected through a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for predicting fatigue sea conditions of flexible pipes based on a BP neural network is implemented. The display screen of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device housing, or an external keyboard, touchpad or mouse, etc.

[0131] Those skilled in the art will understand that Fig.11 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0132] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps of flexible pipe fatigue sea condition prediction based on BP neural network when executing the computer program.

[0133] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of realizing the flexible pipe fatigue sea condition prediction based on the BP neural network are implemented.

[0134] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0135] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0136] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent of the present application shall be subject to the attached claims.

Claims

1. A method for predicting flexible pipe fatigue sea conditions based on BP neural network, characterized in that: The method comprises: Collecting sea condition data of the target sea area through sensors, and generating a wave scatter diagram according to the sea condition data; According to the wave scatter diagram, sample points are extracted using a statistical method to screen out target sea condition data; wherein the target sea condition data includes wave height, period and wave direction; The experimental sea condition data is input into the basic hydrodynamic analysis model, and the tension and bending radius of the flexible pipe under the experimental sea condition are analyzed and obtained as the standard tensile and bending performance of the flexible pipe, and the ultimate load-bearing tensile and bending capacity curve of the flexible pipe is obtained; the tensile-bending load bearing capacity of the flexible pipe is obtained by stress analysis of the armor layer unit; during the analysis, the compressive armor layer and the tensile armor layer are analyzed as a whole; the combination of tensile and bending loads corresponding to the flexible pipe when the ultimate load is reached is calculated, and the ultimate load-bearing tensile and bending capacity curve of the flexible pipe is obtained; Taking the target sea state data as input and the standard tensile bending performance of the flexible pipe as output, a BP neural network is used for model construction training to obtain a proxy model; wherein the proxy model is used to predict the maximum value of the tension and the minimum value of the bending radius of the flexible pipe; An optimization model is established, and a multi-objective genetic algorithm is used to globally optimize the proxy model to obtain a global optimal solution, and the fatigue prediction of the flexible pipe is completed according to the global optimal solution; the maximum tension value and the minimum bending radius are used as optimization targets, and the ultimate load-bearing bending capacity curve is designed, the design variables, objective functions and constraints are determined, and the optimization model is established; wherein, the global optimal solution represents the final representative sea condition, under which the flexible pipe is most likely to fail due to fatigue, and the fatigue prediction of the flexible pipe is completed.

2. The method for predicting flexible pipe fatigue sea conditions based on BP neural network according to claim 1 is characterized in that: The method of collecting sea condition data of the target sea area by using a sensor and generating a wave scatter diagram according to the sea condition data includes: Installing a buoy in the target sea area, placing the sensor on the buoy, and collecting sea condition data of the target sea area; The sensor is provided with a collection unit, through which the sea condition data is regularly transmitted to a signal receiver of a sea condition observation station, and the wave scatter diagram is obtained through processing by a processor of the sea condition observation station.

3. The method for predicting flexible pipe fatigue sea conditions based on BP neural network according to claim 1 is characterized in that: The method of extracting sample points by using a statistical method according to the wave scatter diagram to screen out target sea condition data includes: According to the wave scatter diagram, in the sea condition data, a sample interval range is set using statistics to determine a number of sample points; The marine environment information is obtained, each of the sample points is screened according to the marine environment information, and the screened sample points are used as target sea condition data.

4. The method for predicting flexible pipe fatigue sea conditions based on BP neural network according to claim 1 is characterized in that: The experimental sea condition data is input into the basic hydrodynamic analysis model, and the tension and bending radius of the flexible pipe under the experimental sea condition are analyzed and obtained as the standard tensile bending performance of the flexible pipe, including: The Latin hypercube method is used to sample and select the experimental sea state data from a number of data; By using the basic hydrodynamic analysis model and combining the marine engineering analysis software, the tension and bending radius of the flexible pipe under the experimental sea conditions are obtained according to the experimental sea condition data analysis; The tension and bending radius are used as standard bending performance of the flexible pipe, and the ultimate load-bearing bending capacity curve of the flexible pipe is obtained.

5. The method for predicting flexible pipe fatigue sea conditions based on BP neural network according to claim 1 is characterized in that: The target sea condition data is used as input, the standard bending performance of the flexible pipe is used as output, and a BP neural network is used for model building training to obtain a proxy model, including: Inputting the target sea state data into the input layer of the BP neural network, and adjusting the connection weight between the hidden layer and the output layer in the BP neural network under the stimulation of the target sea state data; Obtain output data through the output layer, and compare the output data with the standard bending performance of the flexible pipe to obtain error data; The weight and bias coefficient of the BP neural network are updated according to the error data until the error data between the output data and the standard bending performance of the flexible pipe is minimized.

6. The method for predicting flexible pipe fatigue sea conditions based on BP neural network according to claim 1 is characterized in that: The method further comprises: The consistency and accuracy of the proxy model are evaluated by using the average error, upper error limit, and correlation coefficient method to obtain a first evaluation result; The Pearson product-moment correlation coefficient is introduced as a quality evaluation criterion to perform quality evaluation on the proxy model to obtain a second evaluation result.

7. The method for predicting flexible pipe fatigue sea conditions based on BP neural network according to claim 1 is characterized in that: The multi-objective genetic algorithm is used to globally optimize the agent model to obtain a global optimal solution, including: Encoding the parameters to be optimized in the proxy model in binary format and establishing an initial population; The initial population undergoes non-dominated sorting, selection, crossover and mutation to generate a progeny population; The crowding degree of individuals in each non-dominated layer is calculated, and the parent population is determined based on the non-dominated relationship and the crowding degree of the individuals; Based on the multi-objective genetic algorithm, the offspring population and the parent population are merged to form a new population until a global optimization is achieved.

8. The method for predicting flexible pipe fatigue sea conditions based on BP neural network according to claim 1 is characterized in that: The method of completing the fatigue prediction of the flexible pipe according to the global optimal solution includes: Decoding the global optimal solution to obtain current sea condition data; The current sea condition data is used as the flexible pipe fatigue data to complete the flexible pipe fatigue prediction.

9. A flexible pipe fatigue sea condition prediction system based on BP neural network, characterized in that: The system comprises: A data acquisition module, used to collect sea condition data of a target sea area through sensors, and generate a wave scatter diagram according to the sea condition data; A data screening module is used to extract sample points using a statistical method according to the wave scatter diagram to screen out target sea condition data; wherein the target sea condition data includes wave height, period and wave direction; The flexible pipe performance calculation module is used to input the experimental sea condition data into the basic hydrodynamic analysis model, analyze and obtain the tension and bending radius of the flexible pipe under the experimental sea condition as the standard tensile and bending performance of the flexible pipe, and obtain the ultimate load-bearing tensile and bending capacity curve of the flexible pipe; the tensile-bending load bearing capacity of the flexible pipe is obtained by stress analysis of the armor layer unit; during the analysis, the compressive armor layer and the tensile armor layer are analyzed as a whole; the combination of tensile and bending loads corresponding to the flexible pipe when the ultimate load is reached is calculated, and the ultimate load-bearing tensile and bending capacity curve of the flexible pipe is obtained; A model training module is used to use the target sea state data as input, the standard tensile bending performance of the flexible pipe as output, and use a BP neural network to perform model construction training to obtain a proxy model; wherein the proxy model is used to predict the maximum value of the flexible pipe tension and the minimum value of the bending radius; The model optimization module is used to establish an optimization model, use a multi-objective genetic algorithm to globally optimize the proxy model, obtain the global optimal solution, and complete the fatigue prediction of the flexible pipe according to the global optimal solution; take the maximum tension and the minimum bending radius as optimization targets, design a curve that satisfies the ultimate load-bearing bending capacity, determine the design variables, objective functions and constraints, and establish an optimization model; wherein the global optimal solution represents the final representative sea condition, under which the flexible pipe is most likely to fail due to fatigue, and completes the fatigue prediction of the flexible pipe.

10. The flexible pipe fatigue sea condition prediction system based on BP neural network according to claim 9, characterized in that: The data acquisition module is also used to install a buoy in the target sea area, set the sensor on the buoy, and collect the sea condition data of the target sea area; the sensor is provided with a collection unit, and the sea condition data is transmitted to the signal receiver of the sea condition observation station at a regular interval through the collection unit, and the wave scatter diagram is obtained by processing through the processor of the sea condition observation station.

Images