A safety evaluation method for pigging operation of submarine crude oil pipeline based on reliability analysis

Abstract

This invention discloses a safety evaluation method for subsea crude oil pipeline cleaning operations based on reliability analysis, relating to the field of subsea pipeline cleaning operations and safety assessment technology. The method includes: acquiring raw data of key parameters affecting the operational safety of the cleaning tool during the cleaning operation; fitting the raw data to a probability distribution to determine the optimal probability distribution model for each key parameter; based on the optimal probability distribution model, using Latin hypercube sampling to perform structured random sampling of the input variables, generating a multidimensional sample matrix of the input variables; constructing a limit state function to evaluate the operating status of the cleaning tool and determine whether the tool has failed; inputting the sample values ​​from the multidimensional sample matrix into the limit state function to calculate the failure probability of the cleaning tool, which is then used as an evaluation index for the safety of the cleaning operation. This invention can quantify the impact of parameter uncertainty on the safety of the cleaning operation, providing a more scientific and accurate basis for safety evaluation.

Description

Technical Field

[0001] This invention relates to the field of subsea pipeline cleaning operations and safety assessment technology, specifically to a safety assessment method for subsea crude oil pipeline cleaning operations based on reliability analysis. Background Technology

[0002] Subsea pipelines are vital arteries for transporting oil and gas in the ocean, and their safe operation is crucial for ensuring energy supply and preventing marine pollution. During the long-term operation of subsea crude oil pipelines, a wax layer gradually accumulates on the pipe walls, leading to decreased pipeline transport efficiency and increased flow resistance. Therefore, regular pipeline cleaning operations to remove the wax layer deposited on the pipe walls are necessary measures to ensure the safe and economical operation of subsea pipelines.

[0003] Currently, safety assessments for subsea pipeline cleaning operations typically employ deterministic analysis methods. This approach assumes key safety parameters, such as temperature, pressure, flow rate, and fluid properties, to be constant, calculates target parameters based on theoretical models, and then compares these target parameters with predetermined safety factors to determine the system's safety status. However, these parameters often exhibit fluctuations and uncertainties under actual operating conditions. For example, pipeline inlet pressure fluctuates with changes in upstream platform operating conditions, crude oil properties change with temperature, and the thickness and mechanical properties of the wax layer also exhibit spatial distribution uncertainties.

[0004] Traditional methods typically add a "safety factor" to the deterministic analysis results to reflect the pipeline's safety and reliability. The safety factor is generally derived from extensive simulations and practical experience, reflecting certain statistical characteristics, but its value range varies depending on the region and the type of pig. Furthermore, the determination of the safety factor lacks systematic theoretical analysis and derivation, relying heavily on personal experience, and is highly subjective and variable. Therefore, relying solely on traditional deterministic analysis methods cannot accurately describe the impact of parameter changes on the safety of pig operation, nor can it provide a scientific and accurate assessment of pipeline failure risks, potentially leading to overly conservative design or unsafe operational decisions.

[0005] In contrast, the reliability-based limit state analysis method is a probabilistic design approach. Its core idea is to treat the uncertainties affecting system safety as random variables, quantify failure risk by assessing the failure probability of the system under limit states, and thus use reliability indicators to measure system safety. This method establishes a corresponding probability distribution model through the analysis of parameter uncertainties; constructs a limit state function by identifying system failure modes; and verifies system reliability by calculating the failure probability and comparing it with the target safety level. However, how to effectively apply reliability analysis methods to the safety evaluation of subsea pipeline cleaning operations, especially how to efficiently and accurately calculate the failure probability of the pipeline cleaning tool, remains a pressing technical problem to be solved in this field. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the prior art and provide a safety evaluation method for subsea crude oil pipeline cleaning operations based on reliability analysis, so as to quantify the impact of parameter uncertainty on the safety of cleaning operations and provide a more scientific and accurate basis for safety evaluation.

[0007] To achieve the above objectives, the technical solution adopted by this invention is as follows: A safety evaluation method for subsea crude oil pipeline cleaning operations based on reliability analysis, comprising the following steps:

[0008] Obtain raw data of key parameters that affect the safe operation of the pipeline cleaning tool during pipeline cleaning operations;

[0009] The original data of the acquired key parameters are fitted with a probability distribution to determine the optimal probability distribution model for each key parameter.

[0010] Based on the determined optimal probability distribution model, the Latin hypercube sampling method is used to perform structured random sampling of the input variables, generating a multidimensional sample matrix of the input variables;

[0011] A limit state function is constructed to evaluate the operating status of the pipeline pig and determine whether the pig has failed. The limit state function defines the failure condition based on the difference between the driving force and the resistance of the pig.

[0012] The sample values ​​in the multidimensional sample matrix are input into the limit state function, and the failure probability of the pig is calculated based on the stress-intensity interference principle, which is used as an evaluation index of the safety of pigging operations.

[0013] Furthermore, the key parameters include pipeline inlet pressure, pipeline outlet pressure, pig operating pressure, and wax layer physical properties.

[0014] Furthermore, the limiting state function is expressed as:

[0015]

[0016] Furthermore, the destructive force F of the wax layer on the pipe wall w The calculation formula is:

[0017]

[0018] Where A represents the dimensionless correction factor; d represents the pipe inner diameter; δ w Indicates the thickness of the wax layer; B represents the yield shear stress of the wax layer; B represents the additional pressure generated by the contact between the pig cup and the edge of the pipe wall.

[0019] Furthermore, probability distribution fitting is performed on the original data of the key parameters, including:

[0020] The parameter data were fitted using multiple candidate distribution models, including normal distribution, log-normal distribution, Gamma distribution, and Weibull distribution.

[0021] The fitting accuracy of each candidate distribution model is evaluated by combining mean squared error and coefficient of determination.

[0022] The model with the highest fitting accuracy is selected as the optimal probability distribution model for the corresponding key parameters.

[0023] Furthermore, the Latin hypercube sampling method is used to perform structured random sampling of the input variables, including:

[0024] On the [0, 1] probability axis, the range of the cumulative probability distribution function of each input variable is divided into N non-overlapping sub-intervals.

[0025] A probability point is randomly generated within each sub-interval;

[0026] The order of the probability points for each variable is independently and randomly shuffled;

[0027] The shuffled probability points are mapped to N sample values ​​through the inverse cumulative distribution function of the corresponding variables, thus forming an N×d multidimensional sample matrix, where d is the dimension of the input variables.

[0028] Furthermore, based on the stress-intensity interference principle, the failure probability of the pipeline pig is calculated, including:

[0029] Statistical histogram of the probability distribution of samples of the maximum allowable working pressure of the pig, calculated from the limit state function;

[0030] Histogram of probability distribution of statistical pipeline operating pressure samples;

[0031] By comparing the two histograms, the probability that the sample value of the pipeline operating pressure is greater than the sample value of the maximum allowable working pressure of the corresponding pig is calculated within the interference region. This probability is then used as the failure probability of the pig.

[0032] Furthermore, after calculating the failure probability of the pig, the Bald Eagle Search algorithm is used to optimize the number of samples required to generate the multidimensional sample matrix, including:

[0033] The number of samples is used as the optimization variable, and the model performance calculated based on the current number of samples is used as the fitness function;

[0034] The Bald Eagle Search algorithm iteratively searches within a preset sample size range through selection, search, and dive phases to find the optimal number of samples for the fitness function.

[0035] The present invention also provides a safety evaluation system for subsea crude oil pipeline cleaning operations based on reliability analysis, including a memory, a processor, and computer program instructions stored in the memory and executable by the processor. When the processor executes the computer program instructions, it can implement the above-mentioned method.

[0036] The present invention also provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the above-described method.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] (1) The evaluation method based on probability analysis in this invention takes into account the uncertainty of relevant parameters and avoids the limitations of artificially assigning quantitative values ​​to random variables. By calculating the pipeline failure probability, the safety of the replacement process of the subsea crude oil pipeline can be quantitatively described, thereby providing a more scientific and accurate basis for the assessment of pipeline flow safety, which has important engineering application value.

[0039] (2) This invention uses LHS sampling. Compared with the traditional Monte Carlo method, LHS can obtain a more uniform and comprehensive sample distribution with the same sample size. The fluctuations are smaller when calculating the expected value, variance, or failure probability, resulting in more stable results and improved computational efficiency. By dividing the value range of each variable equally and sampling in strata, it ensures that samples are selected in each stratum, thereby avoiding sample concentration or omissions. Under the premise of ensuring accuracy, the number of samples required by LHS is much smaller than that of traditional random sampling, significantly reducing computational costs.

[0040] (3) This invention uses the interferometric method to solve for the failure probability of the pig. Based on the overlapping region of the load effect and the resistance distribution, the failure probability is calculated by analyzing the overlapping part of their probability density functions. The calculation process of this method is intuitive and can clearly reflect the physical meaning of structural failure. The interferometric method can obtain high-precision results without a large number of samples, and has high computational efficiency. At the same time, this method is applicable to various probability distribution forms, has good versatility and visualization effect, and can provide a theoretical basis for reliability analysis and safety assessment.

[0041] (4) This invention uses the BES algorithm to optimize the algorithm for solving failure probability based on the interference principle. It can combine the intelligent search characteristics of the BES algorithm to improve the efficiency of solving failure probability and reduce the computational cost. By adopting a balanced strategy of "local encirclement" and "global exploration", the accuracy of solving failure probability is improved. Through the population iteration process, the random error of a single calculation is reduced, making the failure probability calculation result more stable. Attached Figure Description

[0042] Figure 1 This is a block diagram illustrating the implementation principle of the safety evaluation method for subsea crude oil pipeline cleaning operations based on reliability analysis provided in this embodiment of the invention.

[0043] Figure 2 This is a schematic diagram of the interference principle in this embodiment;

[0044] Figure 3 This is a schematic diagram of the LHS sampling principle in an embodiment of the present invention;

[0045] Figure 4 This is an example of an embodiment of the present invention showing the pressure distribution at the station.

[0046] Figure 5 This is a sampling statistical histogram of the wax layer yield stress LHS in an embodiment of the present invention;

[0047] Figure 6 This is a diagram showing the distribution of destructive force on the wax layer in an embodiment of the present invention;

[0048] Figure 7 This is the LHS statistical histogram of the pressure difference across the pig in this embodiment of the invention;

[0049] Figure 8 This is a diagram showing the distribution of the driving force of the pig in an embodiment of the present invention;

[0050] Figure 9 This is a schematic diagram of the interference failure region in an embodiment of the present invention. Detailed Implementation

[0051] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0052] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0053] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0054] like Figure 1 As shown in the figure, this embodiment provides a safety evaluation method for subsea crude oil pipeline cleaning operations based on reliability analysis, and its implementation steps are as follows.

[0055] S1. Obtain raw data of key parameters affecting the safe operation of the pipeline pig during the pipeline cleaning operation. The key parameters include pipeline inlet pressure, pipeline outlet pressure, pipeline pig operating pressure, and wax layer physical properties.

[0056] S2. Fit the probability distribution of the acquired key parameter raw data to determine the optimal probability distribution model for each key parameter; specifically including:

[0057] 1) The parameter data are fitted using multiple candidate distribution models, including normal distribution, log-normal distribution, Gamma distribution and Weibull distribution;

[0058] 2) The fitting accuracy of each candidate distribution model is evaluated by combining the mean squared error and the coefficient of determination;

[0059] 3) Select the model with the highest fitting accuracy as the optimal probability distribution model for the corresponding key parameters.

[0060] S3. Based on the determined optimal probability distribution model, the Latin hypercube sampling method is used to perform structured random sampling of the input variables, generating a multidimensional sample matrix of the input variables; specifically including:

[0061] 1) On the [0, 1] probability axis, divide the range of the cumulative probability distribution function of each input variable into N non-overlapping sub-intervals;

[0062] 2) Randomly generate a probability point within each sub-interval;

[0063] 3) The order of probability points for each variable is independently and randomly shuffled;

[0064] 4) The shuffled probability points are mapped to N sample values ​​through the inverse cumulative distribution function of the corresponding variables, thus forming an N×d multidimensional sample matrix, where d is the dimension of the input variable.

[0065] S4. Construct a limit state function to evaluate the operating status of the pig and determine whether the pig has failed. The limit state function defines the failure condition based on the difference between the driving force and the resistance of the pig.

[0066] S5. Input the sample values ​​in the multidimensional sample matrix into the limit state function, and calculate the failure probability of the pig based on the stress-strength interference principle, so as to use it as an evaluation index of the safety of pigging operation.

[0067] Among them, the failure probability of the pipeline pig is calculated based on the stress-intensity interference principle, including:

[0068] 1) Statistically analyze the probability distribution histogram of the maximum allowable working pressure of the pig, calculated from the limit state function;

[0069] 2) Statistical histogram of probability distribution of pipeline operating pressure samples;

[0070] 3) Compare the two histograms and calculate the probability that the sample value of the pipeline operating pressure is greater than the sample value of the maximum allowable working pressure of the corresponding pig within the interference region. Use this probability as the failure probability of the pig.

[0071] After calculating the failure probability of the pig, the Bald Eagle Search algorithm is used to optimize the number of samples required to generate the multidimensional sample matrix, including:

[0072] 1) Use the number of samples as the optimization variable, and the model performance calculated based on the current number of samples as the fitness function;

[0073] 2) Through the selection phase, search phase and dive phase of the vulture search algorithm, iterative search is performed within a preset sample size range to find the sample size that optimizes the fitness function.

[0074] This invention utilizes reliability analysis to assess the failure probability of subsea pipeline pigs, addressing the impact of pipeline parameter uncertainties on safety. Traditional deterministic analysis methods neglect fluctuations in factors such as temperature, pressure, and flow rate, failing to accurately reflect actual operating conditions. Through reliability analysis, this invention quantifies these uncertainties, establishes corresponding probability distribution models, identifies pig failure modes, and calculates the failure probability. This method not only provides a more accurate safety assessment but also offers a scientific basis for the pipeline cleaning process through the quantification of failure risks. This invention helps optimize pipeline safety management, reduce the probability of accidents, and improve the long-term stability and reliability of pipelines.

[0075] The relevant technical content of this invention will be further explained below.

[0076] 1. Basic Principles

[0077] 1.1 Limit State Equations for Submarine Pipeline Pigs

[0078] When fluid is transported in a pipeline, the safety of a pipeline pig operation is typically determined by whether the operating pressure exceeds the pig's maximum permissible working pressure. The pig is considered to have failed when the operating pressure exceeds its maximum permissible working pressure. According to reliability theory, the limit state equation for pig failure is:

[0079] (1)

[0080] In the formula, x is a vector of random parameters related to the operation of the pipeline pig; C is the maximum allowable working pressure of the pipeline pig, Pa; and S is the pipeline operating pressure, Pa.

[0081] When Z(x)≤0, that is, when the operating pressure of the pig exceeds the maximum allowable working pressure, the pig faces failure.

[0082] (1) Determination of the maximum allowable working pressure of the pig

[0083] The advance of a pipeline pig in a pipeline relies primarily on the driving force generated by the pressure difference between the front and rear ends to overcome various resistances encountered during its operation. This driving force is the decisive factor in whether the pig can operate smoothly and remove the wax layer. To calculate the failure probability of the pig, a force analysis is performed on it. It is assumed that the pig moves at a constant speed in a single-phase crude oil pipeline, the wax layer thickness is uniformly distributed along the pipe length, the wax deposit is firmly bonded to the pipe wall, and the force exerted by the pig's cup on the wax layer is uniformly distributed perpendicularly to the wax layer's cross-section along the wax layer's axis. The force-bearing area of ​​the cup is the same as the wax layer's cross-section. During the pigging process, the damage to the wax layer mainly occurs within the wax deposit (i.e., adhesion failure). Considering the pressures P1 and P2 at both ends of the pig and the destructive force F of the wax layer on the pipe wall during its movement... w The combined effect establishes the limit state function, which can be represented by equation (2). When g(x)≤0, it indicates system failure, and when g(x)>0, it indicates system safety.

[0084] (2)

[0085] In the formula, g(x) represents the force exerted on the pig; A p The cross-sectional area of ​​the pig is m. 2 P1 is the pressure at the rear end of the pig, in Pa; P2 is the pressure at the front end of the pig, in Pa; F w Destructive force of wax layer on pipe wall, N.

[0086] (2) Calculation of destructive force of wax layer

[0087] Assuming that the destructive force of the pig cup on the wax layer of the pipe wall is applied vertically in the direction of the axial direction of the wax layer, the area of ​​action is the same as the cross-section of the wax layer, and the wax layer thickness is evenly distributed along the pipe length, the failure stress of the wax layer can be calculated by equation (3).

[0088] (3)

[0089] In the formula, For the failure stress of the wax layer, Pa; F w δ represents the destructive force of the wax layer, in N; d represents the inner diameter of the pipe, in m; and δ represents the thickness of the wax layer on the pipe wall, in m.

[0090] F w The value of directly affects the calculation result of the pipe blockage probability of the pig. For each specific type and hardness of the pig cup, there is a unique linear relationship between the wax layer failure stress and the wax layer yield strength. The wax layer failure stress and wax layer yield strength under the action of pigs with different cup types and hardness can be expressed by equation (4).

[0091] (4)

[0092] In the formula, The failure stress of the wax layer is expressed in Pa. denoted as σa, where σa is the yield stress of the wax layer, in Pa; σa is a coefficient, dimensionless; and σb is a coefficient, in Pa.

[0093] Substituting equation (4) into equation (3), we get:

[0094] (5)

[0095] In the formula, F w The destructive force of the wax layer is N; δ is the yield stress of the wax layer, Pa; d is the inner diameter of the pipe, m; w denoted as the thickness of the wax layer on the pipe wall, in meters (m); a is a coefficient, dimensionless; b is a coefficient, in Pa.

[0096] The above formula indicates that the wax layer peeling force consists of two parts: the first part represents the yield shear strength of the wax layer that the pig needs to overcome to cause shear failure along the pipe wall, and its magnitude is related to the circumferential area of ​​the pipe, the thickness of the wax layer, and the yield stress; the second part represents the adhesion force or interfacial separation force between the wax layer and the pipe wall that the pig needs to overcome to completely peel off the wax layer, and its magnitude is related to the circumferential area of ​​the pipe and the thickness of the wax layer. The sum of the two parts is the total destructive force required for the wax layer to be completely peeled off. The wax layer can only be effectively peeled off when the pushing force applied by the pig is greater than this destructive force.

[0097] Studies have shown that the thickness of the wax layer directly affects the change in the destructive force of the wax layer with the hardness of the pig cup. Furthermore, the type of pig cup also significantly impacts wax removal. Therefore, the influence of the pig cup characteristics, including cup hardness and cup type, on the destructive force of the wax layer can be attributed to the second term on the right side of the equation.

[0098] (6)

[0099] In the formula, F w Wax layer destructive force, N; A dimensionless correction factor, taking into account the influence of cup shape, contact method, etc.; d pipe inner diameter, in meters; δ w Wax layer thickness, mm; B is the yield shear stress of the wax layer, Pa; B is the additional pressure generated by the contact between the pig cup and the edge of the pipe wall, N.

[0100] in:

[0101] (7)

[0102] In the formula, b is an empirical coefficient describing the influence of the edge characteristics (such as stiffness and inclination angle) of the pig's cup on the wax layer.

[0103] (3) Pipeline operating pressure

[0104] Based on existing data, the pipeline operating pressure is not a constant value for different replacement flow rates, but varies over time. Therefore, the actual pipeline pressure and the maximum allowable operating pressure are statistical values, not definitive numerical values. Thus, the failure of the subsea pipeline replacement process cannot be determined simply by comparing the magnitudes of these values. Currently, the widely used method is the Monte Carlo method, also known as the stochastic simulation method, a numerical calculation method guided by probability and statistics theory. This method uses random numbers to solve computational problems. When the number of samplings is sufficiently large, the obtained results can be considered accurate solutions, making it a powerful tool for analyzing multivariate complex systems and uncertain processes. To ensure the convergence and accuracy of its simulation results, it is usually necessary to increase the number of samplings, which significantly increases the computational load. To address this issue, stratified sampling (LHS) is used to sample from a known probability distribution, thereby reducing the required number of samplings.

[0105] (4) Calculation of driving force

[0106] The driving force for the pig to move along the pipeline mainly comes from the pressure difference between the two ends of the pipeline, while the resistance mainly comes from frictional resistance (and other factors such as sealing friction and wax peeling force). If only the pressure difference and frictional resistance are considered, the resultant force in the uniaxial direction can be written as:

[0107] (8)

[0108] In the formula, F p The thrust provided by the pressure difference, N; F fric The resistance is N, generated by friction along the friction path.

[0109] The propulsive force generated by the pressure difference can be expressed as:

[0110] (9)

[0111] In the formula, P up With P down The static pressures at the front and rear ends of the pig, respectively, are in MPa; A p The effective windward area of ​​the pipeline pig under pressure, in meters. 2 Typically, the projection of the pipe cross-sectional area of ​​the front of the pig is taken.

[0112] Frictional losses in long-distance pipelines between stations are mainly frictional losses along the pipeline route, with local frictional losses accounting for only 1% to 2%. Crude oil pipelines mostly operate in the hydraulically smooth region, and the Repinzon formula can be used to calculate frictional losses along the pipeline route.

[0113] (10)

[0114] In the formula, Frictional loss along the path, m; Q volumetric flow rate, m³ 3 / s; kinematic viscosity, m 2 / s; d is the inner diameter of the oil flow path, in meters; Mileage, m.

[0115] Therefore, the pressure drop between stations caused by frictional resistance along the route is:

[0116] (11)

[0117] In the formula, The density of the oil is expressed in kg / m³. 3 .

[0118] Friction along the friction line generates resistance:

[0119] (12)

[0120] 1.2 Failure Probability Calculation

[0121] (1) Solving the failure probability using the interference principle

[0122] The stress-strength interference model posits that system failure can occur when there is interference between the stress experienced by the system and its permissible strength. Within the interference region, failure occurs when the stress exceeds the permissible strength. The core idea of ​​the interference principle in calculating failure probability is to treat structural resistance and load effect as two independent random variables. When the load effect exceeds the resistance, structural failure occurs. Therefore, the failure probability can be expressed as the probability value corresponding to the "overlapping portion" (i.e., the interference region) of the probability distributions of resistance and load effect. Compared to analytical methods, numerical simulation interferometry is applicable to arbitrary distributions and nonlinear models, not limited to normal distributions, thus having a wider range of applications. The conditions for failure can be expressed by the limit state function:

[0123] (13)

[0124] In the formula, R is the structural resistance, representing the maximum load the system can withstand; S is the load effect, representing the structural response caused by the actual loading; and Z is the safety margin function.

[0125] The structure is safe when Z>0, and fails when Z≤0. In the study of pipeline failure probability, the structural resistance R is the maximum allowable working pressure of the pipeline; the load effect S is the pipeline operating pressure. The pipeline failure probability can be solved using the interferometric method through the probability density functions of the maximum allowable working pressure and the operating pressure.

[0126] like Figure 2 As shown, assume the probability density distributions of pipeline operating pressure and theoretical pressure-bearing capacity (i.e., maximum allowable working pressure) are g(s) and f(c), respectively. Neither of these distribution curves has upper or lower boundaries. Since these two probability density distribution curves overlap, this overlapping area is called the "interference zone." Within the interference zone, failure occurs when the pipeline operating pressure exceeds its pressure-bearing capacity. Clearly, the failure probability is related to the size of the interference zone, but not equal to the area integral of the interference zone. This is because within the interference zone, it is possible for the pipeline operating pressure to be lower than or higher than its pressure-bearing capacity. The red area in the figure represents the interference zone defined by the two probability density functions. In pipeline system failure probability analysis, the limit state function is first defined, and the input variables and their probability distributions are identified (listing the key parameters affecting g(s) and f(c) and assigning them corresponding probability distributions). Then, a corresponding mathematical model is established, the joint distribution of the input variables is determined, and the Latin hypercube sampling (LHS) method is used to solve for the pipeline failure probability through interferometry.

[0127] (2) Optimization of the vulture search algorithm

[0128] The vulture search algorithm is a type of swarm intelligence optimization algorithm inspired by the behavior of vulture groups hunting fish. By mimicking this simple and efficient hunting method, the vulture algorithm possesses characteristics such as strong convergence, few parameters, and ease of implementation. Based on the vulture's hunting strategy and social behavior, the algorithm mainly consists of three stages: the selection stage, the search stage, and the dive stage.

[0129] ① Selection stage

[0130] During the selection phase, individual vultures first randomly distribute themselves within the search space. Then, based on prey density, they determine the approximate hunting location, prioritizing areas with high prey concentration. This strategy ensures the randomness of the search range, improves search efficiency, and accelerates the search for ideal capture locations. This behavior can be represented by the following formula:

[0131] (14)

[0132] In the formula, α is a parameter used to control position changes, ranging from 1.5 to 2, and r is a random number between 0 and 1. P best This represents the search area chosen by the vulture based on the best location found in the previous search. (P) mean This is the average distribution location calculated based on the positions of all bald eagles after the last search. P i It is the current position of the i-th vulture.

[0133] ② Search Phase

[0134] During the search phase, the vulture searches the target area using a spiral flight pattern. This process utilizes the concepts of spiral and group center, so this behavior can be viewed as a combination of whale optimization and particle swarm optimization algorithms. Individual vultures move spirally around a center point, which not only sweeps across the entire search space, increasing algorithmic diversity and escaping local optima, but also avoids computational redundancy, greatly improving search efficiency. The mathematical model for spiral flight is represented using polar coordinates. The mathematical model for the vulture's position update during this process is shown in the following equation:

[0135] (15)

[0136] , (16)

[0137] , (17)

[0138] , (18)

[0139] In the formula, θ(i) and r(i) are the polar angle and polar radius of the spiral equation in polar coordinates, respectively; the ranges of a and R are (5,10) and (0.5,2), respectively, which are mainly used to control the spiral trajectory and determine the search cycle; rand is a random number between (0,1); x(i) and y(i) represent the current position of the vulture in polar coordinates, with values ​​ranging from (-1,1), respectively.

[0140] ③ Dive phase

[0141] During the dive phase, the bald eagle swiftly swoops from its optimal position in the search space towards its prey. Other individuals in the population, influenced by the environment, move towards their optimal positions and simultaneously attack the prey. The following equation illustrates this behavior:

[0142] (19)

[0143] , (20)

[0144] , (twenty one)

[0145] , (twenty two)

[0146] c1 and c2 increase the intensity of the vulture's movement towards the optimal point and the center point, with values ​​ranging from [1,2].

[0147] The three stages above demonstrate that intensification and diversity are the two most prominent characteristics of the Vulture Search algorithm. The concept of a group center ensures high search efficiency, reflecting the intensification characteristic of the Vulture Search algorithm. The spiral flight pattern provides sufficient diversity to help the Vulture Search algorithm escape local optima. This enables the Vulture Search algorithm to obtain the optimal solution efficiently and accurately.

[0148] 1.3 LHS Random Sampling Algorithm

[0149] Latin hypercube sampling is an efficient multidimensional random sampling method commonly used in uncertainty analysis and Monte Carlo simulations to reduce the number of samples while maintaining good representativeness. Latin hypercube sampling ensures a uniform distribution of samples across each dimension while preserving randomness between sampling points. To improve sampling efficiency and avoid sample clustering in the parameter space, the Latin hypercube sampling method is used to sample input variables. This method ensures uniform distribution across each input dimension while exhibiting randomness, making it suitable for reliability analysis and uncertainty propagation problems. In its implementation, each input variable interval is divided into N equally probable sub-intervals. A sampling point is randomly generated in each sub-interval, and the sampling order for each dimension is independently shuffled. By combining the samples from each dimension, an N×d sample matrix that satisfies both uniformity and randomness requirements is obtained, where N is the number of samples and d is the dimension of the input variable. This method is effectively used for reliability analysis and uncertainty propagation problems in pipeline failures.

[0150] In pipeline failure probability calculations, Latin hypercube sampling can improve the efficiency and accuracy of Monte Carlo simulations. Its core idea is to sample a specified random variable M times, divide the 0-1 interval into M non-overlapping equal intervals, and then randomly select a value from each interval, ensuring that each value has a probability of 1 / M. Figure 3 This is a schematic diagram illustrating the LHS sampling principle. The median of the sub-interval is used as the random number, or it is calculated according to a formula.

[0151] (twenty three)

[0152] In the formula, i = 1 to M; V represents a uniformly distributed random number between 0 and 1; U i For any given subinterval, there is one and only one random number.

[0153] (twenty four)

[0154] In the formula, (i-1) / M and i / M are the lower and upper boundaries of the i-th subinterval, respectively.

[0155] 2. Complete technical solution of the present invention

[0156] In the process of calculating the failure probability of subsea pipeline pigs, the relevant parameters of pipeline operation are first collected; then the data are fitted to obtain its probability distribution to meet the needs of subsequent analysis; then Latin hypercube sampling is used to more comprehensively cover the variable space without increasing the number of samples; finally, based on the distribution characteristics of random variables, the failure probability is calculated by analyzing the overlapping area (interference zone) of load effect and resistance distribution through the principle of interference.

[0157] This invention constructs a model for calculating the failure probability of subsea pipeline pigs based on reliability principles. Compared with traditional methods, the probabilistic analysis-based evaluation method considers the uncertainty of relevant parameters and avoids the limitations of artificially assigning quantitative values ​​to random variables. By calculating the failure probability of the pig, the safety of the subsea crude oil pipeline replacement process can be quantitatively described, thus providing a more scientific and accurate basis for assessing the flow safety of pipelines, and has significant engineering application value.

[0158] The implementation steps for solving the failure probability model of subsea pipelines based on reliability are as follows:

[0159] (1) Fitting the probability distribution of pipeline-related parameters: Using MATLAB's fitdist function, normal, log-normal, Gamma, and Weibull distributions were fitted to the data respectively; their probability density functions were calculated and compared with the sample histograms. The kernel density estimation method was adopted, and the smoothed probability density was adaptively calculated using a Gaussian kernel and smoothing bandwidth. The mean squared error (MSE) and the coefficient of determination (R²) were used to determine the probability density. 2 The model fitting accuracy is comprehensively evaluated, and the optimal distribution model is selected.

[0160] (2) The input variables are structured random sampling using the Latin hypercube sampling method.

[0161] ① Divide the probability axis into N equal intervals (construct a layered structure) to divide [0,1].

[0162] ② Generate random probability points within each interval.

[0163] For each variable, a value is randomly selected from the i-th sub-interval, using a uniform distribution.

[0164] ③ Shuffle the order of each variable (or generate a permutation).

[0165] For each variable j, generate a random permutation (of length N). Then, reindex the randomly selected values ​​from the previous step according to the permutation, and assign different intervals to different sample rows for each variable using an independent random permutation. This ensures that the stratification of different variables is "interleaved" among the samples.

[0166] ④ Mapping to the original variable space, apply the inverse cumulative distribution function of the corresponding variable to each probability value to obtain an N×M sample matrix.

[0167] ⑤ Adjust correlation or post-process the samples to match the target correlation matrix. Optimize sample filling quality to improve uniformity.

[0168] ⑥ Check sample quality. Plot the marginal distribution of each variable to confirm that the marginal distributions are met. Calculate the sample correlation coefficients between variables to confirm that they meet the requirements. Visualize the two-dimensional projection to check for clustering or gaps.

[0169] (3) The key parameters of the limit state function are determined to be the operating pressure of the pig and the maximum allowable working pressure. The required probability density sample is obtained by using LHS sampling, the joint distribution of pipeline failure probability is established, and finally the probability of the pipeline operating pressure-maximum allowable working pressure interference histogram is solved.

[0170] ① Calculate the probability distribution histogram of the maximum allowable working pressure of the pipeline pig, find the maximum and minimum critical safety pressure, divide the interval into smaller intervals, and count the number of samples falling within each smaller interval.

[0171] ② Statistical histogram of the probability distribution of pipeline operating pressure: Record the nodes of the equal probability intervals and the number of samples between each interval.

[0172] ③ Compare the first interval of the maximum allowable working pressure of the pipeline pig with each interval of the pipeline operating pressure, and count the number of samples where the pipeline operating pressure is less than the maximum allowable working pressure of the pipeline. Then the failure probability of the sample can be calculated.

[0173] (4) Optimize the number of samples based on the Vulture Optimization Algorithm (BES).

[0174] ① Data preparation and preprocessing: Input the original dataset, separate the features and labels in the data, standardize the features, and finally integrate the preprocessed data of the processed features and the original labels to provide standardized input for subsequent model training.

[0175] ② Sample size-dependent model evaluation: The preprocessed data is validated for reasonableness. The number of input samples is checked to see if it is within the valid range. A specified number of samples are randomly selected from the preprocessed samples. The sampled data is divided into training set and test set. The model is trained using the training set and the performance index is calculated on the test set. Finally, the model performance corresponding to the number of samples is returned.

[0176] ③ The BES optimizer searches for the optimal number of samples: using the algorithm's output model performance as fitness, it searches for the number of samples that achieves the optimal performance. Initial candidate values ​​are randomly generated within the sample size range for fitness calculation. The number of samples with the highest fitness and their performance are tracked during iteration, and the global optimal solution is updated. Based on the BES search strategy, the number of candidate samples is updated and boundary conditions are handled. After reaching the maximum number of iterations, the globally optimal number of samples and their corresponding performance are output.

[0177] ④ Process linkage and result output: The first three steps are iteratively optimized to obtain the optimal number of samples and corresponding accuracy for model performance.

[0178] 3 Examples

[0179] (1) Raw data processing

[0180] A certain subsea crude oil distribution branch pipeline is 15.5 km long, designed with a pressure of 6.3 MPa, a diameter of φ355.6 mm, a designed throughput of 600 × 10⁴ t / a, and a minimum winter throughput of 285 m³. 3 / h. The pipeline cleaning tool selected is a butterfly-shaped cleaning tool with a cup hardness of 80HA. The cleaning tool correction factor is 3.57, which is the contact tension between the edge of the cleaning tool cup and the pipe wall of 31.45 N / m.

[0181] Uncertainty analysis was performed on various operating parameters. Before conducting uncertainty analysis on the outflow pressure, the data was organized and grouped to obtain a sample for uncertainty analysis. The uncertainty analysis of the outflow pressure was carried out by determining the random probability distribution of the sample and statistically analyzing the maximum, minimum, mean, standard deviation, kurtosis, skewness, and coefficient of variation of the sample.

[0182] During the pigging process, considering the pipeline inlet pressure, outlet pressure, wax layer yield stress, and pig friction along the pipeline, the failure probability of the pig during pigging is predicted. The LHS sampling algorithm is used to randomly sample the pipeline inlet pressure, outlet pressure, and wax layer yield stress. 1000 sample values ​​of inlet pressure, outlet pressure, and wax layer yield stress are randomly selected using LHS, and an approximate probability distribution of the samples is fitted.

[0183] In this embodiment, the outbound pressure distribution is as follows: Figure 4 As shown.

[0184] Solution to the probability distribution of the maximum allowable working pressure of the pipeline pig:

[0185] The limit state equation for pigging was constructed using a wax layer peeling model. The pipeline used a disc-type pig with a cup hardness of 80HA. Taking into account the influence of cup shape, contact method, etc., the dimensionless correction factor was 3.57. Through regression of experimental data, the contact tension between the edge of the pig cup and the pipe wall was obtained as 31.45 N / m.

[0186] The destructive force of the wax layer follows a normal distribution, and the wax layer thickness follows a log-normal distribution. 1000 wax layer yield stress samples were randomly selected using the LHS method, and an approximate probability distribution of the samples was fitted. Figure 5 A sampling statistical histogram of the yield stress LHS of the wax layer is presented.

[0187] The distribution of destructive force of the wax layer is obtained from the yield stress and thickness of the wax layer as follows: Figure 6 As shown.

[0188] The frictional resistance along the pipeline was calculated using the Repin general formula, and pressure samples were obtained using the limit state equation of the pipeline pig and the LHS sampling algorithm at Yumen and Zhangye stations. The statistical histogram of the pressure difference distribution across the pipeline pig was then obtained, as shown below. Figure 7 As shown, the probability distribution is fitted.

[0189] The distribution of the driving force of the pig is as follows Figure 8 As shown.

[0190] (2) Solving for failure probability

[0191] As we know from the basics of mathematical statistics, probability distribution histograms are a commonly used method to characterize the statistical properties of data. Based on this, the numerical solution algorithm for the interferometric model is as follows.

[0192] ① The probability distribution histogram of the maximum allowable working pressure S of a statistical pipeline: If the total number of samples is p, first find the maximum value S of the critical safety pressure. max Minimum value S min , the interval [S min ,S max Divide the data into m equal intervals and record the nodes S between each interval. j-1 S j And count the values ​​falling within the small interval (S) j-1 ,S j+1 The number of samples within the range is denoted as g. j .

[0193] ② Histogram of probability distribution of pipeline operating pressure C: If the total number of samples is q, record the nodes C of its m equally probable intervals. j-1 C j And the number of samples f between each cell j .

[0194] ③ Compare the first interval [S1, S2] of the maximum allowable working pressure of the pipeline with each interval of the pipeline operating pressure, and count the number of samples whose pipeline operating pressure is less than the maximum allowable working pressure [S1, S2], denoted as x1. Then the failure probability of the sample in the interval [S1, S2] can be calculated, and the interference failure region can be obtained.

[0195] The failure probability of the pipeline is solved by using the limit state equation of the pipeline pig and the principle of interference.

[0196] Based on the above information, the probability of the pipeline operating pressure - maximum allowable working pressure interference histogram is calculated, from... Figure 9As can be seen from the schematic diagram of the interference failure area, the failure probability of restarting the inter-station pipeline after shutdown is 0.00003. This means that under the set average pressure and allowable working pressure conditions, the actual operating pressure is far below the design limit, the system is very safe, and the failure probability is extremely low.

[0197] (3) BES algorithm optimization

[0198] As an optimization algorithm, BES can more effectively explore the data space, find more representative frequent itemsets and association rules, and the automatic optimization mentioned above selects the optimal minimum support and minimum confidence for our system without the need for parameter comparison. The optimal number of samples and model performance output by the BES algorithm after running are shown in the following table.

[0199] Table 1. Optimization results of the BES algorithm

[0200]

[0201] As shown in Table 1, the optimal sample size for the failure probability algorithm optimized by BES is 800, which is within a reasonable range. The iteration trajectory shows an increasing trend from 320 to 800, indicating that model performance improves with increasing sample size, and accuracy increases, although the growth rate slows down. There is a positive correlation between sample size and model accuracy, but the marginal benefit diminishes. From 0.8765 after 10 iterations to 0.9223 after 50 iterations, the total improvement is approximately 5.23%, significantly optimizing model robustness. This demonstrates that a sample size of 800 strikes a balance between data representativeness and computational efficiency, thus optimizing model performance.

[0202] This embodiment also provides a safety evaluation system for subsea crude oil pipeline cleaning operations based on reliability analysis, including a memory, a processor, and computer program instructions stored in the memory and executable by the processor. When the processor executes the computer program instructions, it can implement the above-mentioned method.

[0203] This embodiment also provides a computer-readable storage medium storing computer program instructions that, when executed by a processor, implement the above-described method.

[0204] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0205] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0206] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0207] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0208] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A safety evaluation method for subsea crude oil pipeline cleaning operations based on reliability analysis, characterized in that, Includes the following steps: Obtain raw data of key parameters that affect the safe operation of the pipeline cleaning tool during pipeline cleaning operations; The original data of the acquired key parameters are fitted with a probability distribution to determine the optimal probability distribution model for each key parameter. Based on the determined optimal probability distribution model, the Latin hypercube sampling method is used to perform structured random sampling of the input variables, generating a multidimensional sample matrix of the input variables; A limit state function is constructed to evaluate the operating status of the pipeline pig and determine whether the pig has failed. The limit state function defines the failure condition based on the difference between the driving force and the resistance of the pig. The sample values ​​in the multidimensional sample matrix are input into the limit state function, and the failure probability of the pig is calculated based on the stress-intensity interference principle, which is used as an evaluation index of the safety of pigging operations.

2. The safety evaluation method for subsea crude oil pipeline cleaning operations based on reliability analysis according to claim 1, characterized in that, The key parameters include pipeline inlet pressure, pipeline outlet pressure, pig operating pressure, and wax layer physical properties.

3. The method for safety evaluation of subsea crude oil pipeline cleaning operations based on reliability analysis according to claim 1, characterized in that, The limit state function is expressed as: Where g(x) represents the force on the pig; A p P1 represents the cross-sectional area of ​​the pig; P2 represents the pressure at the rear end of the pig; F represents the pressure at the front end of the pig; w This indicates the destructive force on the wax layer of the pipe wall; When g(x)≤0, the pig is considered to be ineffective; when g(x)>0, the pig is considered to be safe.

4. The safety evaluation method for subsea crude oil pipeline cleaning operations based on reliability analysis according to claim 3, characterized in that, The destructive force F of the wax layer on the pipe wall w The calculation formula is: Where A represents the dimensionless correction factor; d represents the pipe inner diameter; δ w Indicates the thickness of the wax layer; B represents the yield shear stress of the wax layer; B represents the additional pressure generated by the contact between the pig cup and the edge of the pipe wall.

5. The method for safety evaluation of subsea crude oil pipeline cleaning operations based on reliability analysis according to claim 1, characterized in that, Fitting probability distributions to the raw data of key parameters, including: The parameter data were fitted using multiple candidate distribution models, including normal distribution, log-normal distribution, Gamma distribution, and Weibull distribution. The fitting accuracy of each candidate distribution model is evaluated by combining mean squared error and coefficient of determination. The model with the highest fitting accuracy is selected as the optimal probability distribution model for the corresponding key parameters.

6. The method for safety evaluation of subsea crude oil pipeline cleaning operations based on reliability analysis according to claim 1, characterized in that, The Latin hypercube sampling method is used to perform structured random sampling of the input variables, including: On the [0, 1] probability axis, the range of the cumulative probability distribution function of each input variable is divided into N non-overlapping sub-intervals. A probability point is randomly generated within each sub-interval; The order of the probability points for each variable is independently and randomly shuffled; The shuffled probability points are mapped to N sample values ​​through the inverse cumulative distribution function of the corresponding variables, thus forming an N×d multidimensional sample matrix, where d is the dimension of the input variables.

7. The safety evaluation method for subsea crude oil pipeline cleaning operations based on reliability analysis according to claim 1, characterized in that, Based on the stress-intensity interference principle, the failure probability of the pipeline pig is calculated, including: Statistical histogram of the probability distribution of samples of the maximum allowable working pressure of the pig, calculated from the limit state function; Histogram of probability distribution of statistical pipeline operating pressure samples; By comparing the two histograms, the probability that the sample value of the pipeline operating pressure is greater than the sample value of the maximum allowable working pressure of the corresponding pig is calculated within the interference region. This probability is then used as the failure probability of the pig.

8. The method for safety evaluation of subsea crude oil pipeline cleaning operations based on reliability analysis according to claim 1, characterized in that, After calculating the failure probability of the pig, the Bald Eagle Search algorithm is used to optimize the number of samples required to generate the multidimensional sample matrix, including: The number of samples is used as the optimization variable, and the model performance calculated based on the current number of samples is used as the fitness function; The Bald Eagle Search algorithm iteratively searches within a preset sample size range through selection, search, and dive phases to find the optimal number of samples for the fitness function.

9. A safety evaluation system for subsea crude oil pipeline cleaning operations based on reliability analysis, characterized in that, It includes a memory, a processor, and computer program instructions stored in the memory and executable by the processor, wherein when the processor executes the computer program instructions, it can implement the method as described in any one of claims 1-8.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by a processor, the method described in any one of claims 1-8 is implemented.

Images