Valve inner leakage prediction device based on BP neural network and model training method
By constructing a valve internal leakage prediction device based on BP neural network, the problem of low detection accuracy of internal leakage in nuclear power plant valves is solved, efficient internal leakage prediction is achieved, and the safety of nuclear power plant is improved.
Patent Information
- Application Number
- CN202510311658.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-14
- Publication Date
- 2025-07-18
AI Technical Summary
The existing valve internal leakage detection technology has low detection accuracy and poor efficiency in nuclear power plants, and cannot effectively predict the internal leakage of valves, affecting the safety and efficiency of nuclear power plants.
A valve internal leakage prediction device based on BP neural network is constructed, including a working condition simulation module, a data acquisition module and a model training module. By simulating the internal leakage state of the valve, acoustic emission signals are collected and an internal leakage prediction model is established.
It realizes efficient and accurate prediction of the degree of leakage of the valves in nuclear power plant, helping staff to detect valves in real time, and improving the safety of nuclear power plant.
Smart Images
Figure CN120337706A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of nuclear power plants, and in particular, to a valve internal leakage prediction device and a model training method based on a BP neural network. Background Art
[0002] In a nuclear power plant, as a key fluid control device, the performance of a valve directly affects the safety and efficiency of the system. In a high-temperature and high-pressure working environment, internal leakage of the valve becomes one of the common faults in the nuclear power plant, and any internal leakage of an isolation valve may pose a threat to the nuclear safety of the nuclear power plant. Therefore, accurately measuring the internal leakage of the valve is crucial for fault diagnosis, economic operation, and maintenance strategy formulation of the power station. However, the existing valve internal leakage detection technologies currently have defects such as low detection accuracy and poor efficiency. Summary of the Invention
[0003] The technical problem to be solved by the present invention is to provide a valve internal leakage prediction device and a model training method based on a BP neural network.
[0004] The technical solution adopted by the present invention to solve its technical problem is to construct a valve internal leakage prediction device based on a BP neural network, including:
[0005] A working condition simulation module, configured to access the valve to be tested, and control the opening degree of the valve to be tested based on multiple leakage amounts to simulate multiple internal leakage states of the valve to be tested;
[0006] A data acquisition module, configured to sense the acoustic emission signals of the valve to be tested in each of the internal leakage states, and generate training data according to the corresponding relationship between each acoustic emission signal and each internal leakage state;
[0007] A model training module, configured to establish a valve internal leakage amount prediction model according to the training data and the BP neural network algorithm.
[0008] Preferably, the working condition simulation module includes:
[0009] A test loop;
[0010] A fluid storage container, arranged on the test loop, for providing and recovering the test fluid;
[0011] A valve connection seat, arranged on the test loop, for accessing the valve to be tested;
[0012] A pressure measuring instrument, arranged on the test loop, for measuring the pre-valve fluid pressure of the valve to be tested;
[0013] A pressure control mechanism is provided on the test loop and is used to control the fluid pressure in the test loop during the test, so that the fluid pressure in front of the test valve remains greater than the target fluid pressure in front of the valve.
[0014] A flowmeter is provided on the test loop and is located in the downstream pipeline of the valve connection seat, and is used to measure the internal leakage of the valve under test.
[0015] Preferably, the pressure control mechanism includes:
[0016] A booster pump is provided on the test loop and is used to increase the fluid pressure in the test loop to be greater than the set target pressure.
[0017] A pressure stabilizing tank is mechanically connected to the test loop.
[0018] A pressure reducing valve is provided on the test loop and is used to cooperate with the booster pump and the pressure stabilizing tank, so that the fluid pressure in front of the test valve remains greater than the target fluid pressure in front of the valve.
[0019] The pressure measuring instrument includes:
[0020] A first pressure gauge is provided on the test loop and is located in the upstream pipeline of the valve connection seat, and is used to measure the fluid pressure in front of the test valve of the valve under test.
[0021] Preferably, the data acquisition module includes:
[0022] A characteristic signal acquisition unit is used to sense the initial characteristic signal of the valve under test in each internal leakage state to obtain acoustic emission data.
[0023] A background signal acquisition unit is used to sense the background signal at the upstream or downstream pipeline of the valve under test when internal leakage occurs in each internal leakage state of the valve under test to obtain background noise data.
[0024] A data generation unit is used to perform denoising processing on the acoustic emission data according to the background noise data to obtain the denoised acoustic emission signal of the valve under test in each internal leakage state, and generate training data according to the corresponding relationship between each denoised acoustic emission signal and each internal leakage state.
[0025] Preferably, the characteristic signal acquisition unit includes a first acoustic emission sensor, and the first acoustic emission sensor is used to be fixed at a predetermined position of the valve under test to sense the acoustic emission signal when the valve under test has internal leakage.
[0026] For the prediction of internal leakage of the valve, the background signal acquisition unit includes a second acoustic emission sensor, which is used to be fixed on the test loop or the pipeline at the working site of the valve to be tested, so as to sense the acoustic emission signal at the upstream or downstream pipeline of the valve to be tested when internal leakage occurs.
[0027] When the valve to be tested is installed at the working site, the second acoustic emission sensor is fixed at the first set position on the pipeline at the working site.
[0028] When the valve to be tested is installed on the valve connection seat, the second acoustic emission sensor is fixed at the second set position on the test loop.
[0029] Wherein, the installation methods of the second acoustic emission sensor when it is installed at the first set position and the second set position are the same, and the relative positions of the second acoustic emission sensor when it is installed at the first set position or the second set position and the valve to be tested are the same.
[0030] The present invention also constructs a model training method, including:
[0031] Obtain a plurality of acoustic emission signals; wherein, each of the acoustic emission signals is an acoustic emission signal collected when the valve to be tested controls its opening degree based on a certain leakage amount to simulate internal leakage of the valve to be tested.
[0032] Generate training data according to the corresponding relationship between each acoustic emission signal and the corresponding internal leakage state.
[0033] Establish a prediction model for the internal leakage amount of the valve according to the training data and the BP neural network algorithm.
[0034] Preferably, the method for obtaining the training data includes:
[0035] Control the opening degree of the valve to be tested based on a variety of leakage amounts through the working condition simulation module to simulate a variety of internal leakage states of the valve to be tested.
[0036] Obtain the initial characteristic signals of the valve to be tested in each internal leakage state.
[0037] Obtain the background signals at the upstream or downstream pipeline of the valve to be tested when internal leakage occurs in each internal leakage state of the valve to be tested.
[0038] Perform denoising operations respectively based on the initial characteristic signals and background signals corresponding to each internal leakage state to obtain a plurality of denoised acoustic emission signals corresponding to various internal leakage states one by one.
[0039] Preferably, the step of establishing a prediction model for the internal leakage of the valve according to the training data and the BP neural network algorithm includes:
[0040] Feature extraction is respectively performed on each of the denoised acoustic emission signals to obtain a plurality of feature data corresponding one by one to the tested valve and each of the denoised acoustic emission signals; wherein, each of the feature data includes a ring count and a signal strength eigenvalue, and the signal strength eigenvalue includes at least one of an amplitude value, an electrical average value, an effective value, and an energy value;
[0041] An initial model is established based on the BP neural network algorithm;
[0042] Using the plurality of feature data as input data and the internal leakage amount as output data, training the initial model to obtain the prediction model for the internal leakage of the valve.
[0043] Preferably, in the step of feature extraction according to the training data, the calculation method of the ring count includes: for each of the denoised acoustic emission signals, calculating the number of times greater than a set threshold level in the denoised acoustic emission signal to obtain the ring count of the denoised acoustic emission signal;
[0044] The expression of the amplitude value is: R represents the amplitude value, and X represents the denoised acoustic emission signal;
[0045] The expression of the effective value is: RMS represents the effective value, and X i represents the signal value at the i-th moment within the acquisition time period of the denoised acoustic emission signal;
[0046] The expression of the electrical average value is: ASL represents the electrical average value;
[0047] The expression of the energy value is: E represents the energy value, R represents a correction coefficient, t1 represents the occurrence time, t2 represents the cut-off time, and A(t) represents the amplitude of the denoised acoustic emission signal.
[0048] Preferably, the step of training the initial model includes:
[0049] Setting the number of hidden layer nodes of the initial model with the effective value that minimizes the prediction leakage error as the training target.
[0050] Implementing the present invention has the following beneficial effects: A valve internal leakage prediction device based on a BP neural network is provided. By establishing a valve internal leakage amount prediction model that can efficiently and accurately predict the internal leakage degree of valves in nuclear power plants, it helps staff to detect the valves in nuclear power plants in real time and effectively improves the safety of nuclear power plants. BRIEF DESCRIPTION OF THE DRAWINGS
[0051] The present invention will be further described below in conjunction with the drawings and embodiments. In the drawings:
[0052] Figure 1 is a schematic structural diagram of a valve internal leakage prediction device based on a BP neural network in an embodiment of the present invention;
[0053] Figure 2 is a schematic structural diagram of a working condition simulation module in an embodiment of the present invention;
[0054] Figure 3 is a schematic structural diagram of a data acquisition module in an embodiment of the present invention;
[0055] Figure 4 is a program flow chart of a model training method in an embodiment of the present invention;
[0056] Figure 5 is a program flow chart of step S10 in an embodiment of the present invention;
[0057] Figure 6 is a program flow chart of step S30 in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0058] In order to have a clearer understanding of the technical features, objectives, and effects of the present invention, the specific embodiments of the present invention will now be described in detail with reference to the drawings.
[0059] It should be noted that the flow charts shown in the drawings are only illustrative and do not necessarily include all the contents and operations / steps, nor do they necessarily need to be executed in the described order. For example, some operations / steps can be decomposed, while some operations / steps can be combined or partially combined. Therefore, the actual execution order may change according to the actual situation.
[0060] The block diagrams shown in the drawings are only functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.
[0061] Figure 1It is a schematic structural diagram of a valve internal leakage prediction device based on a BP neural network in an embodiment of the present invention. The valve internal leakage prediction device is used to establish a valve internal leakage amount prediction model. The valve internal leakage amount prediction model can predict the internal leakage amount based on the acoustic emission signals of nuclear power plant valves (which can be various types of valves in a nuclear power plant), and can help staff to detect nuclear power plant valves in real time and efficiently and accurately determine the internal leakage degree of nuclear power plant valves.
[0062] Please refer to Figure 1 , the valve internal leakage prediction device based on a BP neural network may include a working condition simulation module 1, a data acquisition module 2, and a model training module 3.
[0063] The working condition simulation module 1 is used to access the valve to be tested and control the opening degree of the valve to be tested based on various leakage amounts to simulate various internal leakage states of the valve to be tested. Specifically, the leakage amount of nuclear power plant valves is random, and artificially controlling its leakage amount may affect the normal operation of some systems in the nuclear power plant. Therefore, it is not convenient to collect the characteristics of the valve when internal leakage occurs at the nuclear power plant site. For this reason, the solution of the present invention is to simulate the working site of the valve to be tested through the working condition simulation module 1, and enable the valve to be tested to simulate internal leakage states with various different leakage amounts, providing an effective way to obtain the characteristics of various internal leakage states.
[0064] Since the disassembly of nuclear power plant valves may involve major repairs, system shutdowns, etc., valves with the same model as the nuclear power plant valves to be detected can be used as the valves to be tested, so that the training data can be collected without shutdown.
[0065] In one embodiment, as Figure 2 shown, the working condition simulation module 1 may include a test loop 11, a fluid storage container 12, a pressure control mechanism 13, a valve connection seat 14, a pressure measuring instrument 15, a flow meter 16, and a control unit 17.
[0066] The test loop 11. Further, as Figure 2 shown, the test loop 11 may be a loop composed of pipelines so that the test fluid can be recycled. To improve the test accuracy, it is preferably composed of pipelines with the same pipe diameter and material as those in the working site where the nuclear power plant valves to be detected are located to form the test loop 11.
[0067] The fluid storage container 12 is arranged on the test loop 11, and the fluid storage container 12 is used to provide and recycle the test fluid. Further, as Figure 2 shown, the fluid storage container 12 may be a container such as a water tank or a water storage tank.
[0068] Since the liquid viscosity may be one of the factors affecting the internal leakage of the valve, in order to improve the prediction accuracy of the internal leakage volume prediction model of the valve, the fluid used for testing is preferably the same as the fluid at the working site of the valve to be tested.
[0069] The pressure control mechanism 13 is provided on the test loop 11. The pressure control mechanism 13 is used to control the fluid pressure in the test loop 11 during the test, so that the fluid pressure in front of the test valve is kept greater than the target fluid pressure in front of the valve. Further, as Figure 2 shown, the specific position of the pressure control mechanism 13 is preferably set on the pipeline between the fluid storage container 12 and the valve connection seat 14, and the fluid flows from the pressure control mechanism 13 to the valve connection seat 14, as Figure 2 shown in the embodiment, the fluid flows in the test loop 11 in the clockwise direction, so as to ensure that the pressure of the fluid of the valve to be tested input into the valve connection seat 14 is stable.
[0070] It can be understood that controlling the fluid pressure can make the fluid pressure received by the valve to be tested in the working condition simulation module 1 as consistent as possible or even equal to the fluid pressure received by the nuclear power plant valve (the valve model is the same as the valve to be tested) at the working site under different internal leakage states, which helps to improve the confidence level of the training data.
[0071] Further, in one embodiment, as Figure 2 shown, the pressure control mechanism 13 may include a booster pump 131, a pressure stabilizing tank 132 and a pressure reducing valve 133.
[0072] The booster pump 131 is provided on the test loop 11. The booster pump 131 is used to increase the fluid pressure in the test loop 11 to be greater than the set target pressure. Further, since the operation of the booster pump 131 may cause the pipeline close to it to vibrate, affecting the valve to be tested and further affecting the test effect, the booster pump 131 is preferably set at a position as far away from the valve connection seat 14 as possible, as Figure 2 shown, the booster pump 131 can be arranged close to the fluid storage container 12. In addition, the booster pump 131 can be an existing electric water pump or pneumatic pump. Of course, to meet the test requirements, after being boosted by the booster pump 131, the fluid pressure needs to be constantly greater than the target fluid pressure in front of the valve.
[0073] The pressure stabilizing tank 132 is mechanically connected to the test loop 11. Further, the pressure stabilizing tank 132 can be an existing pressure stabilizing tank (such as a bladder type pressure stabilizing tank). The function of the pressure stabilizing tank 132 is to store energy in the air in it after the booster pump 131 works, so as to ensure that the fluid pressure in the test loop 11 remains as stable as possible after the valve to be tested leaks internally. As Figure 2 shown, the pressure stabilizing tank 132 can be arranged on the pipeline between the booster pump 131 and the pressure reducing valve 133.
[0074] A pressure reducing valve 133 is provided on the test loop 11. The pressure reducing valve 133 is used in cooperation with the booster pump 131 and the pressure stabilizing tank 132 to keep the fluid pressure in front of the test valve greater than the target fluid pressure in front of the valve. Among them, the pressure reducing valve 133 can be an existing electric valve or manual valve.
[0075] A valve connection seat 14 is provided on the test loop 11. The valve connection seat 14 is used to connect the valve to be tested. Further, the valve connection seat 14 can be a reserved flange capable of connecting the valve to be tested. In addition, the way of installing the valve to be tested on the valve connection seat 14 is preferably the same as the installation method of the nuclear power plant valve to be detected at the work site.
[0076] A pressure measuring instrument 15 is provided on the test loop 11. The pressure measuring instrument 15 is used to measure the fluid pressure in front of the test valve and the fluid pressure behind the test valve of the valve to be tested.
[0077] Further, in one embodiment, as Figure 2 shown, the pressure measuring instrument 15 can include a first pressure gauge 151 and a second pressure gauge 152.
[0078] The first pressure gauge 151 is provided on the test loop 11 and is located on the upstream pipeline of the valve connection seat 14. The first pressure gauge 151 is used to measure the fluid pressure in front of the test valve of the valve to be tested.
[0079] The second pressure gauge 152 is provided on the test loop 11 and is located on the downstream pipeline of the valve connection seat 14. The second pressure gauge 152 is used to measure the fluid pressure behind the test valve of the valve to be tested.
[0080] In this embodiment, both the first pressure gauge 151 and the second pressure gauge 152 can be existing pressure gauges as long as they can measure the fluid pressure.
[0081] A flow meter 16 is provided on the test loop 11 and is located on the downstream pipeline of the valve connection seat 14. The flow meter 16 is used to measure the internal leakage of the valve to be tested. The flow meter 16 can be an existing flow meter.
[0082] In one embodiment, the working condition simulation module 1 can further include a control unit 17. The control unit 17 is electrically connected to the pressure control mechanism 13, the valve to be tested, the pressure measuring instrument 15, and the flow meter 16. The control unit 17 is used to control the opening degrees of the booster pump 131, the pressure reducing valve 133, and the valve to be tested according to the target fluid pressure in front of the valve, the target fluid pressure behind the valve, and the target leakage amount, so as to keep the fluid pressure in front of the test valve greater than the target fluid pressure in front of the valve, the fluid pressure behind the test valve greater than the target fluid pressure behind the valve, and the internal leakage amount as close as possible to the target leakage amount.
[0083] Understandably, in this embodiment, the control unit 17, as a control device, can automate the control of the booster pump 131, the pressure reducing valve 133, and the valve under test, which helps to improve the acquisition efficiency of training data.
[0084] In one embodiment, the control unit 17 may include an existing control module (such as a PLC controller) and an existing valve control module. Correspondingly, the working principle of the control unit 17 may include: when it is necessary to simulate a certain internal leakage state, the control module first controls the booster pump 131 to start, so that the pressure before the pressure reducing valve 133 increases to the target fluid pressure before the valve. Then, the control module will, according to the measured test fluid pressure before the valve by the first pressure gauge 151 and the measured test fluid pressure after the valve by the second pressure gauge 152, simultaneously control the opening degrees of the pressure reducing valve 133 and the valve under test through the valve control module, so that the test fluid pressure before the valve can be maintained greater than the target fluid pressure before the valve for a long time, the test fluid pressure after the valve can be maintained greater than the target fluid pressure after the valve for a long time, and the internal leakage amount is as close as possible to the target leakage amount.
[0085] Understandably, during the experiment, since the valve under test is leaking, the fluid pressure in the test loop 11 will gradually decrease. Although the leakage amount will also change accordingly, the acoustic emission signals collected during this process will also change accordingly, and accurately record the change characteristics of the leakage amount as the characteristic signals corresponding to this internal leakage state.
[0086] In one embodiment, the working condition simulation module 1 may further include a first isolation valve 171, a second isolation valve 172, a third isolation valve 173, a fourth isolation valve 174, and a fifth isolation valve 175. The first isolation valve 171 is arranged on the pipeline between the booster pump 131 and the pressure reducing valve 133, and is closed when the fluid pressure in the test loop 11 increases to be greater than the set target pressure by the booster pump 131, so that the booster pump 131 is temporarily shut down, and it will not cause the fluid pressure in the test loop 11 to drop suddenly, that is, there is no need for the booster pump 131 to start and stop frequently. The second isolation valve 172 is arranged on the pipeline between the pressure stabilizing tank 132 and the test loop 11, and is used to close when the pressure stabilizing tank 132 does not need to perform the pressure stabilizing work, isolate the pressure stabilizing tank 132, and reduce the pressure relief risk in the test loop 11 during the test. The third isolation valve 173 and the fifth isolation valve 175 are respectively arranged on the pipelines between the first pressure gauge 151 and the second pressure gauge 152 and the test loop 11, and are used to close when pressure measurement is not required, further reducing the pressure relief risk during the test. The fourth isolation valve 174 is arranged on the pipeline between the pressure reducing valve 133 and the valve connection seat 14, and is closed before the valve connection seat 14 is connected to the valve under test, to avoid affecting the installation of the valve under test.
[0087] Among them, the first isolation valve 171, the second isolation valve 172, the third isolation valve 173, the fourth isolation valve 174, and the fifth isolation valve 175 can be existing electric valves or manual valves. When each isolation valve is an electric valve, it can also be electrically connected to the control unit 17, so that the control unit 17 can automatically control the first to fifth isolation valves to close or open according to requirements, which helps to improve the efficiency of training data collection.
[0088] The data acquisition module 2 is used to sense the acoustic emission signals of the valve under test in each internal leakage state, and generate training data according to the corresponding relationship between each acoustic emission signal and each internal leakage state.
[0089] Furthermore, in one embodiment, as Figure 3 shown, the data acquisition module 2 may include a characteristic signal acquisition unit 21, a background signal acquisition unit 22, and a data generation unit 23.
[0090] The characteristic signal acquisition unit 21 is used to sense the initial characteristic signals of the valve under test in each internal leakage state to obtain acoustic emission data. It should be noted that the acoustic emission data includes the acoustic emission signals collected by the characteristic signal acquisition unit 21 in a variety of internal leakage states respectively.
[0091] Furthermore, in one embodiment, the characteristic signal acquisition unit 21 may include a first acoustic emission sensor, and the first acoustic emission sensor is used to be fixed at a predetermined position of the valve under test to sense the acoustic emission signals when the valve under test has internal leakage.
[0092] In order to be able to realize the effect of real-time monitoring of nuclear power plant valves through the valve internal leakage amount prediction model, after the valve internal leakage amount prediction model is trained, it is necessary to collect the acoustic emission signals of the detected nuclear power plant valves during the commissioning process. Therefore, a characteristic signal acquisition unit 21 also needs to be set at the site of the detected nuclear power plant valves to obtain the detected nuclear power plant valves as data sources. Therefore, in order to improve the model prediction accuracy, when monitoring nuclear power plant valves, the first acoustic emission sensor can be installed at the predetermined position of the detected nuclear power plant valves, and it is ensured that the installation position and installation method of the first acoustic emission sensor on the valve under test and the detected nuclear power plant valves are kept consistent. In addition, the first acoustic emission sensor can be an existing acoustic emission sensor.
[0093] Preferably, the first acoustic emission sensor and the body of the detected nuclear power plant valve or the valve under test can be ensured to be tightly connected by means of instant dry glue, filling coupling agent, etc.
[0094] The background signal acquisition unit 22 is used to sense the background signal at the upstream or downstream pipeline of the valve under test when internal leakage occurs in each internal leakage state of the valve under test, so as to obtain background noise data. It should be noted that the background noise data includes the acoustic emission signals (i.e., background signals) collected by the background signal acquisition unit 22 in various internal leakage states respectively.
[0095] Further, in one embodiment, the background signal acquisition unit 22 may include a second acoustic emission sensor, which is used to be fixed on the test loop 11 or the pipeline at the working site of the valve under test to sense the acoustic emission signal at the upstream or downstream pipeline when the valve under test has internal leakage, so as to serve as the background signal.
[0096] Furthermore, when the valve under test is installed at the working site, the second acoustic emission sensor is fixed at the first set position on the pipeline at the working site; when the valve under test is installed on the valve connector 14, the second acoustic emission sensor is fixed at the second set position on the test loop 11; wherein, the installation methods of the second acoustic emission sensor when it is installed at the first set position and the second set position are the same, and the relative positions of the second acoustic emission sensor when it is installed at the first set position or the second set position and the valve under test are the same. It can be understood that this can ensure that the installation method of the second acoustic emission sensor when it is installed at the working site or the test loop 11 and its relative position with the corresponding valve (the nuclear power plant valve to be detected or the valve under test) are as consistent as possible, which helps to improve the prediction accuracy of the model.
[0097] The data generation unit 23 is used to perform denoising processing on the acoustic emission data according to the background noise data to obtain the denoised acoustic emission signals of the valve under test in each internal leakage state, and generate training data according to the corresponding relationship between each denoised acoustic emission signal and each internal leakage state. Specifically, the steps of performing denoising processing on the acoustic emission data according to the background noise data may include: performing denoising operations respectively based on the acoustic emission signals and background signals corresponding to each leakage amount to obtain a plurality of denoised acoustic emission signals corresponding one by one to various internal leakage states. Each internal leakage state corresponds one by one to each leakage amount, so each denoised acoustic emission signal has a corresponding relationship with the corresponding leakage amount, and the training data includes a plurality of denoised acoustic emission signals and a plurality of leakage amounts associated with each denoised acoustic emission signal.
[0098] The model training module 3 is used to establish a valve internal leakage amount prediction model according to the training data and the BP neural network algorithm.
[0099] Understandably, after the valve internal leakage prediction model is trained, the data acquisition module 2 can also be installed at the work site to collect the characteristic data of the valves of the nuclear power plant to be detected in real time. By inputting the characteristic data into the valve internal leakage prediction model, the internal leakage amount of the valves of the nuclear power plant to be detected can be predicted.
[0100] In order to improve the warning effect, in one embodiment, the valve internal leakage prediction device based on the BP neural network may further include an alarm module. The alarm module is used to obtain the output data of the valve internal leakage prediction model to determine the internal leakage amount of the valves of the nuclear power plant to be detected, and also judge whether the internal leakage amount of the valves of the nuclear power plant to be detected is greater than a set threshold. If so, an alarm signal is output to a preset alarm terminal. Among them, the preset alarm terminal may include an audible and visual alarm, the mobile terminals of relevant staff, the monitoring equipment in the control room of the nuclear power plant, etc.
[0101] Implementing the technical solution of the present invention can establish a valve internal leakage prediction model that can efficiently and accurately predict the internal leakage degree of the valves of the nuclear power plant, which can help the staff to detect the valves of the nuclear power plant in real time and improve the safety of the nuclear power plant.
[0102] Figure 4 It is the program flowchart of the model training method in an embodiment of the present invention. The model training method may include step S10, step S20, and step S30.
[0103] Step S10 includes: obtaining a plurality of acoustic emission signals; where each acoustic emission signal is an acoustic emission signal collected when the valve under test controls its opening degree based on a certain leakage amount to simulate the internal leakage state of the valve under test.
[0104] In one embodiment, as Figure 5 shown, the acquisition of the acoustic emission signals can be realized by executing step S101 to step S104.
[0105] Step S101 includes: controlling the opening degree of the valve under test by the working condition simulation module based on a variety of leakage amounts to simulate a variety of internal leakage states of the valve under test. Among them, the specific structure of the working condition simulation module can refer to the working condition simulation module provided in the embodiment of the present invention, which will not be elaborated here.
[0106] Step S102 includes: obtaining the initial characteristic signals of the valve under test in each internal leakage state. In this step, the initial characteristic signals of the valve under test in each internal leakage state can be collected by the characteristic signal acquisition unit provided in the embodiment of the present invention.
[0107] Step S103 includes: obtaining the background signal at the upstream or downstream pipeline of the valve under test when internal leakage occurs in each internal leakage state. In this step, the background signal at the upstream or downstream pipeline of the valve under test when internal leakage occurs in each internal leakage state can be collected by the background signal acquisition unit provided in the embodiment of the present invention.
[0108] Step S104 includes: performing denoising operations respectively based on the initial characteristic signal and the background signal corresponding to each internal leakage state to obtain a plurality of denoised acoustic emission signals corresponding one by one to various internal leakage states. In this step, the denoising operation can be a subtraction operation, that is, subtracting the acoustic emission signal corresponding to each internal leakage state from its corresponding background signal, and the corresponding denoised acoustic emission signal (equivalent to the acoustic emission signal) can be obtained.
[0109] Step S20 includes: generating training data according to the correspondence between each acoustic emission signal and the corresponding internal leakage state. Specifically, since various internal leakage states are in one-to-one correspondence with various leakage amounts, each denoised acoustic emission signal has a corresponding relationship with the corresponding leakage amount, and the training data includes a plurality of denoised acoustic emission signals and a plurality of leakage amounts associated with each denoised acoustic emission signal.
[0110] Step S30 includes: establishing a prediction model for the internal leakage amount of the valve according to the training data and the BP neural network algorithm.
[0111] In one embodiment, as Figure 6 shown, the prediction model for the internal leakage amount of the valve can be established by executing Step S301 to Step S302.
[0112] Step S301 includes: respectively extracting features from each denoised acoustic emission signal to obtain a plurality of feature data corresponding one by one to the valve under test and each denoised acoustic emission signal; wherein, each feature data includes a ringing count and a signal strength feature value, and the signal strength feature value includes at least one of an amplitude value, an electrical average value, an effective value, and an energy value.
[0113] In one embodiment, the ringing count can be determined by performing the following steps: for each denoised acoustic emission signal, calculate the number of times greater than the set threshold level in the denoised acoustic emission signal to obtain the ringing count of the denoised acoustic emission signal.
[0114] In one embodiment, the expression of the amplitude value can be: R represents the amplitude value, and X represents the denoised acoustic emission signal.
[0115] In one embodiment, the expression of the effective value can be: RMS represents the effective value, X iIt represents the signal value of the AE signal after denoising at the i-th moment within the acquisition time period.
[0116] In one embodiment, the expression of the electrical average value can be: ASL represents the electrical average value.
[0117] In one embodiment, the expression of the energy value can be: E represents the energy value, R represents the correction coefficient, t1 represents the occurrence time, t2 represents the cut-off time, and A(t) represents the amplitude of the AE signal after denoising.
[0118] Step S302 includes: establishing an initial model based on the BP neural network algorithm.
[0119] Step S303 includes: using multiple feature data as input data and the internal leakage amount as output data to train the initial model to obtain a prediction model for the internal leakage amount of the valve.
[0120] It can be understood that the ring count, amplitude value, electrical average value, effective value, and energy value all belong to the signal intensity eigenvalue, which are important parameters capable of characterizing the size of the internal leakage amount. The initial feature signal or background signal is usually the signal collected by the AE sensor at a set sampling rate (such as 1 MHz) within a certain time period. The ring count can accurately represent the time domain and frequency domain characteristics of the AE signal by calculating the pulse ratio of the AE signal after noise greater than the set threshold level (which can be adjusted according to the actual situation during the training process). Compared with the prior art, the present invention predicts the internal leakage amount based on the signal intensity eigenvalue and the ring count, with better prediction accuracy, and the ring count determination scheme is relatively simple and does not occupy a large amount of processor computing resources.
[0121] In one embodiment, the step of training the initial model may include: setting the number of hidden layer nodes of the initial model with the effective value that minimizes the prediction leakage amount error as the training target. In this embodiment, the most suitable number of hidden layer nodes can be set through multiple calculations.
[0122] It should be noted that the specific algorithm of the BP neural network algorithm can refer to the existing algorithm and will not be elaborated here.
[0123] In this specification, each embodiment is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the description of the method part.
[0124] Those skilled in the art may further realize that the units and algorithm steps of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.
[0125] The steps of the methods or algorithms described in combination with the embodiments disclosed herein can be directly implemented by hardware, software modules executed by a processor, or a combination of the two. The software modules can be placed in a random access memory (RAM), internal memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium well-known in the technical field.
[0126] It can be understood that the above embodiments only represent the preferred embodiments of the present invention, and the description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, the above technical features can be freely combined, and several deformations and improvements can also be made, which all belong to the protection scope of the present invention. Therefore, all equivalent transformations and modifications made to the scope of the claims of the present invention shall fall within the scope covered by the claims of the present invention.
Claims
1. A valve internal leakage prediction device based on a BP neural network, characterized in that, Including: A working condition simulation module (1), which is used to connect to the valve to be tested and control the opening degree of the valve to be tested based on various leakage amounts to simulate various internal leakage states of the valve to be tested; A data acquisition module (2), which is used to sense the acoustic emission signals of the valve to be tested in each of the internal leakage states and generate training data according to the corresponding relationship between each acoustic emission signal and each internal leakage state; A model training module (3), which is used to establish a valve internal leakage amount prediction model according to the training data and the BP neural network algorithm.
2. The valve internal leakage prediction device based on BP neural network according to claim 1, wherein The working condition simulation module (1) includes: A test loop (11); A fluid storage container (12), which is arranged on the test loop (11) and is used to provide and recover the fluid for testing; A valve connector (14), which is arranged on the test loop (11) and is used to connect to the valve to be tested; A pressure measuring instrument (15), which is arranged on the test loop (11) and is used to measure the fluid pressure in front of the test valve of the valve to be tested; A pressure control mechanism (13), which is arranged on the test loop (11) and is used to control the fluid pressure in the test loop (11) during the test so that the fluid pressure in front of the test valve is kept greater than the target fluid pressure in front of the valve; A flowmeter (16), which is arranged on the test loop (11) and is located in the downstream pipeline of the valve connector (14) and is used to measure the internal leakage amount of the valve to be tested.
3. The valve internal leakage prediction device based on a BP neural network according to claim 2, characterized in that The pressure control mechanism (13) includes: A booster pump (131), which is arranged on the test loop (11) and is used to increase the fluid pressure in the test loop (11) to be greater than the set target pressure; A pressure stabilizing tank (132), which is mechanically connected to the test loop (11); A pressure reducing valve (133), which is arranged on the test loop (11) and is used to cooperate with the booster pump (131) and the pressure stabilizing tank (132) so that the fluid pressure in front of the test valve is kept greater than the target fluid pressure in front of the valve; The pressure measuring instrument (15) includes: A first pressure gauge (151), which is arranged on the test loop (11) and is located in the upstream pipeline of the valve connector (14) and is used to measure the fluid pressure in front of the test valve of the valve to be tested.
4. The valve internal leakage prediction device based on BP neural network according to any one of claims 1 to 3, characterized in that The data acquisition module (2) includes: A characteristic signal acquisition unit (21), which is used to sense the initial characteristic signals of the valve to be tested in each of the internal leakage states to obtain acoustic emission data; A background signal acquisition unit (22), which is used to sense the background signals at the upstream or downstream pipeline of the valve to be tested when internal leakage occurs in each of the internal leakage states of the valve to be tested to obtain background noise data; A data generation unit (23), which is used to perform denoising processing on the acoustic emission data according to the background noise data to obtain the denoised acoustic emission signals of the valve to be tested in each of the internal leakage states, and generate training data according to the corresponding relationship between each denoised acoustic emission signal and each internal leakage state.
5. The valve internal leakage prediction device based on a BP neural network according to claim 4, wherein The feature signal acquisition unit (21) includes a first acoustic emission sensor, which is used to be fixed at a predetermined position of the valve under test to sense the acoustic emission signal when the valve under test has internal leakage; For valve internal leakage prediction, the background signal acquisition unit (22) includes a second acoustic emission sensor, which is used to be fixed on the test loop (11) or the pipeline at the working site of the valve under test to sense the acoustic emission signal at the upstream or downstream pipeline when the valve under test has internal leakage; When the valve under test is installed at the working site, the second acoustic emission sensor is fixed at a first set position on the pipeline at the working site; When the valve under test is installed on the valve connection seat (14), the second acoustic emission sensor is fixed at a second set position on the test loop (11); Wherein, the installation methods of the second acoustic emission sensor when it is installed at the first set position and the second set position are the same, and the relative positions of the second acoustic emission sensor when it is installed at the first set position or the second set position and the valve under test are the same.
6. A model training method, characterized in that, It includes: Obtaining a plurality of acoustic emission signals; wherein, each of the acoustic emission signals is an acoustic emission signal collected when the valve under test controls its opening based on a certain leakage amount to simulate the internal leakage of the valve under test; Generating training data according to the corresponding relationship between each of the acoustic emission signals and the corresponding internal leakage state; Establishing a valve internal leakage amount prediction model according to the training data and the BP neural network algorithm.
7. The model training method according to claim 6, wherein The method for obtaining the training data includes: Controlling the opening of the valve under test based on a variety of leakage amounts through the working condition simulation module to simulate a variety of internal leakage states of the valve under test; Obtaining the initial feature signals of the valve under test in each of the internal leakage states; Obtaining the background signals at the upstream or downstream pipeline of the valve under test when the valve under test has internal leakage in each of the internal leakage states; Performing denoising operations respectively based on the initial feature signals and background signals corresponding to each of the internal leakage states to obtain a plurality of denoised acoustic emission signals corresponding one by one to various internal leakage states.
8. The model training method according to claim 7, wherein The step of establishing a valve internal leakage amount prediction model according to the training data and the BP neural network algorithm includes: Performing feature extraction on each of the denoised acoustic emission signals respectively to obtain a plurality of feature data corresponding one by one to the valve under test and each of the denoised acoustic emission signals; wherein, each of the feature data includes ring count and signal intensity feature values, and the signal intensity feature values include at least one of amplitude value, electrical average value, effective value and energy value; Establishing an initial model based on the BP neural network algorithm; Using the plurality of feature data as input data and the internal leakage amount as output data to train the initial model to obtain the valve internal leakage amount prediction model.
9. The model training method according to claim 8, wherein In the step of feature extraction according to the training data, the calculation method of the ringing count includes: for each of the denoised acoustic emission signals, calculating the number of times greater than a set threshold level in the denoised acoustic emission signal to obtain the ringing count of the denoised acoustic emission signal; The expression of the amplitude value is as follows: R represents the amplitude value, and X represents the denoised acoustic emission signal; The expression for the effective value is as follows: RMS represents the effective value, and X i represents the signal value of the denoised acoustic emission signal at the i-th moment within the acquisition time period; The expression of the electrical average value is as follows: ASL represents the electrical average value; The expression of the energy value is as follows: E represents the energy value, R represents the correction coefficient, t1 represents the occurrence time, t2 represents the cut-off time, and A(t) represents the amplitude of the denoised acoustic emission signal.
10. The model training method according to claim 9, wherein The step of training the initial model includes: Setting the number of hidden layer nodes of the initial model to the effective value with the smallest prediction leakage error as the training target.