Nuclear power device fault diagnosis method and system based on CNN and sliding window verification

Abstract

The invention discloses a nuclear power device fault diagnosis method and system based on CNN and sliding window verification, and belongs to the technical field of nuclear power system fault diagnosis. The method comprises the following steps: acquiring sensor and actuator signals under normal and fault working conditions, and converting the sensor and actuator signals into three-channel color pictures as a training set through discrete sampling and normalization processing; a multi-layer CNN network including an input layer, a 2D convolutional layer and the like is constructed, and a fault diagnosis model is obtained through training; and collecting real-time signals to generate a test set, inputting the test set into the model in batches through a sliding time window, calculating label scores in the window, and selecting the label with the highest score as a fault type. The problems of high cost, low precision and poor real-time performance in the prior art are solved, the accuracy and the real-time performance of fault diagnosis are improved, and the method is suitable for safe operation monitoring of a nuclear power device.

Description

Technical Field

[0001] This invention belongs to the field of nuclear power system fault diagnosis technology, specifically relating to a nuclear power plant fault diagnosis method and system based on CNN and sliding window verification. Background Technology

[0002] Faults pose a serious threat to the safe operation of nuclear energy systems. Timely and accurate monitoring and diagnosis of fault information can provide effective guidance for operators' corrective actions and have significant reference value for subsequent decision-making. This is a crucial guarantee for improving the economic efficiency and safety of nuclear energy utilization. Fault diagnosis technology can monitor the system's operating status, determine whether a fault has occurred, and simultaneously determine the time, location, magnitude, and type of the fault. In other words, it completes fault detection, isolation, and estimation, and is an important method for improving the reliability of complex industrial systems and reducing system operational risks.

[0003] The proportional-integral-derivative (PI) control technology commonly used in the nuclear energy field has significant limitations in dealing with complex and abnormal operating conditions. To improve system safety and fault tolerance and reduce the risk of human-machine interaction errors, current technical solutions mainly focus on the following two dimensions: First, the fault tolerance mechanism at the device level is enhanced by adding multiple hardware redundancy configurations; Secondly, optimize the personnel training system and implement a multi-position collaborative duty system to enhance the emergency response capabilities for sudden work situations.

[0004] However, such traditional improvement paths have two technical drawbacks: First, the integrated deployment of redundant hardware modules will directly lead to an exponential increase in equipment complexity, significantly increasing the maintenance cost throughout the entire life cycle; second, nuclear reactor device failures exhibit multidimensional uncertainty characteristics (including spatial distribution, temporal evolution, and intensity fluctuations), and conventional personnel training models are difficult to cover all potential failure modes. Even with a highly qualified operating team, there is still a risk of cognitive overload, which can easily lead to misjudgment of operating conditions and improper operation.

[0005] It is noteworthy that existing diagnostic models based on conventional neural networks also have inherent technical bottlenecks. Due to the highly nonlinear characteristics of the parameter coupling mechanism in nuclear energy systems, traditional artificial neural network algorithms suffer from insufficient modeling accuracy and limited generalization ability in identifying dynamic and random faults, failing to meet the stringent requirements of high-reliability nuclear safety standards for fault diagnosis accuracy. This technical deficiency is particularly prominent when dealing with complex fault modes with strong time-varying characteristics, seriously affecting the operational safety and economic indicators of nuclear energy plants. Summary of the Invention

[0006] The technical problem to be solved by this invention is to address the shortcomings of the prior art by providing a nuclear power plant fault diagnosis method and system based on CNN and sliding window verification. By establishing a multi-layer CNN neural network and calculating fault type scores based on a sliding window, the method accurately diagnoses the fault type and severity of real-time operating nuclear power plants. This addresses the limitations of existing nuclear power plant fault diagnosis technologies, such as the limited ability of proportional-integral-derivative adjustment technology to handle complex and abnormal operating conditions, its inability to adapt to the multidimensional uncertainty of faults, the exponential increase in equipment complexity and maintenance costs due to redundant hardware configurations, the inability of personnel training to cover all fault modes, and the susceptibility to misjudgments. Furthermore, conventional neural networks suffer from insufficient modeling accuracy and weak generalization ability due to the highly nonlinear characteristics of nuclear energy system parameter coupling.

[0007] The present invention adopts the following technical solution: A fault diagnosis method for nuclear power plants based on CNN and sliding window verification includes the following steps: S1. Collect sensor and actuator signals of the nuclear power plant under normal and fault conditions; S2. Discretely sample, preprocess and normalize the raw signal collected in step S1, and convert the processed data into 3-channel color images as training set; S3. Construct a multi-layer CNN network, which includes an input layer, a 2D convolutional layer, a batch normalization layer, a ReLU layer, a global average pooling layer, a Dropout layer, a fully connected layer, a Softmax layer, and a classification layer. S4. Input the training set obtained in step S2 into the multi-layer CNN neural network constructed in step S3 for training to obtain the fault diagnosis model. S5. Collect sensor signals and actuator signals from the real-time operation of the nuclear power plant. After preprocessing and normalization as described in step S2, convert them into 3-channel color images as a test set. Use a sliding time window to sequentially input the test set into the fault diagnosis model trained in step S4 in batches to obtain the validation set labels for each batch. S6. Calculate the score of each validation set label within each sliding time window, select the label with the highest score as the fault type at the last time point of that window, and realize real-time fault diagnosis of nuclear power plant.

[0008] Preferably, in step S1, the sensor signals include at least one of the following: reactor power setpoint, actual reactor output power, coolant inlet temperature, coolant outlet temperature, steam generator feedwater flow rate, steam generator steam flow rate, steam pressure, turbine inlet steam flow rate, control rod speed, and feedwater valve opening.

[0009] Preferably, the fault conditions include constant deviation, constant gain and jamming fault types, wherein constant deviation and constant gain faults correspond to different fault degrees, totaling 23 fault conditions and one normal condition.

[0010] Preferably, in step S2, the preprocessing and normalization specifically include: The original time series data is discretely sampled, and each sampled value is normalized. The normalization formula is as follows:

[0011] in, This represents the normalized value; For input values; These are the minimum and maximum values ​​in each row of data, respectively.

[0012] Preferably, the 3-channel color image is an RGB image, and the generation method includes: Three grayscale images are input into the R, G, and B channels of an RGB image, respectively, with each grayscale image obtained by converting normalized data.

[0013] Preferably, in step S3, the 2D convolutional layer uses multiple convolutional kernels for feature extraction, and the kernel sizes include 32, 64, 64, and 128. The convolution operation formula for the 2D convolutional layer is:

[0014] in, It is the input feature map. It is a convolution kernel (filter). It is the position index of the output feature map. and It is the size of the convolution kernel. Indicates the convolution operation; The random dropout probability of the Dropout layer is 0.5; the fully connected layer outputs 24 elements, which correspond to 23 fault conditions and 1 normal operating state. The classification layer optimizes the model by calculating cross-entropy loss, the formula for which is:

[0015] in, The number of samples; Number of classes; For the first Class weight; For the first The sample belongs to the first Class indicators; For the first The sample belongs to the first The probability of a class.

[0016] Preferably, in step S4, the training parameters of the multi-layer CNN neural network are set as follows: The maximum number of iterations is 300, and the number of signals sampled in each type of training sample is no less than 3000; the operation formula of the global average pooling layer is:

[0017] in, It is the input feature map. It outputs the feature map. It is the size of the pooling window. , It outputs the feature map location index. It is a channel index.

[0018] Preferably, in step S5, the sliding time window has a window length of 25 images, a sliding step of 1 image, a corresponding time window length of 2.5 seconds, and a sliding step of 0.1 seconds.

[0019] Preferably, in step S6, calculating the score of each validation set label within each sliding time window specifically involves: For each window, the number of times each fault label appears within the window is calculated and divided by the window length to obtain the score for each fault category.

[0020] Secondly, embodiments of the present invention provide a nuclear power plant fault diagnosis system based on CNN and sliding window verification, comprising: The data acquisition module is used to collect sensor and actuator signals of the nuclear power plant under normal and fault conditions. The preprocessing module is used to perform discrete sampling, preprocessing, and normalization on the acquired raw signals, and convert the processed data into 3-channel color images as a training set. The CNN building module is used to build a multi-layer CNN network, which includes an input layer, a 2D convolutional layer, a batch normalization layer, a ReLU layer, a global average pooling layer, a Dropout layer, a fully connected layer, a Softmax layer, and a classification layer. The model training module is used to train the multi-layer CNN neural network constructed from the training set input to obtain the fault diagnosis model; The test set processing module is used to collect sensor signals and actuator signals from the real-time operation of the nuclear power plant. After preprocessing and normalization, the signals are converted into 3-channel color images as the test set. The test set is sequentially input into the trained fault diagnosis model in batches using a sliding time window to obtain the validation set labels for each batch. The fault diagnosis module is used to calculate the score of each validation set label within each sliding time window, select the label with the highest score as the fault type at the last time point of that window, and realize real-time fault diagnosis of nuclear power plants.

[0021] Thirdly, a computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the above-described method for diagnosing nuclear power plant faults based on CNN and sliding window verification.

[0022] Fourthly, embodiments of the present invention provide a computer-readable storage medium including a computer program, which, when executed by a processor, implements the steps of the above-described method for diagnosing nuclear power plant faults based on CNN and sliding window verification.

[0023] Fifthly, a chip includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the above-described method for diagnosing nuclear power plant faults based on CNN and sliding window verification.

[0024] In a sixth aspect, embodiments of the present invention provide an electronic device including a computer program, which, when executed by the electronic device, implements the steps of the above-described method for diagnosing nuclear power plant faults based on CNN and sliding window verification.

[0025] Compared with the prior art, the present invention has at least the following beneficial effects: A fault diagnosis method for nuclear power plants based on CNN and sliding window verification is proposed. This method achieves real-time fault diagnosis of nuclear power plants by combining a CNN network with sliding window verification. Time-series signals are converted into images, and the powerful feature extraction capabilities of CNN are utilized to identify complex fault modes, solving the problem of insufficient identification of nonlinear fault features in traditional methods. The sliding window verification mechanism effectively improves the stability and reliability of the diagnostic results by performing statistical decision-making on the diagnostic results over continuous time periods, avoiding random errors in single-point diagnosis.

[0026] Furthermore, the key sensor signal types are clearly listed, covering the core operating parameters of nuclear power plants. Through multi-parameter fusion analysis, comprehensive system state changes can be captured, improving fault detection coverage. Compared to single-parameter monitoring, multi-signal collaborative diagnosis can effectively distinguish similar fault symptoms and reduce false alarm rates.

[0027] Furthermore, refined fault classification enables the diagnostic system to accurately identify the nature and severity of faults, providing detailed information for subsequent maintenance decisions. Coverage of 23 fault conditions ensures the system's comprehensive diagnostic capabilities for common faults.

[0028] Furthermore, normalization eliminates the influence of signals with different dimensions, improving training stability; image transformation converts a temporal problem into a spatial recognition problem, fully leveraging the advantages of CNNs in image recognition. This data preprocessing method significantly improves the model's convergence speed and generalization ability.

[0029] Furthermore, the RGB three-channel image generation method enhances the richness of feature representation by carrying complementary information in different channels. Compared to single-channel images, the three-channel structure can retain more original information and improve the model's sensitivity to subtle fault features.

[0030] Furthermore, the multi-layer convolutional structure enables layer-by-layer extraction from local to global features; the Dropout mechanism prevents overfitting; and the cross-entropy loss function ensures the correct direction of model optimization. This network design improves training efficiency while maintaining accuracy.

[0031] Furthermore, 300 iterations and 3000 samples ensured sufficient training; global average pooling reduced the number of parameters and improved the model's generalization ability. These parameters were optimized and validated to achieve the best balance between training efficiency and diagnostic accuracy.

[0032] Furthermore, the 2.5-second window length balances real-time performance and statistical significance; the 0.1-second step size ensures the continuity of diagnosis. This timescale design satisfies both real-time monitoring requirements and guarantees the reliability of diagnostic results.

[0033] Furthermore, a frequency-based scoring method is employed, and the final fault type is determined through a majority voting mechanism. This method is simple and effective, can smooth out momentary misjudgments, and improve the stability of diagnostic results.

[0034] It is understood that the beneficial effects of the second to sixth aspects mentioned above can be found in the relevant descriptions in the first aspect mentioned above, and will not be repeated here.

[0035] In summary, this invention achieves accurate and real-time diagnosis of nuclear power plant faults by combining CNN deep feature extraction and sliding window statistical verification. It overcomes the shortcomings of traditional methods in identifying complex nonlinear faults and significantly improves the accuracy, stability, and practicality of fault diagnosis.

[0036] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0037] Figure 1 A schematic diagram of the MATLAB / Simulink simulation platform for the nuclear power plant in a nuclear steam supply system; Figure 2 A diagram illustrating the fault label index for fault information; Figure 3 The image shows the working condition obtained from the fault data at the time of the fault, where (a) is CD_+0.3_ph, (b) is CG_+1.2_cf, (c) is ST_ph, and (d) is ST_cf. Figure 4 A schematic diagram illustrating the accuracy cross-validation of the trained network using the training set. Figure 5 The following are real-time score graphs for each label obtained from batch verification of the test set: (a) is the real-time score graph for normal operating condition (NO), (b) is the real-time score graph for constant deviation fault of steam pressure sensor (CD_+0.3_ph), (c) is the real-time score graph for constant gain fault of steam pressure sensor (CG_+1.2_ph), (d) is the real-time score graph for stuck fault of steam pressure sensor (ST_ph), (e) is the real-time score graph for constant deviation fault of feedwater valve opening actuator (CD_+0.3_cf), (f) is the real-time score graph for constant gain fault of feedwater valve opening actuator (CG_+1.2_cf), (g) is the real-time score graph for stuck fault of feedwater valve opening actuator (ST_cf), and (h) is the real-time score graph for specific fault conditions, such as CG_-0.8_ph. Figure 6 A schematic diagram of a computer device provided in an embodiment of the present invention; Figure 7 This is a block diagram of a chip according to an embodiment of the present invention; Figure 8 This is a flowchart of the method of the present invention.

[0038] Among them, 60. Computer equipment; 61. Processor; 62. Memory; 63. Computer program; 600. Electronic device; 610. Processing unit; 620. Storage unit; 6201. Random access memory unit; 6202. Cache memory unit; 6203. Read-only memory unit; 6204. Program / utility; 6205. Program module; 630. Bus; 640. Display unit; 650. Input / output interface; 660. Network adapter; 700. External device. Detailed Implementation

[0039] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0040] In the description of this invention, it should be understood that the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0041] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0042] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes such combinations. For example, A and / or B can represent three cases: A alone, A and B simultaneously, and B alone. Additionally, the character " / " in this invention generally indicates that the preceding and following objects have an "or" relationship.

[0043] It should be understood that although terms such as first, second, third, etc., may be used in the embodiments of the present invention to describe the preset range, these preset ranges should not be limited to these terms. These terms are only used to distinguish the preset ranges from one another. For example, without departing from the scope of the embodiments of the present invention, the first preset range may also be referred to as the second preset range, and similarly, the second preset range may also be referred to as the first preset range.

[0044] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0045] The accompanying drawings illustrate various structural schematic diagrams according to embodiments disclosed in this invention. These drawings are not to scale, and some details have been enlarged for clarity, and some details may have been omitted. The shapes of the various regions and layers shown in the drawings, as well as their relative sizes and positional relationships, are merely exemplary and may deviate from reality due to manufacturing tolerances or technical limitations. Furthermore, those skilled in the art can design regions / layers with different shapes, sizes, and relative positions as needed.

[0046] This invention provides a fault diagnosis method for nuclear power plants based on CNN and sliding window verification, covering different types and degrees of faults such as constant gain, constant deviation, and jamming in actuators and sensors. Utilizing existing nuclear steam supply system simulation platforms and data observed in actual systems, multiple signals are collected. A multi-layered CNN neural network is established and trained on historical data of these signals to accurately predict the type and severity of historical faults occurring during nuclear power plant operation. A dynamic sliding sampling window is used to extract online monitoring data, and a weighted voting algorithm is used to obtain the confidence distribution of each fault category, thus determining the fault category at each time point and achieving real-time fault diagnosis. This invention significantly improves the accuracy and efficiency of diagnosing different types and degrees of faults in nuclear power plants.

[0047] Please see Figure 8 This invention discloses a fault diagnosis method for nuclear power plants based on CNN and sliding window verification, comprising the following steps: S1. Collect sensor and actuator signals from each control system of the reactor primary loop under normal operating conditions. Introduce faults at different locations according to the fault type and degree to be classified, and collect sensor signals from the reactor primary loop pressure control system, liquid level control system and power control system. Please see Figure 1The figure shows the modeling and simulation boundary of the nuclear steam supply system upon which this invention is based, including the aforementioned control system and the process system shown in the figure. This allows for a relatively complete and comprehensive simulation of the primary loop system of a nuclear power plant, meeting the verification requirements of this invention. Based on practical engineering considerations, existing sensor signals from the nuclear power plant are selected for collection, such as reactor power setpoint and actual output power, coolant inlet and outlet temperatures, steam generator feedwater flow rate, steam flow rate and steam pressure, turbine inlet steam flow rate, control rod speed, and feedwater valve opening.

[0048] The steam pressure sensor and feedwater valve opening actuator, which are relatively prone to failure in actual engineering, were selected as the fault points. Twelve signals were selected for acquisition, including the reactor power setpoint (POWER) and measured value (P), core control rod speed (vrod), coolant inlet / outlet temperature (Tcold, Thot), steam flow rate (SF), feedwater flow rate (FF), steam pressure (SP), feedwater valve opening and pressure drop (Cfv, DP), and turbine flow rate (TF). Considering the data volume, the sampling time was set to 0.1 seconds. The fault types considered in this example include common sensor and actuator fault types, namely constant deviation, constant gain, and jamming. Including different degrees of deviation and gain, a total of 23 fault conditions were identified.

[0049] Please see Figure 2 Fault information is mapped using labels as shown in the figure. No fault is uniformly mapped as "NO". The specific format for mapping deviation and gain conditions is "CD / CG_fault severity_fault location". The stuck condition is mapped as "ST_fault location". The fault location label corresponding to the steam pressure sensor fault is "Ph" and the fault location label corresponding to the feedwater valve opening actuator fault is "Cf".

[0050] S2. Based on the fault type and degree to be classified, the original time series data of the sensor signals collected in steps S1 and S2 are discretely sampled, preprocessed and normalized, and the data is converted into 3-channel color images as training sets. The system collects sensor and actuator signals under normal and fault conditions in the primary loop of the nuclear reactor and performs normalization processing on the data signals. The collected signals mainly include power signals, steam pressure signals, and feedwater valve opening signals. The normalization process specifically involves:

[0051] in, This represents the normalized value; For input values; These are the minimum and maximum values ​​in each row of data, respectively.

[0052] Please see Figure 3 These are the grayscale and RGB color images corresponding to some of the working conditions obtained through conversion.

[0053] Specifically, the normalized data is converted into integer values ​​in the range [0, 255], suitable for grayscale image representation. Next, images are generated. Each image input consists of 12×12 data points; that is, the first image uses the data from the first to the twelfth columns (12 rows), the second image uses the data from the second to the thirteenth columns (12 rows), and so on. Based on the additive mixing principle of the RGB color model, the three grayscale images are input into the R, G, and B channels of the RGB image respectively, thus obtaining an RGB color image. Specifically, the additive mixing process of the RGB color model is as follows:

[0054] in, , , These are three grayscale images at the pixel level. The grayscale value.

[0055] S3. Construct a multi-layer CNN network as a neural network training model for fault data of sensors and actuators in nuclear power plants; Establish a multi-layer CNN network. Specifically, a multi-layer CNN network includes an input layer, 2D convolutional layers, batch normalization layers, ReLU layers, global average pooling layers, Dropout layers, fully connected layers, softmax layers, and a classification layer.

[0056] Furthermore, the input layer receives 12×12 color image training samples from three channels.

[0057] 2D convolutional layers: These extract features by sliding a series of filters (or kernels) across the input image, calculating a weighted sum of local regions, and generating a feature map. Kernel sizes are typically 32, 64, 64, and 128. The convolution operation is as follows:

[0058] in, It is the input feature map. It is a convolution kernel (filter). It is the position index of the output feature map. and It is the size of the convolution kernel. This indicates a convolution operation.

[0059] Batch normalization layer: Normalizes the output of the previous layer, that is, normalizes the output of each feature channel.

[0060] ReLU layer: Performs a threshold operation on each element of the input, setting all values ​​less than zero to zero, thus alleviating the overfitting problem of the network.

[0061] Global average pooling layer: It performs average pooling on the entire feature map to extract global feature representations, specifically as follows:

[0062] in, It is the input feature map. It outputs the feature map. It is the size of the pooling window. , It outputs the feature map location index. It is a channel index.

[0063] Dropout layer: Given a probability of 0.5, each element of the input is randomly multiplied by a Bernoulli distribution value to prevent neurons from developing overly complex dependencies.

[0064] Fully connected layer: Takes the output of the front layer as input value and finally outputs 24 elements, representing 23 types of sensor and actuator faults and one normal operating state.

[0065] Softmax layer: Ensures that the output value is between 0 and 1, and that the sum of all output values ​​is 1, forming an effective probability distribution.

[0066] Classification layer: Calculates the error between the model output and the true label, and calculates the cross-entropy loss, specifically expressed as:

[0067] in, The number of samples; Number of classes; For the first Class weight; For the first The sample belongs to the first Class indicators; For the first The sample belongs to the first The probability of a class.

[0068] S4. Put the training sample image set obtained in step S2 into the multi-layer CNN network constructed in step S3 for neural network training. Please see Figure 4During network training, the maximum number of iterations was set to 300, and the number of signal samples for each type of training sample was greater than 3000. The aforementioned historical training data was then fed into the CNN network for training. Since neural networks need to continuously extract features from a large number of samples to learn and achieve ideal classification results, it is recommended that the number of training samples for each type be more than 3000. Multiple samples can be used to set different fault occurrence times, different types and degrees of sensor and actuator faults to enrich the diversity of the samples and broaden the applicability of the network.

[0069] S5. Re-collect sensor signals of different fault types, degrees and occurrence times in each system. After preprocessing and normalization, the sampled sensor signals are converted into 3-channel color images as test sample image sets. The test set images are put into the CNN network trained in step S4 in batches using a sliding time window to obtain the validation set labels of the test sample set. The test set images are grouped into sets of 25, with a time window length of 25 (corresponding to 2.5s) and a window sliding step of 1 (corresponding to 0.1s). The test set is validated in batches.

[0070] Specifically, scores are calculated for each set of validation labels obtained from the network. The scoring method is: the number of times each label appears within the window length is divided by the window length, and finally, the label with the highest score in each window is taken as the fault type at the last time point of each window.

[0071] S6. Calculate the score for each group using the validation set labels obtained in step S5. Take the label with the highest score in each window as the fault type at the last time point of each window, and finally realize fault diagnosis.

[0072] Real-time score graphs are plotted for each label score obtained in batches, such as... Figure 5 As shown.

[0073] Specifically, based on the real-time score graph, the probability of each fault at each time point can be determined. The system can select the best and second-best scores to assign fault types, issue warnings and indicate fault probabilities, and allow relevant personnel to determine the actual type of fault and take corresponding control strategies.

[0074] In another embodiment of the present invention, a nuclear power plant fault diagnosis system based on CNN and sliding window verification is provided. This system can be used to implement the above-mentioned nuclear power plant fault diagnosis method based on CNN and sliding window verification. Specifically, the nuclear power plant fault diagnosis system based on CNN and sliding window verification includes a data acquisition module, a preprocessing module, a CNN construction module, a model training module, a test set processing module, and a fault diagnosis module.

[0075] The data acquisition module is used to collect sensor signals and actuator signals of the nuclear power plant under normal and fault conditions. The preprocessing module is used to perform discrete sampling, preprocessing, and normalization on the acquired raw signals, and convert the processed data into 3-channel color images as a training set. The CNN building module is used to build a multi-layer CNN network, which includes an input layer, a 2D convolutional layer, a batch normalization layer, a ReLU layer, a global average pooling layer, a Dropout layer, a fully connected layer, a Softmax layer, and a classification layer. The model training module is used to train the multi-layer CNN neural network constructed from the training set input to obtain the fault diagnosis model; The test set processing module is used to collect sensor signals and actuator signals from the real-time operation of the nuclear power plant. After preprocessing and normalization, the signals are converted into 3-channel color images as the test set. The test set is sequentially input into the trained fault diagnosis model in batches using a sliding time window to obtain the validation set labels for each batch. The fault diagnosis module is used to calculate the score of each validation set label within each sliding time window, select the label with the highest score as the fault type at the last time point of that window, and realize real-time fault diagnosis of nuclear power plants.

[0076] This invention provides a terminal device comprising a processor and a memory. The memory stores a computer program, which includes program instructions. The processor executes the program instructions stored in the computer storage medium. The processor may be a Central Processing Unit (CPU), or other general-purpose processors, graphics processing units (GPUs), tensor processing units (TPUs), digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. It is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions to achieve a corresponding method flow or corresponding function. The processor described in this embodiment can be used in the operation of a nuclear power plant fault diagnosis method based on CNN and sliding window verification, including: Sensor and actuator signals from the nuclear power plant under normal and fault conditions are collected. The raw signals are discretely sampled, preprocessed, and normalized, and the processed data is converted into 3-channel color images as the training set. A multi-layer CNN network is constructed, comprising an input layer, 2D convolutional layers, batch normalization layers, ReLU layers, global average pooling layers, Dropout layers, fully connected layers, Softmax layers, and a classification layer. The training set is input into the constructed multi-layer CNN network for training to obtain a fault diagnosis model. Sensor and actuator signals from the nuclear power plant during real-time operation are collected, preprocessed, and normalized, then converted into 3-channel color images as the test set. A sliding time window is used to sequentially input the test set into the trained fault diagnosis model in batches, obtaining validation set labels for each batch. The score of each validation set label within each sliding time window is calculated, and the label with the highest score is selected as the fault type at the last time point of that window, achieving real-time fault diagnosis of the nuclear power plant.

[0077] Please see Figure 6 The terminal device is a computer device. In this embodiment, the computer device 60 includes a processor 61, a memory 62, and a computer program 63 stored in the memory 62 and executable on the processor 61. When executed by the processor 61, the computer program 63 implements the nuclear power plant fault diagnosis method based on CNN and sliding window verification as described in this embodiment. To avoid repetition, details are omitted here. Alternatively, when executed by the processor 61, the computer program 63 implements the functions of each model / unit in the nuclear power plant fault diagnosis system based on CNN and sliding window verification as described in this embodiment. To avoid repetition, details are omitted here.

[0078] Computer device 60 can be a desktop computer, laptop, handheld computer, cloud server, or other computing device. Computer device 60 may include, but is not limited to, a processor 61 and a memory 62. Those skilled in the art will understand that... Figure 6 This is merely an example of computer device 60 and does not constitute a limitation on computer device 60. It may include more or fewer components than shown, or combine certain components, or different components. For example, computer device may also include input / output devices, network access devices, buses, etc.

[0079] The processor 61 may be a Central Processing Unit (CPU), or other general-purpose processors, graphics processing units (GPUs), tensor processing units (TPUs), digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor.

[0080] The memory 62 can be an internal storage unit of the computer device 60, such as a hard disk or RAM of the computer device 60. The memory 62 can also be an external storage device of the computer device 60, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., provided on the computer device 60.

[0081] Furthermore, the memory 62 may include both internal storage units of the computer device 60 and external storage devices. The memory 62 is used to store computer programs and other programs and data required by the computer device. The memory 62 can also be used to temporarily store data that has been output or will be output.

[0082] Please see Figure 7 The terminal device is an electronic device 600, which is manifested in the form of a general-purpose computing device. The components of the electronic device may include, but are not limited to: at least one processing unit 610, at least one storage unit 620, a bus 630 connecting different platform components (including storage unit 620 and processing unit 610), a display unit 640, etc.

[0083] The storage unit stores program code, which can be executed by the processing unit 610 to perform the steps described in the method section of this specification according to various exemplary embodiments of the present invention. For example, the processing unit 610 can perform actions such as... Figure 1 The steps are shown in the figure.

[0084] Storage unit 620 may include a readable medium in the form of a volatile storage unit, such as random access memory (RAM) 6201 and / or cache memory 6202, and may further include a read-only memory (ROM) 6203.

[0085] Storage unit 620 may also include a program / utility 6204 having a set (at least one) program module 6205, such program module 6205 including but not limited to: operating system, one or more application programs, other program modules and program data, each or some combination of these examples may include an implementation of a network environment.

[0086] Bus 630 can represent one or more of several types of bus structures, including a memory cell bus or memory cell controller, a peripheral bus, a graphics acceleration port, a processing unit, or a local bus using any of the multiple bus structures.

[0087] Electronic device 600 can also communicate with one or more external devices 700 (e.g., keyboard, pointing device, Bluetooth device, etc.), and with one or more devices that enable a user to interact with electronic device 600, and / or with any device that enables electronic device 600 to communicate with one or more other computing devices (e.g., router, modem). This communication can be performed via input / output interface 650. Furthermore, electronic device 600 can also communicate with one or more networks (e.g., local area network, wide area network, and / or public network, such as the Internet) via network adapter 660. Network adapter 660 can communicate with other modules of electronic device 600 via bus 630. It should be understood that, although not shown in the figures, other hardware and / or software modules can be used in conjunction with electronic device 600, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage platforms.

[0088] Example 4 This invention also provides a storage medium, specifically a computer-readable storage medium, which is a memory device in a terminal device for storing programs and data. It is understood that the computer-readable storage medium here can include both built-in storage media in the terminal device and extended storage media supported by the terminal device; it can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. The computer-readable storage medium provides storage space that stores the terminal's operating system. Furthermore, the storage space also stores one or more instructions suitable for loading and execution by a processor, which can be one or more computer programs (including program code). More specific examples of the computer-readable storage medium include: an electrical connection with one or more wires, a portable disk, a hard disk, random access memory, read-only memory, erasable programmable read-only memory, optical fiber, portable compact disk read-only memory, optical storage device, magnetic storage device, or any suitable combination thereof.

[0089] Computer-readable storage media also include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A readable storage medium can also be any readable medium other than a readable storage medium that can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the readable storage medium can be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, radio frequency, etc., or any suitable combination thereof.

[0090] Program code for performing the operations of this invention can be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, and conventional procedural programming languages ​​such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device (e.g., via the Internet using an Internet service provider).

[0091] One or more instructions stored in a computer-readable storage medium can be loaded and executed by a processor to implement the corresponding steps of the nuclear power plant fault diagnosis method based on CNN and sliding window verification in the above embodiments; one or more instructions in the computer-readable storage medium are loaded and executed by the processor to perform the following steps: Sensor and actuator signals from the nuclear power plant under normal and fault conditions are collected. The raw signals are discretely sampled, preprocessed, and normalized, and the processed data is converted into 3-channel color images as the training set. A multi-layer CNN network is constructed, comprising an input layer, 2D convolutional layers, batch normalization layers, ReLU layers, global average pooling layers, Dropout layers, fully connected layers, Softmax layers, and a classification layer. The training set is input into the constructed multi-layer CNN network for training to obtain a fault diagnosis model. Sensor and actuator signals from the nuclear power plant during real-time operation are collected, preprocessed, and normalized, then converted into 3-channel color images as the test set. A sliding time window is used to sequentially input the test set into the trained fault diagnosis model in batches, obtaining validation set labels for each batch. The score of each validation set label within each sliding time window is calculated, and the label with the highest score is selected as the fault type at the last time point of that window, achieving real-time fault diagnosis of the nuclear power plant.

[0092] The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0093] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0094] 1. Simulation Experiment Conditions A simulation platform for a nuclear steam supply system was built using MATLAB / Simulink to simulate the operation of the pressure, level, and power control systems of the primary loop of a nuclear power plant. Twelve types of signals, including reactor power setpoint and coolant inlet and outlet temperatures, were collected with a sampling time of 0.1 seconds. Twenty-three fault conditions were set (constant deviation ±0.3 / ±0.5 / ±0.7MPa, constant gain 0.7 / 0.8 / 0.9 / 1.1 / 1.2 / 1.3 times, and jamming fault, covering steam pressure sensors and feedwater valve actuators). For each condition, 3500 training samples and 800 test samples were collected.

[0095] 2. Experimental Data Results Model training: The maximum number of iterations was 300, the accuracy of the training set reached 98.5%, and the loss value converged to 0.08; the accuracy of the test set reached 96.2%, which is 14.2 percentage points higher than that of the conventional BP neural network (82%).

[0096] Real-time diagnostic performance: The sliding window has a length of 25 images (2.5s) and a step size of 1 image (0.1s). The time taken for a single batch of diagnostics is 0.03 seconds, and the total response time is ≤0.13 seconds, which meets the real-time monitoring requirements of nuclear power plants.

[0097] Fault identification accuracy: 97.1% accuracy for constant deviation faults, 95.8% accuracy for constant gain faults, 98.3% accuracy for jammed faults, 99.5% accuracy for normal operating conditions, and a false positive rate of only 1.8%.

[0098] Cost Comparison: Compared to hardware redundancy solutions, system deployment costs are reduced by 62%, and total lifecycle maintenance costs are reduced by 58%; compared to manual diagnostics, the false positive rate is reduced from 15.3% to 1.8%.

[0099] In summary, this invention provides a nuclear power plant fault diagnosis method and system based on CNN and sliding window verification. Relying on an established simulation platform for the nuclear steam supply system of a nuclear power plant and observable data from the actual system, multiple signals are selected for data acquisition. By establishing a multi-layer CNN neural network and training it on historical data of the aforementioned signals, the method accurately predicts the type and severity of historical faults occurring during the operation of the nuclear power plant. Furthermore, the sliding window method for calculating fault type scores based on this invention can accurately diagnose the type and severity of faults in real-time operating nuclear power plants.

[0100] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0101] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0102] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed in this invention can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software 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 implementations should not be considered beyond the scope of this invention.

[0103] In the embodiments provided by this invention, it should be understood that the disclosed devices / terminals and methods can be implemented in other ways. For example, the device / terminal embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0104] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0105] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0106] If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random-access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc. It should be noted that the content included in the computer-readable medium can be appropriately added or removed according to the requirements of legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, computer-readable media do not include electrical carrier signals and telecommunication signals.

[0107] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus, and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0108] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0109] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0110] The above content is only for illustrating the technical concept of the present invention and should not be construed as limiting the scope of protection of the present invention. Any modifications made to the technical solution based on the technical concept proposed in this invention shall fall within the scope of protection of the claims of this invention.

Claims

1. A method for fault diagnosis of nuclear power plants based on CNN and sliding window verification, characterized in that, Includes the following steps: S1. Collect sensor and actuator signals of the nuclear power plant under normal and fault conditions; S2. Discretely sample, preprocess and normalize the raw signal collected in step S1, and convert the processed data into 3-channel color images as training set; S3. Construct a multi-layer CNN network, which includes an input layer, a 2D convolutional layer, a batch normalization layer, a ReLU layer, a global average pooling layer, a Dropout layer, a fully connected layer, a Softmax layer, and a classification layer. S4. Input the training set obtained in step S2 into the multi-layer CNN neural network constructed in step S3 for training to obtain the fault diagnosis model. S5. Collect sensor signals and actuator signals from the nuclear power plant during real-time operation, and convert them into 3-channel color images after preprocessing and normalization as described in step S2, which will be used as a test set. A sliding time window is used to sequentially input the test set into the fault diagnosis model trained in step S4 in batches to obtain the validation set labels for each batch. S6. Calculate the score of each validation set label within each sliding time window, select the label with the highest score as the fault type at the last time point of that window, and realize real-time fault diagnosis of nuclear power plant.

2. The nuclear power plant fault diagnosis method based on CNN and sliding window verification according to claim 1, characterized in that, In step S1, the sensor signals include at least one of the following: reactor power setpoint, actual reactor output power, coolant inlet temperature, coolant outlet temperature, steam generator feedwater flow rate, steam generator steam flow rate, steam pressure, turbine inlet steam flow rate, control rod speed, and feedwater valve opening.

3. The nuclear power plant fault diagnosis method based on CNN and sliding window verification according to claim 2, characterized in that, The fault conditions include constant deviation, constant gain and jamming fault types, with constant deviation and constant gain faults corresponding to different fault degrees, totaling 23 fault conditions and one normal condition.

4. The nuclear power plant fault diagnosis method based on CNN and sliding window verification according to claim 1, characterized in that, In step S2, the preprocessing and normalization specifically include: The original time series data is discretely sampled, and each sampled value is normalized. The normalization formula is as follows: in, This represents the normalized value; For input values; These are the minimum and maximum values ​​in each row of data, respectively.

5. The nuclear power plant fault diagnosis method based on CNN and sliding window verification according to claim 4, characterized in that, The 3-channel color image is an RGB image, and the generation method includes: Three grayscale images are input into the R, G, and B channels of an RGB image, respectively, with each grayscale image obtained by converting normalized data.

6. The nuclear power plant fault diagnosis method based on CNN and sliding window verification according to claim 1, characterized in that, In step S3, the 2D convolutional layer uses multiple convolutional kernels for feature extraction, and the kernel sizes include 32, 64, 64, and 128. The convolution operation formula for the 2D convolutional layer is: in, It is the input feature map. It is a convolution kernel (filter). It is the position index of the output feature map. and It is the size of the convolution kernel. Indicates the convolution operation; The random dropout probability of the Dropout layer is 0.5; the fully connected layer outputs 24 elements, which correspond to 23 fault conditions and 1 normal operating state. The classification layer optimizes the model by calculating cross-entropy loss, the formula for which is: in, The number of samples; Number of classes; For the first Class weight; For the first The sample belongs to the first Class indicators; For the first The sample belongs to the first The probability of a class.

7. The nuclear power plant fault diagnosis method based on CNN and sliding window verification according to claim 1, characterized in that, In step S4, the training parameters of the multi-layer CNN neural network are set as follows: The maximum number of iterations is 300, and the number of signals sampled in each type of training sample is no less than 3000; the operation formula of the global average pooling layer is: in, It is the input feature map. It outputs the feature map. It is the size of the pooling window. , It outputs the feature map location index. It is a channel index.

8. The nuclear power plant fault diagnosis method based on CNN and sliding window verification according to claim 1, characterized in that, In step S5, the sliding time window has a window length of 25 images and a sliding step of 1 image, corresponding to a time window length of 2.5 seconds and a sliding step of 0.1 seconds.

9. The nuclear power plant fault diagnosis method based on CNN and sliding window verification according to claim 1, characterized in that, In step S6, the calculation of the score for each validation set label within each sliding time window specifically involves: For each window, the number of times each fault label appears within the window is calculated and divided by the window length to obtain the score for each fault category.

10. A fault diagnosis system for nuclear power plants based on CNN and sliding window verification, characterized in that, include: The data acquisition module is used to collect sensor and actuator signals of the nuclear power plant under normal and fault conditions. The preprocessing module is used to perform discrete sampling, preprocessing, and normalization on the acquired raw signals, and convert the processed data into 3-channel color images as a training set. The CNN building module is used to build a multi-layer CNN network, which includes an input layer, a 2D convolutional layer, a batch normalization layer, a ReLU layer, a global average pooling layer, a Dropout layer, a fully connected layer, a Softmax layer, and a classification layer. The model training module is used to train the multi-layer CNN neural network constructed from the training set input to obtain the fault diagnosis model; The test set processing module is used to collect sensor signals and actuator signals from the nuclear power plant in real time. After preprocessing and normalization, the signals are converted into 3-channel color images as the test set. A sliding time window is used to sequentially input the test set into the trained fault diagnosis model in batches to obtain the validation set labels for each batch. The fault diagnosis module is used to calculate the score of each validation set label within each sliding time window, select the label with the highest score as the fault type at the last time point of that window, and realize real-time fault diagnosis of nuclear power plants.

Images