Early warning method and system for leakage risk of atmospheric and vacuum distillation unit
By building a fault tree for leakage events of normal pressure reducing device and converting it into Bayesian network, the problem of difficulty in accurately predicting leakage and positioning leakage sources in the prior art is solved, and comprehensive management of leakage risks of normal pressure reducing device is achieved and Quantitative analysis.
Patent Information
- Application Number
- CN202311027198.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2023-08-15
- Publication Date
- 2025-02-21
AI Technical Summary
The prior art is difficult to accurately predict the possibility of leakage from normal pressure reducing devices and locate the leakage source, and cannot effectively quantify the impact of the leakage.
By constructing a fault tree for leakage events of normal pressure reducing devices and converting them into Bayesian networks, Bayesian networks are used for predictive reasoning, diagnostic reasoning and sensitivity analysis, judging the possibility and risk level of leakage, locate the source of leakage and Quantify the effects of leakage.
It realizes comprehensive management of leakage risks of normal pressure reducing devices, can detect weak links in advance, accurately locate the leakage source and quantify the leakage impact, thereby reducing fires, explosions and poisoning accidents in dangerous chemical companies.
Smart Images
Figure CN119494526A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of dangerous chemical leakage risk warning, and in particular relates to a leakage risk warning method and system for a constant pressure reduction device. Background Art
[0002] Dangerous chemical leakage risk warning is an important research topic in the petrochemical industry. Domestic and foreign researchers have achieved certain research results and engineering practice experience in different sub-segments. The relevant research on real-time monitoring and risk root cause analysis is as follows:
[0003] (1) Monitoring of equipment defects
[0004] The data collection technology required for hidden danger investigation has matured and been commercialized. Tank wall and pipeline wall corrosion monitoring technology includes ultrasonic thickness detection, tank magnetic leakage detection, video monitoring and online monitoring systems. Although the means of monitoring equipment defects are becoming more and more mature, and the monitoring results can be used as an effective input for leakage prevention, the cost of excellent monitoring equipment remains high, making it difficult to fully install and use;
[0005] (2) Monitoring of oil and gas leaks
[0006] The research and equipment production of combustible gas detectors, toxic gas detectors, infrared imaging technology, and gas cloud imaging technology have gradually matured and stabilized. For example, spectral imaging cameras can identify and quantify gas releases in real time and can be used to detect leaks of harmful gases. The gas cloud imaging system uses proprietary hyperspectral imaging technology and advanced detection algorithms to assist in locating the source of the leak and measure the volume and concentration of the leaked gas. By monitoring the leaked oil and gas, it is possible to determine in which area the leak exists. However, it is impossible to determine which specific device or equipment has a leak, nor can accurate management and control decisions be made based on this;
[0007] (3) Risk root cause analysis
[0008] In the literature “Research on LNG Station Leakage Risk Assessment Model Based on Random Forest” and the literature “Airport Safety Risk Analysis Based on Bowtie Model”, decision trees (or fault trees) were used to analyze the causes of risk events; however, in terms of risk root cause classification, current research results have failed to adopt the SCL (safety checklist method), HAZOP (hazard and operability analysis), JHA (job hazard analysis) and other methods commonly used by chemical companies, and model them in the root cause analysis. Summary of the invention
[0009] In view of the above problems, the present invention takes the atmospheric and vacuum devices in the chemical industry as the research object, and proposes a leakage risk warning method and system for atmospheric and vacuum devices. The purpose of the present invention is to predict the possibility of leakage in atmospheric and vacuum devices based on observation data, process parameters, and management data, and to find weak links in advance; when the device leaks, diagnose the equipment in the device that may leak, and assist in locating the source of the leak; and quantify the impact of the leak. Combine leakage prediction, leakage consequences, and quantitative risk assessment to form a comprehensive risk management model, promote the application of expert experience in daily operations, reduce fires, explosions, and poisoning accidents in hazardous chemical companies from the source, reasonably allocate inspection and maintenance resources, and promote the balance between risk management and production and operation.
[0010] The technical solution adopted by the present invention is as follows: a method for early warning of leakage risk of a constant pressure reduction device, the method comprising the following steps: constructing a fault tree of leakage events of a constant pressure reduction device; converting the fault tree of leakage events of a constant pressure reduction device into a Bayesian network; performing predictive reasoning based on the Bayesian network to determine the possibility of leakage and the risk level of the constant pressure reduction device; and / or performing diagnostic reasoning based on the Bayesian network to locate the source of leakage; and / or performing sensitivity analysis based on the Bayesian network to study the response degree of leakage events of the constant pressure reduction device to changes in various parameters in the Bayesian network.
[0011] Furthermore, the construction of a fault tree for a leakage event of a constant pressure reduction device includes: according to a list of leakage risk points of a constant pressure reduction device, taking leakage of the constant pressure reduction device as a top event, leakage at each process node as an intermediate event, and equipment for storing or transporting oil and gas media as a basic event, a fault tree for leakage of the constant pressure reduction device is constructed.
[0012] Furthermore, the pruning strategy adopted by the fault tree includes: grouping all leakage events at the sealing points of the same equipment into leakage events of the equipment; and describing multiple equipment of the same type in the same process node through one leakage event.
[0013] Furthermore, the conversion of the fault tree of the leakage event of the constant pressure reduction device into the Bayesian network includes: corresponding the top event, intermediate event and basic event in the fault tree of the leakage event of the constant pressure reduction device to the leaf node, intermediate node and root node of the Bayesian network respectively, and converting to obtain the Bayesian network according to the equivalent conditional probability corresponding to the logic gate relationship.
[0014] Furthermore, the prior probability calculation process of the root node of the Bayesian network includes: according to the equipment category in the leakage risk point list of the constant pressure reduction device, comparing with the average failure probability of similar equipment, obtaining the frequency of small hole leakage, medium hole leakage, large hole leakage and rupture of the equipment of the constant pressure reduction device; and using the frequency of small hole leakage, medium hole leakage, large hole leakage and rupture of the equipment as the prior probability of each root node of the Bayesian network.
[0015] Furthermore, the equipment failure probability calculation formula is: F = F G *F E *F M *F L ; Where F represents the probability of equipment failure; F G Represents the average failure probability of similar equipment; F E Indicates the equipment correction factor; F M Indicates the management system evaluation coefficient; F L Indicates the influence coefficient of excessive defects; F E 、F L and F M These are all impact coefficients related to the application scenario.
[0016] Furthermore, the equipment correction factor F E The calculation formula is: Among them, x i represents the correction factor, x i =∑(Technology Module Factor i , general condition factor, mechanical factor i , process factor i ); The excess defect influence coefficient F L The calculation formula is: L =(T n / T SL )*100%, where T n Indicates the planned continuous service time of equipment or pipeline; T SL Indicates the remaining life; the management system evaluation coefficient F M The calculation formula is: M =10 1-X / 500 , X represents the management system evaluation score obtained by referring to GB26610.4-2014; the average failure probability of the same type of equipment F G It indicates that different types of equipment are assigned values with reference to GB26610.4-2014.
[0017] Furthermore, the joint distribution of random variables in the Bayesian network is:
[0018]
[0019] Where k represents the number of nodes; x 1 ,x 2 ,x 3 ,...,x k Respectively represent the 1st, 2nd, 3rd, ... kth nodes; P(X i |X pa(i) ) indicates that node X is i The conditional probability of
[0020] The Noisy-OR model is used to reduce the number of parameters required for the conditional probability table from 2 n To reduce the number of nodes to n, the Noisy-OR model needs to meet the following conditions:
[0021] Each variable is a binary variable, that is, each variable has two states, assumed to be true (1) and false (0);
[0022] Node Y's parent node X 1 ,X 2 ,…,X n Independent of each other;
[0023] Each variable is sufficient to cause node Y to occur when the other variables are false;
[0024] From the above conditions of the Noisy-OR model, we can see that the probability of Y not happening is equal to the probability of all X i The product of the probability of non-occurrence is:
[0025]
[0026] so:
[0027]
[0028] Among them, p i For the parent node X i =1 when the probability
[0029] Furthermore, the predictive reasoning based on the Bayesian network to judge the possibility of leakage and risk level of the atmospheric and vacuum distillation device includes: determining the value range of the Bayesian network node and establishing a conditional probability table;
[0030] According to the prior probability and conditional probability table of each root node of the Bayesian network, by observing the equipment status, inspection and maintenance records and environmental conditions, when the equipment failure frequency changes, the possibility of leakage and the risk level of the constant pressure reduction device are judged.
[0031] Furthermore, the diagnostic reasoning based on the Bayesian network to locate the leakage source includes: setting the state of the leakage node of the constant pressure reduction device to 100%, performing the reasoning operation of the Bayesian network according to the prior probability of each root node of the Bayesian network, and locating the root node with the highest leakage probability and its leakage type.
[0032] Furthermore, the sensitivity analysis is performed based on the Bayesian network to study the response of the leakage event of the atmospheric and vacuum device to the changes of each parameter in the Bayesian network, including:
[0033] According to the different values of the prior probability of each root node of the Bayesian network, the posterior probability is sorted; and the sensitivity of the change of the prior probability of the root node is determined according to the sorting result;
[0034] According to the sensitivity of the root node prior probability to change, the response of the leakage event of the atmospheric and vacuum device to the changes of various parameters in the Bayesian network is studied.
[0035] In addition, the present invention also proposes a leakage risk warning system for a constant pressure reduction device, which is used to implement the above-mentioned leakage risk warning method for a constant pressure reduction device. The system includes a Bayesian reasoning network module, an equipment failure probability calculation module, and a leakage real-time monitoring module; wherein,
[0036] The real-time leakage monitoring module is used to monitor the equipment parameters, process parameters and environmental parameters of the atmospheric and vacuum distillation device in real time, and dynamically adjust the prior probability of the root node of the Bayesian network according to the monitoring results;
[0037] The equipment failure probability calculation module is used to calculate the failure probability of a specific equipment based on the average failure probability of similar equipment, the equipment correction coefficient, the management system evaluation coefficient and the excess defect impact coefficient;
[0038] The Bayesian reasoning network module is used to determine the possibility and risk level of leakage in the constant pressure reduction device according to the probability of equipment failure.
[0039] Furthermore, the system also includes a task scheduling module, which is used to schedule the coordinated operation of the Bayesian reasoning network module, the equipment failure probability calculation module and the real-time leakage monitoring module.
[0040] The leakage risk warning system for the constant pressure reduction device proposed in the present invention refers to advanced experience, and covers the entire scope of risk management, including hazard source identification, risk level assessment, and consequence assessment, and comprehensively utilizes equipment archives, process parameters, state observation data, and management records. At the same time, based on the method of establishing an independent Bayesian network node for each device, and defining the state of the node as polymorphic, corresponding to a variety of leakage types, it is possible to assist in locating the leakage source accurately to the specific device and the specific leakage type when a leakage occurs. The present invention establishes a Bayesian network for engineering applications, and the prior probability relies less on expert experience and is more based on national standards and the operating data of the enterprise, avoiding the uncertainty of parameters and the randomness of network reasoning results caused by excessive reliance on manual experience, and solving the problem that the risk assessment process is difficult to quantify human factors, environmental factors, and management factors.
[0041] Other features and advantages of the present invention will be described in the following description, and part of them will become obvious from the description, or be understood by practicing the present invention. The purpose and other advantages of the present invention can be realized and obtained through the description and the accompanying drawings. BRIEF DESCRIPTION OF THE DRAWINGS
[0042] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0043] Figure 1 A schematic diagram of a leakage risk early warning method for a constant pressure reduction device;
[0044] Figure 2 This is a schematic diagram of the fault tree for leakage in a constant pressure reducing device;
[0045] Figure 3a and Figure 3b They are respectively the schematic diagrams of the leakage fault tree of the atmospheric and vacuum device and the AND gate and OR gate conversion rules of the Bayesian network;
[0046] Figure 4 A schematic diagram of the structure of converting the leakage fault tree of the atmospheric and vacuum distillation unit into a Bayesian network;
[0047] Figure 5 A schematic diagram of a Bayesian network composed of multiple random variables;
[0048] Figure 6 Schematic diagram of the Bayesian network using the Noisy-OR model;
[0049] Figure 7 This is a schematic diagram of the structure of the Bayesian network prediction reasoning for leakage risk warning of atmospheric and vacuum distillation equipment;
[0050] Figure 8 This is a schematic diagram of the structure of Bayesian network diagnostic reasoning for leakage risk warning of atmospheric and vacuum distillation equipment;
[0051] Fig. 9 The figure is a schematic diagram of the overall structure of a leakage risk early warning system for a constant pressure reduction device. DETAILED DESCRIPTION
[0052] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0053] The present invention proposes a leakage risk early warning method for a constant pressure reduction device, such as Figure 1 As shown, the method comprises the following steps:
[0054] Construct fault tree for leakage event of atmospheric and vacuum unit;
[0055] Convert the fault tree of atmospheric and vacuum distillation unit leakage event into Bayesian network;
[0056] Based on the Bayesian network, predictive reasoning is performed to determine the possibility and risk level of leakage in the atmospheric and vacuum distillation device;
[0057] and / or performing diagnostic reasoning based on a Bayesian network to locate the source of the leak;
[0058] And / or perform sensitivity analysis based on the Bayesian network to study the response of the leakage event of the atmospheric and vacuum device to the changes of each parameter in the Bayesian network.
[0059] Among them, before constructing the fault tree of the atmospheric and vacuum device leakage event, the research object is determined to be the atmospheric and vacuum device of a petrochemical enterprise; the research object is concretized as a set of typical devices, which is helpful to set specific design goals, adopt the current mature hazard source identification methods, collect specific relevant data, and verify the research results in a targeted manner. At the same time, information and data on atmospheric and vacuum devices are collected, mainly including: atmospheric and vacuum device process flow chart, atmospheric and vacuum device floor plan, atmospheric and vacuum device related equipment ledger (mainly towers, heat exchange equipment, pressure vessels, and pipelines for transporting toxic and hazardous media), atmospheric and vacuum device process operation flow, atmospheric and vacuum device safety operation procedures, atmospheric and vacuum device area weather conditions, and atmospheric and vacuum device related equipment failure probability data standards. Then, refer to the hazard source identification methods commonly used by chemical enterprises for atmospheric and vacuum devices to identify the risk points of atmospheric and vacuum device leakage. According to the atmospheric and vacuum device leakage risk point list, a fault tree of atmospheric and vacuum device leakage events is established, and the causes of leakage accidents and the consequences of leakage accidents are qualitatively analyzed.
[0060] Specifically, the list of leakage risk points of the atmospheric and vacuum distillation device is established with reference to the safety checklist method for equipment and facilities, as shown in Table 1:
[0061] Table 1 List of leakage risk points of atmospheric and vacuum distillation devices
[0062]
[0063]
[0064]
[0065]
[0066] Specifically, according to the list of leakage risk points of the atmospheric and vacuum device, the fault tree of the atmospheric and vacuum device leakage event is established, including: according to the list of leakage risk points of the atmospheric and vacuum device, such as Figure 2 As shown in the figure, the leakage of the atmospheric and vacuum device is taken as the top event, and the leakage of each process node is taken as the intermediate event (such as Figure 2 The intermediate events shown in are partial leakage of crude oil fuel, leakage of electric desalting system, leakage of atmospheric tower system, etc.), and the equipment for storing or transporting oil and gas media is taken as the basic event (the basic event is each equipment constituting the whole set of equipment, such as Figure 2 The basic events shown in are crude oil pump, flushing oil pump, crude oil first-stage electric desalting tank, atmospheric pressure tower, atmospheric first-line stripping tower, etc.). Construct a leakage fault tree for atmospheric and vacuum distillation devices. At the same time, in order to simplify the fault tree, the pruning strategies adopted include: (1) A typical atmospheric and vacuum distillation device has as many as tens of thousands of sealing points. Considering that the process parameters (temperature, pressure) have the same impact on the sealing points of a device, the leakage events of all sealing points belonging to the same device can be grouped as leakage events of the device; (2) For multiple devices of the same type in the same process node, they are described as one leakage event. For example Figure 2 As shown, taking "crude oil pump leakage" as an example, all the sealing points ( Figure 2 There are multiple sealing points in the fault tree, such as sealing point 1, sealing point 2, etc.), and other crude oil pumps belonging to the "crude oil feed part" are described by a leakage event in the fault tree. According to the above pruning strategy, the fault tree of "normal and vacuum device leakage" is obtained as follows Figure 2 shown.
[0067] Then, according to Figure 3a-3b The conversion rules between the leakage fault tree of the atmospheric and vacuum device and the Bayesian network are shown in Fig. Figure 3a In is the AND gate conversion rule, Figure 3a P(T=1|X 1 =1,X 2 =1)=1, P(T=1|others)=0; Figure 3b is the OR gate conversion rule, Figure 3b P(T=1|X 1 =0,X 2 =0)=0, P(T=1|others)=1; the top event (T), intermediate event and basic event (X 1, X 2 ) correspond to the leaf nodes, intermediate nodes and root nodes of the Bayesian network respectively, and the logic gate relationship corresponds to the equivalent conditional probability. The Bayesian network obtained by converting the leakage fault tree of the constant pressure reduction device according to the above rules has the overall structure as follows Figure 4 shown. Figure 4 The leaf node "leakage of atmospheric and vacuum unit" in the middle Bayesian network corresponds to the top event in the fault tree, which represents the leakage of the entire atmospheric and vacuum unit; the intermediate events represent the leakage of each process node of the atmospheric and vacuum unit, such as "crude oil feed part", "leakage of electric desalination system", etc.; the root nodes represent the failure of each device. For example, the root node "atmospheric pressure tower" represents the failure of the atmospheric pressure tower. The root node in the Bayesian network, that is, the failure probability of each device, uses the following process to set the prior probability:
[0068] a) Referring to relevant national standards, the aperture sizes of leakage points are divided into four types as shown in Table 2:
[0069] Table 2 Classification of aperture size of leak points
[0070] Aperture Type Small hole Middle hole Large hole rupture Representative size 6mm 25mm 100mm 400mm
[0071] b) Referring to relevant national standards, the average failure probability of similar equipment is shown in Table 3 (DN 20mm pipe in Table 3 means a pipe with a nominal diameter of 20mm):
[0072] Table 3 Data on average failure probability of similar equipment
[0073]
[0074] c) Based on the leakage risk list of the atmospheric and vacuum device obtained above, according to the equipment category, and in comparison with the average failure probability of similar equipment defined in the national standard (as shown in Table 3), the frequency of oil and gas leakage of various apertures in the atmospheric and vacuum device equipment is obtained, which is used as the prior probability of each root node of the Bayesian network, as shown in Table 4:
[0075] Table 4 Prior probability related data of each root node of the Bayesian network
[0076]
[0077]
[0078]
[0079]
[0080]
[0081] The calculation formula of the equipment failure probability F in the Bayesian network is as follows: F = FG *F E *F M *F L , where: F represents the probability of equipment failure; F E Indicates the equipment correction factor; F G Represents the average failure probability of similar equipment; F L Indicates the influence coefficient of excessive defects; F M represents the management system evaluation coefficient; where F E 、F L 、F M These are all influence coefficients related to time changes.
[0082] The equipment correction factor F E The calculation formula is: Among them, x i represents the correction factor, x i =∑(Technology Module Factor i , general condition factor, mechanical factor i , process factor i );
[0083] The excess defect influence coefficient F L The calculation formula is: L =(T n / T SL )*100%, where T n Indicates the planned continuous service time of equipment or pipeline, in years; T SL Remaining life, in years. This invention does not involve the study of the degradation mechanism of equipment related to time. L =1.0;
[0084] The management system evaluation coefficient F M The calculation formula is: M =10 1-X / 500 , X represents the management system evaluation score (refer to GB26610.4-2014);
[0085] The average failure probability F of the same type of equipment G , refer to GB26610.4-2014 to assign values to different types of equipment.
[0086] It can be seen that these changing factors affect the equipment failure probability of the Bayesian network, thereby affecting the inference results.
[0087] like Figure 5 The figure shows a Bayesian network composed of multiple random variables. The joint distribution of multiple random variables in the Bayesian network is:
[0088]
[0089] Where: k is the number of nodes; x 1 ,x 2 ,x 3 ,...,x k Respectively represent the 1st, 2nd, 3rd, ... kth nodes; P(X i |X pa(i) ) indicates that node X is i conditional probability. However, it is unrealistic to find a complete risk case library and obtain data from it for enterprise safety production risks. The Noisy-OR model is a method to solve this type of problem. The Noisy-OR model can be used to describe the intrinsic relationship between n variables and the impact they produce. The Noisy-OR model has three conditions: (1) Each variable is a binary variable, that is, each variable has two states, assuming they are true (1) and false (0); (2) The parent node X of node Y 1 , X 2 , …, X n are independent of each other, such as Figure 6 As shown; (3) Each variable is sufficient to cause the occurrence of result Y when other variables are false. From the conditions satisfied by Noisy-OR, we can see that:
[0090]
[0091] Where: p i Yes X i =1 when the probability
[0092] It can be seen that the Noisy-OR model can reduce the number of parameters required for the conditional probability table from 2 to n The number of conditional probability tables is reduced from 1 to n, which greatly reduces the amount of calculation required for the model. Considering that the parameter settings of the conditional probability table mostly involve expert experience in manual adjustment, this means that the manual cost of parameter adjustment is also greatly reduced, while also improving the accuracy of model derivation calculations. The GeNIe3.0 software designed and developed by the University of Pittsburgh provides support for the Noisy-OR model and supports the Noisy-MAX model that can define more than two states.
[0093] According to the different settings of evidence variables and query variables, Bayesian network reasoning mainly includes four types: predictive reasoning (reasoning from cause to effect); diagnostic reasoning (reasoning from effect to cause); associative reasoning (reasoning between different causes under the same result); and mixed reasoning containing multiple forms of the above. Typical application scenarios of the results of the present invention are predictive reasoning and diagnostic reasoning. The engineering significance of this is: by observing the equipment status, inspection and maintenance records, and environmental conditions, when the frequency of equipment failure changes, the possibility of leakage in the device is predicted; or, when a hazardous chemical leakage is observed in the area where the device is located, reasoning is used to determine the area of the device where the leakage may occur, which type of equipment, and the specific equipment position number.
[0094] Specifically, in the embodiment of the present invention, predictive reasoning, diagnostic reasoning or sensitivity analysis is performed based on the Bayesian network to determine the specific device where the leakage occurs.
[0095] (1) Predictive Reasoning
[0096] Predictive reasoning analysis refers to the impact of basic events (causes) on the probability of occurrence of the final result (consequence) of the event. In the leakage risk model of the constant pressure reduction device, the basic event is each device that constitutes the entire device. In order to simplify the research problem and remove excessive details in order to reduce the computing resources occupied by the model in the experimental environment, similar equipment of the constant pressure reduction device is merged into a node of a Bayesian network. At the same time, equipment whose storage medium is not hazardous chemicals is temporarily not included in the model. The GeNIe 4.0 software of the University of Pittsburgh is used for network modeling. Based on the prior probability and conditional probability table of the basic events previously determined, the Bayesian network is predicted and reasoned to obtain the possibility change of the leakage risk event of the constant pressure reduction device. Figure 7 As shown: For example Figure 7 In the Bayesian network, the risk probability of leakage of the leaf node "leakage of constant pressure reduction device" is 86%; the risk probabilities of leakage of the intermediate nodes "cold and hot oil pump room area", "normal top pump area" and "reduction one and reduction two pump area" are 96%, 35% and 31% respectively; the risk probabilities of small hole leakage of the root nodes "normal top circulating oil pump P_1004B", "normal top circulating oil pump P_1004A", "normal top circulating oil pump P_1004C", "isolation liquid flushing pump P_1045" and "normal pressure residue oil pump P_1101B" are 12%, 6%, 6%, 6% and 6% respectively.
[0097] (2) Diagnostic Reasoning
[0098] When a leak of hazardous chemicals is observed in the area where the atmospheric and vacuum device is located, the state of the "atmospheric and vacuum device leakage" node in the leakage risk warning model is set to 100% (that is, the observed event has occurred and the probability is 100%). After executing the inference operation of the Bayesian network, Figure 8 As shown, it can be seen that "P_1004A" (leakage risk probability is 12%) and "Constant top circulating oil pump P_1004B" (leakage risk probability is 12%) have the highest probability of small hole leakage. The reasoning principle is that, referring to the prior probability provided by the national standard (as shown in Table 4), it can be seen that the "failure probability of similar equipment" for small hole leakage in pump equipment is the highest, and the two equipments have adjusted the prior probability according to the actual working conditions, and the equipment failure probability is higher than the average failure probability of similar equipment. Therefore, the possibility of leakage is the greatest in the entire device. In actual working conditions, each pump equipment is in different service life, inspection and maintenance status, process parameter fluctuations and other conditions, and the actual failure probability will be relatively different.
[0099] (3) Sensitivity analysis
[0100] For the leakage risk warning Bayesian network of the atmospheric and vacuum distillation device, the prior probability and posterior probability of each node are compared as shown in Table 5:
[0101] Table 5 Data related to the prior probability and posterior probability of the nodes of the Bayesian network for leakage risk warning of the atmospheric and vacuum distillation device
[0102]
[0103]
[0104]
[0105]
[0106]
[0107] Combined with the analysis results in Table 5, it can be seen that the prior probability and posterior probability of equipment such as the crude oil pump, primary bottom oil pump, primary top air cooler in the crude oil feed part, the atmospheric top circulating oil pump, and atmospheric top oil-gas air cooler in the atmospheric tower system are all ranked at the top, and are one of the main reasons for the leakage of the atmospheric pressure reduction device.
[0108] In addition, the present invention also proposes a leakage risk early warning system for a constant pressure reduction device such as Fig. 9 As shown in the figure, the architecture modules of the system are divided into the following parts: Bayesian reasoning network module (G), equipment failure probability calculation module (F E 、F G 、F L 、F M ), a real-time leakage monitoring module (M) and a task scheduling module (T) for coordinating the operation of various modules, as well as initialization data and other external input data. The real-time leakage monitoring module is used to monitor the initialization data of the constant pressure reduction device in real time and set the evidence variables, for example, Fig. 9As shown in FIG. 1 , the real-time leakage monitoring module is used to monitor the concentration of combustible and toxic gases, infrared spectrum, video monitoring, and gas cloud imaging, etc., and then the real-time monitored data is processed and set as observation evidence for use by the Bayesian reasoning network module; the equipment failure probability calculation module is used to calculate the average failure probability F of similar equipment. G , Equipment correction factor F E , Management system evaluation coefficient F M And the influence coefficient of the defect exceeding the standard F L , calculate the failure probability of similar equipment; equipment correction factor F E It is obtained by summing up the technical module factor, general condition factor, mechanical factor and process factor. Among them, the technical module factor is related to mechanical fatigue damage, high temperature hydrogen corrosion, furnace tube damage, brittle fracture, etc.; the general condition factor is related to factory conditions, seismic activities, etc.; the mechanical factor is related to safety factor, life cycle, etc.; the process factor is related to process continuity, stability, etc.; the over-standard defect influence coefficient F L Related to whether there is a time-related degradation mechanism in the equipment; management system evaluation coefficient F M Related to operating procedures, personnel training, accident investigation, production safety management, etc. The Bayesian reasoning network module is used to judge the possibility of leakage and risk level of the atmospheric and vacuum distillation device based on the evidence variables and the failure probability of similar equipment. The task scheduling module is used to schedule the coordinated operation of the Bayesian reasoning network module, the equipment failure probability calculation module and the real-time leakage monitoring module. The output data of the system is the leakage possibility prediction result and the leakage risk level.
[0109] Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent substitutions for some of the technical features therein; and these modifications or substitutions do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A leakage risk warning method for a constant pressure reduction device, characterized in that: The method comprises the following steps: Construct fault tree for leakage event of atmospheric and vacuum unit; Convert the fault tree of atmospheric and vacuum distillation unit leakage event into Bayesian network; Based on the Bayesian network, predictive reasoning is performed to determine the possibility and risk level of leakage in the atmospheric and vacuum distillation device; and / or performing diagnostic reasoning based on a Bayesian network to locate the source of the leak; And / or perform sensitivity analysis based on the Bayesian network to study the response of the leakage event of the atmospheric and vacuum device to the changes of each parameter in the Bayesian network.
2. A leakage risk early warning method for a normal and low pressure device according to claim 1, characterized in that: The construction of the fault tree of the leakage event of the atmospheric and vacuum distillation device comprises: According to the list of leakage risk points of the atmospheric and vacuum device, a leakage fault tree of the atmospheric and vacuum device is constructed, with the leakage of the atmospheric and vacuum device as the top event, the leakage of each process node as the intermediate event, and the equipment for storing or transporting oil and gas media as the basic event.
3. A leakage risk early warning method for a constant pressure reduction device according to claim 2, characterized in that: The pruning strategies adopted by the fault tree include: All leakage events at the sealing points of the same equipment are grouped as leakage events of the equipment; Multiple devices of the same type in the same process node are described by one leakage event.
4. A leakage risk early warning method for a normal and low pressure device according to claim 2 or 3, characterized in that: The method of converting the fault tree of the atmospheric and vacuum distillation device leakage event into a Bayesian network comprises: The top event, intermediate event and basic event in the fault tree of the leakage event of the constant pressure reduction device correspond to the leaf node, intermediate node and root node of the Bayesian network respectively, and the equivalent conditional probability corresponding to the logic gate relationship is converted into a Bayesian network.
5. A leakage risk early warning method for a normal and low pressure device according to claim 4, characterized in that: The prior probability calculation process of the root node of the Bayesian network includes: According to the equipment categories in the leakage risk point list of the constant and distillation device and compared with the average failure probability of similar equipment, the frequencies of small hole leakage, medium hole leakage, large hole leakage and rupture of the equipment of the constant and distillation device are obtained; and the frequencies of small hole leakage, medium hole leakage, large hole leakage and rupture of the equipment are used as the prior probability of each root node of the Bayesian network.
6. A leakage risk early warning method for a constant pressure reduction device according to claim 5, characterized in that: The equipment failure probability calculation formula is: F = F G *F E *F M *F L ; Where F represents the probability of equipment failure; F G Represents the average failure probability of similar equipment; F E Indicates the equipment correction factor; F M Represents the management system evaluation coefficient; F L Indicates the influence coefficient of excessive defects; F E 、F L and F M These are all impact coefficients related to the application scenario.
7. A leakage risk early warning method for a normal and low pressure device according to claim 6, characterized in that: The equipment correction factor F E The calculation formula is: Among them, x i represents the correction factor, x i =∑(Technology Module Factor i , general condition factor, mechanical factor i , process factor i ); The excess defect influence coefficient F L The calculation formula is: L =(T n / T SL )*100%, where T n Indicates the planned continuous service time of equipment or pipeline; T SL Indicates the remaining life; The management system evaluation coefficient F M The calculation formula is: M =10 1-X / 500 , X represents the management system evaluation score obtained by referring to GB26610.4-2014; The average failure probability F of similar equipment G It indicates that different types of equipment are assigned values with reference to GB26610.4-2014.
8. A leakage risk early warning method for a normal and low pressure distillation device according to claim 7, characterized in that: The joint distribution of random variables in the Bayesian network is: Where k represents the number of nodes; x1,x2,x3,...,x k Respectively represent the 1st, 2nd, 3rd, ... kth nodes; P(X i |X pa(i) ) indicates that node X is i The conditional probability of The Noisy-OR model is used to reduce the number of parameters required for the conditional probability table from 2 n To reduce the number of nodes to n, the Noisy-OR model needs to meet the following conditions: Each variable is a binary variable, that is, each variable has two states, assumed to be true (1) and false (0); The parent nodes X1, X2, ..., X of node Y n Independent of each other; Each variable is sufficient to cause node Y to occur when the other variables are false; From the above conditions of the Noisy-OR model, we can see that the probability of Y not happening is equal to the probability of all X i The product of the probability of non-occurrence is: so: Among them, p i For the parent node X i =1 when the probability 9. A leakage risk early warning method for a normal and low pressure device according to claim 1, characterized in that: The prediction and reasoning based on the Bayesian network to judge the possibility and risk level of leakage of the atmospheric and vacuum distillation device includes: Determine the value range of the Bayesian network nodes and establish a conditional probability table; According to the prior probability and conditional probability table of each root node of the Bayesian network, by observing the equipment status, inspection and maintenance records and environmental conditions, when the equipment failure frequency changes, the possibility of leakage and the risk level of the constant pressure reduction device are judged.
10. A leakage risk early warning method for a normal and low pressure device according to claim 1, characterized in that: The diagnostic reasoning based on the Bayesian network to locate the leakage source includes: The state of the leakage node of the atmospheric and vacuum distillation device is set to 100%. According to the prior probability of each root node of the Bayesian network, the reasoning operation of the Bayesian network is performed to locate the root node with the highest leakage probability and its leakage type.
11. A leakage risk early warning method for a normal and low pressure distillation device according to claim 1, characterized in that: The sensitivity analysis based on the Bayesian network is used to study the response of the leakage event of the atmospheric and vacuum device to the changes of various parameters in the Bayesian network, including: According to the different values of the prior probability of each root node of the Bayesian network, the posterior probability is sorted; and the sensitivity of the change of the prior probability of the root node is determined according to the sorting result; According to the sensitivity of the root node prior probability to change, the response of the leakage event of the atmospheric and vacuum device to the changes of various parameters in the Bayesian network is studied.
12. A leakage risk warning system for a constant pressure reduction device, the system being used to implement a leakage risk warning method for a constant pressure reduction device according to any one of claims 1 to 11, characterized in that: The system includes a Bayesian reasoning network module, an equipment failure probability calculation module and a leakage real-time monitoring module; wherein, The real-time leakage monitoring module is used to monitor the equipment parameters, process parameters and environmental parameters of the atmospheric and vacuum distillation device in real time, and dynamically adjust the prior probability of the root node of the Bayesian network according to the monitoring results; The equipment failure probability calculation module is used to calculate the failure probability of a specific equipment based on the average failure probability of similar equipment, the equipment correction factor, the management system evaluation factor, and the excess defect impact factor; The Bayesian reasoning network module is used to determine the possibility and risk level of leakage in the constant pressure reduction device according to the probability of equipment failure.
13. A leakage risk warning system for a normal and distillation device according to claim 12, characterized in that: The system also includes a task scheduling module, which is used to schedule the coordinated operation of the Bayesian reasoning network module, the equipment failure probability calculation module and the leakage real-time monitoring module.