Method, apparatus, and medium for fatigue life prediction

Abstract

Embodiments of the present disclosure relate to a method, device and medium for predicting the fatigue life of key components and vulnerable components of a circulating water pump of a nuclear power plant. The device involved in the present disclosure is also integrated with a data acquisition and processing module related to the key components and vulnerable components of the circulating water pump, a storage module, an industrial computing module, a software program module, a fault algorithm module, a prediction algorithm module and a communication transmission module.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of industrial technology, and in particular to a method and device for predicting the fatigue life of key components and vulnerable components of a circulating water pump, and a medium. BACKGROUND

[0002] With the rapid development of industrialization and modernization, nuclear power plants, as one of the important sources of clean energy, have attracted worldwide attention for their safety and reliability. Among the many devices in a nuclear power plant, the circulating water pump plays a crucial role in the safe operation of the entire nuclear power plant due to its key role. The key components of the circulating water pump include bearings, bushings, gearboxes, impeller rings and other key components. Due to long-term work in high load and harsh environment, they are prone to fatigue damage, leading to a decline in device performance and even failure.

[0003] Traditional methods for predicting the life of vulnerable components mainly rely on empirical formulas, physical models or statistical analysis. These methods often require a large amount of historical data support, and have certain limitations in terms of prediction accuracy and real-time performance. In particular, when faced with complex working conditions and changing environments, the prediction accuracy and adaptability of traditional methods are even more inadequate. In addition, these methods usually cannot monitor and evaluate the device status in real time, resulting in maintenance decisions that lack pertinence and timeliness.

[0004] In order to solve the above problems, in recent years, machine learning and artificial intelligence technologies have been introduced into the field of monitoring and prediction of industrial equipment. By using advanced algorithm models such as recurrent neural networks (LSTM), historical and real-time data of equipment can be analyzed in depth, so as to realize more accurate and real-time life prediction. However, how to effectively collect and process equipment data, build and train accurate prediction models, and how to quickly and effectively apply prediction results to maintenance decisions, are still challenges faced by the field. SUMMARY

[0005] In view of the above problems, the present disclosure discloses a method and device for predicting the fatigue life of key components and vulnerable components of a circulating water pump, a computing device and a computer readable storage medium. The present application aims to provide a nuclear power plant circulating water pump key component and vulnerable component fatigue life prediction system and method, which combines advanced data acquisition technology, powerful computing platform, efficient data processing algorithm and machine learning model to realize accurate prediction of the fatigue life of vulnerable components, improve the operation reliability and safety of the equipment, reduce maintenance costs, and provide a scientific basis for preventive maintenance of the equipment.

[0006] According to a first aspect of the present disclosure, a method for predicting fatigue life of key components and vulnerable components of a circulating water pump is provided, wherein the circulating water pump is used in a three-loop circulating cooling water system of a nuclear power plant, and the method comprises: obtaining a vibration signal data set including vibration accelerations at a plurality of monitoring positions of a reduced scale model of the circulating water pump based on a plurality of one-way vibration acceleration sensors arranged on the reduced scale model of the circulating water pump, wherein the plurality of monitoring positions at least include positions within a predetermined range from a setting position of a simulated vulnerable component on the reduced scale model of the circulating water pump; performing normalization preprocessing on the obtained vibration signal data set so as to divide the data set into a training set, a validation set and a test set; training a long short-term memory network including a plurality of fully connected layers based on the training set, thereby obtaining a trained fatigue life prediction model; optimizing the trained fatigue life prediction model based on the validation set, and determining an optimized target fatigue life prediction model satisfying a performance evaluation index based on the test set; and inputting real-time vibration acceleration monitoring data of the circulating water pump in the three-loop circulating cooling water system of the nuclear power plant into the target fatigue life prediction model, thereby determining a predicted fatigue life of the key components and the vulnerable components of the circulating water pump.

[0007] In one embodiment, the normalization preprocessing performed on the obtained vibration signal data set comprises: obtaining full life cycle vibration signal data about the simulated vulnerable component arranged on the reduced scale model of the circulating water pump, wherein the reduced scale model of the circulating water pump is a reduced scale model of the circulating water pump in a same scale, and the simulated vulnerable component is a simulated component of the vulnerable component in a same scale; selecting a vibration acceleration performance index of the simulated vulnerable component in a root mean square value vector; and calculating a normalized root mean square value of the vibration signal data set based on the vibration acceleration performance index.

[0008] In one embodiment, the normalization preprocessing performed on the obtained vibration signal data set further comprises: determining a time window length and a time step to determine a sliding time window for dividing the vibration signal data set; selecting a time point in the vibration signal data set, dividing samples before the selected time point into the training set and dividing samples after the selected time point into the test set; and dividing the validation set based on the training set using a sliding window segmentation.

[0009] In one embodiment, training the long short-term memory network comprising a plurality of fully connected layers based on the training set comprises: segmenting the training set into a plurality of time series segments based on the sliding time window and the time steps; determining input features, target outputs and single training input data batches of the long short-term memory network based on the plurality of time series segments and predefined remaining life labels; inputting the input features and target outputs into the real-time long short-term memory network based on the single training input data batches; and training the long short-term memory network comprising a plurality of fully connected layers based on the input features and target outputs, thereby obtaining the trained fatigue life prediction model.

[0010] In one embodiment, optimizing the trained fatigue life prediction model comprises: optimizing the fatigue life prediction model based on Adam algorithm and setting an abort command; performing recurrent neural network parameter optimization using grid search and cross-validation, thereby updating the recurrent neural network; in response to the validation set not improving the accuracy of the fatigue life prediction model, stopping training based on the abort command and saving the recurrent neural network under the current parameters, thereby constructing the optimized fatigue life prediction model.

[0011] In one embodiment, performing recurrent neural network parameter optimization using grid search and cross-validation comprises: determining the maximum depth, learning rate and parameter quantity of the search space in the grid search; optimizing the recurrent neural network based on the determined grid search; determining the objective function based on the root mean square error; performing cross-validation on the optimized recurrent neural network based on the determined objective function and test set.

[0012] In one embodiment, determining the optimized fatigue life prediction model that meets the performance evaluation index comprises: using the root mean square error as the performance evaluation index for determining the performance of the model.

[0013] In one embodiment, inputting the real-time vibration acceleration of the key parts and vulnerable parts of the circulating water pump into the fatigue life prediction model that meets the performance evaluation index comprises: determining the real-time vibration acceleration of the key parts and vulnerable parts of the circulating water pump corresponding to one sliding time window; calculating the root mean square value of the real-time vibration acceleration of the key parts and vulnerable parts of the circulating water pump; and inputting the calculated root mean square into the optimized fatigue life prediction model, thereby determining the predicted fatigue life of the key parts and vulnerable parts of the circulating water pump.

[0014] According to another aspect of the present disclosure, a device for predicting fatigue life of key components and wearing parts of circulating water pumps of nuclear power plants is disclosed, characterized in that the device comprises a data acquisition and processing unit, a calculation unit and a communication transmission unit, wherein the data acquisition and processing unit is configured to receive and process raw data from sensors to obtain a standardized vibration signal data set for the calculation unit; the calculation unit, wherein the calculation unit is configured to perform the method of any one of claims 1-8 based on the standardized vibration signal data set obtained by the data acquisition and processing module to determine the predicted fatigue life of the key components and wearing parts of the circulating water pumps; a database unit configured to store and manage data used and generated by the data acquisition and processing unit, the calculation unit and the communication transmission unit; and the communication transmission unit configured to transmit the data stored and managed in the database unit.

[0015] According to a third aspect of the present disclosure, a computing device is provided, comprising: a multi-core processor with at least four cores; and a memory communicatively connected with the at least one processor; the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method of the first aspect of the present disclosure.

[0016] In a fourth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to make a computer execute the method of the first aspect of the present disclosure.

[0017] It should be understood that the content described in this section is not intended to identify key or important features of embodiments of the present disclosure, nor is it used to limit the scope of the present disclosure. Other features of the present disclosure will become apparent from the following description. BRIEF DESCRIPTION OF DRAWINGS

[0018] The above and other features, advantages, and aspects of embodiments of the present disclosure will become more apparent by describing in detail exemplary embodiments thereof with reference to the attached drawings in which:

[0019] Figure 1 A schematic diagram of a system 100 for predicting fatigue life of key components and wearing parts of circulating water pumps according to embodiments of the present disclosure is shown.

[0020] Figure 2 A flowchart of predicting fatigue life of key components and wearing parts of circulating water pumps 200 according to embodiments of the present disclosure is shown.

[0021] Figure 3 A schematic diagram of a device 300 for predicting fatigue life of key components and wearing parts of circulating water pumps of nuclear power plants is disclosed.

[0022] Figure 4 A block diagram of an electronic device 400 for implementing embodiments of the present disclosure is shown. DETAILED DESCRIPTION

[0023] Exemplary embodiments of the present disclosure are described herein with reference to the accompanying drawings, in which various details of embodiments of the present disclosure are set forth to assist in the understanding of the present disclosure. It should be appreciated that the present disclosure can be embodied in various forms, and should not be construed as being limited only to the embodiments set forth herein. Rather, the exemplary embodiments of the present disclosure are provided so that this disclosure will be thorough and complete, and will fully convey the concept of the present disclosure to those skilled in the art.

[0024] As used herein, the term "includes" and its variants are meant to be open-ended and mean "comprising." Unless specifically stated, the term "or" means "and / or." The term "based on" means "based, at least in part, on." The term "one example embodiment" and "an embodiment" means "at least one example embodiment." The term "another embodiment" means "at least one additional embodiment." The terms "a first," "a second," etc. can refer to different or the same objects. Other explicitly and implicitly recited definitions can also be possible.

[0025] Figure 1 A schematic diagram of a system 100 for implementing fatigue life prediction of circulating water pump critical components and wearing parts according to embodiments of the present disclosure is shown. As shown in FIG. 1, the system 100 includes a computing device 110 and a circulating water pump system critical components and wearing parts fatigue life prediction system 130 and a network 140. The computing device 110, the water pump critical components and wearing parts fatigue life prediction system 130 can interact data through the network 140 (e.g., the Internet). Figure 1

[0026] The fatigue life prediction system 130, which can collect data of various operating condition parameters of the circulating water pump system and process the data of the operating condition parameters into a format that can be processed by the computing device 110. For example, the circulating water pump critical components and wearing parts fatigue life prediction system 130 can collect parameters such as vibration acceleration of the wearing parts of the circulating water pump system during operation.

[0027] ​Regarding the computing device 110, it is used to obtain data of the operating condition parameters of the circulating water pump system from the circulating water pump key component and vulnerable component fatigue life prediction system 130, for example. The computing device 110 can have one or more processing units, including special-purpose processing units such as GPUs, FPGAs, and ASICs, and general-purpose processing units such as CPUs. In addition, one or more virtual machines can also be running on each computing device 110. In some embodiments, the computing device 110 and the water pump key component and vulnerable component fatigue life prediction system 130 can be integrated together or can be separately arranged from each other. The computing device 110 can include the following modules:

[0028] The module unit 112, the obtaining module 112 is configured to obtain a vibration signal data set including vibration accelerations at each monitoring position of the circulating water pump based on a plurality of one-way vibration acceleration sensors on the circulating water pump;

[0029] The normalization module 114, the normalization module 114 is configured to normalize the obtained vibration signal data set for dividing the data set into a training set, a validation set, and a test set;

[0030] The training module 116, the training module 116 is configured to train a long short-term memory network including a plurality of fully connected layers based on the training set, thereby obtaining a trained fatigue life prediction model;

[0031] The verification module 118, the verification module 118 is configured to optimize the trained fatigue life prediction model based on the validation set and determine an optimized fatigue life prediction model meeting a performance evaluation index based on the test set; and

[0032] The input module 120, the input module 120 is configured to input real-time vibration accelerations of the circulating water pump key component and vulnerable component to the fatigue life prediction model meeting the performance evaluation index, thereby determining a predicted fatigue life of the circulating water pump key component and vulnerable component.

[0033] Figure 2 A flowchart of fatigue life prediction 200 of the circulating water pump key component and vulnerable component according to an embodiment of the present disclosure is shown. It should be understood that the method 200 can also include additional blocks not shown and / or can omit blocks shown, and the scope of the present disclosure is not limited in this regard.

[0034] At step 202, the computing device 110 can obtain a vibration signal data set including vibration accelerations at a plurality of monitoring positions of the circulating water pump reduced scale model based on a plurality of one-way vibration acceleration sensors arranged on the circulating water pump reduced scale model, the plurality of monitoring positions at least including positions within a predetermined range from a set position of a simulated vulnerable component on the circulating water pump reduced scale model.

[0035] First, full life cycle vibration signal data of the vulnerable parts on the scaled down model of the circulating water pump is acquired. The scaled down model of the circulating water pump can be a model water pump that is scaled down by a constant ratio (e.g., 10: 1) from the circulating water pump. The scaled down model of the circulating water pump has the same structure and vulnerable parts as the real circulating water pump. The vulnerable parts of the circulating water pump include rotating parts such as gears, bearings, impellers, etc. in the water pump. The key parts of the circulating water pump include key components such as bearings, bearing shells, gearboxes, impeller rings, etc. of the circulating water pump.

[0036] The scaled down model of the circulating water pump can have multiple sensors installed at multiple monitoring locations to acquire vibration data of the circulating water pump. The full life cycle vibration signal data about the simulated vulnerable parts provided on the scaled down model of the circulating water pump is acquired. The scaled down model of the circulating water pump is a scaled down model of the circulating water pump by a constant ratio. The simulated vulnerable parts are simulated components of the vulnerable parts by a constant ratio. The multiple monitoring locations include at least locations within a predetermined range from the provided locations of the simulated vulnerable parts on the scaled down model of the circulating water pump, so that the sensors can feed back the health of the vulnerable parts, i.e., the expected service life.

[0037] The data is collected by multiple one-way vibration acceleration sensors installed on the circulating water pump. In an embodiment, three one-way vibration acceleration sensors can be installed on key parts or vulnerable parts such as bearings of the circulating water pump of a nuclear power plant, and the three sensors are distributed at various key monitoring locations of the pump.

[0038] In an embodiment, the sensors can sample at a frequency of 10240 Hz, and the number of sampling samples can be 32768 points, so that the continuous sampling time is 3.2 seconds, ensuring that sufficient vibration signal data is collected. The collected vibration data (e.g., a single sample number of 32768 points) can be fed into the water pump vulnerable part failure algorithm or prediction algorithm provided below to determine the failure and life value of the vulnerable parts.

[0039] In an embodiment, the vibration acceleration represented by a root mean square value vector of the simulated vulnerable parts can be selected as a performance indicator. Based on the vibration acceleration performance indicator, the normalized root mean square value of the vibration signal data set is calculated.

[0040] In step 204, the computing device 110 can perform normalization preprocessing on the acquired vibration signal data set, so as to divide the data set into a training set, a validation set, and a test set.

[0041] In one embodiment, the full-life vibration signal data of the circulating water pump key components and wearing parts can be acquired. Then, the vibration acceleration represented by the root mean square value (RMS) is selected as the performance indicator. Based on these performance indicators, the normalized RMS value of the vibration signal data set can be calculated so that the data is comparable under different equipment and operating conditions.

[0042] Before deep learning, the data set needs to be divided. The division can adopt a sliding time window method. The time window length and time step are determined, for example, 64 data points are selected as the time window length, and a specific time point is selected in the vibration signal data set. The samples before the time point are divided into a training set, and the samples after the time point are divided into a test set. The sliding window segmentation method is used to further divide a validation set from the training set. The vibration acceleration represented by the root mean square value (RMS) is selected as the performance indicator.

[0043] Specifically, the RMS value of the vibration signal collected by each sensor can be calculated and normalized by the following formula:

[0044]

[0045] In formulas (1) and (2), n is the size of the single sample data, is the i-th RMS value, X rms is the RMS value vector. The training data set is obtained by a 64-length sliding time window.

[0046] For example, the time window length (such as 64 data points) and the time step (such as 1 data point) are determined to determine the sliding time window for dividing the vibration signal data set. A time point can be selected in the vibration signal data set, and the samples before the time point are divided into a training set, and the samples after the time point are divided into a test set. Through the sliding window segmentation method, a validation set can also be divided from the training set. Finally, the training set, the validation set and the test set with proportions of 0.7, 0.2 and 0.1 can be divided. In the present disclosure, the algorithm for life prediction does not need to be data enhanced and shuffled.

[0047] In step 206, the computing device 110 can train a long short-term memory network including a plurality of fully connected layers based on the training set, thereby obtaining a trained fatigue life prediction model.

[0048] In one embodiment, a long short-term memory network (LSTM) can be used as a prediction model. Based on the sliding time window and the time step, the data in the training set is cut into a plurality of time sequence segments. Each time sequence segment and its corresponding RUL label constitute the input feature and the target output of the LSTM network.

[0049] The long short-term memory network including a plurality of fully connected layers is a classical recurrent neural network LSTM. The long short-term memory network LSTM including a plurality of fully connected layers mainly comprises an input gate, a forgetting gate and an output gate.

[0050] Specifically, the long short-term memory network LSTM including a plurality of fully connected layers can be expressed by the following formula,

[0051] f t =σ(W f ·[h t-1 ,x t ]+b f ) (3)

[0052]

[0053] o t =σ(W o ·[h t-1 ,x t ]+b o ) (5)

[0054]

[0055] h t =o t *tanh(C t ) (7)

[0056] In formula (3), (4), (5), (6), (7), W f ,W C ,W o respectively represent weight matrices, b f ,b C ,b o respectively represent bias matrices, sigma (·) is a sigmoid activation function or a RuLU activation function, * represents element multiplication operation. x t ,C t ,h t respectively represent input, cell state and output state at t time, f t , o t , is a temporary calculation value at t time.

[0057] For the full life cycle experimental data set of the circulating pump, the LSTM model is not optimized. The kernel size of the LSTM model is 64, and then dropout = 0.4 is used to prevent overfitting, and 3 fully connected layers are used after the LSTM layer, and the neuron size is 64-32-1.

[0058] At step 208, the computing device 110 can optimize the trained fatigue life prediction model based on the validation set, and determine an optimized target fatigue life prediction model that meets the performance evaluation indicators based on the test set.

[0059] In one embodiment, the parameter update can be performed using the mini-batch stochastic gradient descent method with a batch size of 256, and the Adam optimizer can be used for optimization of the model parameters. Through grid search and cross-validation, the hyperparameters of the LSTM network, such as learning rate, number of layers, number of neurons, etc., can be adjusted during the training process to find the best network configuration. The optimization goal is to minimize the root mean square error (RMSE), which is a performance evaluation indicator for determining the performance of the model.

[0060] Specifically in the experiment, a batch size of 256 can be used in the feature extractor, and the Adam optimizer can be used to train 100 epochs to optimize the minimum loss.

[0061] The model parameters can be updated using the mini-batch stochastic gradient descent method. Mini-batch gradient descent is a compromise between batch gradient descent and stochastic gradient descent. In each iteration, it uses a small batch of training samples. The mini-batch stochastic gradient descent method can be represented by the following formula.

[0062] θ = θ - ηΔJ(θ; x (i:i+n) ; y (i:i+n) ) (8)

[0063] In formula (8), x (i:i+n) , y (i:i+n) is the i-th to the i+n-th sample. η is the learning rate, which is updated as follows:

[0064]

[0065] In formula (9), η0 is the initial learning rate, which is set to 0.05 in the experiment. k is a hyperparameter used to control the learning rate reduction amplitude, n epoc h is the number of training rounds. Mini-batch gradient descent can take advantage of the parallelism of matrix operations and avoid the computational cost problem when the number of training samples is large.

[0066] When the accuracy on the validation set no longer increases, the training is stopped, and the LSTM network under the current parameters is saved, thereby constructing an optimized fatigue life prediction model.

[0067] Specifically, in order to quantitatively illustrate the performance advantage of the proposed hybrid method, the root mean square error (RMSE) is selected to evaluate the prediction performance of the remaining life.

[0068]

[0069] y n and denote the true and predicted values of the remaining useful life, respectively, and N is the number of samples.

[0070] The scores are converted to a percentage of the prediction error based on the remaining useful life results of the different datasets. In equation (12), y n and denote the true and predicted values of the remaining useful life, respectively, and the percentage error for experiment i is defined as:

[0071]

[0072] In equation (12), y n and denote the true and predicted values of the remaining useful life, respectively.

[0073] The underestimation and overestimation of the prediction results can not be considered in the same way: good performance of the estimation is related to the early prediction of the remaining useful life (i.e. the case of %ERI>0), and the life generally predicted should not exceed the actual life. A more severe score deduction will be made for the case where the predicted value of the remaining useful life exceeds the actual remaining useful life (i.e. the case of %ERI<0). Therefore, the accuracy score of the estimation of the remaining useful life is defined as follows:

[0074]

[0075] The final score of all RUL estimations will be defined as the average of all experiment scores:

[0076]

[0077] The loss function in the experiment is:

[0078]

[0079] Further, the performance on the validation set is constantly monitored, and the RMSE is used as the performance evaluation index. When the performance on the validation set no longer improves, the early stopping rule is used to stop training to avoid overfitting, and the model parameters at this time are saved as the final prediction model.

[0080] At step 210, the computing device 110 can input the acquired real-time vibration acceleration monitoring data of the circulating water pump in the three-loop cooling water system of the nuclear power plant into the target fatigue life prediction model, so as to determine the predicted fatigue life of the key parts and vulnerable parts of the circulating water pump.

[0081] In one embodiment, the vibration signals of the circulating water pump are collected in real time, and the RMS value is calculated. These real-time data are input into the trained LSTM model, and the model will output the predicted fatigue life of the vulnerable parts.

[0082] Specifically, the RMS value of the vibration signal can be used as a health monitoring indicator. Then, the mapping learning of the monitoring indicator and the residual life value is completed through model training. Finally, the model parameter update is realized based on the loss function, and the model training is completed. In practical application, the RMS value of the monitoring signal at the current time (a time window) is input into the trained model, and the model outputs the residual life.

[0083] These prediction results obtained by the above technical means can be used to guide the maintenance team to carry out preventive maintenance, for example, if the predicted fatigue life is close to the critical value, the vulnerable parts can be replaced in time, so as to avoid potential failure and downtime risk.

[0084] Figure 3 A schematic diagram of an apparatus 300 for predicting the fatigue life of key parts and vulnerable parts of a circulating water pump of a nuclear power plant is disclosed. The apparatus 300 for predicting the fatigue life of key parts and vulnerable parts of a circulating water pump of a nuclear power plant comprises a data acquisition and processing unit 302, a calculation unit 304, a database unit 306 and a communication transmission unit 308. As shown, the data acquisition and processing unit is configured to receive and process raw data from sensors, thereby obtaining a standardized vibration signal data set for use by the calculation unit. Figure 3 As shown, the data acquisition and processing unit is configured to receive and process raw data from sensors, thereby obtaining a standardized vibration signal data set for use by the calculation unit.

[0085] In one embodiment, the data acquisition and processing unit is responsible for collecting, conditioning, converting, processing and outputting data to provide accurate and reliable data support for subsequent decision-making, control or analysis. The data acquisition and processing unit receives raw data from vibration sensors and eddy current sensors. It can collect data such as acceleration vibration, pump shaft surface jump and rotational speed on the device in real time. The data acquisition and processing unit processes the collected raw signals through hardware using Fourier transform to ensure the quality and accuracy of the signals. The data acquisition and processing unit converts the processed data into standardized protocols or formats, and currently mainly uses mqtt protocol for conversion.

[0086] The calculation unit is configured to perform the method as described above based on the standardized vibration signal data set obtained by the data acquisition and processing module, thereby determining the predicted fatigue life of the key parts and vulnerable parts of the circulating water pump.

[0087] In one embodiment, the computing unit 304 is mainly for the hardware platform requirements and software environment requirements needed to build the device. The computing unit can include a 16-core or 32-core CPU. Based on this CPU, the system can handle multiple threads at the same time, each core can handle a separate data stream, which can improve the performance of the system running concurrent applications. The computing unit can include a large memory such as 32G, thereby improving system performance, multitasking ability, reducing page file swapping, speeding up data processing, and avoiding phenomena such as freezing and crashing. The computing unit can include a large solid state drive such as not less than 2TB, thereby avoiding system congestion and meeting high-speed read-write capability. The computing unit can include a graphics card unit with cuda, thereby meeting the requirements of learning and computing using GPU in the algorithm. The computing unit can include computer equipment such as a motherboard, which has built-in usb interfaces, Ethernet interfaces, hdmi interfaces, vga interfaces, dvi interfaces, rs-232 interfaces, and also supports wireless communication interfaces such as wifi, 5G, and Bluetooth. The computing unit can use CentOS 7 operating system, for example. Based on an open source database, a database environment is built to achieve data storage.

[0088] In one embodiment, the computing unit 304 can be configured as an i7 and above multi-core processor, with 8 cores being optimal. The memory of the computing unit 304 is at least 32GB, and the hard disk is at least 2TB. The operating system can use CentOS V7.8. The monitoring and key parts and wear part life prediction part in the computing unit can run on linux, and the software will store a part of the collected vibration data in the database and transmit a part of the data to the algorithm program of the fault algorithm / life prediction device through the API interface, and feedback the results to other online monitoring software. The algorithm program of the fault algorithm / life prediction device provided by the present disclosure can also be operated by GPU.

[0089] The computing unit can include program tools for performing methods specifically for predicting the fatigue life of nuclear power plant circulating water pump super-high reliability equipment wear parts under specific working conditions, as described above in the present disclosure. The computing unit can analyze data input and processing by executing program tools for determining the predicted fatigue life of circulating water pump key parts and wear parts, which have received relevant data input, with the main data being vibration, rotating speed, and other original data during the operation of the circulating water pump. These data are processed by the algorithm inside the module to provide a basis for subsequent fatigue life prediction.

[0090] The computing unit can execute the fatigue life prediction algorithm by executing the program tool of the method for determining the predicted fatigue life of the circulating water pump key parts and vulnerable parts: according to the selection and calling of the algorithm interface, the data preprocessing is completed by the algorithm through the import of the basic data, and then the algorithm model analysis is performed to complete the diagnosis or life prediction of a certain fixed component. After the algorithm calculation, the module outputs the predicted fatigue life of the vulnerable parts, which is usually expressed in hours. For some specific needs, the algorithm results trend chart of the device part device situation can be realized, such as holographic spectrum and health state trend change. For the fault condition, a diagnosis report is generated in time, and the relevant personnel fill in the diagnosis opinion, which provides support for subsequent fault analysis and personnel training.

[0091] The database unit is configured to store and manage the data used and generated by the data acquisition and processing unit, the computing unit, and the communication transmission unit.

[0092] The database unit is a core component that provides data storage, retrieval, and management functions to ensure data integrity, security, and efficiency. The database unit is configured for data storage and management: the database module is responsible for storing various types of data, including structured data, unstructured data, and multimedia data, such as vibration data, with timestamp information; provides basic operations for adding, deleting, modifying, and querying data; supports data indexing. The database unit is configured for data integrity assurance: by defining data table structure, constraints, and triggers, etc., to ensure data integrity and accuracy. The database unit is configured for data security assurance: using encryption technology to protect data, supporting user authentication and permission management, providing data backup and recovery functions to prevent data loss or damage. The database unit can also be configured for data migration and synchronization: supporting data import and export functions, and the exported data does not have the problem of packet loss and data loss.

[0093] The communication transmission unit is configured to transmit the data stored and managed in the database unit.

[0094] In one embodiment, the data acquisition and processing unit 302 configured to integrate data acquisition and processing, the database unit 306 configured to store, the computing unit 304 configured to industrial computing, software programs, fault algorithms, and prediction algorithms, and the communication transmission unit 308 configured to 5G transmission can be integrated in the same host chassis. The chassis can correspond to the black box size commonly used in the industrial field, for example, it can be 500x400x300(mm). Those skilled in the art can adjust the black box size according to actual needs to integrate the above units.

[0095] In one embodiment, the communication transmission unit is to have common support for high-speed, low-latency data transmission, ensure the security of data transmission, and adapt to various application and service requirements. The communication transmission unit supports high-speed data upload and download, and lower latency, and for high-frequency vibration data, the use of 5G transmission technology can achieve data integrity and timeliness. The communication transmission unit can be connected to multiple devices, thereby supporting multiple devices to be connected and transmit data at the same time, and realizing the transmission of multi-measurement-point vibration and rotating speed data.

[0096] Figure 4 A block diagram of an electronic device 400 for implementing embodiments of the present disclosure is shown. The device 400 can be used to implement the computing device 110 of Figure 1 Figure 1. As shown, the device 400 includes a central processing unit (CPU) 401, which can perform various appropriate actions and processes according to computer program instructions stored in a read-only memory (ROM) 402 or loaded from a storage unit 408 into a random access memory (RAM) 403. In the RAM 403, various programs and data required for operation of the device 400 can also be stored. The CPU 401, the ROM 402, and the RAM 403 are connected to each other through a bus 404. An input / output (I / O) interface 405 is also connected to the bus 404.

[0097] Various components in the device 400 are connected to the I / O interface 405, including an input unit 406, such as a keyboard, a mouse, etc., an output unit 407, such as various types of displays, a speaker, etc., a storage unit 408, such as a magnetic disk, an optical disk, etc., and a communication unit 409, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 409 allows the device 400 to exchange information / data with other devices through a computer network, such as the Internet, and / or various telecommunications networks.

[0098] The central processing unit 401 performs various methods and processes described above, such as performing the method 200. For example, in some embodiments, the method 200 can be implemented as a computer software program, which is stored in a machine-readable medium, such as the storage unit 408. In some embodiments, part or all of the computer program can be loaded and / or installed on the device 400 via the ROM 402 and / or the communication unit 409. When the computer program is loaded into the RAM 403 and executed by the CPU 401, one or more operations of the method 200 described above can be performed. Alternatively, in other embodiments, the CPU 401 can be configured to perform one or more actions of the method 200 by any other appropriate means, such as by means of firmware.

[0099] The present disclosure can be a method, apparatus, system, and / or computer program product. The computer program product can include a computer-readable storage medium (or media) having computer readable program instructions thereon for performing various aspects of the present disclosure.

[0100] The computer readable storage medium can be a tangible device that can retain and store instructions for use by an instruction execution device. The computer readable storage medium can be, for example, but is not limited to, an electronic storage device, a magnetic storage device, an optical storage device, an electromagnetic storage device, a semiconductor storage device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium include the following: a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), a static random access memory (SRAM), a portable compact disc read-only memory (CD-ROM), a digital versatile disk (DVD), a memory stick, a floppy disk, a mechanically encoded device such as punch-cards or punched tape, a

[0101] The computer readable program instructions described herein can be downloaded to respective computing / processing devices from a computer readable storage medium or to an external computer or external storage device via a network, for example, the Internet, a local area network, a wide area network and / or a wireless network. The network can comprise copper transmission cables, optical transmission fibers, wireless transmission, routers, firewalls, switches, gateway computers and / or edge servers. A network adapter card or network interface in each computing / processing device receives computer readable program instructions from the network and forwards the computer readable program instructions for storage in a computer readable storage medium within the respective computing / processing device.

[0102] Computer readable program instructions for carrying out operations of the present disclosure can be assembler instructions, instruction-set-architecture (ISA) instructions, machine instructions, machine dependent instructions, microcode, firmware instructions, state-setting data, or either source code or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The computer readable program instructions can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider). In some embodiments, electronic circuitry including, for example, programmable logic circuitry, field-programmable gate array (FPGA), or programmable logic array (PLA) can execute the computer readable program instructions by utilizing state information of the computer readable program instructions to personalize the electronic circuitry, in order to perform aspects of the present disclosure.

[0103] The computer readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other device to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0104] The computer readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable apparatus or other device to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0105] The computer readable program instructions can also be loaded onto a computer, other programmable data processing apparatus, or other device to cause a series of operational steps to be performed on the computer, other programmable data processing apparatus or other device to produce a computer implemented process such that the instructions which execute on the computer, other programmable data processing apparatus, or other device implement the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0106] The flow diagrams and the block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each block in the flow diagrams and the block diagrams can represent a module, segment, or portion of instructions, which comprises one or more executable instructions for implementing the specified logical functions (‘instructions’). In some alternative implementations, the functions noted in the blocks can occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently or the blocks can sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and / or flowchart illustrations, and combinations thereof, can be implemented by special purpose hardware-based systems that perform the specified functions or acts, or combinations of special purpose hardware and

[0107] Embodiments of the present disclosure have been described above, and the description is intended to be illustrative of the embodiments of the disclosure and not exhaustive or limiting. Numerous modifications and adaptations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments. The scope of the disclosure is defined by the appended claims, rather than the description of the embodiments. The choice of words in the description is intended to best explain the principles of the embodiments of the disclosure, the practical application or technical improvement over the existing technology, or to enable others skilled in the art to understand the embodiments of the disclosure disclosed herein.

[0108] The above merely preferred embodiments of the present disclosure and are not intended to limit the present disclosure. The present disclosure can have various modifications and various changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present disclosure shall be included in the scope of the present disclosure.

Claims

1. A method for predicting fatigue life of key components and wearing parts of a circulating water pump, characterized in that: The circulating water pump is used in a three-loop circulating cooling water system of a nuclear power plant, and the method comprises: Based on a plurality of unidirectional vibration acceleration sensors disposed on a reduced-scale model of the circulating water pump, a vibration signal dataset including vibration accelerations at a plurality of monitoring positions on the reduced-scale model of the circulating water pump is obtained, wherein the plurality of monitoring positions at least include positions within a predetermined range from a location where a simulated wearing part on the reduced-scale model of the circulating water pump is disposed; Performing normalization preprocessing on the acquired vibration signal data set so as to divide the data set into a training set, a validation set, and a test set; Training a long short-term memory network including a plurality of fully connected layers based on the training set, thereby obtaining a trained fatigue life prediction model; Optimizing the trained fatigue life prediction model based on the validation set, and determining an optimized target fatigue life prediction model that meets the performance evaluation index based on the test set; and The real-time vibration acceleration monitoring data of the circulating water pump in the three-loop circulating cooling water system of the nuclear power plant is input into the target fatigue life prediction model to determine the predicted fatigue life of the key components and wearing parts of the circulating water pump. Among them, the normalization preprocessing of the obtained vibration signal data set includes: obtaining the full life cycle vibration signal data of the simulated wearing parts set on the reduced-scale model of the circulating water pump, the reduced-scale model of the circulating water pump is a proportionally reduced-scale model of the circulating water pump, and the simulated wearing parts are proportionally reduced simulated components of the wearing parts.

2. The method according to claim 1, characterized in that The normalization preprocessing of the acquired vibration signal data set also includes: Selecting a vibration acceleration performance metric expressed as a vector of RMS values ​​for the simulated consumables; and A normalized root mean square value of the vibration signal data set is calculated based on the vibration acceleration performance index.

3. The method according to claim 2, characterized in that The normalization preprocessing of the acquired vibration signal data set also includes: Determine the time window length and time step to determine the sliding time window for partitioning the vibration signal data set; Selecting a time point in the vibration signal data set, dividing the samples before the selected time point into a training set and dividing the samples after the selected time point into a test set; and The validation set is divided based on the training set using a sliding window split.

4. The method according to claim 3, characterized in that Training a long short-term memory network including multiple fully connected layers based on the training set includes: Dividing the training set into a plurality of time series segments based on the sliding time window and the time step; Determining input features, target outputs, and a single training input data batch for a long short-term memory network based on the multiple time series segments and the predefined remaining life labels; Inputting the input features and target output into a real-time long short-term memory network based on a single training data batch; and Based on the input features and target output, a long short-term memory network including multiple fully connected layers is trained to obtain a trained fatigue life prediction model.

5. The method according to claim 4, characterized in that Optimizing the trained fatigue life prediction model includes: Optimize the fatigue life prediction model based on the Adam algorithm and set the abort command; Use grid search and cross-validation to optimize the parameters of the recurrent neural network to update the recurrent neural network; In response to the fact that the accuracy of the validation set relative to the fatigue life prediction model does not increase, the training is stopped based on an abort command and the recurrent neural network under the current parameters is saved, thereby constructing an optimized target fatigue life prediction model.

6. The method according to claim 5, characterized in that Using grid search and cross validation to optimize recurrent neural network parameters includes: Determine the maximum depth of the search space, the learning rate, and the number of parameters in the grid search; Optimizing the recurrent neural network based on the determined grid search; Determine the objective function based on the root mean square error; Based on the determined objective function and test set, cross-validation is performed on the optimized recurrent neural network.

7. The method according to claim 1 or 6, characterized in that The optimized target fatigue life prediction model that meets the performance evaluation indicators is determined to include: The root mean square error was used as the performance evaluation metric to determine the model performance.

8. The method according to claim 7, characterized in that Inputting the real-time vibration acceleration of key components and wearing parts of the circulating water pump into the target fatigue life prediction model that meets the performance evaluation indicators includes: Determine the real-time vibration acceleration of the key components and wearing parts of the circulating water pump corresponding to a sliding time window; Calculate the real-time RMS value of vibration acceleration of key components and wearing parts of circulating water pumps; and The calculated RMS is input into the optimized target fatigue life prediction model to determine the predicted fatigue life of the key components and wearing parts of the circulating water pump.

9. A device for predicting fatigue life of key components and vulnerable parts of circulating water pumps in nuclear power plants, characterized in that: The device includes a data acquisition and processing unit, a calculation unit, a database unit and a communication transmission unit, wherein: a data acquisition and processing unit configured to receive and process raw data from the sensor, thereby obtaining a standardized vibration signal data set for use by the calculation unit; A calculation unit, wherein the calculation unit is configured to execute the method according to any one of claims 1 to 8 based on the standardized vibration signal data set obtained by the data acquisition and processing module, thereby determining the predicted fatigue life of key components and wearing parts of the circulating water pump; a database unit configured to store and manage data used and generated by the data acquisition and processing unit, the computing unit, and the communication and transmission unit; and The communication transmission unit is configured to transmit the data stored and managed in the database unit.

10. The device according to claim 9, characterized in that The calculation unit is further configured to provide diagnosis results of the circulating water pump wearing parts and key parts based on the fault algorithm and life prediction algorithm for the circulating water pump wearing parts and key parts, as well as an API interface for externally calling the diagnosis results.

11. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 8.

Images