Drum filter screen condition monitoring system, method and apparatus for nuclear power plants

Abstract

The application provides a drum filter screen state monitoring system, method and equipment for a nuclear power plant, the system comprising: a sensor module for monitoring the running state of the drum filter screen in real time; a data acquisition module connected to the sensor module for receiving real-time data from the sensor module; a data analysis module connected to the data acquisition module for receiving preprocessed data, normalizing the preprocessed data and extracting features to form to-be-detected data; a fault detection module for calculating the anomaly score of the to-be-detected data according to the isolation forest algorithm, comparing the anomaly score with a preset threshold, and generating abnormal data when the anomaly score is higher than the preset threshold, wherein the abnormal data comprises the anomaly score and a fault type; and an alarm module for sending an alarm signal when the abnormal data is received, which can realize accurate detection and prediction of the filter screen state, send an early warning signal before a fault occurs, and significantly improve the safety and reliability of the operation of the nuclear power plant equipment.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of nuclear power plant cold source equipment monitoring, and in particular to a drum filter screen state monitoring system, method and device for a nuclear power plant. BACKGROUND

[0002] The drum filter screen plays a crucial role in the cold source system of a nuclear power plant. The cold source system is used to provide cooling water for the nuclear reactor, ensuring that the nuclear reactor operates within a safe temperature range and preventing risks such as overheating. The cooling water usually contains a large amount of impurities, including plant residues, silt, fish and shrimp biological substances in the water, especially in complex water sources or in environments contaminated by external factors. These impurities can affect the normal operation of the cooling system. The drum filter screen effectively filters these impurities to ensure the purity of the cooling water, thereby ensuring the safety and stability of the cooling system of the nuclear power plant. However, over time, the drum filter screen is affected by various factors such as physical wear, corrosion, and scaling during use, leading to a gradual decline in performance and even failure. Especially in complex working environments, the failure of the filter screen is usually difficult to predict and detect, greatly affecting the safety of the nuclear power plant.

[0003] Currently, the drum filter screen monitoring method in the related art mainly relies on manual inspection and regular maintenance, and real-time monitoring and failure prediction cannot be achieved. Although obvious failures (such as overload, abnormal vibration, etc.) can be detected by manual operation, it is difficult to discover potential failures in a timely manner due to its reliance on manual judgment. SUMMARY

[0004] Therefore, it is necessary to provide a drum filter screen state monitoring system, method and device for a nuclear power plant, which can overcome at least one of the above defects.

[0005] In a first aspect, the embodiments of the present application provide a drum filter screen state monitoring system for a nuclear power plant, which is applied to monitor the state of the drum filter screen, and the system comprises:

[0006] a sensor module for real-time monitoring of the operating state of the drum filter screen, the sensor module comprising: an axial laser sensor for monitoring the axial displacement and deformation of the drum filter screen; a radial laser sensor for monitoring the radial displacement and tooth ring runout of the drum filter screen; an axial eddy current sensor and a radial eddy current sensor for monitoring the bearing operating state and displacement change of the drum filter screen; a left bearing liquid level sensor and a right bearing liquid level sensor for monitoring the change in lubricating oil level in the main shaft oil tank of the drum filter screen;

[0007] a data acquisition module for receiving real-time data from the sensor module and sampling and preprocessing the real-time data to form preprocessed data;

[0008] The data analysis module is used to receive the preprocessed data, and to normalize and extract features from the preprocessed data to form the data to be detected.

[0009] The fault detection module is used to calculate the anomaly score of the data to be detected according to the isolated forest algorithm, compare the anomaly score with a preset threshold, and generate abnormal data when the anomaly score is higher than the preset threshold. The abnormal data includes the anomaly score and the fault type.

[0010] An alarm module is used to issue an alarm signal when the abnormal data is received;

[0011] The fault prediction module is used to acquire the time sequence information in the data to be detected, and process the data to be detected according to the time sequence information to form a feature sequence, wherein the data to be detected includes displacement information, vibration information and liquid level information;

[0012] Construct a long short-term memory network model;

[0013] The feature sequence is input into the long short-term memory network model to obtain the displacement prediction value, vibration prediction value and liquid level prediction value of the drum filter screen;

[0014] When the predicted displacement value, the predicted vibration value, or the predicted liquid level value exceeds the corresponding preset range, predicted abnormal data is generated.

[0015] The alarm module is also used to issue an alarm signal when it receives the predicted abnormal data.

[0016] In one embodiment, the fault prediction module is further configured to:

[0017] A hybrid prediction network is constructed based on time-series features at multiple scales to output a prediction sequence, which includes the displacement prediction value, the vibration prediction value, or the liquid level prediction value and the corresponding confidence level.

[0018] Based on the predicted sequence and historical fault data of the nuclear power plant, the expected fault occurrence time and corresponding confidence interval are output through the reliability model, wherein the confidence interval is dynamically adjusted according to the real-time operating parameters of the nuclear power plant.

[0019] In one embodiment, the dynamic adjustment of the confidence interval includes:

[0020] The real-time operating parameters of the nuclear power plant are obtained, including load, cooling water inlet temperature, and flow rate.

[0021] The real-time operating parameters are normalized and operating risk indicators are calculated according to preset weights;

[0022] determining an adjustment coefficient of the confidence interval according to the operation risk index through a nonlinear mapping;

[0023] adjusting a reference confidence interval according to the adjustment coefficient to obtain the confidence interval.

[0024] In an embodiment, the fault detection module constructs an Isolation Forest according to multiple Isolation Trees, each of which is generated by the following steps:

[0025] randomly selecting a data feature;

[0026] randomly selecting a split point within the selected feature range;

[0027] dividing the data into two parts according to the split point, and recursively repeating until a stop condition is reached.

[0028] In an embodiment, the anomaly score is calculated by the average path length of all the Isolation Trees, and the path length of each Isolation Tree is the number of steps required from the root node to the leaf node of the data point.

[0029] In an embodiment, the fault detection module calculates the anomaly score according to an anomaly score formula, which is:

[0030]

[0031] wherein, is the anomaly score, is the average path length of point x, is a normalization factor.

[0032] In an embodiment, the fault prediction module is further configured to:

[0033] obtain the change trend of the displacement prediction value, the vibration prediction value and the liquid level prediction value, and calculate a fault occurrence probability according to the change trend;

[0034] generate a fault warning information when the fault occurrence probability exceeds a preset probability threshold, the fault warning information including a fault type and an expected occurrence time.

[0035] In an embodiment, the fault prediction module is further configured to:

[0036] obtain a current operation state parameter of the nuclear power plant;

[0037] calculate a dynamic probability threshold through an S-shaped function according to the current operation state parameter;

[0038] generate the fault warning information according to the dynamic probability threshold.

[0039] In a second aspect, the embodiments of the present application provide a drum filter state monitoring method for nuclear power plants, applied to the drum filter state monitoring system for nuclear power plants as described in the first aspect, and the method comprises:

[0040] monitoring the running state of the drum filter in real time;

[0041] receiving real-time data from the sensor module, sampling and preprocessing the real-time data to form preprocessed data;

[0042] receiving the preprocessed data, normalizing and extracting features of the preprocessed data to form to-be-detected data;

[0043] calculating an anomaly score of the to-be-detected data according to the Isolation Forest algorithm, comparing the anomaly score with a preset threshold, and generating anomaly data when the anomaly score is higher than the preset threshold, wherein the anomaly data comprises the anomaly score and a fault type;

[0044] when the anomaly data is received, triggering an alarm signal.

[0045] In a third aspect, the embodiments of the present application provide an electronic device, comprising:

[0046] a processor;

[0047] a memory for storing processor-executable instructions;

[0048] wherein the processor is configured to implement the drum filter state monitoring method for nuclear power plants as described in the second aspect when executing the instructions.

[0049] The drum filter state monitoring system, method and device for nuclear power plants provided by the embodiments of the present application can monitor the key running state of the drum filter in real time through the sensor module, and realize accurate detection and prediction of the filter state by combining the Isolation Forest algorithm and the long short-term memory network model, thereby issuing a warning signal before a fault occurs, and significantly improving the safety and reliability of the operation of the nuclear power plant equipment. BRIEF DESCRIPTION OF DRAWINGS

[0050] Figure 1 The drum filter state monitoring system module schematic diagram for nuclear power plants provided by an embodiment of the present application.

[0051] Figure 2 The installation schematic diagram of the sensor module provided by an embodiment of the present application.

[0052] Figure 3 The drum filter state monitoring method flowchart for nuclear power plants provided by an embodiment of the present application.

[0053] Figure 4The electronic device schematic diagram provided in an embodiment of the present application.

[0054] Main element symbol description

[0055] Drum screen state monitoring system 1 for nuclear power plant

[0056] Sensor module 10

[0057] Axial laser sensor 100

[0058] Radial laser sensor 200

[0059] Radial eddy current sensor 300

[0060] Axial eddy current sensor 400

[0061] Right bearing liquid level sensor 500

[0062] Left bearing liquid level sensor 600

[0063] Data acquisition module 20

[0064] Data analysis module 30

[0065] Fault detection module 40

[0066] Alarm module 50

[0067] Fault prediction module 60

[0068] Electronic device 2

[0069] Processor 21

[0070] Memory 22

[0071] Method steps S100-S500 DETAILED DESCRIPTION

[0072] The technical solutions in the embodiments of the present application will be described clearly and completely below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are some of the embodiments of the present application, but not all the embodiments of the present application.

[0073] It should be noted that “at least one” in the embodiments of the present application means one or more, and more means two or more than two. Unless otherwise defined, all technical and scientific terms used in the present application have the same meaning as understood by those skilled in the art to which the present application belongs. The terms used in the specification of the present application are only for the purpose of describing the specific embodiments of the present application, and are not intended to limit the present application.

[0074] It should be noted that in the embodiments of the present application, the terms "first", "second", etc. are used only for the purpose of distinguishing description and cannot be understood as indicating or implying relative importance, nor can they be understood as indicating or implying sequence. The features limited by "first", "second" can explicitly or implicitly include one or more of the features. In the description of the embodiments of the present application, the terms "exemplary" or "for example" are used to represent an example, illustration or description. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as more preferred or more advantageous than other embodiments or design schemes. Rather, the use of "exemplary" or "for example" is intended to present the relevant concept in a specific manner.

[0075] Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor are within the scope of protection of the present application.

[0076] Drum-shaped filter screens play a crucial role in the cooling source system of nuclear power plants. The cooling source system is used to provide cooling water for the nuclear reactor, ensuring that the nuclear reactor operates within a safe temperature range and preventing risks such as overheating. The cooling water often contains a large amount of impurities, including plant residues, silt, fish and other biological substances in the water, especially in complex water sources or environments contaminated by external factors. These impurities can affect the normal operation of the cooling system. Drum-shaped filter screens ensure the purity of cooling water by efficiently filtering these impurities, thereby ensuring the safety and stability of the cooling system of nuclear power plants. However, over time, drum-shaped filter screens are affected by various factors such as physical wear, corrosion, and fouling during use, leading to a gradual decline in performance and even failure. Especially in complex working environments, the failure of the filter screen is usually difficult to predict and detect, greatly affecting the safety of the nuclear power plant.

[0077] The cooling source system of a nuclear power plant is usually exposed to a high-humidity environment, and impurities in the water, such as salt, minerals, etc., can accelerate the corrosion process of the drum-shaped filter screen, especially the metal part. Regular inspections are difficult to detect problems when small corrosion spots form on the surface of the filter screen.

[0078] The filter screen is affected by factors such as water flow and flow rate changes during operation, resulting in a certain degree of vibration and wear. Over a long period of operation, the structure of the drum-shaped filter screen may undergo slight deformation or wear, which may not be identified during routine inspections.

[0079] Currently, the drum-shaped filter screen monitoring method in the related art mainly relies on manual inspection and regular maintenance, and real-time monitoring and failure prediction cannot be achieved. Although obvious failures (such as overload, abnormal vibration, etc.) can be detected by manual operation, it is difficult to detect potential failures in a timely manner due to its reliance on manual judgment.

[0080] Therefore, an embodiment of the present application provides a drum filter screen state monitoring system, method and device for nuclear power plants. The key operating states (such as axial displacement, radial displacement, gear ring runout and lubricating oil level) of the drum filter screen are monitored in real time by a sensor module. In combination with an isolation forest algorithm and a long short-term memory (LSTM) model, accurate detection and prediction of the filter screen state are realized, an early warning signal is sent before a fault occurs, and the safety and reliability of the operation of the nuclear power plant equipment are significantly improved.

[0081] Figure 1 The drum filter screen state monitoring system for nuclear power plants is provided by an embodiment of the present application. As shown in Figure 1 The drum filter screen state monitoring system 1 for nuclear power plants includes a sensor module 10, a data acquisition module 20, a data analysis module 30, a fault detection module 40, an alarm module 50 and a fault prediction module 60.

[0082] In the embodiment of the present application, the sensor module 10 is used to monitor the operating state of the drum filter screen in real time, so as to provide high-precision operating parameters and provide reliable data support for subsequent data analysis and fault detection.

[0083] It can be understood that the sensor module 10 is reasonably arranged according to the actual operating conditions of the drum filter screen, so as to comprehensively cover the key monitoring parameters and ensure the monitoring effect and fault recognition ability of the system.

[0084] Specifically, please refer to Figure 2 , Figure 2 The installation schematic diagram of the sensor module 10 is provided by an embodiment of the present application. As shown in Figure 2 The sensor module 10 includes an axial laser sensor 100, a radial laser sensor 200, a radial eddy current sensor 300, an axial eddy current sensor 400, a right bearing liquid level sensor 500 and a left bearing liquid level sensor 600.

[0085] Specifically, the axial laser sensor 100 is used to monitor the axial displacement and deformation of the drum filter screen; the radial laser sensor 200 is used to monitor the radial displacement and gear ring runout of the drum filter screen; the axial eddy current sensor 400 and the radial eddy current sensor 300 are used to monitor the bearing operating state and displacement change of the drum filter screen; and the left bearing liquid level sensor 600 and the right bearing liquid level sensor 500 are used to monitor the change of the lubricating oil level in the main shaft oil tank of the drum filter screen.

[0086] It can be understood that the axial laser sensor 100 is arranged in the axial direction of the drum-shaped filter screen, close to the axial movement area of the filter screen, to ensure real-time capture of axial changes. Through high-precision laser technology, subtle displacement and deformation in the axial direction are obtained, providing key data for filter screen stability analysis. The radial laser sensor 200 is arranged in the radial direction of the filter screen, close to the movement track of the gear ring. The radial displacement change and the jumping condition of the gear ring in operation are detected to analyze whether there is abnormal vibration or gear ring wear problem. The axial eddy current sensor 400 is close to the axial movement direction of the bearing to ensure the capture of subtle displacement signals of the bearing. Through real-time feedback of the eddy current signal, it is monitored whether the bearing running state is stable, and data support is provided for judging the bearing performance. The radial eddy current sensor 300 is located in the radial direction of the bearing to detect the radial displacement and possible abnormal signals. It is used to timely find abnormal phenomena (such as wear or deviation) in the operation of the bearing, to provide early warning for preventing bearing failure. The right bearing liquid level sensor 500 is installed on the right side of the main shaft oil tank, close to the lubricating oil circuit. It is used to monitor the real-time change of the lubricating oil level to ensure the sufficiency and stability of the lubricating oil supply and prevent bearing damage due to oil shortage. The left bearing liquid level sensor 600 is installed on the left side of the main shaft oil tank, arranged opposite to the right liquid level sensor. Together with the right liquid level sensor, it completes dynamic monitoring of the lubricating oil level on both sides of the main shaft oil tank to determine whether the lubricating system is operating normally.

[0087] It can be understood that the arrangement scheme and functional design of the sensor module 10 are fully optimized, which can meet the multi-dimensional monitoring requirements and efficient data acquisition performance.

[0088] In addition, each sensor component adopts high-precision measurement technology and has good anti-interference ability, which can adapt to the operating environment of the drum-shaped filter screen of the nuclear power plant under complex working conditions. At the same time, the design of the sensor module 10 has good expansibility, and in the future, other types of sensors such as temperature sensors or pressure sensors can be added according to actual needs to further improve the monitoring ability and adaptability of the system.

[0089] The multi-source data obtained by the sensor module 10 can provide comprehensive support for the subsequent data analysis module of the system, and combined with the isolated forest algorithm and the long short-term memory network model, the running state of the drum-shaped filter screen can be accurately detected and fault prediction can be realized, to improve the safety and stability of the nuclear power plant operation.

[0090] In the embodiments of the present application, the data acquisition module 20 is connected to the sensor module 10, and is used to receive real-time data from the sensor module, and sample and preprocess the real-time data to form preprocessed data.

[0091] Specifically, the data acquisition module 20 uses a high-precision sampling mechanism to uniformly process the multi-source data (such as axial displacement, radial displacement, vibration information, and liquid level information) transmitted by the sensor module 10. The preprocessing steps include data cleaning, noise reduction, and format standardization to ensure that the input data quality meets the requirements of subsequent data analysis.

[0092] Understandably, the data acquisition module 20 achieves efficient data transmission and processing through optimized hardware interfaces and communication protocols, enabling it to process concurrent data from multiple sensors in real time and possessing a certain redundancy mechanism to cope with data loss or abnormal situations.

[0093] In this embodiment of the application, the data analysis module 30 is connected to the data acquisition module 20 and is used to receive preprocessed data, normalize the preprocessed data and extract features to form data to be detected.

[0094] Specifically, the data analysis module 30 normalizes various sensor data to a unified numerical range, eliminating the impact of different data scales. During feature extraction, the module extracts key features based on the statistical characteristics of the signal, such as the signal mean, standard deviation, and spectral characteristics, to reflect the operating status of the drum filter.

[0095] Understandably, the data analysis module 30 employs efficient algorithms and optimized hardware resource management to ensure the real-time performance and accuracy of data processing, providing stable input for subsequent fault detection.

[0096] In this embodiment, the fault detection module 40 is connected to the data analysis module 30 and is used to calculate the anomaly score of the data to be detected according to the isolated forest algorithm, compare the anomaly score with a preset threshold, and generate abnormal data when the anomaly score is higher than the preset threshold. The abnormal data includes the anomaly score and the fault type.

[0097] Specifically, the fault detection module 40 performs anomaly analysis on the data to be detected by constructing an isolated forest model. The module determines the degree of anomaly based on preset thresholds and generates anomaly data including anomaly scores and specific fault types. For example, anomaly scores above the threshold may indicate specific problems such as excessive axial displacement of the filter screen or insufficient lubricating oil level.

[0098] Understandably, the fault detection module 40 uses a distributed algorithm to optimize the operating efficiency of the isolated forest model, enabling it to quickly and accurately identify abnormal situations in the complex operating environment of a nuclear power plant.

[0099] In this embodiment, the fault detection module 40 calculates an anomaly score according to the anomaly scoring formula, which is:

[0100]

[0101] wherein, is the anomaly score, is the average path length of point x, is the normalization factor.

[0102] Specifically, the Isolation Forest model evaluates the difference between the data point and the normal data distribution by calculating the path length of each isolated tree, and finally obtains the anomaly score. The higher the score, the more likely the data point is an anomaly.

[0103] It can be understood that the formula can effectively adapt to multi-dimensional data distribution in combination with the characteristics of the Isolation Forest model, and is especially suitable for anomaly detection requirements in complex systems of nuclear power plants.

[0104] In the embodiments of the present application, the alarm module 50 is connected to the fault detection module 40, and is configured to send an alarm signal when receiving the abnormal data.

[0105] Specifically, the alarm module 50 generates a corresponding alarm signal according to the abnormal data, including audible and visual alarms, system log records, remote notifications and the like, so that the operation and maintenance personnel can take timely countermeasures.

[0106] It can be understood that the alarm module 50 has multi-level alarm capability, and can flexibly adjust the alarm level according to the severity of the anomaly score, thereby maximizing the reduction of false alarm interference.

[0107] In the embodiments of the present application, the drum filter state monitoring system 1 for nuclear power plants further comprises: a fault prediction module 60 connected to the data analysis module 30 and the alarm module 50, configured to obtain time sequence information in the to-be-detected data, process the to-be-detected data according to the time sequence information to form a feature sequence, wherein the to-be-detected data includes displacement information, vibration information and liquid level information; construct a long short-term memory network model; input the feature sequence into the long short-term memory network model to obtain displacement prediction value, vibration prediction value and liquid level prediction value of the drum filter; generate prediction abnormal data when the displacement prediction value, the vibration prediction value or the liquid level prediction value exceeds the corresponding preset interval; and the alarm module is further configured to send an alarm signal when receiving the prediction abnormal data.

[0108] Specifically, the fault prediction module 60 extracts the time sequence features of the to-be-detected data, constructs a feature sequence, uses the LSTM model for time series prediction, generates future operation state parameters, and compares the prediction results with the preset threshold to determine whether there is a potential fault risk.

[0109] It can be understood that the fault prediction module 60 can identify the possible fault trend of the drum filter in advance, provide a scientific basis for preventive maintenance, and improve the operation reliability and service life of the equipment.

[0110] In the embodiment of the present application, the output formula of the fault prediction module 60 is:

[0111]

[0112] wherein, is the predicted value of the future moment, and the long short-term memory network model.

[0113] Specifically, the LSTM model generates high-precision predicted values of the future running state of the drum filter by memorizing and processing the long and short-term dependencies of the data.

[0114] It can be understood that the LSTM model can effectively capture the time series characteristics of the drum filter running data and is suitable for prediction requirements in complex dynamic environments.

[0115] In the embodiment of the present application, the fault prediction module 60 optimizes the long short-term memory network model through a mean square error loss function, and the formula of the loss function is:

[0116]

[0117] wherein, is the loss function value, is the value of the i-th sample predicted by the model at the k-th future moment, is the value of the i-th sample at the k-th future moment, and N is the total number of samples.

[0118] Specifically, by minimizing the loss function value, the fault prediction module 60 ensures the accuracy and reliability of the prediction results of the LSTM model.

[0119] It can be understood that the mean square error loss function can effectively measure the deviation between the predicted value and the true value, and provide a clear target direction for model optimization.

[0120] In the embodiment of the present application, the fault prediction module 60 is further configured to: construct a hybrid prediction network to output a prediction sequence according to a plurality of scale time sequence characteristics, the prediction sequence comprising displacement predicted values, vibration predicted values or liquid level predicted values and corresponding confidence degrees; and output an expected fault occurrence time and a corresponding confidence interval through a reliability model according to the prediction sequence and historical fault data of the nuclear power plant, wherein the confidence interval is dynamically adjusted according to real-time running parameters of the nuclear power plant.

[0121] Specifically, the fault prediction module 60 is further configured to construct a hybrid prediction network according to time sequence features of multiple scales. By extracting transient, periodic and trend features in short, medium and long time windows respectively, and fusing time domain, frequency domain and time-frequency domain features of each sensor, a unified multi-scale time sequence representation is formed. The multi-scale time sequence representation is input into the hybrid prediction network, which can include prediction sub-models based on recurrent neural networks, time convolution networks or attention mechanisms, and can output multi-step prediction sequences using a model ensemble strategy. Each prediction value is accompanied by a corresponding confidence assessment, which is obtained by methods such as uncertainty measurement of model ensemble, regression interval or Monte Carlo sampling, and improved in interpretability by calibration. Based on the prediction sequence and the historical fault data of the nuclear power plant, the reliability model maps the prediction results to the predicted failure occurrence time and the baseline confidence interval, providing a quantitative reference for subsequent early warning. The above hybrid prediction network and reliability model can be run on edge devices or in the cloud, supporting sliding window online inference and periodic or triggered batch retraining, and saving each prediction result, confidence and predicted failure occurrence time as a historical record for subsequent model evaluation and online learning.

[0122] It can be understood that the implementation of the hybrid prediction network and the confidence and predicted failure occurrence time can be various, for example, end-to-end time convolution or transformer can be used to directly learn the representation, or step-by-step feature extraction and fusion can be used. Confidence estimation can be based on Bayesian approximation or quantile regression, and calibration techniques such as temperature scaling can also be used. The reliability model can be based on empirical lifetime distribution, or survival analysis or risk regressor of machine learning can be used. For scenarios with limited computing power, model pruning, quantization or distilled lightweight predictors can be used, and complex training processes can be moved to the cloud. For high security requirements, a dual mode operation (local lightweight predictor and cloud high precision predictor in parallel) can be set, and the inconsistency between the two results can be used as a trigger condition for manual review. The above parameters such as window length, overlap rate, prediction step, initial width of confidence interval, etc. can be dynamically adjusted or learned online according to the field data to continuously optimize the performance.

[0123] In the embodiment of the present application, the dynamic adjustment of the confidence interval comprises: obtaining real-time running parameters of the nuclear power plant, the real-time running parameters comprising load, cooling water inlet temperature and flow; normalizing the real-time running parameters and calculating a running risk indicator according to a preset weight; determining an adjustment coefficient of the confidence interval through nonlinear mapping according to the running risk indicator; and adjusting the baseline confidence interval according to the adjustment coefficient to obtain the confidence interval.

[0124] Specifically, the dynamic adjustment of the confidence interval includes obtaining current real-time operation parameters of the nuclear power plant, which include but are not limited to unit load, cooling water inlet temperature, cooling water flow and other indicators affecting equipment stress or aging, and performing standardization and normalization processing on these operation parameters, and then combining them into an operation risk indicator according to preset or online learned weights. Based on the operation risk indicator, a confidence interval adjustment coefficient is obtained through a nonlinear mapping function (such as an S-shaped or other monotonically increasing function) to enlarge or shrink the baseline confidence interval, while the adjusted confidence interval is subjected to upper and lower limits to ensure safety. When the real-time operation parameters suddenly change or exceed the safety threshold, the system will immediately recalculate the confidence interval and generate a temporarily elevated risk level, and issue a high-priority alarm if necessary. When real-time parameters are missing or judged to be unreliable, the system will use data interpolation, estimation based on adjacent time periods or more conservative default adjustment coefficients to ensure that the confidence interval is not improperly narrowed. The above dynamic adjustment can be performed on the edge or in the cloud, and the input parameters, calculation process, adjustment coefficient and results of each adjustment are written into the log for traceability and compliance review, while the confidence interval mapping strategy and various weights can be optimized online or periodically based on experience data, so that the adjustment strategy is continuously improved with the aging and operation mode of the equipment.

[0125] It can be understood that the specific implementation of the dynamic adjustment of the confidence interval allows various engineering variants. The weights of the operation risk indicator can be initially set by engineering experts based on experience and fine-tuned on site data with the goal of minimizing false positives or false negatives; the nonlinear mapping function can be manually set or learned using a small neural network or tree model; to meet safety redundancy, a dual-track strategy can be used, i.e. the regular track is used for daily fine-tuning, and the safety track uses a more conservative expansion coefficient in high-risk operating conditions. In order to facilitate operation and maintenance decisions, the system can output the confidence interval before and after adjustment, the adjustment coefficient and the dominant operating parameter leading to the adjustment in the alarm information for quick judgment and disposal by the operation and maintenance personnel; in addition, to prevent false positives caused by frequent oscillation, a minimum time interval can be set for the variation of the adjustment coefficient or a smoothing filter strategy can be used. The above various modifications and parameter settings are conducive to providing sufficient example support in the specification, thereby enhancing the enforceability and anti-rejection ability of the claims.

[0126] In the embodiment of the present application, the fault prediction module 60 is further configured to: obtain the change trend of the displacement prediction value, the vibration prediction value and the liquid level prediction value, calculate the fault occurrence probability according to the change trend; when the fault occurrence probability exceeds a preset probability threshold, generate a fault warning information, and the fault warning information includes a fault type and a predicted occurrence time.

[0127] Specifically, the fault prediction module 60 first performs a multi-step prediction output sequence on the long short-term memory network model The three prediction sequences are organized according to variable categories (axial displacement d, radial vibration v, bearing liquid level l) , , . For each prediction sequence, the module calculates a normalized hyper-limit metric. Then, the normalized hyper-limit metrics are linearly weighted and mapped to a single probability of single fault occurrence through a sigmoid function. The fault type is determined by the variable that contributes the most, for example, when , the fault type is determined as "displacement anomaly"; when the vibration term is the largest, the fault type is determined as "vibration anomaly"; when the liquid level term is the largest or the liquid level deviation direction is negative, the fault type is determined as "insufficient lubrication / leakage", etc.

[0128] For the calculation of the estimated time of occurrence (ETA), the fault prediction module 60 adopts the following process: for each prediction sequence, find the earliest time on the time axis that exceeds the corresponding safety threshold, that is, find the minimum that makes satisfy the hyper-limit condition; if there are multiple variables that exceed the limit at the same time, select the earliest as the overall estimated time of occurrence; if the prediction sequence does not exceed the limit within the given window, the approximate time to reach the threshold can be estimated based on the linear or quadratic fitting extrapolation of the sequence. The module records the estimated time of occurrence in the fault warning information in the form of absolute time (for example, "2025-XX-XX 14:30") and relative time (for example, "estimated in 12 minutes"). When the calculated exceeds the preset probability threshold (or exceeds the dynamic probability threshold, see below), the system generates and issues a fault warning information, which includes: fault type, fault occurrence probability , estimated time of occurrence, main contributing features and confidence interval, so that the operation and maintenance personnel can make quick decisions and take measures.

[0129] It can be understood that this calculation process not only considers the most serious deviation of each monitoring variable within the prediction window, but also fuses the multivariate information into a single probability index through weighting and sigmoid mapping, which is convenient for unified decision-making; by clearly defining the estimated time of occurrence of the warning and the main contributing features, the operation and maintenance personnel can determine the priority and take targeted measures accordingly. In addition, this fault occurrence probability can be used in combination with the anomaly score output by the fault detection module (isolation forest, reconstruction error, etc.) to reduce the false alarm rate, for example, by using a fusion score strategy, and giving the decision basis after fusion in the warning information, to improve the explainability and reliability of the alarm.

[0130] In the embodiments of the present application, the fault prediction module 60 is further configured to: acquire a current operating state parameter of the nuclear power plant; calculate a dynamic probability threshold through a sigmoid function according to the current operating state parameter; and generate a fault warning information according to the dynamic probability threshold.

[0131] Specifically, the fault prediction module 60 obtains a set of operating state parameters of the nuclear power plant in real time from a monitoring system or a data analysis module , including but not limited to cooling water flow, inlet water temperature, rotating speed / load, ambient temperature, power supply voltage, etc. The module first normalizes each parameter:

[0132]

[0133] wherein , are the historical mean and standard deviation (which can be updated periodically) of the parameter respectively. Then the operating severity index (Severity Index) S is calculated:

[0134]

[0135] wherein is the weight coefficient, reflecting the influence of each operating parameter on the risk of the drum filter, which can be determined by historical data regression or expert scoring and adjusted online. Based on the severity index S, the module generates a dynamic probability threshold using a sigmoid function:

[0136]

[0137] wherein and are the lower and upper limits of the threshold respectively, k is the curvature coefficient, is the center point (i.e. when S= , the threshold is the middle value). In actual settings, when the operating severity increases (S increases), k<0 or the weight direction can be set to make decrease, so as to increase the sensitivity of the early warning in high-risk operating conditions; otherwise, the threshold is increased to reduce false positives in low-risk operating conditions. The calculated is compared with the probability of failure : when , the system generates a fault warning; otherwise, no trigger or only a low-level prompt is generated.

[0138] It can be understood that the dynamic probability threshold based on the sigmoid function can make the system threshold no longer a fixed static value, but automatically adjusted according to the on-site operating conditions, so as to balance the needs of detection sensitivity and false alarm suppression. This mechanism allows the trigger threshold to be automatically lowered to provide early warning during high-load, abnormal environment or critical operating phases of the nuclear power plant, and the threshold to be raised to avoid unnecessary interference during stable, low-load phases. The weight coefficient of the operating severity index S, the upper and lower limits of the threshold, and the sigmoid parameters (k, ) can be trained offline or optimized online by historical failure samples and operation experience; meanwhile, the system records the value of each dynamic threshold and the triggering decision, which facilitates post-audit and model improvement. In addition, to ensure safety, the system supports forcing the dynamic threshold to the protection value and triggering the corresponding protection or shutdown instruction under the emergency level, thereby linking the prediction and early warning with safety protection and improving the overall reliability of the nuclear power plant operation.

[0139] In the embodiment of the present application, the fault detection module 40 constructs an isolated forest according to multiple isolated trees, and each isolated tree is generated by the following steps: randomly selecting a data feature; randomly selecting a split point in the selected feature range; dividing the data into two parts according to the split point, and recursively repeating until the stop condition is reached.

[0140] Specifically, the isolated forest generates multiple isolated trees by the above method, and uses the overall structure of the forest to realize the identification of abnormal points.

[0141] It can be understood that this method can efficiently process multi-dimensional data sets and ensure the accuracy of anomaly detection in complex environments.

[0142] In the embodiment of the present application, the anomaly score is calculated by the average path length of all isolated trees, and the path length of each isolated tree is the number of steps required from the root node to the leaf node of the data point.

[0143] Specifically, by calculating the path length, the isolated forest model quantifies the abnormality degree of the data point and generates an abnormality score with reference value.

[0144] It can be understood that this scoring method can intuitively reflect the deviation degree of the data point from the normal distribution, which helps to quickly locate potential problems.

[0145] In one exemplary embodiment, the operation process of the drum filter screen state monitoring system for nuclear power plants is as follows:

[0146] When the drum filter screen starts to operate, the sensor module 10 monitors its operating state in real time. Among them, the axial laser sensor 100 and the radial laser sensor 200 collect the axial displacement, radial displacement and gear ring jump amount data of the filter screen, the axial eddy current sensor 400 and the radial eddy current sensor 300 monitor the operating state and displacement change of the bearing, and the left bearing liquid level sensor 600 and the right bearing liquid level sensor 500 detect the liquid level change of the lubricating oil in the main shaft oil tank. These real-time data are transmitted to the data acquisition module 20 through a high-precision interface.

[0147] The data acquisition module 20 samples and pre-processes the received sensor data. The pre-processing process includes signal denoising, data format standardization, and missing value processing, etc. operations, forming high-quality pre-processed data. Subsequently, the data analysis module 30 normalizes the pre-processed data, unifies each sensor data to the same scale range, and extracts feature values such as the mean value of displacement, the frequency spectrum characteristics of vibration, and the fluctuation trend of liquid level, etc. These features are integrated into the data to be detected for subsequent analysis.

[0148] The fault detection module 40 receives the data to be detected and calculates the anomaly score of each data based on the Isolation Forest algorithm. Assuming that the system detects the axial displacement score at a certain time as 0.85, which exceeds the preset threshold value 0.8 set by the system, it is determined that the data is abnormal data. The fault detection module immediately generates an abnormal data record, which includes the anomaly score, the specific fault type (such as excessive axial displacement) and related information. The alarm module 50 immediately issues an audible and visual alarm after receiving the abnormal data, and sends a remote notification to the monitoring system, prompting the operation and maintenance personnel to check and handle the drum filter.

[0149] In order to further improve the reliability of the equipment operation, the fault prediction module 60 is also deployed in the system. This module uses the time series feature data generated by the data analysis module 30 to build and train a long short-term memory (LSTM) model. The fault prediction module predicts the axial displacement trend of the drum filter within the next hour through the model, generating a displacement prediction value. At a certain time, if the prediction result shows that the axial displacement will reach the dangerous threshold within the next 15 minutes, the module generates a predicted abnormal data and sends a warning signal to the alarm module 50, prompting the operation and maintenance personnel to take preventive measures in advance.

[0150] In the whole process, the system combines real-time monitoring, anomaly detection and trend prediction to comprehensively evaluate the running state of the drum filter. By integrating the Isolation Forest algorithm and the LSTM model, this system not only can timely detect anomalies in operation, but also can effectively predict potential faults, ensuring the stable operation of the nuclear power plant.

[0151] In an exemplary embodiment, the drum filter state monitoring system for nuclear power plants fault prediction module 60 runs as follows:

[0152] During normal operation of the drum filter, the data analysis module 30 continuously receives and processes pre-processed data from the sensor module 10, extracts displacement information, vibration information, and liquid level information, etc. feature data. These data are arranged in chronological order as time series feature sequences to form input data sets for prediction.

[0153] The fault prediction module 60 utilizes a long short-term memory (LSTM) model to train and predict the feature sequence. For example, the data input to the LSTM model includes the axial displacement change of the drum filter, the radial vibration amplitude, and the dynamic change trend of the lubricating oil level in the past 10 minutes. The model outputs the predicted values of the axial displacement, the vibration amplitude, and the liquid level in the next 15 minutes.

[0154] At a certain moment, the LSTM model calculates that the predicted value of the axial displacement in the next 15 minutes is 0.45 mm, and the preset threshold of the system is 0.4 mm. Since the predicted value exceeds the threshold, the fault prediction module 60 generates a predicted abnormal data record, which includes the predicted displacement value, the time point of the abnormality, and the predicted abnormal type (such as excessive axial displacement). The abnormal data is immediately transmitted to the alarm module 50.

[0155] After receiving the predicted abnormal data, the alarm module 50 sends a warning signal, including an audible and visual alarm prompt and a remote push information, to notify the operation and maintenance personnel to pay attention to the risk of the axial displacement possibly exceeding the limit. At the same time, the system interface displays detailed information of the abnormal data, such as the trend chart of the predicted displacement value and the time point when the abnormality may occur in the future.

[0156] In addition, the fault prediction module 60 optimizes the LSTM model through a mean square error loss function to make the prediction result more accurate. In the long-term running process, the system continuously adapts the LSTM model to improve the prediction performance and reduce the possibility of false positives and false negatives.

[0157] Through such a prediction process, when the drum filter is about to have an axial displacement exceeding the limit fault, the system can send a warning 15 minutes in advance to provide sufficient time for the operation and maintenance personnel to intervene, such as adjusting the filter running parameters or performing preventive maintenance, thereby effectively avoiding the occurrence of the fault.

[0158] Figure 3 is a flowchart of a drum filter state monitoring method for a nuclear power plant provided by an embodiment of the present application. As shown in Figure 3 The drum filter state monitoring method for a nuclear power plant at least includes the following steps: S100, real-time monitoring of the running state of the drum filter; S200, receiving real-time data from the sensor module and sampling and preprocessing the real-time data to form preprocessed data; S300, receiving the preprocessed data and performing normalization processing and feature extraction on the preprocessed data to form to-be-detected data; S400, calculating the anomaly score of the to-be-detected data according to the Isolation Forest algorithm, comparing the anomaly score with a preset threshold, and generating abnormal data when the anomaly score is higher than the preset threshold, the abnormal data including the anomaly score and the fault type; S500, triggering an alarm signal when the abnormal data is received.

[0159] S100: Real-time monitoring of the running state of the drum filter.

[0160] In the embodiment of the present application, the drum filter state monitoring method for nuclear power plants in step S100 includes real-time monitoring of the running state of the drum filter. The specific monitoring method can refer to Figure 1 、 Figure 2 and the corresponding description, which will not be repeated here.

[0161] S200: Receive real-time data from the sensor module, and sample and preprocess the real-time data to form preprocessed data.

[0162] In the embodiment of the present application, the drum filter state monitoring method for nuclear power plants in step S200 includes receiving real-time data from the sensor module, and sampling and preprocessing the real-time data to form preprocessed data. The specific processing method can refer to Figure 1 、 Figure 2 and the corresponding description, which will not be repeated here.

[0163] S300: Receive preprocessed data, and normalize and extract features of the preprocessed data to form detection data.

[0164] In the embodiment of the present application, the drum filter state monitoring method for nuclear power plants in step S300 includes receiving preprocessed data, and normalizing and extracting features of the preprocessed data to form detection data. The specific processing method can refer to Figure 1 、 Figure 2 and the corresponding description, which will not be repeated here.

[0165] S400: Calculate the anomaly score of the detection data according to the Isolation Forest algorithm, compare the anomaly score with a preset threshold, and generate anomaly data when the anomaly score is higher than the preset threshold. The anomaly data includes the anomaly score and the fault type.

[0166] In the embodiment of the present application, the drum filter state monitoring method for nuclear power plants in step S400 includes calculating the anomaly score of the detection data according to the Isolation Forest algorithm, comparing the anomaly score with a preset threshold, and generating anomaly data when the anomaly score is higher than the preset threshold. The anomaly data includes the anomaly score and the fault type. The specific processing method can refer to Figure 1 、 Figure 2 and the corresponding description, which will not be repeated here.

[0167] S500: Trigger an alarm signal when receiving the anomaly data.

[0168] In the embodiment of the present application, the drum filter state monitoring method for nuclear power plants in step S500 includes triggering an alarm signal when receiving the anomaly data. The specific processing method can refer toFigure 1 、 Figure 2 and the corresponding description, which are not repeated here.

[0169] Figure 4 is an electronic device 2 provided by an embodiment of the present application. As shown in Figure 4 , the electronic device 2 at least includes the following parts: a processor 21 and a memory 22.

[0170] In an embodiment of the present application, the memory 22 is configured to store processor 21 executable instructions, and the processor 21 is configured to execute the instructions to implement the drum screen state monitoring method for nuclear power plants as shown in Figure 1 .

[0171] In an embodiment of the present application, a computer readable storage medium includes instructions, and the instructions instruct the device to execute the drum screen state monitoring method for nuclear power plants as in the second aspect. For example, the instructions instruct the device to execute the drum screen state monitoring method for nuclear power plants as shown in Figure 1 steps S100 to S500.

[0172] The program working in the electronic device 2 related to an embodiment of the present application can be a program (a program for making a computer function) for controlling a central processing unit (CPU) and the like to realize the functions of the above-mentioned embodiments related to one solution of the present application. Then, the information processed by these devices is temporarily stored in a random access memory (RAM) when it is processed, and then stored in various ROMs such as a read only memory (Flash ROM), a hard disk drive (HDD), and the like, and read out, corrected, and written by the CPU as needed.

[0173] It should be noted that a part of the electronic device 2 of the above-mentioned embodiments can also be realized by a computer. In this case, a program for realizing the control function can be recorded in a computer readable recording medium, and realized by reading the program recorded in the recording medium into the computer and executing it.

[0174] It should be noted that the "computer" mentioned here refers to a computer built into the electronic device 2, which is a computer including an OS, peripheral devices and the like hardware. In addition, the "computer readable recording medium" refers to removable media such as a floppy disk, a magneto-optical disk, a ROM, a CD-ROM, and the like, and storage devices such as a hard disk built into the computer.

[0175] Furthermore, a "computer-readable recording medium" can include: a medium that dynamically stores a program for a short period of time, such as a communication line used when transmitting a program via a network such as the Internet or a communication line such as a telephone line; or a medium that stores a program for a fixed period of time, such as volatile memory inside a computer that serves as a server or client in this case. In addition, the aforementioned program can be a program used to implement the above-mentioned functions, or it can be a program that can implement the above-mentioned functions by combining with programs already recorded in the computer.

[0176] Furthermore, the electronic device 2 in the above embodiments can also be implemented as an assembly (device group) composed of multiple devices. Each device constituting the device group can possess some or all of the functions or functional blocks of the electronic device 2 in the above embodiments. As a device group, it is sufficient to have all the functions or functional blocks of the electronic device 2.

[0177] It is understood that the nuclear power plant drum filter condition monitoring system 1, method, and equipment 2 provided in this application embodiment monitor the key operating states of the drum filter (such as axial displacement, radial displacement, gear ring runout, and lubricating oil level) in real time through the sensor module 10. Combined with the isolated forest algorithm and the Long Short-Term Memory (LSTM) network model, it achieves accurate detection and prediction of the filter state, issuing early warning signals before faults occur, significantly improving the safety and reliability of nuclear power plant equipment operation. The system introduces the isolated forest algorithm to calculate anomaly scores and combines it with the LSTM network to analyze and predict time-series data. This allows for accurate identification of potential fault types and locations of the drum filter, generating predicted anomaly data, supporting advance maintenance planning, and reducing downtime and repair costs caused by equipment failures. The alarm module can not only trigger alarms when the anomaly score exceeds a threshold but also generate multi-level alarm signals based on predicted anomaly data, guiding maintenance personnel to prioritize serious faults, further improving operation and maintenance efficiency and the safety factor of nuclear power plant operation.

[0178] Those skilled in the art should recognize that the above embodiments are only used to illustrate this application and are not intended to limit this application. Any appropriate changes and variations made to the above embodiments within the essential spirit and scope of this application fall within the scope of protection claimed in this application.

Claims

1. A drum filter state monitoring system for nuclear power plants, applied to monitor the state of a drum filter, characterized by, The system comprises: a sensor module for monitoring the running state of the drum filter in real time, the sensor module comprising: an axial laser sensor for monitoring the axial displacement and deformation of the drum filter; a radial laser sensor for monitoring the radial displacement and gear ring runout of the drum filter; an axial eddy current sensor and a radial eddy current sensor for monitoring the bearing running state and displacement change of the drum filter; a left bearing liquid level sensor and a right bearing liquid level sensor for monitoring the change of the lubricating oil level in the main shaft oil tank of the drum filter; a data acquisition module for receiving real-time data from the sensor module and sampling and preprocessing the real-time data to form preprocessed data; a data analysis module for receiving the preprocessed data and performing normalization processing and feature extraction on the preprocessed data to form detection data; a fault detection module for calculating the anomaly score of the detection data according to the isolation forest algorithm, comparing the anomaly score with a preset threshold, and generating abnormal data when the anomaly score is higher than the preset threshold, the abnormal data including the anomaly score and a fault type; an alarm module for issuing an alarm signal when the abnormal data is received; a fault prediction module for obtaining time sequence information in the detection data, processing the detection data according to the time sequence information to form a feature sequence, wherein the detection data includes displacement information, vibration information and liquid level information; constructing a long short-term memory network model, inputting the feature sequence into the long short-term memory network model to obtain displacement prediction values, vibration prediction values and liquid level prediction values of the drum filter; generating prediction abnormal data when the displacement prediction values, the vibration prediction values or the liquid level prediction values exceed corresponding preset intervals; the alarm module is further configured to issue an alarm signal when the prediction abnormal data is received.

2. The drum screen condition monitoring system for nuclear power plants according to claim 1, characterized in that, The fault prediction module is further configured to: construct a hybrid prediction network according to a plurality of scale time sequence features to output a prediction sequence, the prediction sequence including the displacement prediction values, the vibration prediction values or the liquid level prediction values and corresponding confidence levels; output an expected fault occurrence time and a corresponding confidence interval through a reliability model according to the prediction sequence and historical fault data of the nuclear power plant, wherein the confidence interval is dynamically adjusted according to real-time running parameters of the nuclear power plant.

3. The drum screen condition monitoring system for nuclear power plants according to claim 2, characterized in that, The dynamic adjustment of the confidence interval comprises: obtaining the real-time running parameters of the nuclear power plant, the real-time running parameters including load, cooling water inlet temperature and flow rate; normalizing the real-time running parameters and calculating a running risk indicator according to a preset weight; determining an adjustment coefficient of the confidence interval through a nonlinear mapping according to the running risk indicator; adjusting a reference confidence interval according to the adjustment coefficient to obtain the confidence interval.

4. The drum screen condition monitoring system for nuclear power plants according to claim 1, characterized by, The fault detection module constructs an isolation forest according to a plurality of isolation trees, and each isolation tree is generated by the following steps: randomly selecting a data feature; randomly selecting a split point within the selected feature range; dividing the data into two parts according to the split point, and recursively repeating until a stop condition is reached.

5. The drum screen condition monitoring system for nuclear power plants according to claim 4, characterized in that, The anomaly score is calculated by the average path length of all the isolated trees, and the path length of each isolated tree is the number of steps from the root node to the leaf node of the data point.

6. The drum screen condition monitoring system for nuclear power plants according to claim 1, characterized by, The fault detection module calculates the anomaly score according to an anomaly score formula, and the anomaly score formula is: Wherein, s(x) is the anomaly score, E(h(x)) is the average path length of point x, and c(n) is a normalization factor.

7. The drum screen condition monitoring system for nuclear power plants according to claim 1, characterized by, The fault prediction module is further configured to: Obtain the change trend of the displacement prediction value, the vibration prediction value and the liquid level prediction value, and calculate a fault occurrence probability according to the change trend; When the fault occurrence probability exceeds a preset probability threshold, generate a fault warning information, and the fault warning information includes a fault type and an expected occurrence time.

8. The drum screen condition monitoring system for nuclear power plants according to claim 7, characterized in that, The fault prediction module is further configured to: Obtain the current operating state parameters of the nuclear power plant; According to the current operating state parameters, calculate a dynamic probability threshold through an S-shaped function; Generate the fault warning information according to the dynamic probability threshold.

9. A method for monitoring the condition of a drum filter for a nuclear power plant, applied to the system for monitoring the condition of a drum filter for a nuclear power plant according to any one of claims 1 to 8, characterized by, The method comprises: Real-time monitoring the operating state of the drum filter screen; Receiving real-time data from the sensor module, and sampling and preprocessing the real-time data to form preprocessed data; Receiving the preprocessed data, and normalizing and extracting features of the preprocessed data to form to-be-detected data; Calculating the anomaly score of the to-be-detected data according to the isolated forest algorithm, comparing the anomaly score with a preset threshold, and generating abnormal data when the anomaly score is higher than the preset threshold, wherein the abnormal data includes the anomaly score and a fault type; When the abnormal data is received, triggering an alarm signal.

10. An electronic device, comprising: Comprise: A processor; A memory for storing processor-executable instructions; Wherein, the processor is configured to execute the instructions to realize the drum filter screen state monitoring method for nuclear power plants in claim 9.

Images