CN111275195A - Dynamic Bayesian network modeling method based on coal gasification equipment - Google Patents

Dynamic Bayesian network modeling method based on coal gasification equipment Download PDF

Info

Publication number
CN111275195A
CN111275195A CN202010091388.6A CN202010091388A CN111275195A CN 111275195 A CN111275195 A CN 111275195A CN 202010091388 A CN202010091388 A CN 202010091388A CN 111275195 A CN111275195 A CN 111275195A
Authority
CN
China
Prior art keywords
dynamic
bayesian network
node
coal gasification
gate
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202010091388.6A
Other languages
Chinese (zh)
Inventor
杨雪峰
王承彬
孙铁
张素香
多依丽
王志荣
郭品坤
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Liaoning Shihua University
Original Assignee
Liaoning Shihua University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Liaoning Shihua University filed Critical Liaoning Shihua University
Priority to CN202010091388.6A priority Critical patent/CN111275195A/en
Publication of CN111275195A publication Critical patent/CN111275195A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N7/00—Computing arrangements based on specific mathematical models
    • G06N7/02—Computing arrangements based on specific mathematical models using fuzzy logic
    • G06N7/06—Simulation on general purpose computers
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N7/00—Computing arrangements based on specific mathematical models
    • G06N7/01—Probabilistic graphical models, e.g. probabilistic networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063—Operations research, analysis or management
    • G06Q10/0635—Risk analysis of enterprise or organisation activities
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Electricity, gas or water supply

Abstract

The invention discloses a dynamic Bayesian network modeling method based on coal gasification equipment, which is used for carrying out systematic investigation and analysis on coal gasification key equipment risk influence factors. And then accurately identifying the process deviation hazard factors based on the analysis result, qualitatively analyzing possible reasons and consequences generated by the deviation, and constructing a static/dynamic fault tree to describe the time sequence and logic relation of the accident. And converting the fault tree into a dynamic Bayesian network model according to the conversion rule of the static/dynamic logic gate into the dynamic Bayesian network. And finally, importing historical information of the leading event and determining the prior probability of the root node by applying a fuzzy set theory, deducing the posterior accident occurrence probability of the root node by means of the strong reverse reasoning function of the Bayesian network, and sequencing the posterior accident occurrence probability so as to determine the weak link of the system.

Description

Dynamic Bayesian network modeling method based on coal gasification equipment
Technical Field
The invention relates to the technical field of safety risk assessment, in particular to a dynamic Bayesian network modeling method based on coal gasification equipment.
Background
Coal energy is an important energy resource in daily life of people, however, coal gasification is a new industry, and the development of a safe, green and environment-friendly coal gasification industry can effectively supplement the shortage of oil and gas resources in China. However, most coal gasification production processes are harsh in process conditions, complex in production devices, diversified in intermediate reaction, and many unsafe factors such as high temperature and high pressure, and most raw and auxiliary materials and products have dangerous characteristics such as explosiveness, toxicity and corrosiveness. In view of the particularity of the production process of coal gasification enterprises, accidents not only cause property loss of facilities and equipment, but also can cause casualties and even environmental pollution. The existing research mainly aims at carrying out static risk analysis on a specific production device of a coal gasification enterprise, and time factors are rarely considered when the static risk analysis is carried out on the specific production device. Therefore, how to carry out dynamic risk assessment on the operation process of coal gasification process devices and equipment, reveal the inoculation of accidents, the generation dynamics and the disaster causing mechanism, scientifically avoid risks is an important problem which is mainly concerned and urgently needed to be solved in the modern coal gasification industry, and no dynamic analysis model exists in the dynamic analysis process.
Disclosure of Invention
In view of the above drawbacks and deficiencies, the present invention provides a dynamic bayesian network modeling method for a coal gasification facility.
In order to achieve the above purpose, the technical scheme of the invention is as follows:
a coal gasification equipment-based dynamic Bayesian network modeling method comprises the following steps:
analyzing coal gasification equipment, acquiring a plurality of input events of the coal gasification equipment, and establishing a static/dynamic fault tree model of each coal gasification equipment, wherein the static/dynamic fault tree model comprises a root node, an intermediate node and a leaf node;
converting the static/dynamic fault tree model into a dynamic Bayesian network model, and expressing a logical relationship between equipment failures through a topological structure of a Bayesian network;
calculating the prior probability and the root node posterior probability of each node, and inputting the leading event information serving as evidence information into the dynamic Bayesian network model according to the records of field actual production and experimental research data.
Analyzing the coal gasification equipment by a brain storm method or HAZOP analysis to obtain a plurality of input events of the coal gasification equipment, and establishing a static/dynamic fault tree model of each coal gasification equipment.
The analysis of the coal gasification equipment by the HAZOP analysis comprises the following steps: familiar with the process flow, selecting an evaluation node, performing node description according to the selected evaluation node, then determining deviation, performing HAZOP analysis, obtaining an event with high possibility and serious consequence, and obtaining a fault tree model analysis conclusion.
The HAZOP analysis comprises deviation values, event generation reasons, event caused consequences, existing protective measures, risk level evaluation and suggested measures.
The conversion rule for converting the static/dynamic fault tree model into the dynamic Bayesian network model comprises:
converting static logic gates in the static/dynamic fault tree model, wherein the static logic gates comprise an AND gate, an OR gate and a NOT gate;
and converting the dynamic logic gates in the static/dynamic fault tree model, wherein the dynamic logic gates comprise a priority gate, a sequential phase gate, a spare part gate and a function phase gate.
The specific steps for converting the static logic gate in the static/dynamic fault tree model are as follows:
assuming that E-0 means that event E does not occur, E-1 means that event E occurs, fE(t) probability density of event E occurrence timeA degree function;
and the logical relation of the AND gates is corresponding to a dynamic Bayesian network, wherein the conditional probability distribution of the nodes is as follows:
Figure BDA0002383840020000021
P{A(T+ΔT)＝1|A(T)＝1}＝1 (2)
Figure BDA0002383840020000022
P{B(T+ΔT)＝1|A(T)＝1}＝1 (4)
P{TE＝1|A(T+ΔT)＝1,B(T+ΔT)＝1}＝1 (5)
P{TE＝1|else}＝0 (6)
and logically corresponding an OR gate to a dynamic Bayesian network, wherein the conditional probability distribution of a node A (T + delta T) is the same as that of the formula (1) and the formula (2), and the conditional probability distribution of a node TE is as follows:
P{TE＝1|A(T+ΔT)＝1}＝1 (7)
P{TE＝1|B(T+ΔT)＝1}＝1 (8)
P{TE＝1|A(T+ΔT)＝0,B(T+ΔT)＝0}＝0 (9)
and corresponding the NOT gate logic to a dynamic Bayesian network, wherein the conditional probability distribution of the node A (T + delta T) is the same as the conditional probability distribution of the node A (1) and the node TE, and the conditional probability distribution of the node TE is as follows:
P{TE＝1|A(T+ΔT)＝1}＝0 (10)
P{TE＝1|A(T+ΔT)＝0}＝1 (11)
where Δ T is the time interval.
The specific steps of the conversion of the dynamic logic gate in the static/dynamic fault tree model are as follows:
adding two-state nodes FS1 and FS2 according to the sequential logical relationship of priority and gates, where FS 1-1 means that a occurs before B, FS 2-1 means that B occurs before C, FS 1-0 means that a does not occur before B, and FS 2-0 means that B does not occur before C;
the conditional probability distribution of each node of the priority AND gate is as follows:
Figure BDA0002383840020000031
analyzing the functional associated gate of the basic event according to the sequential logical relationship of the sequential associated gates to obtain a corresponding dynamic Bayesian network, wherein the conditional probability distribution of each node is as follows:
Figure BDA0002383840020000032
analyzing the warm-up spare part gates of the main part and the backup part according to the time sequence logic relationship of the warm-up spare part gates to obtain a dynamic Bayesian network corresponding to the warm-up spare part gates, wherein the conditional probability distribution of each node is as follows:
Figure BDA0002383840020000041
in the formula fαS(t) is a function of the failure density of the component S during the backup period;
according to the relationship between the trigger event and the related basic event, analyzing the functional correlation gate including one trigger event and two related basic events in fig. 10, and obtaining a dynamic bayesian network corresponding to the functional correlation gate, wherein the conditional distribution probability of each node is as follows:
Figure BDA0002383840020000042
the classification of coal gasification plants into key, main and general plants comprises: the main production equipment is divided into key equipment, main equipment and general equipment according to the degree of influence and loss on the aspects of production, quality, cost, safety and delivery date after the main production equipment is out of order and when the equipment is stopped for repair.
The prior probability and the root node posterior probability of each node are calculated as follows:
calculating the prior probability of each node by using a fuzzy mathematical method or historical data; and calculating the posterior probability of the root node by using a Bayesian network reverse reasoning function.
The leading events include alarms, non-recovery events and accidents.
Compared with the prior art, the invention has the beneficial effects that:
the invention provides a dynamic Bayesian network modeling method based on coal gasification equipment, which is used for carrying out systematic investigation and analysis on coal gasification key equipment risk influence factors. And then accurately identifying the process deviation hazard factors based on the analysis result, qualitatively analyzing possible reasons and consequences generated by the deviation, and constructing a static/dynamic fault tree to describe the time sequence and logic relation of the accident. And converting the fault tree into a dynamic Bayesian network model according to the conversion rule of the static/dynamic logic gate into the dynamic Bayesian network. And finally, importing historical information of the leading event and determining the prior probability of the root node by applying a fuzzy set theory, deducing the posterior accident occurrence probability of the root node by means of the strong reverse reasoning function of the Bayesian network and sequencing the posterior accident occurrence probability so as to determine the weak link of the system, wherein the calculation result shows that the equipment failure probability and the accident risk show a remarkable growth trend along with the growth of the production time of the coal gasification device and the leading event. The research of the dynamic risk evaluation method solves the difficult problems of the system in describing polymorphism, failure correlation, dynamics, process variables and the like, and provides a theoretical basis for the safety production risk analysis and safety production accident prevention of coal gasification enterprises.
Drawings
FIG. 1 is a flow chart of the dynamic risk assessment of the present invention;
FIG. 2 is a comprehensive idea diagram of the HAZOP-FTA method of the present invention;
FIG. 3 is a diagram of a preferred AND gate and corresponding dynamic Bayesian network of the present invention;
FIG. 4 is a diagram of a dynamic Bayesian network with sequential gates and corresponding gates of the present invention;
FIG. 5 is a diagram of a warm spare part door and corresponding dynamic Bayesian network of the present invention;
FIG. 6 is a diagram of the functionally related gates and corresponding dynamic Bayesian networks of the present invention;
FIG. 7 is a FTA model diagram of the pressurized pulverized coal delivery system of the present invention;
FIG. 8 is a DBN model diagram of the pressurized pulverized coal delivery system of the present invention;
FIG. 9 is a graph of system failure rate change without consideration of maintenance factors in accordance with the present invention;
FIG. 10 is a graph of the failure rate of the system in consideration of maintenance factors according to the present invention.
Detailed Description
The present invention will now be described in detail with reference to the drawings, wherein the described embodiments are only some, but not all embodiments of the invention. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, belong to the scope of the present invention.
Example 1
As shown in fig. 1, the invention provides a coal gasification equipment-based dynamic bayesian network modeling method, which comprises the following steps:
s1, analyzing the coal gasification equipment, acquiring a plurality of input events of the coal gasification equipment, and establishing a static/dynamic fault tree model of each coal gasification equipment, wherein the static/dynamic fault tree model comprises a root node, a middle node and a leaf node;
specifically, the coal gasification equipment can be divided into key equipment, main equipment and general equipment, and main dangers and harmful factors existing in the operation process of each equipment in the coal gasification process and the mutual influence relationship among the factors are analyzed systematically;
coal gasification enterprises need a lot of complex equipment, each device has different height, size, dynamic and static combination, complex process pipelines and harsh process control conditions. Specifically, the main production equipment is divided into key equipment, main equipment and general equipment according to the degree of influence on the aspects of production, quality, cost, safety, delivery date and the like and the loss caused by the main production equipment after failure and during shutdown and repair.
Preferably, a static/dynamic fault tree model of each key device and main device is established by using a brainstorm method or a HAZOP analysis result, and Hazard and Operability analysis (Hazard and Operability Study, abbreviated as HAZOP) is a system safety analysis method which performs comprehensive system identification and analysis on a Hazard source generated by a deviation occurring in a design stage or an existing production process in a comprehensive examination manner to find out potential hazards caused by misoperation of a device facility or a person. The HAZOP analysis can accurately identify the process of a process hazard source, and qualitatively analyzes the danger and operability problems caused by the deviation of the production process. HAZOP and FTA (Fault Tree Analysis) serve as two different safety Analysis methods, and the Analysis processes are obviously different and are strongly related. The HAZOP method is based on the thought of deviation-cause-result, carries out comprehensive qualitative analysis of 'surface' on the harm factors of the production process, and the FTA defines the causal logical relationship of 'point' control of each device by using a quantitative result. The point and the surface are combined, and the quality and the quantity are complementary, so that the analysis of the hazard source is more systematic, complete and objective.
As shown in fig. 2, the analysis of the coal gasification facility by the HAZOP analysis includes: familiar with the process flow, selecting an evaluation node, performing node description according to the selected evaluation node, then determining deviation, performing HAZOP analysis, obtaining an event with high possibility and serious consequence, and obtaining a fault tree model analysis conclusion.
S2, converting the static/dynamic fault tree model into a dynamic Bayesian network model, and expressing the logical relationship between equipment failures through the topological structure of the Bayesian network;
dynamic risk is a classification of risk in the dimension of changing state, and in contrast to static risk, it is emphasized that the presence state of risk changes with time and space, which is an extension of risk definition in considering time-varying features. The fault tree can visually describe the influence of various combinations of the bottom-level events on the occurrence of the top event, and the system can make reverse inference from the top event so as to find out various possible reasons for the occurrence of the top event. The logic symbol describing the causal relationship between the events is called a logic gate, the input event of the logic gate is the 'cause' of the output event, whereas the output event of the logic gate is the 'effect' of the input event, such as an AND gate, an OR gate, a NOT gate and the like, which is different from a dynamic logic gate, and the logic gate can be called a static logic gate; the logic gates capable of representing the occurrence sequence relationship of events, namely dynamic logic gates, mainly include a preferential Gate (PAND), a sequential Gate (SEQ), a SPARE Gate (SPARE), and a functional Gate (FDEP), wherein the SPARE gates may be further divided into a Cold SPARE Gate (CSP), a Warm SPARE Gate (WSP), and a Hot SPARE Gate (Hot SPARE, HSP).
(1) Priority and gate (PAND): a logical relationship is a number of input events that occur when they occur in a particular order, and output events occur. As shown in fig. 3, when both A, B and C have occurred, and an a event occurred before a B event and a B event occurred before a C event, the system fails, i.e., the output logic value of the "priority and gate" is true. If all three input events do not occur, or event B occurs before event A, or event C occurs before event B, then the system will not exhibit a fault condition and the "priority AND gate" output value is "FALSE". It is specifically noted that if the input events occur simultaneously, the priority AND gate logic value is considered to be "true".
(2) Sequential door closing (SEQ): the input events of the system are forced to occur in sequence according to a specific sequence (from left to right), and when the DFT occurs in the SEQ gate, the system does not output fault events when the input events fail in a non-specific sequence. The SEQ gate can only trigger an output fault when the basic events fail in a particular order, and other event fault sequences cannot be replaced, but the PAND gate does not impose a requirement on the assumptions, and the system will output a fault as soon as the corresponding fault sequence is detected.
(3) SPARE part door (SPARE) comprising a main part and a plurality of backup parts, wherein after the main part fails, the 1 st backup part starts to operate to replace the main part, and after the 1 st backup part fails, the 2 nd backup part starts to be started, and so on. When all components fail, the system fails and the logic value is true.
① cold spare gate has a basic input and more than one selectable input, all input events being basic events the basic input initially goes into an active state, and the selectable inputs being those replacement spare parts that initially do not operate but are the basic input.
② Warm spare part door the warm spare part door differs from the cold spare part door in that the warm spare part has a failure rate greater than zero before entering the operating state and the cold spare part failure rate is zero.
③ Hot spare door in some systems with higher reliability requirements, the hot spare (the failure rate of the component is the same whether running or stored) is often used to switch to the working state at any time.
(4) Functionally closed door (FDEP): the function-wanting door closing comprises a triggering event (which can be a basic event or the output event of other logic gates) and a plurality of related basic events, wherein the related basic events are repeated events. The related basic events are related to the trigger event function, when the trigger event occurs, the output events occur simultaneously, and the related basic events are forced to all occur.
Dynamic Bayesian Networks (DBNs) have evolved from Bayesian Networks, in which at each point in time each factor of the environment is represented by a random variable, in such a way that the changing environment is modeled. The relationship between these variables describes how the state evolves over time. The process of changing state can be regarded as a series of snapshots, each of which describes the state of the environment at a particular Time, each snapshot is called a Time Slice (Time Slice) and contains a set of random variables, one part of which is observable, called observed variables, and the other part of which is not observable, called hidden variables. It consists of an initial network and a transition network, where each time segment corresponds to a static Bayes network. The whole network contains finite (N) time segments (not set to N >1), each time segment is composed of a directed acyclic graph GT ═ VT, ET > and conditional probability distributions satisfying the conditional independence assumption, where VT and ET distributions are the set of nodes and the set of directed edges of the time segment T. The segments are connected through directed edges, the directed edges are called a transition network, and the transition network of the time segment T is represented by ETtmp:
Figure BDA0002383840020000081
in the formula: t0 is the initial time slice.
The dynamic bayesian network satisfies the first order markov assumption: the state in time segment T is only related to the state in time segment T- Δ T and not to the state in time segments before T- Δ T, i.e.
Figure BDA0002383840020000082
Let the dynamic Bayesian network G be < V, E >, then
Figure BDA0002383840020000083
In the formula:
Figure BDA0002383840020000084
Figure BDA0002383840020000085
is the collective cartesian product.
The conversion rule for converting the static/dynamic fault tree model into the dynamic Bayesian network model comprises:
converting static logic gates in the static/dynamic fault tree model, wherein the static logic gates comprise an AND gate, an OR gate and a NOT gate;
and converting the dynamic logic gates in the static/dynamic fault tree model, wherein the dynamic logic gates comprise a priority gate, a sequential phase gate, a spare part gate and a function phase gate.
The specific steps of converting the static logic gate in the static/dynamic fault tree model are as follows:
assuming that E-0 means that event E does not occur, E-1 means that event E occurs, fE(t) is a probability density function of the time of occurrence of event E;
and the logical relation of the AND gates is corresponding to a dynamic Bayesian network, wherein the conditional probability distribution of the nodes is as follows:
Figure BDA0002383840020000086
P{A(T+ΔT)＝1|A(T)＝1}＝1 (2)
Figure BDA0002383840020000087
P{B(T+ΔT)＝1|A(T)＝1}＝1 (4)
P{TE＝1|A(T+ΔT)＝1,B(T+ΔT)＝1}＝1 (5)
P{TE＝1|else}＝0 (6)
and logically corresponding an OR gate to a dynamic Bayesian network, wherein the conditional probability distribution of a node A (T + delta T) is the same as that of the formula (1) and the formula (2), and the conditional probability distribution of a node TE is as follows:
P{TE＝1|A(T+ΔT)＝1}＝1 (7)
P{TE＝1|B(T+ΔT)＝1}＝1 (8)
P{TE＝1|A(T+ΔT)＝0,B(T+ΔT)＝0}＝0 (9)
and corresponding the NOT gate logic to a dynamic Bayesian network, wherein the conditional probability distribution of the node A (T + delta T) is the same as the conditional probability distribution of the node A (1) and the node TE, and the conditional probability distribution of the node TE is as follows:
P{TE＝1|A(T+ΔT)＝1}＝0 (10)
P{TE＝1|A(T+ΔT)＝0}＝1 (11)
where Δ T is the time interval.
The specific steps of the conversion of the dynamic logic gate in the static/dynamic fault tree model are as follows:
(1) priority and gate: the priority and gate includes a number of input events that occur when they occur in a particular order, and the output events occur. Such as the priority and gate shown in fig. 3. If events A, B and C both occur, and the A event occurs before the B event, which occurs before the C event, the output event will not occur. If all three input events do not occur, or event B occurs before event A or event C occurs before event B, an output event will not occur. It should be noted that if the input events occur simultaneously, the output event of the priority AND gate is considered to occur here.
Adding two-state nodes FS1 and FS2 according to the sequential logical relationship of priority and gates, where FS 1-1 means that a occurs before B, FS 2-1 means that B occurs before C, FS 1-0 means that a does not occur before B, and FS 2-0 means that B does not occur before C;
the analysis of the priority and gate can establish a dynamic bayesian network as shown in fig. 4, where the conditional probability distribution of each node is:
Figure BDA0002383840020000101
(2) closing the doors sequentially: the sequential closing door includes a number of input events that require the input events to occur sequentially in a particular order (left to right). Unlike the priority gate, sequential gate closing forces its input events to occur only in a specific order.
According to the sequential logical relationship of sequential closing, analyzing the functional closing of the three basic events in fig. 4, and obtaining a dynamic bayesian network corresponding to the functional closing, wherein the conditional probability distribution of each node is as follows:
Figure BDA0002383840020000102
(3) according to the size of a sleep factor α (ratio of failure rate during backup to failure rate during operation), the backup gates can be divided into three types, namely a cold backup gate (α is 0), a warm backup gate (0 < α < 1) and a hot backup gate (α is 1), and only the warm backup is discussed here, and other types of backup gates can be used as special examples of the backup gates.
According to the sequential logic relationship of the warm-up spare part gate, analyzing the warm-up spare part gate comprising a main component and a backup component in fig. 5, a dynamic bayesian network corresponding to the warm-up spare part gate can be obtained, wherein the conditional probability distribution of each node is as follows:
Figure BDA0002383840020000111
in the formula fαS(t) is a function of the failure density of the component S during the backup period;
(4) the functions are closed: the function-related gate comprises a trigger event (which can be a basic event or the output event of other logic gates) and a plurality of related basic events, wherein the related basic events are repeated events. The related basic events are related to the trigger event function, when the trigger event occurs, the output events occur simultaneously, and the related basic events are forced to all occur.
According to the relationship between the trigger event and the related basic event, analyzing the functional correlation gate including one trigger event and two related basic events in fig. 6, and obtaining a dynamic bayesian network corresponding to the functional correlation gate, wherein the conditional distribution probability of each node is:
Figure BDA0002383840020000112
and S3, calculating the prior probability and the posterior probability of the root node of each node, and inputting the leading event information serving as evidence information into the dynamic Bayesian network model according to the records of field actual production and experimental research data.
The prior probability and the root node posterior probability of each node are calculated as follows:
calculating the prior probability of each node by using a fuzzy mathematical method or historical data; and calculating the posterior probability of the root node by using a Bayesian network reverse reasoning function. The leading events include alarms, non-recovery events and accidents.
Example 2
The coal gasification device consists of a coal grinding and drying unit, a pulverized coal pressurizing and conveying unit, a gasification and washing unit, a deslagging unit, an ash water processing unit and a gasification public engineering system. Meanwhile, an inner pipe gallery, a slag conveying and slag gasifying bin system of the coal gasification device is constructed in a matched manner. The main function of the coal powder pressurizing and conveying system is to provide stable and sufficient coal powder for the gasification furnace. If the system has faults of coal powder leakage, unsmooth coal powder blanking, large pressure fluctuation of a feeding tank, failure of a solid flow meter and the like, unstable coal powder conveying can be caused, and a series of problems of partial jetting of a burner in a gasification furnace, overtemperature of a combustion chamber and the like are caused. Taking a pulverized coal pressurized conveying system as an example, risk factors of the system are analyzed through a brain storm method so as to construct a fault tree, and a dynamic risk analysis model is established according to a conversion rule from the fault tree to dynamic Bayes. And then, introducing maintenance factors and time sequences, and determining the weak link and the dynamic change rule of the reliability of the pulverized coal pressurized conveying system by using the two-way reasoning capability of the dynamic Bayesian network. The fault tree established with pulverized coal transport instability as the top event is shown in fig. 7.
Table 4 shows the intermediate events, and table 5 shows the basic events and the corresponding failure rates and maintenance rates, and the reliability parameters are derived from the data guidelines for the reliability analysis of chemical equipment.
TABLE 4 FTA model intermediate event definition
Figure BDA0002383840020000121
TABLE 5 FTA model basic event reliability parameters
Figure BDA0002383840020000131
According to the rule of converting the FTA model to the DBN, a DBN model of the pulverized coal pressurized conveying system is established as shown in FIG. 8. The arc arrow appearing in the model represents the influence of the node leading event evidence information on the next moment; the straight-line arrows indicate that the information of the previous node propagates to the next node, and the propagation direction is from the root node to the intermediate nodes and then to the leaf nodes.
When the pulverized coal pressurized conveying system does not consider the maintenance factors, the fault rate change curve of the system is shown in fig. 9. When t is 50h, the failure rate of the system is significantly increased compared to when maintenance considerations are taken into account. If it is assumed that the life and maintenance time of each device in the system follows an exponential distribution. The system is operated for 50h, a GeNIe software is used for constructing a DBN model by combining the reliability parameters of the basic events in the table 4, the software can rapidly calculate the occurrence probability of system faults at any moment by using a combined tree algorithm, and the fault rate change curve of the system is shown in figure 10 when maintenance factors are considered.
Therefore, in the operation process of the system, in order to ensure the safe operation of the system, the key equipment is necessary to be maintained and overhauled in time.
It will be appreciated by those skilled in the art that the above embodiments are merely preferred embodiments of the invention, and thus, modifications and variations may be made in the invention by those skilled in the art, which will embody the principles of the invention and achieve the objects and objectives of the invention while remaining within the scope of the invention.

Claims (10)

1. A coal gasification equipment-based dynamic Bayesian network modeling method is characterized by comprising the following steps:
analyzing coal gasification equipment, acquiring a plurality of input events of the coal gasification equipment, and establishing a static/dynamic fault tree model of each coal gasification equipment, wherein the static/dynamic fault tree model comprises a root node, an intermediate node and a leaf node;
converting the static/dynamic fault tree model into a dynamic Bayesian network model, and expressing a logical relationship between equipment failures through a topological structure of a Bayesian network;
calculating the prior probability and the root node posterior probability of each node, and inputting the leading event information serving as evidence information into the dynamic Bayesian network model according to the records of field actual production and experimental research data.
2. The coal gasification equipment dynamic Bayesian network modeling method based on the claim 1 is characterized in that a coal gasification equipment is analyzed through a brainstorm method or a HAZOP analysis, a plurality of input events of the coal gasification equipment are obtained, and a static/dynamic fault tree model of each coal gasification equipment is established.
3. The dynamic bayesian network modeling method for coal gasification-based plant according to claim 2, wherein the HAZOP analysis analyzing the coal gasification plant comprises: familiar with the process flow, selecting an evaluation node, performing node description according to the selected evaluation node, then determining deviation, performing HAZOP analysis, obtaining an event with high possibility and serious consequence, and obtaining a fault tree model analysis conclusion.
4. The dynamic Bayesian network modeling method for coal gasification facility-based according to claim 3, wherein the HAZOP analysis comprises deviation values, event causes, event consequences, existing protective measures, evaluation risk levels, and suggested measures.
5. The coal gasification facility-based dynamic bayesian network modeling method of claim 1, wherein said transformation rules for transforming static/dynamic fault tree models into dynamic bayesian network models comprise:
converting static logic gates in the static/dynamic fault tree model, wherein the static logic gates comprise an AND gate, an OR gate and a NOT gate;
and converting the dynamic logic gates in the static/dynamic fault tree model, wherein the dynamic logic gates comprise a priority gate, a sequential phase gate, a spare part gate and a function phase gate.
6. The coal gasification facility-based dynamic bayesian network modeling method of claim 5, wherein the specific steps of converting the static logic gates in the static/dynamic fault tree model are as follows:
assuming that E-0 means that event E does not occur, E-1 means that event E occurs, fE(t) is a probability density function of the time of occurrence of event E;
and the logical relation of the AND gates is corresponding to a dynamic Bayesian network, wherein the conditional probability distribution of the nodes is as follows:
Figure FDA0002383840010000021
P{A(T+ΔT)＝1|A(T)＝1}＝1 (2)
Figure FDA0002383840010000022
P{B(T+ΔT)＝1|A(T)＝1}＝1 (4)
P{TE＝1|A(T+ΔT)＝1,B(T+ΔT)＝1}＝1 (5)
P{TE＝1|else}＝0 (6)
and logically corresponding an OR gate to a dynamic Bayesian network, wherein the conditional probability distribution of a node A (T + delta T) is the same as that of the formula (1) and the formula (2), and the conditional probability distribution of a node TE is as follows:
P{TE＝1|A(T+ΔT)＝1}＝1 (7)
P{TE＝1|B(T+ΔT)＝1}＝1 (8)
P{TE＝1|A(T+ΔT)＝0,B(T+ΔT)＝0}＝0 (9)
and corresponding the NOT gate logic to a dynamic Bayesian network, wherein the conditional probability distribution of the node A (T + delta T) is the same as the conditional probability distribution of the node A (1) and the node TE, and the conditional probability distribution of the node TE is as follows:
P{TE＝1|A(T+ΔT)＝1}＝0 (10)
P{TE＝1|A(T+ΔT)＝0}＝1 (11)
where Δ T is the time interval.
7. The coal gasification facility-based dynamic bayesian network modeling method according to claim 5, wherein the steps of converting the dynamic logic gates in the static/dynamic fault tree model are as follows:
adding two-state nodes FS1 and FS2 according to the sequential logical relationship of priority and gates, where FS 1-1 means that a occurs before B, FS 2-1 means that B occurs before C, FS 1-0 means that a does not occur before B, and FS 2-0 means that B does not occur before C;
the conditional probability distribution of each node of the priority AND gate is as follows:
Figure FDA0002383840010000031
analyzing the functional associated gate of the basic event according to the sequential logical relationship of the sequential associated gates to obtain a corresponding dynamic Bayesian network, wherein the conditional probability distribution of each node is as follows:
Figure FDA0002383840010000032
analyzing the warm-up spare part gates of the main part and the backup part according to the time sequence logic relationship of the warm-up spare part gates to obtain a dynamic Bayesian network corresponding to the warm-up spare part gates, wherein the conditional probability distribution of each node is as follows:
Figure FDA0002383840010000041
in the formula fαS(t) is a function of the failure density of the component S during the backup period;
according to the relationship between the trigger event and the related basic event, analyzing the functional correlation gate including one trigger event and two related basic events in fig. 10, and obtaining a dynamic bayesian network corresponding to the functional correlation gate, wherein the conditional distribution probability of each node is as follows:
Figure FDA0002383840010000042
8. the dynamic bayesian network modeling method for coal gasification plant based on of claim 1, wherein the dividing of the coal gasification plant into key plant, main plant and general plant comprises: the main production equipment is divided into key equipment, main equipment and general equipment according to the degree of influence and loss on the aspects of production, quality, cost, safety and delivery date after the main production equipment is out of order and when the equipment is stopped for repair.
9. The coal gasification equipment-based dynamic bayesian network modeling method according to claim 1, wherein the prior probability and root node posterior probability of each node are calculated as follows:
calculating the prior probability of each node by using a fuzzy mathematical method or historical data; and calculating the posterior probability of the root node by using a Bayesian network reverse reasoning function.
10. The coal gasification facility-based dynamic bayesian network modeling method of claim 1, wherein said leading events include alarms, failure events, and accidents.
CN202010091388.6A 2020-02-13 2020-02-13 Dynamic Bayesian network modeling method based on coal gasification equipment Pending CN111275195A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010091388.6A CN111275195A (en) 2020-02-13 2020-02-13 Dynamic Bayesian network modeling method based on coal gasification equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010091388.6A CN111275195A (en) 2020-02-13 2020-02-13 Dynamic Bayesian network modeling method based on coal gasification equipment

Publications (1)

Publication Number Publication Date
CN111275195A true CN111275195A (en) 2020-06-12

Family

ID=71003625

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010091388.6A Pending CN111275195A (en) 2020-02-13 2020-02-13 Dynamic Bayesian network modeling method based on coal gasification equipment

Country Status (1)

Country Link
CN (1) CN111275195A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112270128A (en) * 2020-10-29 2021-01-26 电子科技大学 Dynamic fault tree-based drilling pump hydraulic end fault diagnosis method
CN113505448A (en) * 2021-06-09 2021-10-15 上海电力大学 Wind turbine generator dynamic reliability evaluation method based on improved Bayesian network
CN116882548A (en) * 2023-06-15 2023-10-13 中国矿业大学 Tunneling roadway coal and gas outburst prediction method based on dynamic probability reasoning
CN116882548B (en) * 2023-06-15 2024-05-17 中国矿业大学 Tunneling roadway coal and gas outburst prediction method based on dynamic probability reasoning

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110137703A1 (en) * 2004-12-21 2011-06-09 University Of Virginia Patent Foundation Method and system for dynamic probabilistic risk assessment
CN106401597A (en) * 2016-10-27 2017-02-15 华中科技大学 Failure prediction and diagnosis control method applicable to shield tunneling machine
CN107145634A (en) * 2017-04-09 2017-09-08 北京工业大学 A kind of shield cutter and the polymorphic dynamic reliability appraisal procedure of drive system
CN110489898A (en) * 2019-08-26 2019-11-22 北京航空航天大学 A kind of dynamic multilayer grade system modelling and trend prediction method based on mixing cognition

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20110137703A1 (en) * 2004-12-21 2011-06-09 University Of Virginia Patent Foundation Method and system for dynamic probabilistic risk assessment
CN106401597A (en) * 2016-10-27 2017-02-15 华中科技大学 Failure prediction and diagnosis control method applicable to shield tunneling machine
CN107145634A (en) * 2017-04-09 2017-09-08 北京工业大学 A kind of shield cutter and the polymorphic dynamic reliability appraisal procedure of drive system
CN110489898A (en) * 2019-08-26 2019-11-22 北京航空航天大学 A kind of dynamic multilayer grade system modelling and trend prediction method based on mixing cognition

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
赵志博等: "基于动态贝叶斯网络的煤粉加压输送系统可靠性分析" *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112270128A (en) * 2020-10-29 2021-01-26 电子科技大学 Dynamic fault tree-based drilling pump hydraulic end fault diagnosis method
CN112270128B (en) * 2020-10-29 2022-10-11 电子科技大学 Dynamic fault tree-based drilling pump hydraulic end fault diagnosis method
CN113505448A (en) * 2021-06-09 2021-10-15 上海电力大学 Wind turbine generator dynamic reliability evaluation method based on improved Bayesian network
CN113505448B (en) * 2021-06-09 2022-08-23 上海电力大学 Wind turbine generator dynamic reliability evaluation method based on improved Bayesian network
CN116882548A (en) * 2023-06-15 2023-10-13 中国矿业大学 Tunneling roadway coal and gas outburst prediction method based on dynamic probability reasoning
CN116882548B (en) * 2023-06-15 2024-05-17 中国矿业大学 Tunneling roadway coal and gas outburst prediction method based on dynamic probability reasoning

Similar Documents

Publication Publication Date Title
CN111311092B (en) Evaluation method based on dynamic risk of coal gasification equipment
Meel et al. Plant-specific dynamic failure assessment using Bayesian theory
CN102830666B (en) Optimize the system and method for plant operation
CN100444070C (en) Setting method for fault diagnosis and accident prediction
Guimaraes et al. Fuzzy inference to risk assessment on nuclear engineering systems
Travé-Massuyès et al. Gas-turbine condition monitoring using qualitative model-based diagnosis
Jafari et al. Reliability evaluation of fire alarm systems using dynamic Bayesian networks and fuzzy fault tree analysis
CN102262690B (en) Modeling method of early warning model of mixed failures and early warning model of mixed failures
Adedigba et al. Process accident model considering dependency among contributory factors
EP2593844A1 (en) Machine learning for power grids
CN111275195A (en) Dynamic Bayesian network modeling method based on coal gasification equipment
Hu et al. DBN based failure prognosis method considering the response of protective layers for the complex industrial systems
Cooke et al. Reliability databases in perspective
Kodoth et al. Leak frequency analysis for hydrogen-based technology using bayesian and frequentist methods
Zhou et al. A dynamic reliability-centered maintenance analysis method for natural gas compressor station based on diagnostic and prognostic technology
Al-Douri et al. Mitigation of operational failures via an economic framework of reliability, availability, and maintainability (RAM) during conceptual design
Koltsidopoulos Papatzimos et al. Offshore wind turbine fault alarm prediction
Ouazraoui et al. Layers of protection analysis in the framework of possibility theory
Linnerooth‐Bayer et al. Applications of probabilistic risk assessments: the selection of appropriate tools 1
Peng et al. Reliability modelling and assessment of a heterogeneously repaired system with partially relevant recurrence data
CN110598878A (en) Maintenance plan TIER4 assessment technology based on refinery device shutdown overhaul
Li et al. Analysis of risk factors of coal chemical enterprises based on text mining
Suleimenov et al. Synthesis of the equipment health management system of the turbine units' of thermal power stations
Redutskiy Modelling and design of Safety Instrumented Systems for upstream processes of petroleum sector
Shafaghi Equipment failure rate updating—Bayesian estimation

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination