Fuzzy bayesian network-based fault diagnosis method for FPSO single point mooring system

Abstract

The application provides a fuzzy Bayesian network-based FPSO internal turret type single point mooring system fault diagnosis method, which comprises the following steps: 1) identifying the fault mode of the FPSO internal turret type single point mooring system, and constructing a Bayesian network structure model; 2) collecting expert opinions and converting them into failure probabilities, and constructing a Bayesian network parameter model; 3) collecting sensor data of the FPSO internal turret type single point mooring system, determining the state of the observable nodes in the Bayesian network fault diagnosis model of the FPSO internal turret type single point mooring system, and the Bayesian network fault diagnosis model is composed of the Bayesian network structure model and the Bayesian network parameter model, and then relying on the Bayesian network fault diagnosis model to perform posterior reasoning, and realizing fault diagnosis. The application can comprehensively and quantitatively analyze the FPSO internal turret type single point mooring system, realize risk analysis and fault diagnosis, and provide protection and reference for the operation and maintenance of the FPSO internal turret type single point mooring system.

Description

Technical Field

[0001] This invention relates to the field of FPSO fault diagnosis, and specifically to a fault diagnosis method for FPSO single-point mooring systems based on fuzzy Bayesian networks. Background Technology

[0002] Floating Production Storage and Offtake (FPSO) vessels are a crucial component of offshore oil development and are widely used in the Bohai and South China Seas. The single-point mooring system, the core system connecting the FPSO to the production platform / subsea production system, consists of mooring chains, connectors, bearings, slip ring systems, etc. It undertakes core tasks such as FPSO positioning, rotation, well fluid transfer, platform power supply, and communication transmission. A failure in this system can directly lead to reduced oilfield production or even shutdown.

[0003] Traditional oilfield operators typically include obvious aspects such as mechanical equipment and production process equipment in their production and maintenance planning. However, underwater mooring systems are often overlooked or even considered to be indestructible devices. In fact, due to insufficient design redundancy, fatigue, installation damage, and other reasons, single-point mooring systems often have inherent defects. Therefore, risk analysis and fault diagnosis of single-point mooring systems should be given sufficient attention.

[0004] Therefore, in order to ensure the safety of single-point mooring systems, it is necessary to conduct research on the failure mechanisms, risk analysis and fault diagnosis of key components of FPSO single-point mooring systems, so as to provide an important foundation for fault early warning and fault investigation of single-point mooring systems. Summary of the Invention

[0005] To address the technical problems existing in the background art, this invention proposes a fault diagnosis method for FPSO single-point mooring systems based on fuzzy Bayesian networks. Its design is reasonable and can achieve comprehensive quantitative analysis of FPSO internal turret single-point mooring systems, as well as risk analysis and fault diagnosis, providing assurance and reference for the operation, maintenance, and repair of FPSO internal turret single-point mooring systems.

[0006] To address the aforementioned technical problems, this invention provides a fault diagnosis method for FPSO single-point mooring systems based on fault trees and fuzzy Bayesian networks, which mainly includes the following steps:

[0007] 1) Identify the failure modes of the FPSO internal turret single-point mooring system and construct a Bayesian network structure model;

[0008] 2) Collect expert opinions and convert them into failure probabilities to construct a Bayesian network parameter model;

[0009] 3) Collect sensor data from the FPSO internal turret single-point mooring system, determine the state of observable nodes in the Bayesian network fault diagnosis model of the FPSO internal turret single-point mooring system, and the Bayesian network fault diagnosis model consists of a Bayesian network structure model and a Bayesian network parameter model. Then, rely on the Bayesian network fault diagnosis model to perform posterior inference to achieve fault diagnosis.

[0010] The fault diagnosis method for FPSO single-point mooring systems based on fault trees and fuzzy Bayesian networks, wherein step 1) specifically includes the following steps:

[0011] 1.1) Combining the fault database, fault history data, relevant expert experience in the field of marine engineering, and the structure of the FPSO internal turret single-point mooring system, identify and sort out the typical fault modes and fault mechanisms of the FPSO internal turret single-point mooring system.

[0012] 1.2) The overall failure state of the FPSO internal turret single-point mooring system is taken as the top event of the fault tree, the failure modes of each component in the FPSO internal turret single-point mooring system are taken as the intermediate events of the fault tree, and the failure mechanism of the FPSO internal turret single-point mooring system is taken as the bottom event of the fault tree. Based on the causal logic relationship between each event node, the single-point system fault tree model is constructed from top to bottom.

[0013] 1.3) Using the mapping principle, the fault tree is transformed into a Bayesian network, resulting in a Bayesian network structure model composed of fault mechanism nodes, component fault mode nodes, and observation nodes.

[0014] The fault diagnosis method for FPSO single-point mooring systems based on fault trees and fuzzy Bayesian networks, wherein step 2) specifically includes the following steps:

[0015] 2.1) Develop an expert scoring sheet, set weights for different scorers, and collect expert information and scoring results;

[0016] 2.2) Calculate the prior probability of the root node of the Bayesian network model;

[0017] 2.3) Calculate the conditional probability table of the Bayesian network using the Noisy-OR gate method to obtain the parameter model of the Bayesian network.

[0018] The fault diagnosis method for FPSO single-point mooring systems based on fault trees and fuzzy Bayesian networks, wherein step 2.2) specifically includes the following steps:

[0019] 2.2.1) First, calculate the expert's score matrix Θ using the following formula (1):

[0020]

[0021] For the s-th evaluation criterion (s=1,…,l), θ js It is expert E j The score of the evaluation criteria for (j=1,…,m);

[0022] Then, the comprehensive score q of each expert is calculated using the entropy weight method according to the following formula (2). j :

[0023]

[0024] In equation (2) above, p js Let e ​​represent the proportion of the j-th expert in that evaluation indicator under the s-th evaluation criterion. s Let d represent the entropy value of the s-th evaluation indicator. s For information entropy redundancy, w s q represents the weight of each evaluation indicator. j This represents the overall score of each expert;

[0025] Finally, the expert E is calculated using the following formula (3). j The weights W(E) j ):

[0026]

[0027] 2.2.2) The relative consistency RA(E) of the expert ratings is calculated using the following formula (4). j ):

[0028]

[0029] In the above formula (4) Indicated by expert E j The fuzzy scoring language is transformed into triangular intuitionistic fuzzy numbers. (a j ,b j ,c j ,a′ j ,b j ,c′ j ) represents the specific numerical value in the fuzzy number; WA(E) j ) represents the weighted consistency degree;

[0030] Among them, in the above formula (4) The similarity between two rating results is calculated by the following formula (5):

[0031]

[0032] Among them, in the above formula (5) Represents fuzzy numbers The expected assessment is calculated by the following formula (6):

[0033]

[0034] 2.2.3) Taking into account the expert score weight W(E) j ) and relative consistency RA(E) j The expert E is calculated by the following formula (7). j Consistency coefficient CC(E) j ( ), as an expert E in the weighted calculation of fuzzy opinion aggregation. j The weights;

[0035] CC(E j )=β·W(E j )+(1-β)·RA(E j (7);

[0036] In equation (7) above, β (0 < β < 1) is a relaxation factor. This coefficient is used to indicate the relative importance of expert weights and relative consistency. When calculating, β = 0.5 is taken, which means that expert weights and relative consistency are equally important.

[0037] 2.2.4) Calculate the result u after the aggregation of expert fuzzy opinions using the following equations (8) to (9). z (x);

[0038] Trigonometric Intuition Fuzzy Numbers The addition and multiplication formulas are defined by the following formula (8):

[0039]

[0040] In equation (8) above, CC(E) j ) is a constant. It is a triangular intuitive fuzzy number. λ is any constant.

[0041]

[0042] The result of fuzzy opinion aggregation u z (x) is calculated by the following formula (9):

[0043]

[0044] In equation (9) above, u z (x) represents the aggregated result of expert fuzzy opinions at node x, u z (x)=(a x ,b x ,c x ;a x ′,bx ,c x ′);

[0045] 2.2.5) Calculate the fuzzy failure probability FFP of the root node of the Bayesian network model using the following equations (10) to (12), and use the fuzzy failure probability FFP as the prior probability of the root node of the Bayesian network model:

[0046] For the fuzzy opinion aggregation result u z (x), whose centroid FPS is defuzzified using the intuitive fuzzy method is shown in equation (10):

[0047]

[0048] The centroid FPS is converted to the fuzzy failure probability FFP using the following conversion formula (11):

[0049]

[0050] In equation (11), K is as shown in equation (12):

[0051]

[0052] The fault diagnosis method for FPSO single-point mooring systems based on fault trees and fuzzy Bayesian networks, wherein the specific steps for calculating the Bayesian network conditional probability table in step 2.3) are as follows:

[0053] 2.3.1) The probability of a child node failing when only the parent node fails is obtained through expert scoring;

[0054] 2.3.2) The conditional probabilities of all child nodes are calculated using the following formula (13):

[0055]

[0056] In equation (13) above, p i Represents node X i (1≤i≤n) is true, and the remaining nodes X j The probability that Y is true when (1≤j≤n, and j≠i) is false; X={X1,...,X n}

[0057] The fault diagnosis method for FPSO single-point mooring systems based on fault trees and fuzzy Bayesian networks, wherein step 3) specifically includes the following steps:

[0058] 3.1) Collect operational status data of the FPSO's internal turret single-point mooring system, and determine the status of the observation nodes based on the collected data;

[0059] 3.2) By performing posterior reasoning through the Bayesian network fault diagnosis model, the conditional probability of fault occurrence of fault mechanism nodes is obtained, the most likely fault mechanism is determined, and the location where the most likely fault occurs is found.

[0060] The fault diagnosis method for FPSO single-point mooring systems based on fault trees and fuzzy Bayesian networks, wherein the specific process of step 3.2) is as follows:

[0061] 3.2.1) Based on the conditional independence assumption and the chain rule, the variables U = A1, A2, ..., A n The joint probability P(U) can be calculated according to equation (14):

[0062]

[0063] In equation (14) above, A i It is a node in a Bayesian network, Pa(A) i ) is A in a Bayesian network i The parent node;

[0064] 3.2.2) The posterior probability of the root node of the Bayesian network model is obtained by performing backpropagation of the Bayesian network using the following equation (15):

[0065]

[0066] In equation (15) above, E = {e1,…,e m} is a given set of evidence, determined by the collected data;

[0067] 3.2.3) Define the fault identification rules, specifically as follows:

[0068] ① When the probability of failure of a component is greater than 70%, the component fails;

[0069] ② When the probability of failure of a component is greater than 15% but less than 70%, the component may fail;

[0070] ③ When the failure probability of a component is less than 15%, the component is considered fault-free.

[0071] By adopting the above technical solution, the present invention has the following beneficial effects:

[0072] This invention presents a reasonable fault diagnosis method for FPSO internal turret single-point mooring systems based on fuzzy Bayesian networks. It enables comprehensive quantitative analysis of FPSO internal turret single-point mooring systems, achieving risk analysis and fault diagnosis, thus providing assurance and reference for the operation and maintenance of FPSO internal turret single-point mooring systems. The fault diagnosis method in this invention is a data-driven approach. Compared to traditional model- or signal-based fault diagnosis methods, this method does not rely on accurate mathematical models or precise monitoring signals, making it more suitable for complex industrial systems like FPSO internal turret single-point mooring systems. Compared to other data-driven fault diagnosis methods, such as FMEA and fault tree methods, this method offers quantitative analysis and fully utilizes real-time data from the sensor monitoring system of the FPSO internal turret single-point mooring system, achieving more real-time, faster, and more accurate fault diagnosis. Specifically, it obtains fuzzy evaluation results from experts through expert scoring, and further aggregates multi-expert fuzzy opinions through similarity aggregation, comprehensively considering the influence of individual expert weights and opinion consistency on the aggregation results. Simultaneously, it uses T... w The operator avoids the accumulation of ambiguity; the Bayesian network conditional probability table is constructed by using the Noisy-OR gate, which saves the number of parameters required for scoring and reduces the amount of computation; after determining the state of the observed node, the failure probability of the component is calculated through posterior inference, and the relationship between the failure probability and the fault identification rules is given, providing a guarantee and reference for risk identification and fault diagnosis of the FPSO internal turret single-point mooring system. Attached Figure Description

[0073] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0074] Figure 1 This is a flowchart illustrating the specific process of the FPSO single-point mooring system fault diagnosis method based on fuzzy Bayesian networks according to the present invention.

[0075] Figure 2 This is a schematic diagram of the fault tree mapping Bayesian network involved in the fault diagnosis method for FPSO single-point mooring system based on fuzzy Bayesian network of the present invention.

[0076] Figure 3 This is a schematic diagram of the Bayesian network fault diagnosis model involved in the fault diagnosis method for FPSO single-point mooring system based on fuzzy Bayesian network of the present invention. Detailed Implementation

[0077] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0078] The present invention will be further explained below with reference to specific embodiments.

[0079] like Figure 1 As shown in this embodiment, the fault diagnosis method for an FPSO internal turret single-point mooring system based on fuzzy Bayesian networks includes the following steps:

[0080] S100. Identify the failure modes of the FPSO internal turret single-point mooring system (specifically including hydraulic slip rings, electric slip rings, bearings and mooring system), and construct a Bayesian network structure model.

[0081] Specifically, by combining the FPSO internal turret single-point mooring system fault database, fault history data, relevant expert experience in the field of marine engineering, and the structure of the FPSO internal turret single-point mooring system, typical fault modes and fault mechanisms of the FPSO internal turret single-point mooring system are identified and sorted out. These are generally common or potential fault modes. Based on the structure and design of the FPSO internal turret single-point mooring system, the causal relationships between each fault mode are determined. The overall failure state of the FPSO internal turret single-point mooring system is used as a fault tree model, the fault modes of each component of the FPSO internal turret single-point mooring system are used as intermediate events in the fault tree, and the fault mechanism of the FPSO internal turret single-point mooring system is used as the bottom event of the fault tree. Using the mapping principle, the fault tree is transformed into a Bayesian network, resulting in a Bayesian network structure model composed of fault mechanism nodes, component fault mode nodes, and observation nodes.

[0082] Among them, the aforementioned Bayesian network fault diagnosis model is as follows: Figure 3 As shown in Table 1, the common or potential fault modes and node descriptions of the FPSO internal turret single-point mooring system, as presented in the Bayesian network fault diagnosis model, are as follows:

[0083] Table 1

[0084]

[0085]

[0086] S200. Collect expert opinions and convert them into failure probabilities to construct a Bayesian network parameter model; specifically, this includes the following steps:

[0087] S210. Develop an expert scoring table, set the weights for different scorers, and collect expert information and expert scoring results.

[0088] S220. Calculate the prior probability of the root node of the Bayesian network model; specifically including the following steps:

[0089] S221. Collect expert information and consider the impact of expert objective weights on the evaluation results. First, calculate the expert score matrix Θ using the following formula (1):

[0090]

[0091] For the s-th evaluation criterion (s=1,…,l), θ js It is expert E j The score of the evaluation criteria for (j=1,…,m);

[0092] Then, the comprehensive score q of each expert is calculated using the entropy weight method according to the following formula (2). j :

[0093]

[0094] In equation (2) above, p js Let e ​​represent the proportion of the j-th expert in that evaluation indicator under the s-th evaluation criterion. s Let d represent the entropy value of the s-th evaluation indicator. s For information entropy redundancy, w ss q represents the weight of each evaluation indicator. j This represents the overall score of each expert;

[0095] Finally, the expert E is calculated using the following formula (3). j The weights W(E) j ):

[0096]

[0097] Specifically, the objective weighting scoring criteria for experts are shown in Table 2:

[0098] Table 2

[0099]

[0100] The objective weighted scoring criteria calculated based on expert information are shown in Table 3.

[0101] Table 3

[0102]

[0103] S222. Collect expert scoring results, convert the scoring results into intuitionistic triangular fuzzy numbers, consider the impact of the relative consistency of expert opinions on the evaluation results, and calculate the relative consistency of expert scores RA(E) using the following formula (4). j ):

[0104]

[0105] Among them, in the above formula (4) Indicated by expert E j The fuzzy scoring language is transformed into triangular intuitionistic fuzzy numbers. (a j ,b j ,c j ,a′ j ,b j ,c′ j ) represents the specific numerical value in the fuzzy number, WA(E) j ) represents the weighted consistency degree;

[0106] Among them, in the above formula (4) The similarity between two rating results is calculated by the following formula (5):

[0107]

[0108] Among them, in the above formula (5) Represents fuzzy numbers The expected assessment is calculated by the following formula (6):

[0109]

[0110] Specifically, the expert scoring criteria and their corresponding intuitionistic triangular fuzzy numbers are shown in Table 4:

[0111] Table 4

[0112]

[0113]

[0114] S223. Considering that the objective weight of experts and relative consistency are equally important to the evaluation results, the expert score weight W(E) is comprehensively considered. j ) and relative consistency RA(E) j ), take β=0.5, and calculate expert E using the following formula (7). j Consistency coefficient CC(E) j ( ), as an expert E in the weighted calculation of fuzzy opinion aggregation. j Weights:

[0115] CC(E j )=β·W(Ej )+(1-β)·RA(E j (7);

[0116] In equation (7) above, β (0 < β < 1) is a relaxation factor. This coefficient is used to indicate the relative importance of expert weights and relative consistency. Usually, β = 0.5 is taken in the calculation to indicate that the two are equally important.

[0117] S224. By aggregating the fuzzy numbers of scores from multiple experts through weighted averaging, T is introduced. w To reduce the accumulation of fuzziness, the operator calculates the result u after the aggregation of expert fuzzy opinions using equations (8) to (9). z (x);

[0118] Trigonometric Intuition Fuzzy Numbers The addition and multiplication formulas are defined by the following formula (8):

[0119]

[0120] In equation (8) above, CC(E) j ) is a constant. It is a triangular intuitive fuzzy number. λ is any constant.

[0121]

[0122] Fuzzy opinion aggregation result u z (x) is calculated by the following formula (9):

[0123]

[0124] In equation (9) above, u z (x) represents the aggregated result of expert fuzzy opinions at node x, u z (x)=(a x ,b x ,c x ;a x ′,b x ,c x ′);

[0125] S225. In order to obtain the failure probability, the transformation formula is used to calculate the fuzzy failure probability FFP of the root node of the Bayesian network model from the following formulas (10) to (12), and the fuzzy failure probability FFP is used as the prior probability of the root node of the Bayesian network model:

[0126] For the fuzzy opinion aggregation result u z (x), whose centroid FPS is defuzzified using the intuitive fuzzy method is shown in equation (10):

[0127]

[0128] The centroid FPS is converted to the fuzzy failure probability FFP using the following conversion formula (11):

[0129]

[0130] In equation (11), K is as shown in equation (12):

[0131]

[0132] Specifically, the failure probabilities of the root nodes of all Bayesian network models are shown in Table 5:

[0133] Table 5

[0134]

[0135]

[0136] S230. Using the fuzzy failure probability FFP as the prior probability of the root node of the Bayesian network model, the conditional probability table of the Bayesian network is calculated using the Noisy-OR gate method to obtain the Bayesian network parameter model; the conditional probability table of the Bayesian network is constructed using the following method:

[0137] S231, obtained by expert scoring, with only parent node B i Given a node that fails, the probability that child node A will fail; where it is assumed that all nodes are Boolean types, such as Figure 2 As shown, there are n reasons X1, X2, ..., X n This leads to an effect Y, which has n parameters p1, p2, ..., p obtained through expert scoring. n ;

[0138] S232. The conditional probability table of all child nodes is obtained by calculating the following formula (13):

[0139]

[0140] In equation (13) above, p i Represents node X i (1≤i≤n) is true, and the remaining nodes X j The probability that Y is true when (1≤j≤n and j≠i) is false; X={X1,…,X n};

[0141] Specifically, taking node M1 as an example, it has 3 parent nodes, and the conditional probability table of node M1 can be calculated as shown in Table 6:

[0142] Table 6

[0143]

[0144]

[0145] S300: Collect sensor data from the FPSO's internal turret single-point mooring system (specifically including fluid pressure, temperature, and leakage of the liquid slip rings; voltage, current, and temperature of the electric slip rings; bearing lubrication; and GPS data from the FPSO hull). Determine the state (normal and failed states) of the observable nodes in the Bayesian network fault diagnosis model of the FPSO's internal turret single-point mooring system. The Bayesian network fault diagnosis model consists of a Bayesian network structure model and a Bayesian network parameter model. Perform posterior inference to achieve fault diagnosis. Specifically, this includes the following steps:

[0146] S301. Collect the working status data of the FPSO internal turret single-point mooring system, and determine the status of the observation node based on the collected data;

[0147] S302. By performing posterior reasoning through the Bayesian network fault diagnosis model, the conditional probability of fault occurrence of fault mechanism nodes is obtained, the most likely fault mechanism is determined, and the location where the most likely fault occurs is found.

[0148] Based on the conditional independence assumption and the chain rule, the variables U = A1, A2, ..., A n The joint probability P(U) can be calculated according to equation (14):

[0149]

[0150] In equation (14) above, A i It is a node in a Bayesian network, Pa(A) i ) is A in a Bayesian network i The parent node;

[0151] The posterior probability of the root node of the Bayesian network model is obtained by performing backpropagation using the following equation (15):

[0152]

[0153] In equation (15) above, E = {e1,...,e} m} is a given set of evidence, determined by the collected data;

[0154] Specifically, node M1 is determined to be faulty, and the posterior probability of the root node of the Bayesian network model is calculated as shown in Table 7.

[0155] Table 7

[0156] X1 0.683 X8 4.373E-04 X2 0.29 X9 3.477E-05 X3 0.0274 X10 3.477E-05 X4 4.373E-04 X11 4.919E-05 X5 3.477E-05 X12 3.348E-05 X6 3.477E-05 X13 9.264E-04 X7 4.919E-05

[0157] Among all the root nodes of the Bayesian network model, the node with the highest failure probability is X1, with a failure probability of 68.3%.

[0158] As the posterior probability of a node increases, the probability of that node failing also increases. The failure identification rule is defined as follows:

[0159] ① When the probability of failure of a component is greater than 70%, the component fails;

[0160] ② When the probability of failure of a component is greater than 15% but less than 70%, the component may fail;

[0161] ③ When the failure probability of a component is less than 15%, the component is considered fault-free.

[0162] Based on the posterior reasoning results, it can be seen that nodes X1 and X2 may fail, with node X1 having a high probability of failure, while the other nodes are in normal condition; therefore, the liquid slip ring sealing system should be inspected and repaired in a timely manner.

[0163] This invention is well-conceived and can identify and assess risks in FPSO internal turret single-point mooring systems, and quickly and accurately diagnose faults, providing assurance and reference for the operation and maintenance of FPSO internal turret single-point mooring systems.

[0164] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A fault diagnosis method for a single-point mooring system of FPSO based on fault tree and fuzzy Bayesian network, characterized in that, The diagnostic method mainly includes the following steps: 1) Identify the failure modes of the FPSO internal turret single-point mooring system and construct a Bayesian network structure model; 2) Collect expert opinions and convert them into failure probabilities to construct a Bayesian network parameter model; specifically, this includes the following steps: 2.1) Develop an expert scoring sheet, set weights for different scorers, and collect expert information and scoring results; 2.2) Calculate the prior probability of the root node of the Bayesian network model; specifically including the following steps: 2.2.1) First, calculate the expert's score matrix using the following formula (1). : （1）； Among them, for the first One evaluation criterion ( ), Expert ( The score based on the evaluation criteria; Then, the comprehensive score of each expert is calculated using the entropy weight method according to the following formula (2). : (2)； In the above formula (2) For the first Under the first evaluation criterion The proportion of experts in this evaluation criterion, Indicates the first The entropy value of the evaluation criteria, For information entropy redundancy, This indicates the weight of each evaluation criterion. This represents the overall score of each expert; Finally, the expert calculated using the following formula (3) weight : （3）； 2.2.2) The relative consistency of expert ratings is calculated using the following formula (4). : (4)； In the above formula (4) Indicated by experts The fuzzy scoring language is transformed into triangular intuitionistic fuzzy numbers. , The specific numerical value in the fuzzy number; Indicates the weighted consistency degree; Among them, in the above formula (4) The similarity between two rating results is calculated by the following formula (5): (5)； Among them, in the above formula (5) Representing fuzzy numbers The expected assessment is calculated by the following formula (6): (6)； 2.2.3) Taking into account the weight of expert scores and relative consistency The expert was calculated using the following formula (7). Consistency coefficient When weighted calculation of expert fuzzy opinion aggregation, experts The weights; （7）； In the above formula (7), This is the relaxation factor, a coefficient used to indicate the relative importance of expert weights and relative consistency. It is calculated using... This indicates that expert weight and relative consistency are equally important; 2.2.4) Calculate the result of the aggregation of expert fuzzy opinions using the following formulas (8) to (9). ; Trigonometric Intuition Fuzzy Numbers The addition and multiplication formulas are defined by the following formula (8): （8）； In the above formula (8), It is a constant. It is a triangular intuitive fuzzy number. , λ is any constant. , , , , ; The result of fuzzy opinion aggregation Calculated by the following formula (9): （9）； In the above formula (9), This represents the aggregated result of expert fuzzy opinions for node x. ; 2.2.5) Calculate the fuzzy failure probability of the root node of the Bayesian network model using the following equations (10) to (12). and fuzzy failure probability Prior probabilities as the root node of a Bayesian network model: For fuzzy opinion aggregation results Its intuitive fuzzy method solves the fuzzy centroid As shown in equation (10): （10）； The centroid is converted using the following formula (11). Convert to fuzzy failure probability : （11）； Among them, in the above formula (11) As shown in equation (12): （12）； 2.3) The conditional probability table of the Bayesian network is calculated using the Noisy-OR gate method to obtain the parameter model of the Bayesian network; 3) Collect sensor data from the FPSO internal turret single-point mooring system, determine the state of observable nodes in the Bayesian network fault diagnosis model of the FPSO internal turret single-point mooring system, and the Bayesian network fault diagnosis model consists of a Bayesian network structure model and a Bayesian network parameter model. Then, rely on the Bayesian network fault diagnosis model to perform posterior inference to achieve fault diagnosis.

2. The fault diagnosis method for FPSO single-point mooring systems based on fault tree and fuzzy Bayesian network as described in claim 1, characterized in that, Step 1) specifically includes the following steps: 1.1) Combining the fault database, fault history data, relevant expert experience in the field of marine engineering, and the structure of the FPSO internal turret single-point mooring system, identify and sort out the typical fault modes and fault mechanisms of the FPSO internal turret single-point mooring system. 1.2) The overall failure of the FPSO internal turret single-point mooring system is taken as the top event of the fault tree, the failure modes of each component in the FPSO internal turret single-point mooring system are taken as the intermediate events of the fault tree, and the failure mechanism of the FPSO internal turret single-point mooring system is taken as the bottom event of the fault tree. Based on the causal logic relationship between each event node, the single-point system fault tree model is constructed from top to bottom. 1.3) Using the mapping principle, the fault tree is transformed into a Bayesian network, resulting in a Bayesian network structure model composed of fault mechanism nodes, component fault mode nodes, and observation nodes.

3. The fault diagnosis method for FPSO single-point mooring systems based on fault trees and fuzzy Bayesian networks as described in claim 1, characterized in that... The specific steps for calculating the conditional probability table of the Bayesian network in step 2.3) are as follows: 2.3.1) The probability of a child node failing when only the parent node fails is obtained through expert scoring; 2.3.2) The conditional probabilities of all child nodes are calculated using the following formula (13): （13）； In the above formula (13), Represents a node If true, the remaining nodes If it is false, The probability of it being true; .

4. The fault diagnosis method for FPSO single-point mooring systems based on fault tree and fuzzy Bayesian network as described in claim 1, characterized in that, Step 3) specifically includes the following steps: 3.1) Collect operational status data of the FPSO's internal turret single-point mooring system, and determine the status of the observation nodes based on the collected data; 3.2) By using the Bayesian network fault diagnosis model to perform posterior reasoning, the conditional probability of fault occurrence of fault mechanism nodes is obtained, the most likely fault mechanism is determined, and the location where the most likely fault occurs is found.

5. The fault diagnosis method for FPSO single-point mooring systems based on fault trees and fuzzy Bayesian networks as described in claim 4, characterized in that... The specific process of step 3.2) is as follows: 3.2.1) Based on the conditional independence assumption and the chain rule, variables joint probability It can be calculated according to formula (14): （14）； In the above formula (14), It is a node in a Bayesian network. In Bayesian networks The parent node; 3.2.2) The posterior probability of the root node of the Bayesian network model is obtained by performing backpropagation of the Bayesian network using the following equation (15): （15）； In the above formula (15), It is a given set of evidence, determined by the collected data; 3.2.3) Define the fault identification rules, specifically as follows: ① When the probability of failure of a component is greater than 70%, the component fails. ② When the probability of failure of a component is greater than 15% but less than 70%, the component may fail. ③ When the failure probability of a component is less than 15%, the component is considered fault-free.

Images