A nuclear power plant secondary circuit high and low pressure heater drain control system

Abstract

The present invention discloses a nuclear power plant secondary circuit high and low pressure heater drain control system, which obtains the eddy current detection signal of the first high pressure heater to be detected collected by the electromagnetic induction detector; extracts the eddy current graphic features of the eddy current detection signal of the first high pressure heater to obtain the eddy current signal graphic feature map; based on the pixel granularity correlation relationship of the eddy current signal graphic feature map, the eddy current signal graphic feature map is strengthened to obtain the strengthened eddy current signal graphic feature map; and based on the strengthened eddy current signal graphic feature map, it is determined whether the first high pressure heater to be detected has defects. In this way, the difficulty of maintenance can be greatly reduced and the efficiency of maintenance can be improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent drainage technology, and in particular to a drainage control system for secondary circuit high and low pressure heaters in a nuclear power plant. Background Art

[0002] The function of the secondary circuit system of a pressurized water reactor nuclear power plant is to convert thermal energy into mechanical energy and form a complete thermal cycle. The thermal equipment in the secondary circuit system will inevitably experience aging and degradation problems. Among them, uniform corrosion, flow-accelerated corrosion, decommissioning corrosion and erosion corrosion are common forms of aging and degradation of conventional secondary circuit system thermal equipment.

[0003] Corrosion monitoring is an important part of the secondary circuit high and low pressure heater drain system and control system of nuclear power plants. Traditionally, thermal equipment is visually inspected and corrosion products or sediments are analyzed by X-ray diffractometer to determine the corrosion status of the main equipment of the secondary circuit system, especially the heater. However, this method of inspecting the drain system equipment requires a lot of manpower and material resources, and it also takes a lot of time to maintain the equipment. The maintenance is difficult and inefficient. Therefore, an optimized solution is expected. Summary of the invention

[0004] The embodiment of the present invention provides a nuclear power plant secondary circuit high and low pressure heater drain control system, which obtains the eddy current detection signal of the first high pressure heater to be detected collected by the electromagnetic induction detector; extracts the eddy current graphic features of the eddy current detection signal of the first high pressure heater to obtain the eddy current signal graphic feature map; based on the pixel granularity correlation relationship of the eddy current signal graphic feature map, the eddy current signal graphic feature map is strengthened to obtain the strengthened eddy current signal graphic feature map; and based on the strengthened eddy current signal graphic feature map, it is determined whether the first high pressure heater to be detected has defects. In this way, the difficulty of maintenance can be greatly reduced and the efficiency of maintenance can be improved.

[0005] An embodiment of the present invention also provides a nuclear power plant secondary circuit high and low pressure heater drain control system, which includes: a detection signal acquisition module, used to obtain the eddy current detection signal of the first high pressure heater to be detected collected by the electromagnetic induction detector; a feature extraction module, used to extract the eddy current graphic features of the eddy current detection signal of the first high pressure heater to be detected to obtain an eddy current signal graphic feature map; an attention enhancement module, used to enhance the attention of the eddy current signal graphic feature map based on the pixel granularity correlation relationship of the eddy current signal graphic feature map to obtain an enhanced eddy current signal graphic feature map; and a detected first high pressure heater judgment module, used to determine whether the detected first high pressure heater has defects based on the enhanced eddy current signal graphic feature map. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0007] In the attached picture: Figure 1 The present invention provides a block diagram of a nuclear power plant secondary circuit high and low pressure heater drain control system according to an embodiment of the present invention.

[0008] Figure 2 The present invention is a flowchart of a method for controlling the drainage of high and low pressure heaters in a secondary circuit of a nuclear power plant provided in an embodiment of the present invention.

[0009] Figure 3 It is a schematic diagram of the system architecture of a method for controlling the hydrophobicity of high and low pressure heaters in a secondary circuit of a nuclear power plant provided in an embodiment of the present invention.

[0010] Figure 4 This is an application scenario diagram of a nuclear power plant secondary circuit high and low pressure heater drain control system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0011] To make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the embodiments of the present invention are further described in detail below in conjunction with the accompanying drawings. Here, the exemplary embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0012] Unless otherwise specified, all technical and scientific terms used in the embodiments of the present invention have the same meanings as those commonly understood by those skilled in the art of the present invention. The terms used in the present invention are only for the purpose of describing specific embodiments and are not intended to limit the scope of the present invention.

[0013] In the description of the embodiments of the present invention, it should be noted that, unless otherwise specified and limited, the term "connection" should be understood in a broad sense. For example, it can be an electrical connection or a connection between two components. It can be a direct connection or an indirect connection through an intermediate medium. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0014] It should be noted that the terms "first\second\third" involved in the embodiments of the present invention are only used to distinguish similar objects, and do not represent a specific order of the objects. It is understandable that the specific order or sequence of "first\second\third" can be interchanged where permitted. It should be understood that the objects distinguished by "first\second\third" can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein.

[0015] The secondary circuit system of a pressurized water reactor nuclear power plant is an important part of the nuclear power plant. It plays the role of converting thermal energy into mechanical energy and forming a complete thermodynamic cycle. The secondary circuit system is mainly composed of the main steam system, reheat system, condenser system and feed water system. The main steam system is the part of the nuclear reactor that produces steam. The fuel elements in the nuclear reactor convert thermal energy into steam. The steam drives the steam turbine under high temperature and high pressure to generate mechanical energy and drive the generator to generate electricity. The reheat system is a link in the steam turbine. After part of the steam is discharged from the high-pressure cylinder, it is heated again and sent to the low-pressure cylinder to improve the efficiency of the steam turbine. The condenser system is the part that condenses the low-pressure steam discharged from the steam turbine into water. In the condenser, the low-pressure steam releases heat and condenses into water through contact with the cooling medium (usually cooling water), and then is pumped back to the feed water system. The feed water system is responsible for sending the condensed water back to the main steam system after treatment to ensure horizontal replenishment and circulation in the nuclear reactor.

[0016] However, the thermal equipment in the secondary circuit system will age and degrade as the service time increases. Common forms of aging and degradation include uniform corrosion, flow-accelerated corrosion, decommissioning corrosion and scouring corrosion. Uniform corrosion refers to the phenomenon that the surface of the equipment is uniformly corroded due to chemical reactions when the surface of the equipment is in contact with the working medium for a long time, which may cause the material of the equipment to become thinner, weaken its strength, and even cause the risk of leakage. Flow-accelerated corrosion is the phenomenon that the medium impacts and corrodes the surface of the equipment when it flows inside the equipment. This corrosion usually occurs at the elbows and pipe connections of the equipment, which may cause local damage and leakage of the equipment. Decommissioning corrosion refers to the phenomenon that the surface of the equipment corrodes due to the residue of the medium or the influence of environmental conditions during the shutdown period of the equipment. This corrosion may cause corrosion damage and performance degradation of the equipment. Scouring corrosion refers to the phenomenon that the surface of the equipment is scourred and corroded when the medium flows inside the equipment. This corrosion usually occurs at the flow channel, nozzle and other locations of the equipment, which may cause damage to the equipment and performance degradation.

[0017] In order to ensure the safe and reliable operation of the secondary circuit system, nuclear power plants need to conduct regular inspections and maintenance, monitor and deal with aging and degradation problems, so as to ensure the normal operation of equipment and the safe operation of nuclear power plants. Corrosion monitoring is a very important link in the secondary circuit high and low pressure heater drain system and control system of nuclear power plants. Although traditional visual inspection and X-ray diffractometer analysis methods can determine the corrosion status of equipment, there are problems such as large consumption of manpower and material resources, long maintenance time and difficulty in maintenance. In order to improve efficiency and accuracy, modern nuclear power plants have adopted some advanced corrosion monitoring technologies.

[0018] Here are some commonly used corrosion monitoring technologies: Ultrasonic technology can be used to detect changes in the thickness of the heater pipe wall to determine whether corrosion exists. By scanning the equipment with an ultrasonic probe, accurate data on the thickness of the pipe wall can be obtained to assess the extent of corrosion. Electrochemical corrosion monitoring is a method of assessing the corrosion of equipment by measuring current and potential. By placing electrodes on the surface of the equipment and measuring changes in current and potential, it can be determined whether the equipment is corroded and the type and extent of corrosion can be determined. Magnetic marking is a method of detecting equipment corrosion using the behavior of magnetic particles in a magnetic field. Magnetic particles are added to the surface of the equipment. When the equipment corrodes, the magnetic particles change. By detecting and analyzing the magnetic particles, the corrosion condition of the equipment can be determined. Infrared thermal imaging uses an infrared thermal imager to detect temperature changes on the surface of the equipment. Corrosion usually causes changes in the surface temperature of the equipment. By scanning the equipment with infrared thermal imaging, the corroded area can be found and the extent of corrosion can be assessed.

[0019] The use of advanced corrosion monitoring technology can improve the accuracy and efficiency of monitoring, reduce dependence on manpower and material resources, and be able to detect and identify corrosion problems early, which helps to take repair and maintenance measures in time to ensure the safe operation of the secondary circuit system.

[0020] In one embodiment of the present invention, Figure 1 FIG. 1 is a block diagram of a nuclear power plant secondary circuit high and low pressure heater drain control system provided in an embodiment of the present invention. Figure 1As shown, a nuclear power plant secondary circuit high and low pressure heater drain control system 100 according to an embodiment of the present invention includes: a detection signal acquisition module 110, used to obtain an eddy current detection signal of a first high pressure heater to be detected collected by an electromagnetic induction detector; a feature extraction module 120, used to extract the eddy current graphic features of the eddy current detection signal of the first high pressure heater to be detected to obtain an eddy current signal graphic feature map; an attention enhancement module 130, used to enhance the attention of the eddy current signal graphic feature map based on the pixel granularity correlation relationship of the eddy current signal graphic feature map to obtain an enhanced eddy current signal graphic feature map; and a detected first high pressure heater judgment module 140, used to determine whether the detected first high pressure heater has a defect based on the enhanced eddy current signal graphic feature map.

[0021] In the detection signal acquisition module 110, the eddy current detection signal of the first high-voltage heater to be detected is acquired by the electromagnetic induction detector. When using this module, ensure that the electromagnetic induction detector is correctly installed and calibrated to obtain accurate detection signals. The effective operation of this module is very critical for the subsequent feature extraction and judgment modules.

[0022] In the feature extraction module 120, the eddy current pattern features of the eddy current detection signal of the first high-voltage heater to be detected are extracted to generate an eddy current signal pattern feature diagram. In the feature extraction process, appropriate feature extraction algorithms and parameters are selected to ensure that the extracted features can accurately reflect the key information of the eddy current signal. Appropriate feature extraction can improve the accuracy of the subsequent judgment module.

[0023] In the attention enhancement module 130, the image is enhanced by using the correlation between pixels in the eddy current signal graphic feature map. The attention mechanism can improve the focus on key information and reduce the interference of noise and irrelevant information. In the process of attention enhancement, an appropriate attention mechanism and weight calculation method are selected to ensure that the enhanced image can highlight the important features of the eddy current signal.

[0024] In the first high-voltage heater detection judgment module 140, based on the enhanced eddy current signal graphic feature map, it is used to determine whether the first high-voltage heater detected has defects. The key to the judgment module is to establish an accurate defect classification model or rule to make judgments based on the defect features in the feature map. The training and verification of the model requires sufficient sample data and accurate annotation information to ensure the accuracy and reliability of the judgment module.

[0025] In this way, the accuracy and efficiency of eddy current testing are improved, and it is possible to quickly determine whether the equipment has defects, provide reference data for equipment repair and maintenance, and thus improve the safety and reliability of the equipment.

[0026] In response to the above technical problems, the technical concept of the present invention is to use the principle of electromagnetic induction to generate eddy currents, and to reflect the defects or property changes on the surface or near the surface of the conductor being detected through the eddy current detection signal, so that the corrosion of the first high-voltage heater can be discovered in time. That is, the electromagnetic induction detector generates eddy currents through electromagnetic induction. When defects or property changes occur on the surface or near the surface of the conductor being detected, it will affect the intensity and distribution of the eddy currents. In the technical solution of the present invention, it is expected to utilize this characteristic and combine it with artificial intelligence technology based on deep learning to learn the eddy current graphic feature information in the eddy current detection signal, so as to determine whether the first high-voltage heater being detected has defects.

[0027] It should be understood that electromagnetic induction detectors can detect defects or property changes on the surface or near the surface of conductors by inducing the intensity and distribution of eddy currents. Combined with artificial intelligence technology based on deep learning, the eddy current graphic features in eddy current detection signals can be learned and analyzed. Deep learning models can learn complex feature representations from large amounts of data, thereby improving the sensitivity of defect detection, helping to discover small or hidden defects, and providing early warning of possible problems.

[0028] Traditional eddy current detection methods usually require manual extraction and analysis of feature information in eddy current signals, which requires professional knowledge and experience. However, artificial intelligence technology based on deep learning can realize automated feature extraction. The deep learning model can learn the abstract representation of eddy current graphic features through training, thereby reducing the burden of manual feature extraction and improving the accuracy and efficiency of feature extraction. By learning the eddy current graphic feature information in eddy current detection signals, artificial intelligence technology based on deep learning can identify different types of defects. The deep learning model can learn the feature representation of different defect types through training, thereby realizing diversified defect recognition, which helps to improve the coverage of defect recognition and can cope with defects of different types and sizes.

[0029] By utilizing the eddy current characteristics generated by the electromagnetic induction detector and combining it with artificial intelligence technology based on deep learning to learn the eddy current graphic feature information in the eddy current detection signal, it is possible to determine whether the first high-voltage heater being detected has defects. This can improve the sensitivity of defect detection, realize automated feature extraction, and achieve diversified defect identification, thereby improving the safety and reliability of the equipment.

[0030] Based on this, in the technical solution of the present invention, the eddy current detection signal of the first high-voltage heater to be detected is first obtained by the electromagnetic induction detector; and the eddy current detection signal of the first high-voltage heater to be detected is passed through an eddy current graphic feature extractor based on a convolutional neural network model to obtain an eddy current signal graphic feature diagram. Among them, the eddy current detection signal is based on the eddy current induction principle and is obtained through the interaction between the induction coil and the alternating electromagnetic field. Specifically, when the alternating electromagnetic field passes through the metal surface, eddy currents are generated in the metal. These eddy currents will in turn generate a magnetic field, which interacts with the magnetic field of the induction coil, resulting in a change in the induced electromotive force in the induction coil. This changing electrical signal is called an eddy current detection signal.

[0031] Here, the eddy current detection signal of the first high-voltage heater being detected can reflect the corrosion condition of the heater. As mentioned above, when corrosion occurs, the intensity and distribution of the eddy current detection signal will change. In the technical solution of the present invention, an eddy current graphic feature extractor based on a convolutional neural network model is used to extract image feature information of eddy current distribution in the eddy current detection signal to learn abstract features in the eddy current signal, such as edges, textures, and shapes. These features are of great significance for the judgment and classification of corrosion conditions.

[0032] In a specific embodiment of the present invention, the feature extraction module is used to: use a deep learning network model to perform feature extraction on the eddy current detection signal of the first high-voltage heater to obtain the eddy current signal graphic feature diagram.

[0033] Among them, the deep learning network model is a vortex graphic feature extractor based on a convolutional neural network model; wherein, the vortex graphic feature extractor based on a convolutional neural network model includes: an input layer, a convolution layer, a pooling layer, an activation layer and an output layer.

[0034] Specifically, in a specific embodiment of the present invention, the feature extraction module is used to: pass the eddy current detection signal of the first high-voltage heater to be detected through the eddy current graphic feature extractor based on the convolutional neural network model to obtain the eddy current signal graphic feature map.

[0035] In traditional eddy current testing, feature extraction usually requires manual participation, expertise and experience. The eddy current graphic feature extractor based on the convolutional neural network model can automatically learn and extract key features from eddy current signals without manual intervention, which can reduce the demand for human resources and improve the efficiency and consistency of feature extraction. The convolutional neural network model has a powerful feature extraction capability and can capture complex patterns and subtle changes in eddy current signals. By using this model for feature extraction, the resulting eddy current signal graphic feature map can more comprehensively and accurately reflect the information of the eddy current signal and provide a richer and more representative feature representation.

[0036] The features extracted from the eddy current signal graphic feature map are learned by the convolutional neural network model and have better discrimination ability. These features can be used for subsequent defect judgment and classification tasks to improve the accuracy and reliability of judgment. Compared with the traditional manual feature extraction method, the feature extractor based on the convolutional neural network can better capture the useful information in the eddy current signal, thereby improving the accuracy of judgment. By extracting the features of the eddy current detection signal through the eddy current graphic feature extractor based on the convolutional neural network model, it is possible to realize automated feature extraction, provide a more comprehensive and accurate feature representation, and improve the accuracy of judgment, which helps to improve the efficiency and reliability of eddy current detection and provide a more reliable basis for the identification and maintenance of equipment defects.

[0037] In one embodiment of the present invention, the attention enhancement module includes: an autocorrelation feature matrix calculation unit, which is used to calculate the activated pixel granularity autocorrelation feature matrix based on the pixel granularity correlation relationship of the eddy current signal graphic feature map; and a feature map calculation unit, which is used to use the activated pixel granularity autocorrelation feature matrix as a class attention weight matrix, and respectively calculate the position point multiplication between each feature matrix along the channel dimension of the eddy current signal graphic feature map and the activated pixel granularity autocorrelation feature matrix to obtain the enhanced eddy current signal graphic feature map.

[0038] Furthermore, in a specific embodiment of the present invention, the autocorrelation feature matrix calculation unit includes: a correlation measurement coefficient calculation subunit, used to calculate the pixel granularity correlation measurement coefficient between every two channel feature vectors in the eddy current signal graphic feature map to obtain a pixel granularity autocorrelation matrix; and an activation subunit, used to input the pixel granularity autocorrelation matrix into a Sigmoid activation function to obtain the activated pixel granularity autocorrelation feature matrix.

[0039] The autocorrelation feature matrix calculation unit calculates the autocorrelation feature matrix of the activated pixel granularity based on the pixel granularity correlation relationship of the eddy current signal graphic feature map. The autocorrelation feature matrix can capture the correlation and repetitive patterns between pixels in the image. By calculating the autocorrelation feature matrix, important autocorrelation information in the eddy current signal graphic feature map can be extracted, which helps to highlight the key features in the eddy current signal and provides a basis for subsequent feature enhancement.

[0040] The feature map calculation unit uses the activated pixel granularity autocorrelation feature matrix as the attention-like weight matrix to calculate the positional dot product between each feature matrix of the eddy current signal graphic feature map along the channel dimension and the activated pixel granularity autocorrelation feature matrix, so as to enhance and weight the feature map. Through the operation of the feature map calculation unit, the response strength of the key features in the eddy current signal graphic feature map can be improved, the influence of noise and irrelevant information can be weakened, and the important features in the eddy current signal can be further highlighted.

[0041] The autocorrelation feature matrix calculation unit and the feature map calculation unit play an important role in the processing of eddy current signal graphic feature maps. By calculating the autocorrelation feature matrix and enhancing the feature map, they can improve the expression and discrimination capabilities of the eddy current signal graphic feature map, help to more accurately capture the key features in the eddy current signal, and improve the accuracy and reliability of eddy current detection.

[0042] Then, the pixel granularity correlation measurement coefficient between every two channel feature vectors in the eddy current signal graphic feature map is calculated to obtain a pixel granularity autocorrelation matrix; and the pixel granularity autocorrelation matrix is ​​input into a Sigmoid activation function to obtain an activated pixel granularity autocorrelation feature matrix. In a specific example of the present invention, the encoding process of calculating the pixel granularity correlation measurement coefficient between every two channel feature vectors in the eddy current signal graphic feature map to obtain the pixel granularity autocorrelation matrix includes: calculating the pixel granularity correlation measurement coefficient between every two channel feature vectors in the eddy current signal graphic feature map using the following correlation measurement formula to obtain the pixel granularity autocorrelation matrix; wherein the correlation measurement formula is: ;in, is the pixel granularity correlation measurement coefficient, is the value of each position in the previous channel feature vector, is the value of each position in the next channel feature vector, is the dimension of the previous channel feature vector and the next channel feature vector, Represents the logarithmic function operation with base 2.

[0043] Here, the pixel granularity correlation measurement coefficient in the form of a standard cross entropy function can reflect the cross-difference between each two channel feature vectors, that is, the difference between the latter channel feature vector and the former channel feature vector, and the difference between the former channel feature vector and the latter channel feature vector, and this cross-difference can extract the difference between the eddy current distribution and eddy current intensity between different local areas in the eddy current detection signal of the first high-voltage heater to be detected, so as to reveal the correlation and difference between the eddy current distribution characteristics in each local area. It should be understood that if the corrosion degree of the first high-voltage heater to be detected is relatively light, the eddy current detection signal should present a uniform eddy current distribution; on the contrary, if the corrosion degree of the first high-voltage heater to be detected is relatively strong, the eddy current detection signal should present an uneven distribution, that is, the eddy current distribution morphology between each local area may have a large difference. At the same time, calculating the pixel granularity correlation measurement coefficient between each two channel feature vectors in the eddy current signal graphic feature map can effectively aggregate the target information by using the correlation between any two pixels in the eddy current signal graphic feature map.

[0044] In particular, the purpose of inputting the pixel granularity autocorrelation matrix into the Sigmoid activation function is to utilize the activation characteristics of the Sigmoid activation function to limit the value range of the eigenvalues ​​of each position in the pixel granularity autocorrelation matrix to between 0 and 1, that is, to integrate global information to highlight more important feature information in the local area and suppress unimportant feature information, thereby obtaining an activated pixel granularity autocorrelation feature matrix.

[0045] Next, the activated pixel granularity autocorrelation feature matrix is ​​used as the attention-like weight matrix, and the positional point multiplication between each feature matrix along the channel dimension of the eddy current signal graphic feature map and the activated pixel granularity autocorrelation feature matrix is ​​calculated to obtain an enhanced eddy current signal graphic feature map. That is, the important feature distribution in the eddy current signal graphic feature map is enhanced by the positional point multiplication method to complete the self-optimization and feature expression correction of the eddy current signal graphic feature map.

[0046] Specifically, the activated pixel granularity autocorrelation feature matrix is ​​used as a class attention weight matrix, and the value of each position indicates the contribution of the feature information at different positions to the final result. By multiplying each feature matrix of the eddy current signal graphic feature map with the activated pixel granularity autocorrelation feature matrix by position, the feature map can be strengthened. In this way, the features of the positions with high attention weights will be enhanced, while the features of the positions with low attention weights will be suppressed.

[0047] In one embodiment of the present invention, the judgment module for the first high-voltage heater under test is used to: pass the enhanced eddy current signal graphic characteristic diagram through a classifier to obtain a classification result, and the classification result is used to indicate whether the first high-voltage heater under test has defects.

[0048] Then, the enhanced eddy current signal graphic feature graph is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the first high-voltage heater under test has a defect. By inputting the enhanced eddy current signal graphic feature graph into the classifier for classification, the learning ability and discrimination ability of the classifier can be used to determine whether the first high-voltage heater under test has a defect. The classifier can make accurate classification decisions based on the input feature graph, thereby improving the accuracy of defect detection, helping to discover and identify potential equipment defects early, and promoting timely repair and maintenance.

[0049] Inputting the enhanced eddy current signal graphic feature map into the classifier for classification can realize automatic defect identification. Traditional defect identification usually requires manual participation, professional knowledge and experience. By using the classifier for automatic classification, the demand for human resources can be reduced, the efficiency and consistency of defect identification can be improved, which can save time and cost, and improve the efficiency of equipment maintenance and management.

[0050] By inputting the enhanced eddy current signal graphic feature map into the classifier for classification, real-time monitoring and early warning of the first high-voltage heater under inspection can be achieved. The classifier can instantly classify and judge the eddy current signal and give the corresponding defect identification results, which helps to take timely measures to prevent equipment defects from further developing and causing irreversible damage, and improve the safety and reliability of the equipment. By inputting the enhanced eddy current signal graphic feature map into the classifier for classification, a classification result can be obtained to indicate whether the first high-voltage heater under inspection has defects. This can improve the accuracy of defect detection, realize automated defect identification, and realize real-time monitoring and early warning, thereby improving the safety and reliability of the equipment.

[0051] In the above technical solution, each feature matrix along the channel dimension of the eddy current signal graphic feature map expresses the image semantic features of the signal waveform image of the eddy current detection signal of the first high-voltage heater being detected, and each feature matrix follows the channel distribution of the convolutional neural network model. Therefore, when calculating the pixel granularity correlation measurement coefficient between every two channel feature vectors to obtain a pixel granularity autocorrelation matrix, and inputting the pixel granularity autocorrelation matrix into the Sigmoid activation function to obtain an activated pixel granularity autocorrelation feature matrix, it is actually to obtain the image semantic feature channel global correlation representation of each feature matrix on the channel dimension. In this way, Taking the activated pixel granularity autocorrelation feature matrix as the class attention weight matrix, the position point multiplication between each feature matrix along the channel dimension of the eddy current signal graphic feature map and the activated pixel granularity autocorrelation feature matrix is ​​calculated respectively to obtain the enhanced eddy current signal graphic feature map. The enhanced eddy current signal graphic feature map as a whole has a spatial distribution attribute feature expression based on the association with the spatial distribution of image semantic features in the spatial distribution dimension of the feature matrix and the channel distribution dimension, so that the enhanced eddy current signal graphic feature map has a regression-like interpretability loss due to insufficient aggregation of image semantic feature distribution, which affects the accuracy of the classification results.

[0052] Based on this, in a preferred embodiment of the present application, the enhanced eddy current signal graphic characteristic diagram is passed through a classifier to obtain a classification result, including: optimizing the enhanced eddy current signal graphic characteristic diagram to obtain an optimized enhanced eddy current signal graphic characteristic diagram; passing the optimized enhanced eddy current signal graphic characteristic diagram through a classifier to obtain a classification result.

[0053] The process of optimizing the enhanced eddy current signal graphic characteristic diagram to obtain an optimized enhanced eddy current signal graphic characteristic diagram comprises the steps of: expanding the enhanced eddy current signal graphic characteristic diagram into an enhanced eddy current signal graphic characteristic vector; and calculating the autocorrelation matrix of the enhanced eddy current signal graphic characteristic vector, namely: ;in, represents the graphical feature vector of the enhanced eddy current signal, represents the transpose of a vector, represents matrix multiplication, Denotes the autocorrelation matrix.

[0054] The enhanced eddy current signal graphic feature vector is multiplied by the autocorrelation matrix in a query manner to obtain an intermediate feature vector, namely: ;in, represents the graphical feature vector of the enhanced eddy current signal, represents the autocorrelation matrix, represents matrix multiplication, Denotes the intermediate eigenvector.

[0055] Calculate the autocorrelation matrix Row vector and Inner product sum of row vectors distance, to obtain the first weight matrix and the second weight matrix , expressed as: ;in, and The autocorrelation matrix is ​​represented by Row vector and row vector, represents the transpose of a vector, represents matrix multiplication, represents the autocorrelation matrix, represents the first weight matrix, represents the first weight matrix The eigenvalues ​​of the positions, represents the second weight matrix, represents the second weight matrix The eigenvalues ​​of the positions, Represents the calculation vector distance.

[0056] The first weight matrix and the second weight matrix After matrix multiplication, further matrix multiplication is performed with the intermediate eigenvector to obtain the optimized enhanced eddy current signal graphic eigenvector, namely: ;in, represents the first weight matrix, represents the second weight matrix, represents the intermediate eigenvector, represents matrix multiplication, Representing the optimized enhanced eddy current signal graphic feature vector; restoring the optimized enhanced eddy current signal graphic feature vector to an optimized enhanced eddy current signal graphic feature diagram.

[0057] That is, for the bit-by-bit scattering correlation response space of the eigenvalues ​​of the enhanced eddy current signal graphic feature map, the deep embedding structure in the high-order correlation space is obtained through the joint embedding space of the vector inner product embedding space and the vector independent embedding space, and the decomposable correlation response sub-dimensions of the bit-by-bit scattering correlation response space of the eigenvalues ​​of the enhanced eddy current signal graphic feature map are represented as a set of spatial sub-dimensions to further perform dependent space mapping of the decomposable correlation response sub-dimensions of the deep embedding structure of the enhanced eddy current signal graphic feature map, so as to improve the interpretability of the solution space of the class regression solution of the enhanced eddy current signal graphic feature map based on the correlation distribution through the complex dimensional representation of the correlation response, thereby improving the accuracy of the classification result obtained by the classifier of the enhanced eddy current signal graphic feature map.

[0058] In one embodiment of the present application, the nuclear power plant secondary circuit high and low pressure heater drain control system also includes a training module for training the eddy current pattern feature extractor based on the convolutional neural network model and the classifier. Wherein, the training module includes: a training data acquisition unit for acquiring training data, wherein the training data includes a training eddy current detection signal of the first high pressure heater to be detected collected by an electromagnetic induction detector; a training feature extraction unit for passing the training eddy current detection signal of the first high pressure heater to be detected through the eddy current pattern feature extractor based on the convolutional neural network model to obtain a training eddy current signal pattern feature map; a training autocorrelation feature matrix calculation unit for calculating a training activation pixel granularity autocorrelation feature matrix based on the pixel granularity correlation relationship of the training eddy current signal pattern feature map; a training feature map calculation unit for using the training activation pixel granularity autocorrelation feature matrix as a class attention weight matrix to calculate the training activation pixel granularity autocorrelation feature matrix respectively. The training enhanced eddy current signal graphic feature map is obtained by multiplying the feature matrices of the training eddy current signal graphic feature map along the channel dimension and the training activated pixel granularity autocorrelation feature matrix by position points; a classification loss unit is used to pass the training enhanced eddy current signal graphic feature map through a classifier to obtain a classification loss function value; a training unit is used to train the eddy current graphic feature extractor based on the convolutional neural network model and the classifier based on the classification loss function value and propagating in the direction of gradient descent, wherein each time the training enhanced eddy current signal graphic feature vector after the training enhanced eddy current signal graphic feature map is expanded passes through the classifier for classification regression iteration, the training enhanced eddy current signal graphic feature vector is optimized.

[0059] In a preferred embodiment of the present application, each time the training enhanced eddy current signal graphic feature vector after the training enhanced eddy current signal graphic feature graph is expanded passes through the classifier for classification regression iteration, the training enhanced eddy current signal graphic feature vector is optimized, which is expressed as: the training enhanced eddy current signal graphic feature vector is optimized by the following formula to obtain the optimized training enhanced eddy current signal graphic feature vector; wherein the formula is: ;in, is the training enhanced eddy current signal graphic feature vector, is the training enhanced eddy current signal graphic feature vector The square of the 1-norm of is the training enhanced eddy current signal graphic feature vector The reciprocal of the square root of the 2-norm of is the first feature vector of the training enhanced eddy current signal graph The eigenvalues ​​at the positions, is the training enhanced eddy current signal graphic feature vector The length of is a scaling hyperparameter, represents the logarithmic function value with base 2, is the first characteristic vector of the optimized training enhanced eddy current signal graph. The feature value at each position.

[0060] Here, the eddy current signal graphic feature vector is strengthened based on the training The low-rank structure of the norm is represented as the training enhanced eddy current signal graph feature vector The voting cluster of the eigenvalue set aggregation is used to strengthen the eddy current signal graphic feature vector of the training Each eigenvalue of performs canonical voting relative to the information perception framework to map the eigenvalues ​​belonging to the same regression class to a similar local canonical coordinate set by aggregating the regression representation of the direction and scale of the feature distribution, thereby improving the training enhanced eddy current signal graphic feature vector The aggregation effect of the feature set is improved to improve the accuracy of the classification result obtained by the classifier of the training enhanced eddy current signal graphic feature map and the training speed of the classifier.

[0061] The present invention installs an electromagnetic induction detector in the device drainage system. The electromagnetic induction detector can detect the side wall corrosion degree in the No. 1 high-pressure heater in real time. The electromagnetic induction detector generates eddy currents through electromagnetic induction. When defects appear on the surface or near the surface of the conductor being detected or the properties change, the intensity and distribution of the eddy currents will be affected, so that defects in the No. 1 high-pressure heater and whether the metal properties have changed can be detected. Subsequently, the No. 1 high-pressure heater can drive the device body to automatically remove the corrosion on the inner wall of the device, which greatly reduces the difficulty of maintenance and improves the efficiency of maintenance.

[0062] In summary, a nuclear power plant secondary circuit high and low pressure heater drain control system 100 based on an embodiment of the present invention is explained, which utilizes the principle of electromagnetic induction to generate eddy currents, and reflects the defects or property changes on the surface or near-surface of the conductor being detected through eddy current detection signals, thereby being able to promptly detect the corrosion of the first high-pressure heater, and combining with artificial intelligence technology based on deep learning to learn the eddy current graphic feature information in the eddy current detection signal, thereby determining whether the first high-pressure heater being detected has defects.

[0063] As described above, the nuclear power plant secondary circuit high and low pressure heater drain control system 100 according to the embodiment of the present invention can be implemented in various terminal devices, such as a server for nuclear power plant secondary circuit high and low pressure heater drain control. In one example, the nuclear power plant secondary circuit high and low pressure heater drain control system 100 according to the embodiment of the present invention can be integrated into the terminal device as a software module and / or a hardware module. For example, the nuclear power plant secondary circuit high and low pressure heater drain control system 100 can be a software module in the operating system of the terminal device, or it can be an application developed for the terminal device; of course, the nuclear power plant secondary circuit high and low pressure heater drain control system 100 can also be one of the many hardware modules of the terminal device.

[0064] Alternatively, in another example, the nuclear power plant secondary circuit high and low pressure heater drain control system 100 and the terminal device may also be separate devices, and the nuclear power plant secondary circuit high and low pressure heater drain control system 100 may be connected to the terminal device via a wired and / or wireless network, and transmit interactive information in accordance with an agreed data format.

[0065] Figure 2 The present invention is a flowchart of a method for controlling the drainage of high and low pressure heaters in a secondary circuit of a nuclear power plant provided in an embodiment of the present invention. Figure 3 Schematic diagram of the system architecture of a method for controlling the drainage of high and low pressure heaters in a secondary circuit of a nuclear power plant provided in an embodiment of the present invention. Figure 2 and Figure 3As shown, a method for controlling hydrophobicity of high and low pressure heaters in a secondary circuit of a nuclear power plant comprises: 210, acquiring an eddy current detection signal of a first high pressure heater to be detected collected by an electromagnetic induction detector; 220, extracting eddy current graphic features of the eddy current detection signal of the first high pressure heater to be detected to obtain an eddy current signal graphic feature map; 230, based on the pixel granularity correlation relationship of the eddy current signal graphic feature map, performing attention enhancement on the eddy current signal graphic feature map to obtain an enhanced eddy current signal graphic feature map; and, 240, based on the enhanced eddy current signal graphic feature map, determining whether the first high pressure heater to be detected has defects.

[0066] Those skilled in the art will appreciate that the specific operations of each step in the above-mentioned nuclear power plant secondary circuit high and low pressure heater drain control method have been referenced above. Figure 1 It has been introduced in detail in the description of the secondary circuit high and low pressure heater drain control system of a nuclear power plant, and therefore, its repeated description will be omitted.

[0067] Figure 4 This is an application scenario diagram of a nuclear power plant secondary circuit high and low pressure heater drain control system provided in an embodiment of the present invention. Figure 4 As shown, in this application scenario, first, an eddy current detection signal of the first high-voltage heater to be detected is acquired by an electromagnetic induction detector (for example, Figure 4 Then, the acquired eddy current detection signal is input to a server (for example, Figure 4 In S) shown in FIG. 1 , the server is capable of processing the eddy current detection signal based on the nuclear power plant secondary circuit high and low pressure heater drain control algorithm to determine whether the detected first high pressure heater has a defect.

[0068] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A nuclear power plant secondary circuit high and low pressure heater drain control system, It is characterized in that include: A detection signal acquisition module, used to acquire an eddy current detection signal of the first high-voltage heater under detection collected by an electromagnetic induction detector; A feature extraction module, used for extracting eddy current graphic features of the eddy current detection signal of the first high-voltage heater to obtain an eddy current signal graphic feature diagram; An attention enhancement module, configured to enhance the attention of the eddy current signal graphic feature map based on the pixel granularity correlation relationship of the eddy current signal graphic feature map to obtain an enhanced eddy current signal graphic feature map; as well as a first high-voltage heater judgment module to be detected, used for determining whether the first high-voltage heater to be detected has a defect based on the enhanced eddy current signal graphical characteristic diagram; Wherein, the attention enhancement module includes: An autocorrelation feature matrix calculation unit is used to calculate an activated pixel granularity autocorrelation feature matrix based on the pixel granularity correlation relationship of the eddy current signal graphic feature map; and A feature map calculation unit, used to use the activated pixel granularity autocorrelation feature matrix as a class attention weight matrix, and respectively calculate the position point multiplication between each feature matrix along the channel dimension of the eddy current signal graphic feature map and the activated pixel granularity autocorrelation feature matrix to obtain the enhanced eddy current signal graphic feature map; Wherein, the autocorrelation feature matrix calculation unit includes: A correlation metric coefficient calculation subunit is used to calculate the pixel granularity correlation metric coefficient between every two channel feature vectors in the eddy current signal graphic feature map to obtain a pixel granularity autocorrelation matrix; and An activation subunit, used for inputting the pixel granularity autocorrelation matrix into a Sigmoid activation function to obtain the activated pixel granularity autocorrelation feature matrix; Wherein, the correlation measurement coefficient calculation subunit is used to: The pixel granularity correlation measurement coefficient between every two channel feature vectors in the eddy current signal graphic feature map is calculated using the following correlation measurement formula to obtain the pixel granularity autocorrelation matrix; Wherein, the correlation measurement formula is: ;in, is the pixel granularity correlation measurement coefficient, is the value of each position in the previous channel feature vector, is the value of each position in the next channel feature vector, is the dimension of the previous channel feature vector and the next channel feature vector, Represents the logarithmic function operation with base 2.

2. The nuclear power plant secondary circuit high and low pressure heater drain control system according to claim 1, It is characterized in that The feature extraction module is used to: A deep learning network model is used to extract features of the eddy current detection signal of the first high-voltage heater to obtain a graphical feature diagram of the eddy current signal.

3. The nuclear power plant secondary circuit high and low pressure heater drain control system according to claim 2, It is characterized in that The deep learning network model is an eddy current graphic feature extractor based on a convolutional neural network model; Wherein, the eddy current graphic feature extractor based on the convolutional neural network model includes: an input layer, a convolution layer, a pooling layer, an activation layer and an output layer.

4. The nuclear power plant secondary circuit high and low pressure heater drain control system according to claim 3, It is characterized in that The feature extraction module is used to: The eddy current detection signal of the first high-voltage heater being detected is passed through the eddy current pattern feature extractor based on the convolutional neural network model to obtain the eddy current signal pattern feature diagram.

5. The nuclear power plant secondary circuit high and low pressure heater drain control system according to claim 4, It is characterized in that The detected first high-voltage heater judgment module is used for: The enhanced eddy current signal graphical feature diagram is passed through a classifier to obtain a classification result, and the classification result is used to indicate whether the first high-voltage heater under inspection has a defect.

6. The nuclear power plant secondary circuit high and low pressure heater drain control system according to claim 5, It is characterized in that It also includes a training module for training the eddy current pattern feature extractor based on the convolutional neural network model and the classifier.

7. The nuclear power plant secondary circuit high and low pressure heater drain control system according to claim 6, It is characterized in that The training module comprises: A training data acquisition unit, used to acquire training data, wherein the training data includes a training eddy current detection signal of the first high-voltage heater to be detected collected by an electromagnetic induction detector; A training feature extraction unit, used for passing the training eddy current detection signal of the first high-voltage heater to be detected through the eddy current pattern feature extractor based on the convolutional neural network model to obtain a training eddy current signal pattern feature map; A training autocorrelation feature matrix calculation unit, used for calculating a training activation pixel granularity autocorrelation feature matrix based on the pixel granularity correlation relationship of the training eddy current signal graphic feature map; A training feature map calculation unit is used to use the training activation pixel granularity autocorrelation feature matrix as a class attention weight matrix, and respectively calculate the position point multiplication between each feature matrix along the channel dimension of the training eddy current signal graphic feature map and the training activation pixel granularity autocorrelation feature matrix to obtain a training enhanced eddy current signal graphic feature map; A classification loss unit, used for passing the training enhanced eddy current signal graphical feature map through a classifier to obtain a classification loss function value; A training unit is used to train the eddy current pattern feature extractor based on the convolutional neural network model and the classifier based on the classification loss function value and through the direction propagation of gradient descent, wherein each time the training enhanced eddy current signal graphic feature vector after the training enhanced eddy current signal graphic feature map is expanded is subjected to classification regression iteration through the classifier, the training enhanced eddy current signal graphic feature vector is optimized.

Images