Toughness-driven underwater production system on-condition maintenance decision optimization method and system

Abstract

The invention relates to the technical field of ocean engineering, and provides a toughness-driven underwater production system on-condition maintenance decision optimization method and system, and the method comprises the steps: work safety analysis, maintenance operation risk assessment, maintenance process primary optimization under the condition of known risk, and toughness-driven maintenance process secondary optimization. From the perspective of maintaining the operation risk and the system stability, the secondary optimization decision is implemented. The first optimization aims to minimize the expected time and the expected risk of the overall maintenance process. And based on a multi-objective optimization result, the second optimization takes maximization of system toughness as an ultimate objective, so that the optimal maintenance time and the lowest risk of the underwater production system are determined, and the maintenance sequence of a plurality of components in the system is determined. By reasonably arranging the time of each step in the maintenance process, the risk of maintenance operation can be effectively reduced, and the recovery capability of the underwater production system is enhanced, so that the overall stability of the system is improved.

Description

Technical Field

[0001] The invention relates to the field of marine engineering technology and proposes a resilience-driven underwater production system condition-based maintenance decision optimization method and system. Background Art

[0002] Offshore oil development is one of the important energy industries in the world today. The maintenance process of underwater equipment is complicated, involving special operations such as lifting operations, high-altitude operations, and confined space operations. There are many dangerous factors and the operation is difficult. Underwater production system failures caused by emergencies such as typhoons and internal waves are inevitable, which will cause huge economic losses and casualties.

[0003] In order to mitigate the adverse effects of sudden system failures and prevent the impact of failures from further increasing, it is very necessary to arrange reasonable and effective maintenance activities to restore system performance to normal or higher levels.

[0004] Maintenance operations involve frequent interactions between personnel, equipment and the environment. There are inevitably many hazards in the maintenance process. Quickly completing a series of maintenance operations within a limited time may bring huge operational risks and threaten the safe realization of maintenance goals. It is necessary to consider maintenance optimization based on operational risks to ensure the smooth implementation of maintenance activities.

[0005] Currently, risk-based maintenance optimization methods are mainly based on the failure risk of the system to minimize the occurrence of high-risk failure modes, while maintenance operation risks are rarely considered.

[0006] For example, patent CN112801390A quantifies the probability and consequences of equipment failure through fault trees and historical data modeling, and prioritizes high-risk failure modes. For example, in deepwater submarine pipeline maintenance, the maintenance priority is determined based on a semi-quantitative risk assessment method, focusing on high-risk events such as pipeline rupture and leakage; patent CN119181014A allocates resources to high-damage equipment through damage level classification, but does not assess the risk of secondary damage or operational errors that may be caused during the maintenance process.

[0007] The aforementioned patent solutions do not consider the risks of maintenance operations, such as: Human factors: Operational errors and fatigue of maintenance personnel in complex underwater environments are not quantified. For example, in offshore drilling platform maintenance, traditional regular maintenance may introduce new defects due to frequent disassembly and assembly, but the existing model does not include this in risk calculation.

[0008] Environmental dynamism: Underwater operations are affected by environmental factors such as ocean currents and pressure, and the risk of maintenance tool failure or operation delay is not dynamically modeled. For example, the optimization model of patent CN112801390A assumes that the maintenance cost and reliability are fixed, and does not consider the surge in maintenance costs caused by sudden changes in the environment. Summary of the invention

[0009] The purpose of the present invention is to address the deficiencies of the prior art and propose a resilience-driven underwater production system condition-based maintenance decision optimization method and system to solve the following technical problems: Existing decision optimization methods ignore maintenance operation risks, resulting in unreasonable maintenance plans and problems such as poor maintenance timeliness.

[0010] In order to achieve the above object, the present invention adopts the following technical solutions: The resilience-driven optimization method for condition-based maintenance decision of subsea production systems includes: S1 Work safety analysis: Develop preliminary maintenance procedures, identify and analyze risk factors in the preliminary maintenance procedures and possible consequences of maintenance operation failures; S2 Maintenance operation risk assessment: Map the work safety analysis results into a Bayesian network, establish an equipment maintenance operation failure probability assessment model and a failure consequence assessment model, and the maintenance operation risk is the product of the maintenance failure probability and the failure consequence; S3 Initial optimization of maintenance process under known risk conditions: Construct a maintenance scenario for the underwater production system, perform initial optimization of the maintenance process from the perspective of maintenance time and maintenance risk, and obtain the optimal solution set for maintenance time and maintenance operation risk; S4 Resilience-driven secondary optimization of maintenance process: Calculate system resilience based on system performance curve, take maximizing resilience as the optimization goal, and perform secondary optimization on the maintenance process.

[0011] Furthermore, the formulation of the preliminary maintenance process includes: Establish a system performance degradation prediction model based on degradation data; Predict the cumulative degradation amount through the system performance degradation prediction model to determine the system status; Develop a preliminary maintenance process based on system status and previous maintenance experience.

[0012] Furthermore, the system performance degradation prediction model is established through the Wiener random process, and the Wiener process model is as follows: ; in, X ( t ) is the system at time t The amount of performance degradation; X (0) is the initial performance degradation, which is usually assumed to be 0;m is the drift coefficient reflecting the individual degradation rate; s 2 is the diffusion coefficient of the Wiener process, which indicates the degree of fluctuation of performance degradation; B ( t ) represents the standard Brownian motion, which is used to describe the uncertainty of the random degradation quantity on the time axis and satisfies B ( t )=0 and B ( t )~ N (0, t ), N (0, t ) is a normal distribution with mean 0 and variance t.

[0013] Furthermore, the cumulative degradation amount is calculated as follows: ; Among them, Δ X ( t ) is the time difference Δ t The cumulative degradation of performance during the period; X ( t +Δ t ) is the system at time t +Δ t The amount of performance degradation.

[0014] Furthermore, the Wiener process also includes estimating a drift coefficient and a diffusion coefficient, the steps are as follows: S101: Initialize the distribution: choose initial parameters m (0) and σ (0) ; S102: Step E, calculate expectation: estimate the expected value of the unknown parameter and give the existing parameter estimate; S103: Step M, update parameters: re-estimate the unknown parameters, determine the values ​​of the model parameters by maximizing the likelihood function, and obtain the expected estimates of the unknown parameters; S104: Convergence judgment: Check whether the parameters have converged. If not, return to step (2); otherwise, stop the iteration.

[0015] Furthermore, the Bayesian network structure of the maintenance failure probability assessment model is divided into four layers, namely, failure factor layer, intermediate node layer, maintenance activity step failure probability assessment layer, and maintenance step failure probability assessment layer.

[0016] Furthermore, the maintenance failure probability includes a priori probability and conditional probability, wherein the a priori probability of the root node is determined by combining the work safety analysis results and expert experience, and the conditional probability is calculated by the Noisy-or model: ; in, P ( H 1, H 2,…, Hn ) represents the conditional probability of the parent node, P Hi Indicates i The realization probability of each child node.

[0017] Furthermore, the failure consequences are evaluated based on physical meta-theory, including: S2021: Determine the Failure Mode Effect Level Set and Failure Effect Characteristic Set Classify the failure modes into different levels and establish the failure mode consequence level set as follows: ; in, N is the set of failure mode effect levels, N 1 , N 2 , N j and N m Represents different levels of failure mode consequences, m is the number of levels divided, The failure consequence feature set is composed of the measures of the impact of failure consequences on different aspects, as shown below: ; in, c is the failure consequence feature set, c 1 , c 2 , c i and c n Represents different failure consequence characteristics, n is the number of failure consequence characteristics; S2022: Determine the matter-element matrix to be evaluated The matter-element matrix to be evaluated is expressed as: ; in, R is the matter-element matrix, N For the object to be evaluated, c i (i= 1, 2, … , n ) is the failure consequence characteristic, V i ( i= 1, 2, … , n ) is the failure consequence evaluation characteristic c i The value of S2023: Determine the classical domain matrix and the section domain matrix For j The failure consequence classical domain matrix of the failure mode can be expressed as: ; in, R j is the classical domain matrix, N j ( j= 1, 2, … , m ) is the j The consequence level, V ji = ( a ji , b ji ) is the failure consequence characteristic c i The value range of R u is the section matrix, j The value ranges of all failure consequence characteristics of the classical domain matrix of all failure consequences of each failure mode are combined to obtain the node domain matter-element matrix; S2024: Determine correlation function and comprehensive correlation The correlation degree indicates the correlation degree between the failure mode consequence level and the failure consequence characteristics. x 0 With finite real interval X =( Xa , Xb ) is: ; Xa and Xb The finite interval X The upper and lower limits of According to the distance calculation method from point to interval, the expression of the correlation function can be obtained as follows: ; in, K j ( V i ) is the correlation function, P ( V i , V ji ) is a numerical value V i With interval V ji = ( a ji , b ji ) distance; The expression of comprehensive correlation is: ; in, K j ( P ) is the comprehensive correlation, p i For evaluation indicators c i The weight of can be obtained by the hierarchical analysis method; S2025: Determine the failure consequence level and failure consequence quantification value The maximum degree of attribution in the comprehensive correlation is taken as the failure consequence level of different failure modes. F j Indicates the failure consequence level; the expression of the failure consequence quantification value is: ; in, F It is the quantitative value of the failure consequence.

[0018] Furthermore, the step S3 specifically includes: E ab ( x ) represents a random variable x In maintenance activities ( a , b ), the expected function of the random variable x It can represent the repair time or repair risk as follows: ; in, p ab ( x ) represents the maintenance activity under the condition of random variable x ( a , b ) is realized ifx For maintenance activities ( a , b ) maintenance time, then p ab ( x ) represents x The probability of completing the maintenance activity within hours; f ab Indicates maintenance activities ( a , b ) in the random variable x The characteristic function is as follows: ; in, s represents any real number; According to the characteristic function and maintenance activities ( a , b ) is the probability of realization p ab , the transfer function can be calculated as: ; in, W ab ( s ) indicates maintenance activities ( a , b )’s transfer function; The equivalent transfer function of the self-loop structure is: ; in, W' ab ( s ) represents the equivalent transfer function of the self-loop structure, f aa Indicates maintenance activities a Random variables during re-execution x The characteristic function of The equivalent transfer function of the entire graphical evaluation and review technology network is expressed as W E ( s ), can be solved by matrix method, when s = 0, the equivalent transfer function of the entire network is: ; in, p E is the equivalent probability of successful execution from the first node to the last node, W E (0) s = 0 when the network equivalent transfer function; The equivalent characteristic function corresponding to the entire graphical evaluation and review technology network is: ; in, f E ( s ) is the equivalent characteristic function corresponding to the entire graphical evaluation and review technology network; According to the basic properties of characteristic function, characteristic function s = 0 n The value of the derivative is the random variable n In order to evaluate the effectiveness of the maintenance process, the expected function of the random variable in the entire process is calculated as: ; in, E ( x ) represents a random variable x The expected function in the whole maintenance process; Assuming that there is a linear relationship between maintenance time and maintenance risk, the relationship between additional operation time, maintenance time and maintenance risk is: ; in, T ab In maintenance activities ( a , b ) maintenance time before additional operating time is invested, R ab During maintenance activities ( a , b ) Risk of maintenance operations before investing additional operating time, t ab is the additional operation time, T' ab During maintenance activities ( a , b ) Maintenance time after investing extra operating time, R' ab During maintenance activities ( a , b ) Risk of maintenance operation after investing extra operating time, z ab is the proportionality coefficient of the effect of additional maintenance time on the risk of maintenance operations; Based on the multi-objective particle swarm optimization algorithm, the optimal maintenance time and optimal maintenance operation risk in the maintenance process are optimized. The multi-objective function of the optimization algorithm is expressed as ; in, E [ T ] represents the expected maintenance time of the maintenance process,E [ R ] represents the expected maintenance risk of the maintenance process.

[0019] Furthermore, a resilience-driven subsea production system condition-based maintenance decision optimization system includes: The data acquisition module collects monitoring data of the underwater production system based on the monitoring data of the master control station; The status analysis module analyzes and determines the status of the subsea production system based on the collected monitoring data; The maintenance process preliminary decision module preliminarily formulates maintenance strategies based on the status of the underwater production system; The work safety analysis module divides the maintenance steps according to the established preliminary maintenance process and analyzes the potential failure causes in the maintenance process; The maintenance operation risk assessment module conducts quantitative assessment of the failure probability and failure consequences of maintenance operations based on the results; The maintenance operation risk output module outputs the maintenance operation failure probability and maintenance operation failure consequences; The maintenance process evaluation module constructs a maintenance process evaluation model based on graphical evaluation and review technology under the condition that the maintenance operation risk is known; The maintenance process evaluation result output module outputs the expression of expected maintenance risk and expected maintenance time of the maintenance process evaluation; The maintenance process initial optimization module establishes a multi-objective initial optimization model under the minimum maintenance risk and minimum maintenance time; The maintenance process secondary optimization module establishes a secondary optimization model under the condition of maximizing resilience based on the Pareto solution set of maintenance risk and maintenance time. The maintenance process decision output module outputs the final maintenance optimization result, and relevant maintenance personnel extract information to formulate maintenance plans to ensure the safety of oil and gas production.

[0020] The beneficial effects of the present invention are as follows: With respect to the maintenance process of the underwater production system, the present invention implements secondary optimization decisions from the perspective of maintenance operation risks and system stability. The first optimization aims to minimize the expected time and expected risk of the overall maintenance process. Based on the results of multi-objective optimization, the second optimization takes maximizing system resilience as the ultimate goal, thereby determining the optimal time and lowest risk for maintenance of the underwater production system, and establishing the maintenance order of multiple components in the system. By reasonably arranging the time of each step in the maintenance process, the risk of maintenance operations can be effectively reduced, and the resilience of the underwater production system can be enhanced, thereby improving the overall stability of the system. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a flow chart of a resilience-driven subsea production system condition-based maintenance decision optimization method; Figure 2 It is the basic structure of the graphical evaluation and review technology model; Figure 3 It is a schematic diagram of the step-by-step recovery of system performance; Figure 4 It is a schematic diagram of the underwater production control system; Figure 5 It is a schematic diagram of the resilience-driven underwater production system condition-based maintenance decision optimization system.

[0022] In the figure: 101. Main control station, 102. Emergency shutdown system, 103. Electric power unit, 104. Hydraulic power unit, 105. Uninterruptible power supply, 106. Upper umbilical cable terminal, 107. Accumulator group, 108. Chemical injection unit, 109. Umbilical cable, 110. Subsea distribution unit, 111. Electric flying line, 112. Subsea control module, 113. Liquid flying line, 114. Subsea Christmas tree, 201. Data acquisition module, 202. Status analysis module, 203. Maintenance process preliminary decision module, 204. Work safety analysis module, 205. Maintenance failure probability analysis unit, 206. Maintenance failure consequence analysis unit, 207. Maintenance operation risk assessment module, 208. Maintenance operation risk output module, 209. Maintenance process evaluation module, 210. Maintenance process evaluation result output module, 211. Expected maintenance risk output unit, 212. Expected maintenance time output unit, 213. Maintenance process primary optimization module, 214. Maintenance process secondary optimization module, 215. Maintenance process decision output module. DETAILED DESCRIPTION

[0023] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0024] Embodiment 1: like Figure 1 As shown in the figure, the resilience-driven maintenance process decision optimization method for underwater production systems mainly consists of six steps: maintenance failure probability assessment, failure consequence assessment based on matter-element theory, maintenance operation risk quantification, maintenance process evaluation based on graphical evaluation and review technology, minimum expected maintenance time and expected maintenance risk optimization based on multi-objective particle swarm algorithm, and maintenance time and maintenance sequence optimization under the condition of maximum system resilience.

[0025] S1: Work safety analysis. After the underwater production system is subjected to external impact, the degradation data is extracted from the main control station 101 to predict the cumulative degradation of each component and then evaluate the current state of the system. The performance degradation prediction model of the system is established using the Wiener random process.

[0026] The Wiener process model is shown as follows: ; in, X ( t ) is the system at time t The amount of performance degradation; X (0) is the initial performance degradation, which is usually assumed to be 0; m is the drift coefficient reflecting the individual degradation rate; s 2 is the diffusion coefficient of the Wiener process, which indicates the degree of fluctuation of performance degradation; B ( t ) represents the standard Brownian motion, which is used to describe the uncertainty of the random degradation quantity on the time axis and satisfies B ( t )=0 and B ( t )~ N (0, t ), N (0, t ) is a normal distribution with mean 0 and variance t.

[0027] According to the properties of the Wiener process, its increments are independent and follow a normal distribution, so: ; Among them, Δ X ( t ) is the time difference Δ t The cumulative degradation of performance during the period; X ( t +Δ t ) is the system at time t +Δ t The amount of performance degradation.

[0028] The Wiener process is a continuous-time random process. The expectation maximization algorithm is used to estimate the drift coefficient and diffusion coefficient based on the extracted degradation data. The specific steps are as follows: S101: Initialize the distribution: choose initial parameters m (0) and σ (0) ; S102: Step E, calculate expectation: estimate the expected value of the unknown parameter and give the existing parameter estimate; S103: Step M, update parameters: re-estimate the unknown parameters, determine the values ​​of the model parameters by maximizing the likelihood function, and obtain the expected estimates of the unknown parameters; S104: Convergence judgment: Check whether the parameters have converged. If not, return to step (2); otherwise, stop the iteration.

[0029] The system status is divided into two categories: one is the normal working status, and the other is the fault status. According to the predicted cumulative degradation, the current system status is judged, and the maintenance method of the underwater production system is initially customized according to previous maintenance experience. The work safety analysis method is used to identify and analyze the risk factors existing in the maintenance operation process, as well as the consequences that may be caused by the failure of the maintenance operation, and the identified risk factors and consequences are classified. The risk influencing factors mainly include human factors, environmental factors, management factors and equipment factors. At the same time, through the work safety analysis method, a complete maintenance operation process is divided into a set of clear steps.

[0030] S2: Maintenance operation risk assessment. The maintenance operation risk is quantitatively calculated as the product of the probability of maintenance failure and the consequence of failure: ; Among them, Risk(a,b) represents the maintenance operation risk of operation activity (a,b), P(a,b) represents the probability of maintenance operation failure of operation activity (a,b), and F(a,b) represents the consequence of maintenance failure of operation activity (a,b).

[0031] S201: Maintenance failure probability assessment.

[0032] By mapping the work safety analysis results into a Bayesian network, the work safety analysis results mainly include task step decomposition, risk factor identification, risk level assessment, control measures, etc., and establishing an equipment maintenance operation failure probability assessment model. According to the results of the work safety analysis method, the risk factors of different maintenance stages are determined from four aspects: personnel, equipment, environment and management, and a maintenance operation failure probability assessment model is established. Through the Bayesian network model, the maintenance failure probability of each step is evaluated. The structure of the Bayesian network is divided into four layers, namely, the failure factor layer, the intermediate node layer, the maintenance activity step failure probability assessment layer, and the maintenance step failure probability assessment layer.

[0033] The prior probability of the root node is determined by combining the work safety analysis results and expert experience. The conditional probability represents the dependency relationship between the parent node and the child node. The occurrence of any root node event will lead to the failure of the sub-step maintenance activity, and the failure of any sub-step maintenance activity will lead to the failure of the maintenance activity of this step. The relationship between the parent node and the child node is an OR gate, and the Noisy-or model can be used to calculate the conditional probability. The Noisy-or model is as follows: ; in, P ( H 1, H 2,…, Hn ) represents the conditional probability of the parent node,P Hi Indicates i The realization probability of each child node.

[0034] S202: Failure consequence assessment based on matter-element theory.

[0035] Matter-element theory is a method to study the internal relationship and change law of matter and quantity, and can describe the state of things from both qualitative and quantitative perspectives. The failure consequence assessment method based on matter-element theory takes the failure mode consequence level set, failure consequence feature set and failure consequence feature value range as the three basic elements of failure consequence assessment. The failure consequence feature set of the failure mode consequence level set is established. According to the failure consequence feature value range, the correlation function is used to calculate the correlation between the failure consequence feature and the failure mode level, and the failure consequence quantitative value of a single failure mode is obtained, thus realizing the quantification of the failure consequence of a single failure mode.

[0036] The specific steps to describe the failure consequences of each failure mode based on matter-element theory are as follows: S2021: Determine the Failure Mode Effect Level Set and Failure Effect Characteristic Set The range of failure consequence characteristic values ​​is determined by the failure mode consequence level set and the failure consequence characteristic set. According to the different accident consequences corresponding to each state of the failure mode, the failure mode is divided into different levels and the failure mode consequence level set is established, as shown below: ; in, N is the set of failure mode effect levels, N 1 , N 2 , N j and N m Represents different levels of failure mode consequences, m is the number of levels divided.

[0037] The failure consequence feature set is composed of the measures of the impact of failure consequences on different aspects, as shown below: ; in, c is the failure consequence feature set, c 1 , c 2 , c i and c n Represents different failure consequence characteristics, n is the number of failure consequence characteristics.

[0038] S2022: Determine the matter-element matrix to be evaluated The matrix in matter-element theory is a table structure used to organize information systematically, which clearly presents multiple attributes of matter-element and their value range in a two-dimensional way, which is convenient for analysis, comparison or comprehensive evaluation. The matrix of matter-element to be evaluated is expressed as: ; in, R is the matter-element matrix, N For the object to be evaluated, c i ( i= 1, 2, … , n ) is the failure consequence characteristic, V i ( i= 1, 2, … , n ) is the failure consequence evaluation characteristic c i The value of .

[0039] S2023: Determine the classical domain matrix and the section domain matrix For j The failure consequence classical domain matrix of the failure mode can be expressed as: ; in, R j is the classical domain matrix, N j ( j= 1, 2, … , m ) is the j The consequence level, V ji = ( a ji , b ji ) is the failure consequence characteristic c i The value range of .

[0040] R u is the section matrix, j The node domain matter-element matrix can be obtained by combining the value ranges of all failure consequence characteristics of the classical domain matrix of all failure consequences of a failure mode.

[0041] S2024: Determine correlation function and comprehensive correlation The correlation degree indicates the degree of correlation between the failure mode effect level and the failure consequence characteristics. x 0 With finite real interval X =( Xa , Xb ) is: ; in, Xa and Xb The finite interval X upper and lower limits.

[0042] According to the distance calculation method from point to interval, the expression of the correlation function can be obtained as follows: ; in, K j ( V i ) is the correlation function, P ( V i , V ji ) is a numerical value V i With interval V ji = ( a ji , b ji ) distance.

[0043] The expression of comprehensive correlation is: ; in, K j ( P ) is the comprehensive correlation, p i For evaluation indicators c i The weight can be obtained by the hierarchical analysis method.

[0044] S2025: Determine the failure consequence level and failure consequence quantification value The maximum degree of attribution in the comprehensive correlation is taken as the failure consequence level of different failure modes. F j Indicates the failure consequence level.

[0045] The expression of the quantitative value of failure consequence is: ; in, F It is the quantitative value of the failure consequence.

[0046] S3: Initial optimization of maintenance process under known risk conditions.

[0047] The maintenance scenario of the underwater production system was constructed using graphical evaluation and review technology. The multi-objective indicators in the maintenance plan were coupled and evaluated from the perspective of maintenance time and maintenance risk, and the expected expressions of maintenance time and maintenance risk in the maintenance process were obtained. Based on the multi-objective particle swarm algorithm, the maintenance time and maintenance operation risk in the maintenance process were optimized to obtain the optimal solution set of maintenance time and maintenance operation risk.

[0048] S301: Maintenance process evaluation based on graphical evaluation and review technology.

[0049] The basic units of the graphical evaluation and review technology model are as follows: Figure 2 As shown. Among them, a and b There are two nodes, representing maintenance activities at different steps in the maintenance process. p ab Indicates that at the node a On the premise of completion, the node a To Node b Probability of realization. Maintenance activities a There is a certain probability that it cannot be achieved. At this time, a self-loop structure is introduced, that is, the maintenance activities are re-executed. a . p aa Indicates that at the node a Re-perform maintenance activities a probability. p ab The value of is obtained by calculating the failure probability of the maintenance activity at each step using the Bayesian network model. p aa and p bb The sum of is 1.

[0050] random variable x Indicates maintenance activities ( a , b ), such as the risk of a maintenance operation or the maintenance time. x , E ab ( x ) represents a random variable x In maintenance activities ( a , b ), as follows: ; in, p ab (x ) represents the maintenance activity under the condition of random variable x ( a , b ) is realized. For example, if x For maintenance activities ( a , b ) maintenance time, p ab ( x ) represents x The probability of completing the maintenance activity within hours.

[0051] f ab Indicates maintenance activities ( a , b ) in the random variable x The characteristic function is as follows: ; in, s Represents any real number.

[0052] According to the characteristic function and maintenance activities ( a , b ) is the probability of realization p ab , the transfer function can be calculated as: ; in, W ab ( s ) indicates maintenance activities ( a , b )’s transfer function.

[0053] The self-loop activity may be executed multiple times, so the self-loop activity is replaced by a set of parallel structures. The equivalent transfer function of the self-loop structure is: ; in, W' ab ( s ) represents the equivalent transfer function of the self-loop structure, f aa Indicates maintenance activities a Random variables during re-execution x The characteristic function of .

[0054] The equivalent transfer function of the entire graphical evaluation and review technology network is expressed as W E ( s ), can be solved by matrix method. s = 0, the equivalent transfer function of the entire network is: ; in, p E is the equivalent probability of successful execution from the first node to the last node, W E (0) s = 0.

[0055] The equivalent characteristic function corresponding to the entire graphical evaluation and review technology network is: ; in, f E ( s ) is the equivalent characteristic function corresponding to the entire graphical evaluation and review technology network.

[0056] According to the basic properties of characteristic function, characteristic function s = 0 n The value of the derivative is the random variable n In order to evaluate the effectiveness of the maintenance process, the expected function of the random variable in the entire process is calculated as: ; in, E ( x ) represents a random variable x Expected function in the entire maintenance process.

[0057] S302: Optimization of minimum expected maintenance time and expected maintenance risk based on multi-objective particle swarm algorithm.

[0058] In order to optimize the maintenance operation risk, additional operation time is introduced to analyze the maintenance operation process. By adjusting the maintenance operation time, rationally allocating maintenance tasks and personnel arrangements, and refining the operations within each maintenance step, the maintenance risk can be reduced. Assuming that there is a linear relationship between maintenance time and maintenance risk, the relationship between additional operation time, maintenance time and maintenance risk is: ; in, T ab In maintenance activities ( a , b ) maintenance time before additional operating time is invested, R ab During maintenance activities ( a , b ) Risk of maintenance operations before investing additional operating time, t ab is the additional operation time, T' abDuring maintenance activities ( a , b ) Maintenance time after investing extra operating time, R' ab During maintenance activities ( a , b ) Risk of maintenance operation after investing extra operating time, z ab It is the proportional coefficient of the impact of additional maintenance time on the risk of maintenance operation.

[0059] Based on the multi-objective particle swarm optimization algorithm, the optimal maintenance time and optimal maintenance operation risk in the maintenance process are optimized. The multi-objective function of the optimization algorithm is expressed as: ; in, E [ T ] represents the expected maintenance time of the maintenance process, E [ R ] represents the expected maintenance risk of the maintenance process.

[0060] S4: Resilience-driven secondary optimization of maintenance processes.

[0061] When the system is affected by external shocks, it may cause the underwater production system to shut down and the system performance to drop to zero. At this time, maintenance activities need to be carried out immediately. For underwater production systems, multiple components are geographically dispersed. Limited maintenance resources make it difficult to restore all components at the same time. In actual operations, maintenance work is often carried out in batches, and the function of a certain component is restored each time, thus forming a step-by-step performance improvement. In this case, the step-by-step performance recovery curve is closer to reality. The system performance step-by-step recovery curve is as follows: Figure 3 shown. Ts Indicates the moment when the system is impacted and repairs are performed immediately at this moment. You Indicates the time when maintenance activities are completed. P 0 represents the initial performance of the system, P ( t )represent t System performance at all times. ΔP i Indicates i The incremental system performance brought by each component T i Indicates i The time required to repair a component.

[0062] The incremental system performance brought by the maintenance of each component is determined based on the comprehensive weight of the component. The comprehensive weight of the component comprehensively considers the impact of the component state distribution probability and the component state transition rate on the system reliability.i The comprehensive weight of the components is calculated as: ; in, IM ( i ) represents i The combined weight of the components, P ( t )represent t System performance at all times, P i ( t )represent t Moment i The performance of each component, l i For the i The failure rate of a component.

[0063] The impact of external shocks on the subsea production system can be quantified as the performance loss from the time the shock occurs to the end of the maintenance activity. The system resilience value can be obtained by calculating the area ratio of the performance curve. Taking the maximum resilience as the optimization goal, the maintenance process is optimized twice. The system resilience can be calculated as ; Among them, Re represents the resilience of the system.

[0064] Embodiment 2: like Figure 4As shown, the subsea production control system includes a main control station 101, an emergency shutdown system 102, an electric power unit 103, a hydraulic power unit 104, an uninterruptible power supply 105, an upper umbilical cable terminal 106, an accumulator group 107, a chemical agent injection unit 108, an umbilical cable 109, a subsea distribution unit 110, an electric flying line 111, a subsea control module 112, a hydraulic flying line 113, and a subsea Christmas tree 114; wherein the main control station 101 is connected to the electric power unit 103, the hydraulic power unit 104 and the upper umbilical cable terminal 106 through cables. 06, for controlling the operation of the underwater production system and collecting monitoring data of various equipment; the emergency shutdown system 102 is connected to the main control station 101 through a cable, for emergency shutdown of the underwater production system; the electric power unit 103 is connected to the upper umbilical cable terminal 106 through a cable, for providing power and communication for the underwater equipment; the hydraulic power unit 104 is connected to the upper umbilical cable terminal 106 through a hydraulic pipeline, for providing high-pressure oil and low-pressure oil for the underwater equipment; the uninterruptible power supply 105 is connected to the electric power unit 103 through a cable, for The upper umbilical cable terminal 106 is directly connected to the umbilical cable 109 to transmit electricity, hydraulic power and control signals to the underwater production system. The accumulator group 107 is connected to the hydraulic power unit 104 through a hydraulic pipeline to maintain the pressure stability of the underwater production system and supply high-pressure oil. The chemical injection unit 108 is connected to the upper umbilical cable terminal 106 through a pipeline to inject chemicals into the underwater production system. The umbilical cable 109 is directly connected to the underwater distribution unit 110 to transmit The underwater distribution unit 110 is connected to the underwater control module 112 via a cable to distribute underwater electric power and hydraulic power; the underwater distribution unit 110 is connected to the underwater oil production tree 114 via an electric flying line 111 to transmit power and communication signals; the underwater control module 112 is connected to the underwater oil production tree 114 via a cable to control and monitor the underwater oil production tree 114; the underwater distribution unit 110 is connected to the underwater oil production tree 114 via a liquid flying line 113 to provide hydraulic power for the underwater oil production tree 114.

[0065] Embodiment three: A resilience-driven underwater production system condition-based maintenance decision optimization method is applied to the resilience-driven underwater production system condition-based maintenance decision optimization system.

[0066] like Figure 5As shown, the resilience-driven underwater production system condition-based maintenance decision optimization system includes a data acquisition module 201, a state analysis module 202, a maintenance process preliminary decision module 203, a work safety analysis module 204, a maintenance failure probability analysis unit 205, a maintenance failure consequence analysis unit 206, a maintenance operation risk assessment module 207, a maintenance operation risk output module 208, a maintenance process evaluation module 209, a maintenance process evaluation result output module 210, an expected maintenance risk output unit 211, an expected maintenance time output unit 212, a maintenance process primary optimization module 213, a maintenance process secondary optimization module 214, and a maintenance process decision output module 215. Among them, the maintenance operation risk assessment module 207 includes a maintenance failure probability analysis unit 205 and a maintenance failure consequence analysis unit 206; the maintenance process evaluation result output module 210 includes an expected maintenance risk output unit 211 and an expected maintenance time output unit 212.

[0067] The data acquisition module 201 is connected to the main control station 101 through a cable, and is used to collect monitoring data of the underwater production system; the state analysis module 202 is connected to the data acquisition module 201 through a cable, and is used to analyze the collected monitoring data and determine the state of the underwater production system; the maintenance process preliminary decision module 203 is connected to the state analysis module 202 through a cable, and preliminarily formulates a maintenance strategy based on the state of the underwater production system; the work safety analysis module 204 is connected to the maintenance process preliminary decision module 203 through a cable, and is used to divide the maintenance steps and analyze the potential failure causes in the maintenance process; the maintenance failure probability analysis unit 205 is connected to the work safety analysis module 204 through a cable, and is used to calculate the failure probability of the maintenance operation; the maintenance failure consequence analysis unit 206 is connected to the work safety analysis module 204 through a cable, and is used to evaluate the failure consequences of the maintenance operation; the maintenance operation risk assessment module 207 includes the maintenance failure probability analysis unit 205 and the maintenance failure consequence analysis unit 206; the maintenance operation risk output module 208 is connected to the maintenance operation risk assessment module 207 through a cable, and is used to output the failure probability of the maintenance operation and the failure consequences of the maintenance operation; The evaluation module 209 is connected to the maintenance operation risk output module 208 through a cable, and is used to construct a maintenance process evaluation model based on graphical evaluation and review technology under the condition of known maintenance operation risk; the expected maintenance risk output unit 211 is connected to the maintenance process evaluation module 209 through a cable, and is used to output the expression of expected maintenance risk; the expected maintenance time output unit 212 is connected to the maintenance process evaluation module 209 through a cable, and is used to output the expression of expected maintenance time; the maintenance process evaluation result output module 210 includes the expected maintenance risk output unit 211 and the expected maintenance time output unit 212; the maintenance process primary optimization module 213 is connected to the maintenance process evaluation result output module 210 through a cable, and is used to establish a multi-objective primary optimization model under the minimum maintenance risk and minimum maintenance time; the maintenance process secondary optimization module 214 is connected to the maintenance process primary optimization module 213 through a cable, and is used to establish a secondary optimization model under the condition of maximum resilience based on the Pareto solution set of maintenance risk and maintenance time; the maintenance process decision output module 215 is connected to the maintenance process secondary optimization module 214 through a cable, and is used to output the final maintenance process.

[0068] In the process of the resilience-driven underwater production system situation-based maintenance decision optimization system working, the data acquisition module 201 collects monitoring data of the underwater production system based on the monitoring data of the main control station 101, the state analysis module 202 analyzes and determines the state of the underwater production system based on the collected monitoring data, the maintenance process preliminary decision module 203 preliminarily formulates a maintenance strategy based on the state of the underwater production system, the work safety analysis module 204 divides the maintenance steps and analyzes the potential failure causes in the maintenance process based on the formulated preliminary maintenance process, the maintenance operation risk assessment module 207 quantitatively evaluates the failure probability and failure consequences of the maintenance operation based on the results, and the maintenance operation risk output module 208 outputs the failure probability and maintenance consequences of the maintenance operation. The maintenance process evaluation module 209 constructs a maintenance process evaluation model based on graphical evaluation and review technology under the condition of known maintenance operation risks. The maintenance process evaluation result output module 210 outputs the expression of expected maintenance risk and expected maintenance time of maintenance process evaluation. The maintenance process primary optimization module 213 establishes a multi-objective primary optimization model under the minimum maintenance risk and minimum maintenance time. The maintenance process secondary optimization module 214 establishes a secondary optimization model under the condition of maximizing resilience based on the Pareto solution set of maintenance risk and maintenance time. The maintenance process decision output module 215 outputs the final maintenance optimization result, and the relevant maintenance personnel extract information to formulate a maintenance plan to ensure the safety of oil and gas production.

[0069] In the present invention, maintenance risk is maintenance operation risk, and maintenance operation risk is quantitatively calculated as the product of maintenance failure probability and failure consequence; maintenance failure probability is quantitatively evaluated by mapping the work safety analysis result into a Bayesian network, and according to the result of the work safety analysis method, the risk factors of different maintenance stages are determined from four aspects of personnel, equipment, environment and management, and a maintenance operation failure probability evaluation model is established; the structure of the Bayesian network is divided into four layers, namely, failure factor layer, intermediate node layer, maintenance activity step failure probability evaluation layer, and maintenance step failure probability evaluation layer; the prior probability of the root node of the Bayesian network is determined by combining the work safety analysis result and expert experience; the relationship between the parent node and the child node is an OR gate, and the Noisy-or model can be used to calculate the conditional probability.

[0070] The basic maintenance operation time is determined by historical data such as actual operation cases. The additional operation time is introduced to optimize the maintenance operation plan. By adjusting the maintenance operation time, the maintenance tasks and personnel arrangements are reasonably allocated, and the operations in each maintenance step are refined, the maintenance risk is reduced. The specific value of the additional operation time is calculated from the multi-objective optimization results.

[0071] Maintenance risk and maintenance time can also be obtained in the following ways: 1. Acquisition of maintenance risk: Based on Fault Tree Analysis (FTA): By establishing a fault tree model of the underwater production system, the minimum cut set of system failures is found, the main failure mode of each component and its impact on the system are determined, and then the failure probability of each component is calculated as a quantitative indicator of maintenance risk.

[0072] Based on reliability analysis: Using the reliability function and failure rate model in reliability theory, combined with the historical operation data and failure data of the equipment, the failure probability of the equipment at different time points is calculated to evaluate the maintenance risk. For example, by analyzing the equipment's failure interval data and fitting the corresponding probability distribution model, such as exponential distribution, Weibull distribution, etc., the failure rate function of the equipment is obtained, and then its maintenance risk is determined.

[0073] 2. Obtaining maintenance time: Degradation model based methods: Gamma process model: The degradation process of the equipment is modeled as a Gamma process. Based on the degradation parameters and failure threshold of the equipment, the time required for the equipment to reach the failure state is calculated to determine the maintenance time.

[0074] Wiener process model: The Wiener process is used to model the performance degradation data of the equipment. By estimating the model parameters, the remaining service life of the equipment is predicted, and the maintenance time is determined.

[0075] Method based on failure data statistics: collect historical failure data of equipment, count the failure intervals of equipment, establish corresponding probability distribution models, such as Weibull distribution model, calculate the failure probability density function and cumulative distribution function of equipment according to model parameters, and thus determine the maintenance time interval of equipment.

[0076] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can make equivalent replacements or changes according to the technical scheme and inventive concept of the present invention within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A resilience-driven optimization method for situation-based maintenance decision-making in a subsea production system, characterized in that: include: Conduct quantitative assessment of maintenance operation risks; Based on the quantitative evaluation results, the maintenance process is initially optimized from the perspective of maintenance time and maintenance risk; Combine system resilience to conduct secondary optimization based on the initial optimization; The system toughness is calculated based on the system performance curve.

2. The resilience-driven subsea production system condition-based maintenance decision optimization method according to claim 1 is characterized in that: Specifically include: S1 Work Safety Analysis: Develop preliminary maintenance procedures, identify and analyze risk factors in the preliminary maintenance procedures and possible consequences of maintenance operation failures; S2 Maintenance operation risk assessment: Map the work safety analysis results into a Bayesian network, establish an equipment maintenance operation failure probability assessment model and a failure consequence assessment model, and the maintenance operation risk is the product of the maintenance failure probability and the failure consequence; S3 Initial optimization of maintenance process under known risk conditions: Construct a maintenance scenario for the underwater production system, perform initial optimization of the maintenance process from the perspective of maintenance time and maintenance risk, and obtain the optimal solution set for maintenance time and maintenance operation risk; S4 Resilience-driven secondary optimization of maintenance process: Calculate system resilience based on system performance curve, take maximizing resilience as the optimization goal, and perform secondary optimization on the maintenance process.

3. The resilience-driven subsea production system condition-based maintenance decision optimization method according to claim 2 is characterized in that: The development of the initial maintenance process includes: Establish a system performance degradation prediction model based on degradation data; Predict the cumulative degradation amount through the system performance degradation prediction model to determine the system status; Develop a preliminary maintenance process based on system status and previous maintenance experience.

4. The resilience-driven subsea production system condition-based maintenance decision optimization method according to claim 3 is characterized in that: The system performance degradation prediction model is established through the Wiener random process. The Wiener process model is as follows: ； in, X ( t ) is the system at time t The amount of performance degradation; X (0) is the initial performance degradation, which is usually assumed to be 0; μ is the drift coefficient reflecting the individual degradation rate; σ 2 is the diffusion coefficient of the Wiener process, which indicates the degree of fluctuation of performance degradation; B ( t ) represents the standard Brownian motion, which is used to describe the uncertainty of the random degradation quantity on the time axis and satisfies B ( t )=0 and B ( t )~ N (0, t ), N (0, t ) is a normal distribution with mean 0 and variance t.

5. The resilience-driven subsea production system condition-based maintenance decision optimization method according to claim 4 is characterized in that: The cumulative degradation amount is calculated as follows: ； Among them, Δ X ( t ) is the time difference Δ t The cumulative degradation of performance during the period; X ( t +Δ t ) is the system at time t +Δ t The amount of performance degradation.

6. The resilience-driven subsea production system condition-based maintenance decision optimization method according to claim 5 is characterized in that: The Wiener process also includes estimating the drift coefficient and the diffusion coefficient, the steps are as follows: S101: Initialize the distribution: choose initial parameters μ (0) and σ (0) ; S102: Step E, calculate expectation: estimate the expected value of the unknown parameter and give the existing parameter estimate; S103: Step M, update parameters: re-estimate the unknown parameters, determine the values ​​of the model parameters by maximizing the likelihood function, and obtain the expected estimates of the unknown parameters; S104: Convergence judgment: Check whether the parameters have converged. If not, return to step (2); otherwise, stop the iteration.

7. The resilience-driven subsea production system condition-based maintenance decision optimization method according to claim 6 is characterized in that: The Bayesian network structure of the maintenance failure probability assessment model is divided into four layers, namely, failure factor layer, intermediate node layer, maintenance activity step failure probability assessment layer, and maintenance step failure probability assessment layer.

8. The resilience-driven subsea production system condition-based maintenance decision optimization method according to claim 7 is characterized in that: The maintenance failure probability includes a priori probability and conditional probability, wherein the a priori probability of the root node is determined by combining the work safety analysis results and expert experience, and the conditional probability is calculated by the Noisy-or model: ； in, P ( H 1, H 2,…, H ) represents the conditional probability of the parent node, P Hi Indicates i The realization probability of a child node.

9. The resilience-driven subsea production system condition-based maintenance decision optimization method according to claim 8 is characterized in that: The failure consequences are evaluated based on physical meta-theory, including: S2021: Determine the Failure Mode Effect Level Set and Failure Effect Characteristic Set Classify the failure modes into different levels and establish the failure mode consequence level set as follows: ； in, N is the set of failure mode effect levels, N 1. N 2. N j and N m Represents different levels of failure mode consequences, m is the number of levels divided, The failure consequence feature set is composed of the measures of the impact of failure consequences on different aspects, as shown below: ； in, c is the failure consequence feature set, c 1. c 2. c i and c n Represents different failure consequence characteristics, n is the number of failure consequence characteristics; S2022: Determine the matter-element matrix to be evaluated The matter-element matrix to be evaluated is expressed as: ； in, R is the matter-element matrix, N For the object to be evaluated, c i ( i= 1,2, … , n ) is the failure consequence characteristic, V i ( i= 1,2, … , n ) is the failure consequence evaluation characteristic c i The value of S2023: Determine the classical domain matrix and the section domain matrix For j The failure consequence classical domain matrix of the failure mode can be expressed as: ； in, R j is the classical domain matrix, N j ( j= 1,2, … , m ) is the j The consequence level, V ji = ( a ji , b ji ) is the failure consequence characteristic c i The value range of R u is the section matrix, j The value ranges of all failure consequence characteristics of the classical domain matrix of all failure consequences of each failure mode are combined to obtain the node domain matter-element matrix; S2024: Determine correlation function and comprehensive correlation The correlation degree indicates the correlation degree between the failure mode consequence level and the failure consequence characteristics. x 0 and finite real interval X =( XA , Xb ) is: ； XA and Xb The finite interval X The upper and lower limits of According to the distance calculation method from point to interval, the expression of the correlation function can be obtained as follows: ； in, K j ( V i ) is the correlation function, P ( V i , V ji ) is a numerical value V i With interval V ji = ( a ji , b ji ) distance; The expression of comprehensive correlation is: ； in, K j ( P ) is the comprehensive correlation, p i For evaluation indicators c i The weight of can be obtained by the hierarchical analysis method; S2025: Determine the failure consequence level and failure consequence quantification value The maximum degree of attribution in the comprehensive correlation is taken as the failure consequence level of different failure modes. F j Indicates the failure consequence level; the expression of the failure consequence quantification value is: ； in, F It is the quantitative value of the failure consequence.

10. The resilience-driven subsea production system condition-based maintenance decision optimization method according to claim 9 is characterized in that: The step S3 specifically includes: E ab ( x ) represents a random variable x In maintenance activities ( a , b ), as follows: ； in, p ab ( x ) represents the maintenance activity under the condition of random variable x ( a , b ) is the probability of realization, x For maintenance activities ( a , b ) maintenance time, p ab ( x ) represents x The probability of completing the maintenance activity within hours; φ ab Indicates maintenance activities ( a , b ) in the random variable x The characteristic function is as follows: ； in, s represents any real number; According to the characteristic function and maintenance activities ( a , b ) is the probability of realization p ab , the transfer function can be calculated as: ； in, W ab ( s ) indicates maintenance activities ( a , b )’s transfer function; The equivalent transfer function of the self-loop structure is: ； in, W' ab ( s ) represents the equivalent transfer function of the self-loop structure, φ aa Indicates maintenance activities a Random variables during re-execution x The characteristic function of The equivalent transfer function of the entire graphical evaluation and review technology network is expressed as W E ( s ), can be solved by matrix method, when s = 0, the equivalent transfer function of the entire network is: ； in, p E is the equivalent probability of successful execution from the first node to the last node, W E (0) s = 0 when the network equivalent transfer function; The equivalent characteristic function corresponding to the entire graphical evaluation and review technology network is: ； in, φ E ( s ) is the equivalent characteristic function corresponding to the entire graphical evaluation and review technology network; According to the basic properties of characteristic functions, characteristic functions are s = 0 n The value of the derivative is the random variable n In order to evaluate the effectiveness of the maintenance process, the expected function of the random variable in the entire process is calculated as: ； in, E ( x ) represents a random variable x The expected function in the whole maintenance process; Assuming that there is a linear relationship between maintenance time and maintenance risk, the relationship between additional operation time, maintenance time and maintenance risk is: ； in, T ab In maintenance activities ( a , b ) maintenance time before additional operating time is invested, R ab During maintenance activities ( a , b ) Risk of maintenance operations before investing additional operating time, t ab is the additional operation time, T' ab During maintenance activities ( a , b ) Maintenance time after investing extra operating time, R' ab During maintenance activities ( a , b ) Risk of maintenance operation after investing extra operating time, z ab is the proportionality coefficient of the effect of additional maintenance time on the risk of maintenance operations; Based on the multi-objective particle swarm optimization algorithm, the optimal maintenance time and optimal maintenance operation risk in the maintenance process are optimized. The multi-objective function of the optimization algorithm is expressed as: ； in, E [ T ] represents the expected maintenance time of the maintenance process, E [ R ] represents the expected maintenance risk of the maintenance process.

11. A resilience-driven subsea production system condition-based maintenance decision optimization system, characterized in that: include: The data acquisition module collects monitoring data of the underwater production system based on the monitoring data of the master control station; The status analysis module analyzes and determines the status of the subsea production system based on the collected monitoring data; The maintenance process preliminary decision module preliminarily formulates maintenance strategies based on the status of the underwater production system; The work safety analysis module divides the maintenance steps according to the established preliminary maintenance process and analyzes the potential failure causes in the maintenance process; The maintenance operation risk assessment module conducts quantitative assessment of the failure probability and failure consequences of maintenance operations based on the results; The maintenance operation risk output module outputs the maintenance operation failure probability and maintenance operation failure consequences; The maintenance process evaluation module constructs a maintenance process evaluation model based on graphical evaluation and review technology under the condition that the maintenance operation risk is known; The maintenance process evaluation result output module outputs the expression of expected maintenance risk and expected maintenance time of the maintenance process evaluation; The maintenance process initial optimization module establishes a multi-objective initial optimization model under the minimum maintenance risk and minimum maintenance time; The maintenance process secondary optimization module establishes a secondary optimization model under the condition of maximizing resilience based on the Pareto solution set of maintenance risk and maintenance time. The maintenance process decision output module outputs the final maintenance optimization result, and relevant maintenance personnel extract information to formulate maintenance plans to ensure the safety of oil and gas production.

Images