CN111664083A - Nuclear power main pump fault diagnosis method based on Bayesian network - Google Patents
Nuclear power main pump fault diagnosis method based on Bayesian network Download PDFInfo
- Publication number
- CN111664083A CN111664083A CN202010474489.1A CN202010474489A CN111664083A CN 111664083 A CN111664083 A CN 111664083A CN 202010474489 A CN202010474489 A CN 202010474489A CN 111664083 A CN111664083 A CN 111664083A
- Authority
- CN
- China
- Prior art keywords
- main pump
- value
- probability
- bayesian network
- symptom
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000003745 diagnosis Methods 0.000 title claims abstract description 41
- 238000012544 monitoring process Methods 0.000 claims description 69
- 208000024891 symptom Diseases 0.000 claims description 65
- 238000004364 calculation method Methods 0.000 claims description 17
- 238000005070 sampling Methods 0.000 claims description 17
- 230000002159 abnormal effect Effects 0.000 claims description 6
- 238000004458 analytical method Methods 0.000 claims description 5
- 238000005259 measurement Methods 0.000 claims description 5
- 238000004445 quantitative analysis Methods 0.000 claims 1
- 238000007689 inspection Methods 0.000 abstract description 2
- 238000012423 maintenance Methods 0.000 abstract description 2
- 208000011580 syndromic disease Diseases 0.000 description 3
- 230000000875 corresponding effect Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000005856 abnormality Effects 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000002405 diagnostic procedure Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Images
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F04—POSITIVE - DISPLACEMENT MACHINES FOR LIQUIDS; PUMPS FOR LIQUIDS OR ELASTIC FLUIDS
- F04B—POSITIVE-DISPLACEMENT MACHINES FOR LIQUIDS; PUMPS
- F04B51/00—Testing machines, pumps, or pumping installations
Abstract
The invention belongs to the field of intelligent diagnosis of equipment faults, and particularly relates to a Bayesian network-based fault diagnosis method for a nuclear power main pump. The invention comprises the following steps: determining information of each node of the Bayesian network; step two: determining the structure of the Bayesian network; step three: determining a conditional probability table of the Bayesian network; step four: inputting a Bayesian diagnosis model; step five: application of a diagnostic model. According to the method, the three major faults, namely main pump sensor faults, vibration abnormity caused by process parameter fluctuation and main pump body faults, are diagnosed by combining the historical data and manual experience of the main pump, the problem of on-site false alarm is reduced, the inspection and maintenance efficiency of the main pump is improved, and the safe and reliable operation of the main pump is guaranteed.
Description
Technical Field
The invention belongs to the field of intelligent diagnosis of equipment faults, and particularly relates to a Bayesian network-based fault diagnosis method for a nuclear power main pump.
Background
The nuclear power main pump is one of key equipment of a pressurized water reactor nuclear power station, once the main pump breaks down, the production is influenced slightly, economic loss is caused, and if the main pump breaks down, machines are damaged and people die, nuclear leakage is caused, and serious environment pollution is caused. Therefore, accurate and rapid diagnosis of the fault of the main pump plays an important role in safe and efficient operation of the nuclear power plant.
A single-parameter threshold alarm method is adopted to monitor the state of a main pump in a nuclear power site, and the method has a large number of false alarms, wherein the two main false alarms are as follows: the sensor fault and the vibration abnormity caused by the fluctuation of the process parameters exceed the alarm threshold, and a method is necessary to identify the two types of faults. It is therefore necessary to establish a diagnostic method for sensor failure, vibration abnormality due to fluctuation of process parameters, and failure of the main pump apparatus body.
Considering that the main pump device accumulates a large amount of history data, if it can be applied to solve the problem, it will be effective to improve the accuracy of diagnosis. Based on the analysis, the invention provides a nuclear main pump fault diagnosis method based on a Bayesian network. According to the method, the three major faults, namely main pump sensor faults, vibration abnormity caused by process parameter fluctuation and main pump body faults, are diagnosed by combining the historical data and manual experience of the main pump, the problem of on-site false alarm is reduced, the inspection and maintenance efficiency of the main pump is improved, and the safe and reliable operation of the main pump is guaranteed.
Disclosure of Invention
The invention provides a main pump fault diagnosis method based on a Bayesian network model, aiming at the problem that a single-parameter threshold alarm method adopted on site has false alarm on whether main pump equipment has a fault or not, and the false alarm caused by sensor faults and process parameter fluctuation is diagnosed. The method uses characteristic indexes monitored by various items of the main pump as sign variables, and deduces the most possible fault type through the Bayesian network.
The main pump fault diagnosis method based on the Bayesian network provided by the invention comprises the following 5 steps:
step 1: and determining the information of each node of the Bayesian network.
The Bayesian network-based main pump fault diagnosis model provided by the invention comprises a fault layer and a symptom layer.
Wherein, the "determining information of each node of the bayesian network" in the step 1 includes: and determining the fault and corresponding symptoms of the main pump according to the relevant information of the main pump, thereby determining the fault layer node and the symptom layer node of the Bayesian network.
The fault layer nodes comprise 3 major faults of sensor faults, abnormal vibration caused by process parameter fluctuation and main pump equipment body faults; the symptom layer nodes include fluctuation indexes, multi-parameter correlation indexes, experience indexes and 4 types of vibration frequency domain indexes, and the symptom nodes of the 4 types of characteristic indexes are described below.
The fluctuation index is used for measuring the fluctuation condition of each monitoring point, the calculation method of the fluctuation index is the absolute value of the Pearson correlation coefficient obtained by monitoring point data and the period of time in a sampling duration, the value range is [0,1], the larger the value is, the more violent the fluctuation of the monitoring point is, and the value is [0,1], the fluctuation index can be regarded as the probability value as the probability of sign occurrence in the Bayesian network. The probability that there will be a fluctuation is represented by the set P (a ═ 1), as shown in equation 1.
P(A＝1)＝{P(A1＝1),P(A2＝1),…,P(Ai＝1),…,P(AN＝1)} (1)
In the formula 1, Ai1 indicates that the i th monitoring measuring point of the main pump fluctuates, P (A)i1) represents the probability of fluctuation of the ith monitoring measuring point of the main pump, N is the number of the monitoring measuring points of the main pump,P(Aithe calculation method of 1) is shown in formula 2.
Furthermore, Ai0 indicates that there is no fluctuation at the ith monitoring point of the main pump, P (A)i0) indicates the probability of no fluctuation at the ith monitoring station of the main pump, due to P (a)i＝0)+P(Ai1-1, so it is not necessary to explain P (a) againi＝0)。
In formula 2, X is vibration data with a sampling duration, and includes n sampling point data Xj(ii) a T is a time sequence formed by one sampling duration, and any time point T in the time sequencejCorresponds to X in XjThe sampling time of (a); e (-) represents the calculated mean.
The multi-parameter correlation index is used for deducing sensor faults according to the correlation among multi-sensor data, specifically, two sensors are mounted at some important parts of a main pump for monitoring, if the correlation of data of monitoring measuring points of the two sensors is larger in a certain time, the data trends collected by the two sensors are consistent, and because the probability that the two sensors simultaneously break down is very small, the probability that the sensor faults are smaller when the correlation is larger is known. The probability of the sensor failure is represented by a set P (B ═ 1), as shown in equation 3.
P(B＝1)＝{P(B1＝1),P(B2＝1),…,P(Bk＝1),…,P(BM＝1)} (3)
In formula 3, Bk1 denotes that the monitoring point sensor at the kth site of the main pump is out of order, P (B)k1) represents the fault probability of the monitoring measuring point sensor at the kth position of the main pump, and takes the value of [0, 1%]M is the number of sites on the main pump with two monitoring stations, P (B)kThe calculation method of 1) is shown in formula 4.
In addition, Bk0 indicates that the monitoring point sensor at the kth position of the main pump is not in fault, and P (B)i0) indicates the probability that the kth monitoring station sensor of the main pump is not in fault, since P (B)i＝0)+P(Bi1-1, so it is not necessary to explain P (B) againi＝0)。
In formula 4, Y is vibration data within a sampling duration collected by one of two identical sensors at a certain position of the main pump, and comprises n sampling point data Yj(ii) a Y ' is vibration data in a section of sampling duration acquired by another sensor at the same part, and n sampling point data Y ' are included in the vibration data 'j(ii) a E (-) represents the calculated mean.
The "experience index" is a symptom node added according to manual experience, and may include one or more. For example, according to field experience, when the data of the monitoring measuring points exceed the preset alarm threshold value, it is indicated that a fault of the main pump equipment body is possible to occur, so that whether the monitoring measuring points reach the alarm threshold value set on the field or not can be set as a symptom node; for another example, it is possible to set a sign node as whether the sensor measurement value is zero (or full scale value) because it is highly likely that the sensor is malfunctioning when the sensor measurement value is zero (or full scale value) by experience. The monitoring point reaching the alarm threshold is represented by a set P (C is 1), as shown in formula 5; let the monitoring point value be an end value (zero value or full scale value) and be represented by a set P (D ═ 1).
In formula 5, Ci1 represents that the ith monitoring measuring point of the main pump exceeds an alarm threshold, P (C)i1) representing the probability that the ith monitoring measuring point of the main pump exceeds the alarm threshold; di1 represents the value of the ith monitoring measuring point as an end value, P (D)i1) representing the probability that the value of the ith monitoring measuring point is an end value; n is the number of monitoring measuring points of the main pump; p (C)i1) and P (D)i1) are both state quantities 0 or 1, 0 indicating no occurrence, probability 0,1 indicating occurrence, and probability 100%.
In addition Ci0 denotes the ith supervision of the main pumpThe measured point does not exceed the alarm threshold, Di0 means that the value of the ith monitoring point is not an end value, since P (C)i＝1)+P(Ci＝0)＝1、P(Di＝1)+P(Di0) is 1, so it is not necessary to explain P (C) againi0) and P (D)i＝0)。
The vibration frequency domain index refers to the main pump vibration data which is subjected to frequency domain analysis to obtain the main frequency, and whether the main frequency is one-time frequency or not is judged. Let the main frequency component of the vibration frequency domain be a frequency doubling represented by a set P (E ═ 1), as shown in equation 6.
P(E＝1)＝{P(E1＝1),P(E2＝1),…,P(Eh＝1),…,P(EH＝1)}(6)
In formula 6, Eh1 indicates that the main frequency component is a doubling of frequency, P (E)h1) represents the probability that the main frequency of the h-th vibration measuring point is one-time frequency, and is represented by a state quantity 0 or 1, wherein 0 represents not one-time frequency, the probability is 0,1 represents one-time frequency, and the probability is 100%; h is the number of vibration measuring points.
Furthermore, EhThe term "0" means that the main frequency component is not a frequency doubling, and P (E ═ 1) + P (E ═ 0) ═ 1, so that it is not necessary to explain P (E ═ 0) again.
Step 2: the structure of the bayesian network is determined. Determining the Bayesian network structure establishes a connection relationship between the fault layer and the symptom layer. For the 4 types of characteristic indexes described in step 1, connections between all symptom nodes and fault nodes are established, as shown in fig. 2, a schematic diagram of the established bayesian network structure is shown, where a connecting line indicates that the corresponding fault node and the symptom node are associated.
And step 3: a conditional probability table for the bayesian network is determined. The purpose of determining the conditional probability table of the bayesian network is to determine the conditional probability between the fault node and the symptom node connected to each other and the prior probability of each fault occurring in the bayesian network structure established in step 2, i.e. the conditional probability between the connection lines in fig. 3.
The method comprises the following specific steps: for the fluctuation index, firstly, the vibration caused by sensor failure and process parameter fluctuation is screened from the historical monitoring dataThe data related to dynamic abnormity and main pump equipment body faults are quantitatively analyzed through the screened data, and the prior probability P of various faults is determinedprior(ii) a Secondly, counting the fluctuation probability of each monitoring measuring point when each fault occurs, and taking the probability as the fluctuation index of each monitoring measuring point and the conditional probability of the fault; setting other three kinds of characteristic indexes according to manual experience and literature data, and setting the conditional probability of the undeterminable symptom nodes as the reciprocal (P) of the number of all the symptom nodesguess) Thereby determining a conditional probability table P of the Bayesian networkcondition。
And 4, step 4: and inputting the Bayesian diagnosis model. Based on the three steps, the establishment of the Bayesian diagnosis model is completed, and the symptom variable value of each symptom node is calculated as the input of the Bayesian diagnosis model.
For different characteristic indexes, the calculation methods are different, and the following description is respectively given:
(1) for the "fluctuation index", the value of the symptom variable of such a node reflects whether there is fluctuation at the monitoring point, and is represented by a probability value, the calculation method is as described in formula 2 in step 1, and the value is a value between [0,1], 0 represents that there is no fluctuation, 1 represents that there is fluctuation, and the value between (0,1) represents the probability value that there is fluctuation, and it can be considered that the closer the value is to 1, the more severe the fluctuation is.
(2) For the "multi-parameter correlation coefficient index", the variable of such a node is the correlation between two sensor monitoring points at the same position, the calculation method is as described in formula 4 in step 1, the value is a value between [0,1], 0 represents that two parameters are correlated, 1 represents that two parameters are uncorrelated, and the value between (0,1) represents the probability value that two parameters have no correlation, and the closer the value is to 1, the more uncorrelated the two parameters are.
(3) For the experience index, the nodes comprise 'whether a monitoring measuring point exceeds an alarm threshold' and 'whether a sensor measuring value reaches an end value'. For the condition that whether the monitoring measuring points exceed the alarm threshold or not, the alarm threshold of each monitoring measuring point is determined according to a main pump manual, when the amplitude of the monitoring measuring points is larger than the alarm threshold, the symptom variable value of the symptom node is 1, otherwise, the symptom variable value is 0; and determining the measuring range according to each sensor parameter if the measured value of the sensor reaches the end value, wherein when the value of the monitoring measuring point is the end value, the symptom variable value of the symptom node is 1, and otherwise, the symptom variable value is 0.
(4) For the "vibration frequency domain index", the node variable is whether the main frequency component is a frequency doubling. And performing frequency domain analysis on the vibration data, wherein if the main characteristic of the analyzed vibration frequency domain characteristic is one frequency multiplication, the symptom variable value of the symptom node is 0, and otherwise, the symptom variable value is 1.
The input to the bayesian network thus obtained is:
P(I)＝(P(A),P(B),P(C),P(D),P(E))
and 5: application of a diagnostic model. And (4) calculating the probability of occurrence of each type of fault by using the symptom variable values of each symptom node obtained in the step (4) as input and reasoning through a Bayesian network diagnosis model, wherein the calculation method is shown as a formula 7, the maximum probability value is used as a diagnosis conclusion, the formula 7 is abstract, and a specific calculation method is described in the embodiment.
The invention provides a Bayesian network-based nuclear main pump fault diagnosis method, which combines historical monitoring data and artificial experience and can realize three major faults, namely main pump sensor faults, vibration abnormity caused by process parameter fluctuation and main pump equipment body faults. In addition, compared with a method for determining whether the main pump equipment has body faults or not based on a single-parameter threshold, the method can effectively reduce the false alarm rate.
Drawings
FIG. 1 is a flow chart of a nuclear main pump fault diagnosis method based on a Bayesian network
FIG. 2 is a schematic structural diagram of a Bayesian network diagnosis model established by the invention
FIG. 3 is the structure of the detailed Bayesian diagnosis model established by the present invention
FIG. 4 is a structure of a Bayesian diagnostic model constructed in accordance with an exemplary embodiment
FIG. 5 is a graph showing the variation trend of the vibration data of the measuring point No. 1 in the embodiment
FIG. 6 is a graph showing the variation trend of the vibration data of the measuring point No. 2 in the embodiment
Detailed Description
The invention will be further explained by taking a main pump of a nuclear power plant as an example and combining the attached drawings.
Fig. 1 shows a flow chart of implementation steps of a bayesian network-based nuclear main pump fault diagnosis method, which specifically includes the following 5 steps:
step 1: and determining the information of each node of the Bayesian network.
According to the related information of the nuclear main pump, the fault layer nodes are three types of sensor faults, abnormal vibration caused by process parameter fluctuation and main pump body faults. The symptom layer nodes include four classes, respectively: fluctuation indexes, multi-parameter correlation indexes, experience indexes and vibration frequency domain indexes. Each type of symptom node comprises a plurality of main pump monitoring measuring points, such as pressure, temperature, flow and the like, and in order to express the method more clearly, only pump shaft vibration monitoring parameters are selected for description.
Taking two vibration measuring points included at the pump shaft as an example, that is, N is 2, three types of faults of the main pump are diagnosed, fig. 5 and fig. 6 respectively show a variation trend graph of the pump shaft No. 1 vibration measuring point and the pump shaft No. 2 vibration measuring point, an alarm threshold value preset on the site of the two vibration measuring points is 250, a horizontal line in the graph is the alarm threshold value, and a symptom layer includes 7 symptom nodes, which are respectively as follows:
(1)2 "fluctuation indexes": fluctuation index P (A) of pump shaft No. 1 vibration measuring point11) and a fluctuation index P (A) of a No. 2 pump shaft vibration measuring point2＝1)；
(2)1 "multi-parameter correlation index": relevance P (B) of No. 1 vibration measuring point and No. 2 vibration measuring point of pump shaft1＝1)；
(3) For "experience indicator", it is chosen whether the alarm threshold is exceeded, so 2 "experience indicators" are included: whether the amplitude of the vibration measuring point No. 1 of the pump shaft exceeds an alarm threshold value P (C)1) Shaft 2 vibrationWhether the amplitude of the dynamic measurement point exceeds an alarm threshold value P (C)2＝1)；
(4)2 "vibration frequency indices": pump shaft No. 1 vibration measuring point whether main frequency is one-time frequency P (E)11), whether the main frequency of the pump shaft No. 2 vibration measuring point is one-time frequency P (E)2＝1)。
A total of 7 symptom nodes are included, then
The node information is shown in fig. 4.
Step 2: and determining the incidence relation of each node.
According to the method described in step 2 of the present disclosure, a bayesian network structure as shown in fig. 4 is established.
And step 3: a conditional probability table for the bayesian network is determined.
According to the method of step 3 of the present invention, a conditional probability table of a bayesian network is determined by analyzing the monitored data of the nuclear main pump-almanac history, as shown in table 1.
TABLE 1 conditional probability table for Bayesian networks
Wherein the code number of each node has the following meaning:
(1)Y1indicating a sensor failure, Y2Representing vibration anomalies due to fluctuations in process parameters, Y3Indicating a main pump apparatus body failure;
(2) for failed node Y1～Y3A variable 0 indicates that the fault does not occur, and a variable 1 indicates that the fault occurs;
(3) for a symptom node, variable 0 indicates that the symptom does not occur and variable 1 indicates that the symptom occurs, e.g., A1＝0，Y1,Y2,Y3At 0, denotes at Y1～Y3Under the condition that no three faults occur, the probability that the pump shaft vibration data do not fluctuate is 0.98, and the probability that the pump shaft vibration data fluctuate is 0.02.
And 4, step 4: the pump shaft vibration data shown in the calculation figures 5 and 6 are used for calculating the symptom node A by adopting the formula 2 in the content of the invention1Is measured in the syndrome variable P (A)11-0.893; symptom node A2Is measured in the syndrome variable P (A)21-0.912; according to the invention, formula 4 is calculated to indicate node B1Is measured in the syndrome variable P (A)31-0.067; the values of the symptom variables of the rest of the symptom nodes are shown in table 2, the values are represented by the step 1 in the summary of the invention, and the symptom in table 2 can be represented as formula 8.
I＝(A1＝1,A2＝1,B1＝1,C1＝1,C2＝0,E1＝0,E2＝0)(8)
TABLE 2 symptom node input values
And 5: application of diagnostic model
In the actual use, the manual calculation is not needed, and the diagnosis result can be directly output by inputting the symptom variable after the Bayesian network is constructed by using a computer.
Applying Bayes' theorem, there are
Wherein the content of the first and second substances,
similarly, P (A) can be obtained1＝1|Y2＝0)＝0.188。
Due to P (A)11-0.893, obtained by interpolation
P(A1＝1|Y2＝1)＝0.188+0.893×(0.812-0.188)＝0.745。
Similarly, other signs can be obtained at Y2Conditional probability at 1, results are shown in table 3, table 3 at Y20 or Y2Conditional probability of a relevant symptom determined when 1 is reached
From the conditional probabilities obtained in Table 3, the results are obtained
Probability of occurrence of abnormal vibration fault caused by fluctuation of process parameters:
P(Y2＝1|I)＝0.751
the probability P (Y) of sensor failure can be obtained by the same principle11| I) ═ 0.219, and the main pump body fault occurrence probability is P (Y)3＝1|I)＝0.028。
According to the calculation results of the Bayesian network diagnosis model, under the condition that the symptoms in the table 2 appear, the probability of the vibration abnormal fault caused by the fluctuation of the process parameters is the highest, and the fault is the most likely fault and accords with the conclusion of manual reasoning.
Claims (9)
1. A nuclear power main pump fault diagnosis method based on a Bayesian network is characterized in that: the method comprises the following steps: determining information of each node of the Bayesian network; step two: determining the structure of the Bayesian network; step three: determining a conditional probability table of the Bayesian network; step four: inputting a Bayesian diagnosis model; step five: application of a diagnostic model;
wherein, the first step: determining information of each node of the Bayesian network, specifically comprising: determining the fault and the corresponding symptom of the main pump according to the related information of the main pump, thereby determining the fault layer node and the symptom layer node of the Bayesian network;
the fault layer nodes comprise sensor faults, abnormal vibration caused by process parameter fluctuation and main pump equipment body faults, namely 3 types of faults; the symptom layer nodes comprise 4 kinds of characteristic indexes including fluctuation indexes, multi-parameter correlation indexes, experience indexes and vibration frequency domain indexes.
2. The Bayesian network-based nuclear power main pump fault diagnosis method according to claim 1, characterized in that: the fluctuation index is used for measuring the fluctuation condition of each monitoring point, the calculation method of the fluctuation index is the absolute value of the Pearson correlation coefficient obtained by monitoring point data and the period of time in a sampling duration, the value range is [0,1], and the fluctuation index can be regarded as the probability value as the probability of symptom occurrence in the Bayesian network because the value is [0,1 ];
the probability of the presence of a fluctuation is represented by the set P (a ═ 1), as shown in equation 1:
P(A＝1)＝{P(A1＝1),P(A2＝1),…,P(Ai＝1),…,P(AN＝1)} (1)
in the formula 1, Ai1 indicates that the i th monitoring measuring point of the main pump fluctuates, P (A)i1) representing the probability of fluctuation of the ith monitoring measuring point of the main pump, wherein N is the number of the monitoring measuring points of the main pump;
Ai0 indicates that there is no fluctuation at the ith monitoring point of the main pump, P (A)i0) indicates the probability of no fluctuation at the ith monitoring station of the main pump, due to P (a)i＝0)+P(Ai1-1, so it is not necessary to explain P (a) againi＝0)；
In formula 2, X is vibration data with a sampling duration, and includes n sampling point data Xj(ii) a T is a time sequence of one sample duration,at any time t thereinjCorresponds to X in XjThe sampling time of (a); e (-) represents the calculated mean.
3. The Bayesian network-based nuclear power main pump fault diagnosis method according to claim 1, characterized in that: the "multi-parameter correlation index" is used to infer a sensor fault according to the correlation between multi-sensor data, and the probability of the sensor fault is represented by a set P (B is 1), as shown in equation 3:
P(B＝1)＝{P(B1＝1),P(B2＝1),…,P(Bk＝1),…,P(BM＝1)} (3)
in formula 3, Bk1 denotes that the monitoring point sensor at the kth site of the main pump is out of order, P (B)k1) represents the fault probability of the monitoring measuring point sensor at the kth position of the main pump, and takes the value of [0, 1%]M is the number of sites on the main pump with two monitoring stations, P (B)kThe calculation method of 1) is shown in formula 4;
in addition, Bk0 indicates that the monitoring point sensor at the kth position of the main pump is not in fault, and P (B)i0) indicates the probability that the kth monitoring station sensor of the main pump is not in fault, since P (B)i＝0)+P(Bi1-1, so it is not necessary to explain P (B) againi＝0)；
In formula 4, Y is vibration data within a sampling duration collected by one of two identical sensors at a certain position of the main pump, and comprises n sampling point data Yj(ii) a Y ' is vibration data in a section of sampling duration acquired by another sensor at the same part, and n sampling point data Y ' are included in the vibration data 'j(ii) a E (-) represents the calculated mean.
4. The Bayesian network-based nuclear power main pump fault diagnosis method according to claim 1, characterized in that: the experience indexes are symptom nodes added according to manual experience; the monitoring measuring point reaching the alarm threshold is represented by a set P (C is 1); let the monitoring measurement point value as an end value (zero value or full scale value) be represented by a set P (D ═ 1), as shown in equation 5:
in formula 5, Ci1 represents that the ith monitoring measuring point of the main pump exceeds an alarm threshold, P (C)i1) representing the probability that the ith monitoring measuring point of the main pump exceeds the alarm threshold; di1 represents the value of the ith monitoring measuring point as an end value, P (D)i1) representing the probability that the value of the ith monitoring measuring point is an end value; n is the number of monitoring measuring points of the main pump; p (C)i1) and P (D)i1) are both state quantities 0 or 1, 0 indicating no occurrence, probability 0,1 indicating occurrence, and probability 100%;
in addition Ci0 represents that the ith monitoring measuring point of the main pump does not exceed the alarm threshold, Di0 means that the value of the ith monitoring point is not an end value, since P (C)i＝1)+P(Ci＝0)＝1、P(Di＝1)+P(Di0) is 1, so it is not necessary to explain P (C) againi0) and P (D)i＝0)。
5. The Bayesian network-based nuclear power main pump fault diagnosis method according to claim 1, characterized in that: the vibration frequency domain index refers to the main pump vibration data which is subjected to frequency domain analysis to obtain main frequency, and whether the main frequency is one-time frequency or not is judged; let the main frequency component of the vibration frequency domain be a frequency doubling represented by a set P (E ═ 1), as shown in equation 6:
P(E＝1)＝{P(E1＝1),P(E2＝1),…,P(Eh＝1),…,P(EH＝1)} (6)
in formula 6, Eh1 indicates that the main frequency component is a doubling of frequency, P (E)h1) represents the probability that the main frequency of the h-th vibration measurement point is one frequency doubling, and is represented by a state quantity 0 or 1, wherein 0 represents not one frequency doubling, the probability is 0,1 represents one frequency doubling, and the probability is 100Percent; h is the number of vibration measuring points;
furthermore, EhThe term "0" means that the main frequency component is not a frequency doubling, and P (E ═ 1) + P (E ═ 0) ═ 1, so that it is not necessary to explain P (E ═ 0) again.
6. The Bayesian network-based nuclear power main pump fault diagnosis method according to claim 1, characterized in that: the step 2: determining the structure of the Bayesian network specifically comprises the following steps: determining a Bayesian network structure to establish a connection relation between a fault layer and a symptom layer; and (4) establishing the connection between all the symptom nodes and the fault node aiming at the 4 large-class characteristic indexes in the step one.
7. The Bayesian network-based nuclear power main pump fault diagnosis method according to claim 1, characterized in that: the step 3: determining a conditional probability table of the Bayesian network; the purpose of determining the conditional probability table of the bayesian network is to determine the conditional probability between the fault node and the symptom node connected to each other and the prior probability of each fault in the bayesian network structure established in step 2, and the specific steps are as follows:
for the fluctuation index, firstly, data related to sensor faults, abnormal vibration caused by process parameter fluctuation and main pump equipment body faults are screened out from historical monitoring data, and the screened data are subjected to quantitative analysis to determine the prior probability P of various faultsprior；
Secondly, counting the fluctuation probability of each monitoring measuring point when each fault occurs, and taking the probability as the fluctuation index of each monitoring measuring point and the conditional probability of the fault;
setting other three kinds of characteristic indexes according to manual experience and literature data, and setting the conditional probability of the undeterminable symptom nodes as the reciprocal (P) of the number of all the symptom nodesguess) Thereby determining a conditional probability table P of the Bayesian networkcondition。
8. The Bayesian network-based nuclear power main pump fault diagnosis method according to claim 1, characterized in that: and 4, step 4: the input of the Bayesian diagnosis model specifically comprises the following steps:
based on the three steps, the establishment of the Bayesian diagnosis model is completed, and the symptom variable value of each symptom node is calculated as the input of the Bayesian diagnosis model as follows:
the calculation methods for different characteristic indexes are different, and the specific method is as follows:
(1) for the 'fluctuation index', the symptom variable value of the node reflects whether the monitoring measuring point has fluctuation or not and is represented by a probability value, the calculation method is as described in formula 2 in step 1, the value is a value between [0 and 1], 0 represents that the fluctuation does not exist, 1 represents that the fluctuation exists, and the value between (0 and 1) represents the probability value of the fluctuation, and the closer the value is to 1, the more severe the fluctuation is;
(2) for the 'multi-parameter correlation coefficient index', the variable of the node is the correlation between two monitoring points of the same sensor, the calculation method is as described in formula 4 in step 1, the value is a numerical value between [0 and 1], 0 represents the correlation between two parameters, 1 represents the irrelevance between two parameters, and the value between (0 and 1) represents the probability value that the two parameters have no correlation, and the closer the value is to 1, the more the two parameters are uncorrelated;
(3) for the experience index, the nodes comprise that whether the monitoring measuring point exceeds an alarm threshold and that whether the sensor measuring value reaches an end value; for the condition that whether the monitoring measuring points exceed the alarm threshold or not, the alarm threshold of each monitoring measuring point is determined according to a main pump manual, when the amplitude of the monitoring measuring points is larger than the alarm threshold, the symptom variable value of the symptom node is 1, otherwise, the symptom variable value is 0; for the fact that whether the measured value of the sensor reaches an end value or not, the measuring range is determined according to the parameters of each sensor, when the value of the monitoring measuring point is the end value, the sign variable value of the sign node is 1, otherwise, the sign variable value is 0;
(4) for the "vibration frequency domain index", the node variable is whether the main frequency component is a frequency doubling. Carrying out frequency domain analysis on the vibration data, wherein if the main characteristic of the analyzed vibration frequency domain characteristic is one frequency multiplication, the symptom variable value of the symptom node is 0, otherwise, the symptom variable value is 1;
the input to the bayesian network thus obtained is:
P(I)＝(P(A),P(B),P(C),P(D),P(E))。
9. the Bayesian network-based nuclear power main pump fault diagnosis method according to claim 1, characterized in that: and 5: and (3) applying a diagnosis model, wherein the symptom variable values of all the symptom nodes obtained in the step (4) are used as input, the Bayesian network diagnosis model is used for reasoning, the probability calculation method for calculating the occurrence probability of all kinds of faults is shown in the formula 7, and the maximum probability value is used as a diagnosis conclusion:
wherein, PpriorThe prior probability of occurrence of various faults; pconditionA conditional probability table of the bayesian network; pfaultThe occurrence probability of each type.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010474489.1A CN111664083A (en) | 2020-05-29 | 2020-05-29 | Nuclear power main pump fault diagnosis method based on Bayesian network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010474489.1A CN111664083A (en) | 2020-05-29 | 2020-05-29 | Nuclear power main pump fault diagnosis method based on Bayesian network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111664083A true CN111664083A (en) | 2020-09-15 |
Family
ID=72385132
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010474489.1A Pending CN111664083A (en) | 2020-05-29 | 2020-05-29 | Nuclear power main pump fault diagnosis method based on Bayesian network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111664083A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112687407A (en) * | 2020-12-28 | 2021-04-20 | 山东鲁能软件技术有限公司 | Nuclear power station main pump state monitoring and diagnosing method and system |
| CN112801171A (en) * | 2021-01-25 | 2021-05-14 | 中国商用飞机有限责任公司北京民用飞机技术研究中心 | Sensor false alarm identification method and device, computer equipment and storage medium |
| CN113780566A (en) * | 2021-06-23 | 2021-12-10 | 核动力运行研究所 | Bayesian network parameter initialization method |
| CN115095534A (en) * | 2022-04-11 | 2022-09-23 | 中核核电运行管理有限公司 | KPCA-based CANDU6 reactor main pump fault diagnosis method |
| CN116992392A (en) * | 2023-09-27 | 2023-11-03 | 中国长江电力股份有限公司 | Simulation sample generation method oriented to equipment failure mechanism |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6785636B1 (en) * | 1999-05-14 | 2004-08-31 | Siemens Corporate Research, Inc. | Fault diagnosis in a complex system, such as a nuclear plant, using probabilistic reasoning |
| CN101915234A (en) * | 2010-07-16 | 2010-12-15 | 西安交通大学 | Method for diagnosing compressor-associated failure based on Bayesian network |
| CN106372330A (en) * | 2016-08-31 | 2017-02-01 | 北京化工大学 | Application of dynamic Bayesian network to intelligent diagnosis of mechanical equipment failure |
| CN110008350A (en) * | 2019-03-06 | 2019-07-12 | 杭州哲达科技股份有限公司 | A kind of pump Ankang knowledge base lookup method based on Bayesian inference |
-
2020
- 2020-05-29 CN CN202010474489.1A patent/CN111664083A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6785636B1 (en) * | 1999-05-14 | 2004-08-31 | Siemens Corporate Research, Inc. | Fault diagnosis in a complex system, such as a nuclear plant, using probabilistic reasoning |
| CN101915234A (en) * | 2010-07-16 | 2010-12-15 | 西安交通大学 | Method for diagnosing compressor-associated failure based on Bayesian network |
| CN106372330A (en) * | 2016-08-31 | 2017-02-01 | 北京化工大学 | Application of dynamic Bayesian network to intelligent diagnosis of mechanical equipment failure |
| CN110008350A (en) * | 2019-03-06 | 2019-07-12 | 杭州哲达科技股份有限公司 | A kind of pump Ankang knowledge base lookup method based on Bayesian inference |
Non-Patent Citations (1)
| Title |
|---|
| 鲍依兰等: "贝叶斯网络用于屏蔽泵系统故障分析方法的研究", 《核科学与工程》, vol. 32, no. 02, pages 103 - 109 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112687407A (en) * | 2020-12-28 | 2021-04-20 | 山东鲁能软件技术有限公司 | Nuclear power station main pump state monitoring and diagnosing method and system |
| CN112687407B (en) * | 2020-12-28 | 2022-05-17 | 山东鲁能软件技术有限公司 | Nuclear power station main pump state monitoring and diagnosing method and system |
| CN112801171A (en) * | 2021-01-25 | 2021-05-14 | 中国商用飞机有限责任公司北京民用飞机技术研究中心 | Sensor false alarm identification method and device, computer equipment and storage medium |
| CN113780566A (en) * | 2021-06-23 | 2021-12-10 | 核动力运行研究所 | Bayesian network parameter initialization method |
| CN115095534A (en) * | 2022-04-11 | 2022-09-23 | 中核核电运行管理有限公司 | KPCA-based CANDU6 reactor main pump fault diagnosis method |
| CN116992392A (en) * | 2023-09-27 | 2023-11-03 | 中国长江电力股份有限公司 | Simulation sample generation method oriented to equipment failure mechanism |
| CN116992392B (en) * | 2023-09-27 | 2023-12-08 | 中国长江电力股份有限公司 | Simulation sample generation method oriented to equipment failure mechanism |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111664083A (en) | Nuclear power main pump fault diagnosis method based on Bayesian network | |
| CN104573850B (en) | A kind of Power Plant Equipment state evaluating method | |
| KR101491196B1 (en) | Fuzzy classification approach to fault pattern matching | |
| JP4046309B2 (en) | Plant monitoring device | |
| CN112083244B (en) | Integrated intelligent diagnosis system for faults of avionic equipment | |
| EP1982301A1 (en) | Method of condition monitoring | |
| JP2000259222A (en) | Device monitoring and preventive maintenance system | |
| JPH08202444A (en) | Method and device for diagnosing abnormality of machine facility | |
| CN113359682B (en) | Equipment fault prediction method, device, equipment fault prediction platform and medium | |
| KR102005138B1 (en) | Device abnormality presensing method and system using thereof | |
| CN113757093B (en) | Flash steam compressor unit fault diagnosis method | |
| CN113221435A (en) | Sensor screening method and device and sensor data reconstruction method and system | |
| CN112288126B (en) | Sampling data abnormal change online monitoring and diagnosing method | |
| CN115034094B (en) | Prediction method and system for operation state of metal processing machine tool | |
| CN110532699B (en) | Fuzzy DCD-based fault diagnosis method for hydrometallurgy dense washing process | |
| CN115600695A (en) | Fault diagnosis method of metering equipment | |
| Guo et al. | Fault Diagnosis Combining Information Entropy with Transfer Entropy for Chemical Processes | |
| CN115511237A (en) | Device operation condition monitoring method and system | |
| Tochev et al. | System condition monitoring through Bayesian change point detection using pump vibrations | |
| CN116342110B (en) | Intelligent fault diagnosis and fault tolerance measurement method for multiple temperature measurement loops of train | |
| CN116738859B (en) | Online nondestructive life assessment method and system for copper pipe | |
| CN117560300B (en) | Intelligent internet of things flow prediction and optimization system | |
| Peršin et al. | A system for automated online oil analysis | |
| CN117972600A (en) | Wind turbine generator set key component abnormality detection method based on multidimensional fault feature learning | |
| CN116720091A (en) | Motor fault analysis method |
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 |