Nuclear power station main pump fault monitoring method

By combining digital instrumentation and control systems and expert knowledge in the fault monitoring of the main pump of nuclear power plants, using deep neural networks and other technologies to extract and analyze vibration signal characteristics, the problem of difficulty in monitoring cavitation and insufficient cooling in the existing technology is solved, and the problem of failures such as cavitation and insufficient cooling is achieved. Accurate monitoring and diagnosis of the main pump failure of nuclear power plant.

CN118934587APending Publication Date: 2024-11-12JIANGSU NUCLEAR POWER CORP +1
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411113150.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-14
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

The prior art is difficult to effectively monitor the cavitation and insufficient cooling of the main pump of nuclear power plants, and because fault-type data is difficult to obtain, it is impossible to directly use data-driven methods to model diagnostic models.

Method used

Use digital instrumentation and control system to obtain data, combine expert knowledge, use digital tracking filtering, time-frequency analysis, and multiple statistical tools to extract vibration signal characteristics, and establish a mapping relationship between feature data and feature description through deep neural networks to reduce dependence on fault data , realize accurate determination of fault types.

Benefits of technology

It realizes accurate monitoring and diagnosis of the main pump fault of the nuclear power plant, reduces the dependence on fault data, and provides a way for the precise maintenance of the main pump.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure BDA0004992868620000041
    Figure BDA0004992868620000041
  • Figure BDA0004992868620000042
    Figure BDA0004992868620000042
Patent Text Reader

Abstract

The invention relates to the field of nuclear power equipment monitoring, in particular to a nuclear power station main pump fault monitoring method. The method comprises the following steps of: 1, acquiring data from a standby port of the digital instrument control system; 2, performing stable rotating speed conversion on the vibration signal data by using a digital tracking filtering tool; step 3, using a time frequency analysis tool and a multivariate statistical tool to carry out feature extraction on the vibration signal under the stable rotating speed; 4, establishing a feature description-category mapping table according to expert knowledge; 5, establishing a mapping relation from the feature data to the feature description by using a deep neural network; and step 6, collecting new data, calculating the obtained feature space for each piece of collected new data, obtaining a metric vector a of the data in accordance with the degree described by each feature, and obtaining the fault type to which the state data belongs by using a distance measurement method. According to the method, an existing digital instrument control system is fused, expert knowledge is combined, main pump faults in the operation process are monitored, dependence on fault data is reduced, and a way is provided for precise maintenance of the main pump.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of nuclear power equipment monitoring, and in particular to a method for monitoring main pump failures in a nuclear power plant. Background Art

[0002] The main pump of a nuclear power plant is an important power source for medium transportation and heat exchange in the nuclear power plant. It is also an important safety facility for the safe shutdown of the reactor. Its safety and reliability directly affect the functional integrity and safety and reliability of the reactor.

[0003] At present, the condition assessment of the main pump of a nuclear power plant only uses the vibration data of the main pump to identify the abnormal conditions of the main pump. This method can identify vibration-characterized faults of components such as bearing wear and rotor eccentricity, but it is helpless against faults such as cavitation and insufficient cooling. At the same time, in actual engineering applications, since fault data is difficult to obtain, it is impossible to directly use data-driven methods to build diagnostic models. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a method for monitoring the main pump failure of a nuclear power plant, integrate the existing digital instrumentation and control system, combine expert knowledge, monitor the main pump failure during operation, reduce the dependence on fault data, and provide a way for accurate maintenance of the main pump.

[0005] The present invention provides a method for monitoring a main pump failure in a nuclear power plant, comprising the following steps:

[0006] Step 1: Obtain data from the backup port of the digital I&C system;

[0007] Step 2: Use digital tracking filter tools to perform steady speed conversion on vibration signal data;

[0008] Step 3: Use time-frequency analysis tools and multivariate statistical tools to extract features of vibration signals at steady speed;

[0009] Step 4: Establish a feature description-category mapping table based on expert knowledge;

[0010] Step 5: Use a deep neural network to establish a mapping relationship from feature data to feature description;

[0011] Step 6: Collect new data, calculate the feature space for each new data collected, obtain the measurement vector a of the degree to which the data conforms to each feature description, and use the distance measurement method to obtain the fault type to which the state data belongs.

[0012] In a specific embodiment of the present invention, in step 1, the digital instrumentation and control system includes a main pump status monitoring system, a main pump speed measurement system and a DCS system.

[0013] Obtain data from the backup port of the digital instrumentation and control system through the network cable and signal cable,

[0014] The main pump monitoring system outputs the main pump vibration displacement signal and the main pump vibration speed signal respectively. The signal type is analog signal. The main pump speed measurement system outputs the main pump speed signal. The signal type is analog signal. The DCS system outputs the main pump operation status data. The signal type is digital signal. The two analog signals are synchronously sampled using a synchronous acquisition card to obtain the digital signal data of the main pump vibration displacement, vibration speed and speed as well as timestamps, which are matched and aligned with the status data and timestamps of the DCS system.

[0015] In a specific embodiment of the present invention, the status data includes: main pump inlet temperature, flow, pressure, bearing temperature, and main oil pump oil flow.

[0016] In a specific embodiment of the present invention, step 2 specifically includes:

[0017] The synchronously acquired pump vibration displacement signal and speed signal are aligned and intercepted, a speed curve is drawn according to the speed pulsation signal, the equal-angle resampling time is determined, and the vibration signal is smoothly converted into speed at the equal-angle resampling time using digital tracking filtering.

[0018] In a specific embodiment of the present invention, in step 3, the time-frequency analysis tool is a fast Fourier transform analysis tool, a wavelet transform analysis tool or a fast Fourier transform.

[0019] In a specific embodiment of the present invention, in step 3, the time-frequency analysis tool is used to transform the time series signal at a steady speed, and a suitable multivariate statistical algorithm is selected according to the signal type to extract sensitive features, and the DCS state data is integrated as features to form a feature space that can characterize the main pump fault type;

[0020] The timing signals are a vibration displacement signal and a vibration velocity signal.

[0021] In a specific embodiment of the present invention, step 4 specifically includes:

[0022] According to the expert's understanding of the main pump failure mechanism, a mapping matrix A of feature description to fault category is established. 1 in the matrix indicates that the fault category has the feature description, and 0 indicates that the fault category does not have the description.

[0023] In a specific embodiment of the present invention, in step 5, n deep neural networks are used to train the mapping relationship of the feature space to the mapping matrix A of the feature description to the fault category;

[0024] n is the same as the number of feature descriptions.

[0025] In a specific embodiment of the present invention, step 5 is specifically as follows: the input of the model is the feature space of each data, the output is whether the feature description exists, and the deep neural network is trained to establish a mapping relationship between the data and the feature description. A .

[0026] In a specific embodiment of the present invention, in step 6, the distance between the metric vector a and the mapping matrix A of the feature description to the fault category is compared, and the fault type corresponding to the closest feature description is taken as the fault type of the status data;

[0027] a=g A (x).

[0028] Compared with the prior art, the method for monitoring the main pump failure of a nuclear power plant of the present invention forms a main pump failure database based on the data obtained by the existing digital instrumentation and control system and combines expert knowledge. The main pump failure during operation is monitored and compared with the main pump failure database to accurately determine the type of failure. The method of the present invention reduces the dependence on failure data and provides a way for accurate maintenance of the main pump. DETAILED DESCRIPTION

[0029] In order to further understand the present invention, the embodiments of the present invention are described below in conjunction with examples, but it should be understood that these descriptions are only for further illustrating the features and advantages of the present invention, rather than limiting the present invention.

[0030] An embodiment of the present invention discloses a method for monitoring a main pump failure in a nuclear power plant, comprising the following steps:

[0031] Step 1: Obtain data from the backup port of the digital I&C system;

[0032] The digital instrumentation and control system includes a main pump status monitoring system, a main pump speed measurement system and a DCS system.

[0033] Obtain data from the backup port of the digital instrumentation and control system through the network cable and signal cable,

[0034] The main pump monitoring system outputs the main pump vibration displacement signal and the main pump vibration speed signal respectively. The signal type is analog signal. The main pump speed measurement system outputs the main pump speed signal. The signal type is analog signal. The DCS system outputs the main pump operation status data. The signal type is digital signal. The two analog signals are synchronously sampled using a synchronous acquisition card to obtain the digital signal data of the main pump vibration displacement, vibration speed and speed as well as timestamps, which are matched and aligned with the status data and timestamps of the DCS system.

[0035] The status data include: main pump inlet temperature, flow, pressure, bearing temperature, and main oil pump oil flow.

[0036] Step 2: Use digital tracking filter tools to perform steady speed conversion on vibration signal data;

[0037] The synchronously acquired pump vibration displacement signal and speed signal are aligned and intercepted, a speed curve is drawn according to the speed pulsation signal, the equal-angle resampling time is determined, and the vibration signal is smoothly converted into speed at the equal-angle resampling time using digital tracking filtering.

[0038] Smooth speed conversion refers to the use of speed information to change non-uniform speed signals (such as increasing speed, decreasing speed) into uniform speed signals.

[0039] Assume that the calculated equal-angle resampling sequence is x(t n ), the digital tracking filter is h s , then:

[0040]

[0041] Where Δt is the sampling time interval, f s is the sampling rate. kΔt is t n The points of the nearby original sampling take 2m+1 values. The larger the m value is, the higher the reconstruction accuracy is.

[0042] Step 3: Use time-frequency analysis tools and multivariate statistical tools to extract features of vibration signals at steady speed;

[0043] The time-frequency analysis tool is a fast Fourier transform analysis tool, a wavelet transform analysis tool, and a fast Fourier transform.

[0044] The time-frequency analysis tool is used to transform the time series signal at steady speed, and the appropriate multivariate statistical algorithm is selected according to the signal type to extract sensitive features. At the same time, the DCS status data is integrated as features to form a feature space that can characterize the main pump fault type.

[0045] The timing signals are a vibration displacement signal and a vibration velocity signal.

[0046] The feature space used in the present invention is shown in Table 1:

[0047] Table 1 Feature space used for diagnosis

[0048]

[0049] Step 4: Establish a feature description-category mapping table based on expert knowledge;

[0050] According to the expert's understanding of the main pump failure mechanism, a mapping matrix A of feature description to fault category is established. 1 in the matrix indicates that the fault category has the feature description, and 0 indicates that the fault category does not have the description.

[0051] The values ​​of part A of the mapping matrix of the feature description to the fault category established by the present invention are shown in Table 2.

[0052] Table 2. Part A values ​​of the mapping matrix of feature description to fault category

[0053] Rotor eccentricity Bearing wear Cavitation Insufficient cooling …… There is a frequency peak near 1 / 2 octave 0 1 0 0 …… 1x frequency is significant 1 1 0 0 …… Bearing temperature is too high 0 1 0 0 …… Main oil pump oil temperature is too high 0 0 0 1 …… The blade passing frequency peak is obvious 0 0 1 0 …… Change of inlet and outlet pressure difference 0 0 1 0 …… …… …… …… …… …… ……

[0054] The mapping matrix A of feature description to fault category is expandable.

[0055] Step 5: Use a deep neural network to establish a mapping relationship from feature data to feature description;

[0056] n deep neural networks are used to train the mapping relationship from feature space to feature description to fault category mapping matrix A;

[0057] n is the same as the number of feature descriptions.

[0058] The input of the model is the feature space of each data, and the output is whether it has the feature description. It is trained through a deep neural network to establish a mapping relationship between data and feature description. A .

[0059] Step 6: Collect new data, calculate the feature space for each new data collected, obtain the measurement vector a of the degree to which the data conforms to each feature description, and use the distance measurement method to obtain the fault type to which the state data belongs.

[0060] Compare the distance between the metric vector a and the mapping matrix A of the feature description to the fault category, and take the fault type corresponding to the closest feature description as the fault type of the status data;

[0061] a=g A (x).

[0062] For example, a mapping matrix A of feature description to fault category is obtained as shown in Table 2.

[0063] The measurement vector a=[0.94, 0.93, 0.82, 0.01, 0.01, 0.03, …] is obtained. According to Table 2 above, the data is most likely to be bearing wear. Therefore, the current state of the main pump is determined to be bearing wear.

[0064] The present invention (1) proposes a nuclear power plant main pump fault monitoring method that integrates data from multiple monitoring systems; (2) proposes a nuclear power plant main pump fault diagnosis method when the state data sample size is insufficient and the fault type coverage is incomplete; (3) proposes a nuclear power plant main pump fault diagnosis method that integrates expert knowledge and data-driven.

[0065] The above embodiments are only used to help understand the method and core idea of ​​the present invention. It should be noted that, for those skilled in the art, several improvements and modifications can be made to the present invention without departing from the principles of the present invention, and these improvements and modifications also fall within the scope of protection of the claims of the present invention.

[0066] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for monitoring a main pump failure in a nuclear power plant, characterized in that: The following steps are involved: Step 1: Obtain data from the backup port of the digital I&C system; Step 2: Use digital tracking filter tools to perform steady speed conversion on vibration signal data; Step 3: Use time-frequency analysis tools and multivariate statistical tools to extract features of vibration signals at steady speed; Step 4: Establish a feature description-category mapping table based on expert knowledge; Step 5: Use a deep neural network to establish a mapping relationship from feature data to feature description; Step 6: Collect new data, calculate the feature space for each new data collected, obtain the measurement vector a of the degree to which the data conforms to each feature description, and use the distance measurement method to obtain the fault type to which the state data belongs.

2. The method for monitoring main pump failure in a nuclear power plant according to claim 1, characterized in that: In step 1, the digital instrumentation and control system includes a main pump status monitoring system, a main pump speed measurement system and a DCS system. Obtain data from the backup port of the digital instrumentation and control system through the network cable and signal cable, The main pump monitoring system outputs the main pump vibration displacement signal and the main pump vibration speed signal respectively. The signal type is analog signal. The main pump speed measurement system outputs the main pump speed signal. The signal type is analog signal. The DCS system outputs the main pump operation status data. The signal type is digital signal. The two analog signals are synchronously sampled using a synchronous acquisition card to obtain the digital signal data of the main pump vibration displacement, vibration speed and speed as well as timestamps, which are matched and aligned with the status data and timestamps of the DCS system.

3. The method for monitoring main pump failure in a nuclear power plant according to claim 2, characterized in that: The status data include: main pump inlet temperature, flow, pressure, bearing temperature, and main oil pump oil flow.

4. The method for monitoring main pump failure in a nuclear power plant according to claim 1, characterized in that: The step 2 specifically includes: The synchronously acquired pump vibration displacement signal and speed signal are aligned and intercepted, a speed curve is drawn according to the speed pulsation signal, the equal-angle resampling time is determined, and the vibration signal is smoothly converted into speed at the equal-angle resampling time using digital tracking filtering.

5. The method for monitoring main pump failure in a nuclear power plant according to claim 1, characterized in that: In step 3, the time-frequency analysis tool is a fast Fourier transform analysis tool, a wavelet transform analysis tool or a fast Fourier transform.

6. The method for monitoring main pump failure in a nuclear power plant according to claim 5, characterized in that: In step 3, the time-frequency analysis tool is used to transform the time series signal at a steady speed, and a suitable multivariate statistical algorithm is selected according to the signal type to extract sensitive features, and the DCS state data is integrated as features to form a feature space that can characterize the main pump fault type; The timing signals are a vibration displacement signal and a vibration velocity signal.

7. The method for monitoring main pump failure in a nuclear power plant according to claim 1, characterized in that: The step 4 specifically includes: According to the expert's understanding of the main pump failure mechanism, a mapping matrix A of feature description to fault category is established. 1 in the matrix indicates that the fault category has the feature description, and 0 indicates that the fault category does not have the description.

8. The method for monitoring main pump failure in a nuclear power plant according to claim 7, characterized in that: In step 5, n deep neural networks are used to train the mapping relationship from feature space to feature description to fault category mapping matrix A; n is the same as the number of feature descriptions.

9. The method for monitoring main pump failure in a nuclear power plant according to claim 8, characterized in that: The step 5 is as follows: the input of the model is the feature space of each piece of data, the output is whether the feature description exists, and the deep neural network is trained to establish a mapping relationship between data and feature description. A .

10. The method for monitoring main pump failure in a nuclear power plant according to claim 9, characterized in that: In step 6, the distance between the metric vector a and the mapping matrix A of the feature description to the fault category is compared, and the fault type corresponding to the closest feature description is taken as the fault type of the status data; a=g A (x)。