Fan service life detection method and device, electronic equipment and storage medium

Abstract

The invention relates to the technical field of nuclear power plant equipment safety, in particular to a fan service life detection method and device, electronic equipment and a storage medium. According to the fan service life detection method provided by the embodiment of the invention, a pre-trained fan service life detection model is acquired; acquiring target running state data of the fan; performing feature extraction on the target operation state data to obtain target operation state features; and performing fan life detection on the target operation state characteristics through a pre-trained fan life detection model to obtain target fan life data. Therefore, the detection accuracy of the service life of the fan can be improved.

Description

Technical Field

[0001] The present application relates to the field of fan detection, and in particular, to a method and device for detecting the life of a fan, an electronic device, and a storage medium. Background Art

[0002] The detection of the fan life is a key technology for nuclear power plants to improve the wind energy utilization efficiency and ensure the safe operation of fan equipment. However, with the complexity of the fan operating environment, the diversity of fan failure modes, and the increase in fan data, the difficulty of detecting the life of the fan equipment in nuclear power plants is increasing continuously, and the safe operation of the fan equipment in nuclear power plants may be affected.

[0003] Currently, the detection of the fan life usually predicts the remaining service life of the fan equipment through a machine learning model (such as a support vector machine, a random forest, or a linear regression, etc.). However, this method requires manual selection of the characteristic indexes of the fan operation data and also ignores the time characteristics of the fan operation, so that the accuracy of the fan life detection is limited by the selection of the characteristic indexes, resulting in a low accuracy of the fan life detection. Therefore, how to improve the accuracy of the fan life detection is still a difficult problem to be solved in the industry. Summary of the Invention

[0004] The present application aims to at least solve one of the technical problems existing in the prior art. For this purpose, the present application provides a method and device for detecting the life of a fan, an electronic device, and a storage medium, which can improve the accuracy of the fan life detection.

[0005] According to the fan life detection method of the first aspect embodiment of the present application, it includes:

[0006] Obtain a pre-trained fan life detection model; wherein, the pre-training process of the fan life detection model includes:

[0007] Obtain the prior operation state data of the fan;

[0008] Perform fan life distribution prediction on the prior operation state data through a preset Weibull cumulative distribution function to obtain prior life distribution data;

[0009] Train the initial fan life detection model based on the prior life distribution data and the prior operation state data to obtain the pre-trained fan life detection model;

[0010] Obtain the target operation state data of the fan;

[0011] Extract features from the target operation state data to obtain target operation state features;

[0012] Performing fan life detection on the target operating state features through the pre-trained fan life detection model to obtain target fan life data.

[0013] According to some embodiments of the present application, training the initial fan life detection model based on the prior life distribution data and the prior operating state data to obtain the pre-trained fan life detection model includes:

[0014] Obtaining the initial fan life detection model;

[0015] Performing fan life prediction on the prior operating state data through the fan life detection model to obtain predicted fan life data;

[0016] Performing fan life distribution prediction on the predicted fan life data through the Weibull cumulative distribution function to obtain predicted life distribution data;

[0017] Obtaining the true fan life data of the prior operating state data, and calculating the target loss value of the predicted fan life data, the predicted life distribution data, the true fan life data, and the prior life distribution data according to a preset loss function;

[0018] Updating the model parameters of the fan life detection model based on the target loss value, and returning to perform fan life prediction on the prior operating state data through the fan life detection model until the fan life detection model meets the preset training conditions to obtain the pre-trained fan life detection model.

[0019] According to some embodiments of the present application, the loss function includes a neural network loss function and a Weibull loss function;

[0020] The calculating the target loss value of the predicted fan life data, the predicted life distribution data, the true fan life data, and the prior life distribution data according to a preset loss function includes:

[0021] Calculating the life loss value between the predicted fan life data and the true fan life data according to the neural network loss function;

[0022] Calculating the life distribution loss value between the predicted life distribution data and the prior life distribution data according to the Weibull loss function;

[0023] Integrating the life loss value and the life distribution loss value through a preset weight hyperparameter to obtain the target loss value.

[0024] According to some embodiments of the present application, the prediction of the fan life distribution for the prior operating state data through a preset Weibull cumulative distribution function to obtain prior life distribution data includes:

[0025] Determine the fan shape characterization parameter based on the prior operating state data;

[0026] Determine the fan scale characterization parameter according to the fan shape characterization parameter and the prior operating state data;

[0027] Predict the fan life distribution for the prior operating state data according to the fan shape characterization parameter and the fan scale characterization parameter to obtain the prior life distribution data.

[0028] According to some embodiments of the present application, the determination of the fan scale characterization parameter according to the fan shape characterization parameter and the prior operating state data includes:

[0029] Obtain the fan failure time, fan failure quantity, and fan failure variable of the prior operating state data;

[0030] Perform scale characterization calculation on the fan failure time, the fan failure quantity, the fan failure variable, and the fan shape characterization parameter to obtain the fan scale characterization parameter.

[0031] According to some embodiments of the present application, the detection of the target fan life for the target operating state feature through the pre-trained fan life detection module to obtain target fan life data includes:

[0032] Detect the fan health state of the target operating state feature through the fan life detection model to obtain fan health degree data;

[0033] Detect the remaining life of the fan for the target operating state feature according to the fan health degree data to obtain the target fan life data.

[0034] According to some embodiments of the present application, the detection of the fan health state of the target operating state feature through the fan life detection model to obtain fan health degree data includes:

[0035] Extract the fan operation health feature of the target operating state feature through the fan life detection model to obtain the fan operation health feature;

[0036] Evaluate the fan health degree of the fan operation health feature according to a preset linear function to obtain a fan health evaluation score;

[0037] Determine the fan health degree data according to the fan health evaluation score.

[0038] According to some embodiments of the present application, extracting the fan operation health characteristics from the target operation state characteristics through the fan life detection model to obtain the fan operation health characteristics includes:

[0039] Performing feature encoding on the target operation state characteristics through the fan life detection model to obtain encoded operation characteristics;

[0040] Performing health characteristic weighting processing on the encoded operation characteristics to obtain weighted healthy operation characteristics;

[0041] Decoding the weighted healthy operation characteristics to obtain the fan operation health characteristics.

[0042] According to some embodiments of the present application, detecting the remaining life of the target fan from the target operation state characteristics according to the fan health degree data to obtain the target fan life data includes:

[0043] Obtaining the health degree weight and health degree offset term of the fan health degree data;

[0044] Performing fan operation fault detection on the target operation state characteristics according to the health degree weight, the health degree offset term, and the fan health degree data to obtain fan operation fault data;

[0045] Determining the target fan life data according to the fan operation fault data.

[0046] According to some embodiments of the present application, extracting the target operation state characteristics from the target operation state data includes:

[0047] Performing Fourier transform processing on the target operation state characteristics to obtain at least one operation state spectrum characteristic;

[0048] Selecting the operation state spectrum characteristic corresponding to the maximum amplitude from at least one of the operation state spectrum characteristics as the target operation state characteristic.

[0049] According to some embodiments of the present application, selecting the operation state spectrum characteristic corresponding to the maximum amplitude from at least one of the operation state spectrum characteristics as the target operation state characteristic includes:

[0050] Performing frequency band division on at least one of the operation state spectrum characteristics to obtain at least one operation state frequency band;

[0051] Selecting the operation state frequency band corresponding to the maximum amplitude from at least one of the operation state frequency bands as the target operation state characteristic.

[0052] According to some embodiments of the present application, before extracting features from the target operating state data to obtain target operating state features, it further includes:

[0053] Obtain initial operating state data, and perform data cleaning on the initial operating state data to obtain cleaned operating state data;

[0054] Perform normalization processing on the cleaned operating state data to obtain normalized operating state data;

[0055] Divide the normalized operating state data through a preset fan operating time period to obtain the target operating state data.

[0056] According to an embodiment of the second aspect of the present application, a fan life detection device includes:

[0057] A fan life detection model acquisition module, configured to acquire a pre-trained fan life detection model;

[0058] A fan operating state data acquisition module, configured to acquire target operating state data of a fan;

[0059] A feature extraction module, configured to extract features from the target operating state data to obtain target operating state features;

[0060] A fan life detection module, configured to perform fan life detection on the target operating state features through the pre-trained fan life detection model to obtain target fan life data.

[0061] In a third aspect, an embodiment of the present application provides an electronic device, including: a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, it implements the fan life detection method according to any one of the embodiments of the first aspect of the present application.

[0062] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, the storage medium stores a program, and when the program is executed by a processor, it implements the fan life detection method according to any one of the embodiments of the first aspect of the present application.

[0063] The fan life detection method, its device, electronic equipment, and storage medium according to the embodiments of the present application have at least the following beneficial effects: obtaining a pre-trained fan life detection model; wherein, the pre-training process of the fan life detection model includes: obtaining prior operating state data of the fan; predicting the fan life distribution for the prior operating state data through a preset Weibull cumulative distribution function to obtain prior life distribution data; training an initial fan life detection model based on the prior life distribution data and the prior operating state data to obtain a pre-trained fan life detection model; obtaining target operating state data of the fan; extracting features from the target operating state data to obtain target operating state features; performing fan life detection on the target operating state features through the pre-trained fan life detection model to obtain target fan life data. In this way, the accuracy of fan life detection can be improved.

[0064] Additional aspects and advantages of the present application will be given in part in the following description, become apparent in part from the following description, or be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] The above and / or additional aspects and advantages of the present application will become apparent and be readily understood from the description of the embodiments in conjunction with the following drawings, wherein:

[0066] Figure 1 is a schematic flowchart of the fan life detection method provided by the embodiments of the present application;

[0067] Figure 2 is Figure 1 a flowchart of step S101 in

[0068] Figure 3 is Figure 2 a flowchart of step S202 in

[0069] Figure 4 is Figure 1 a flowchart of step S101 in

[0070] Figure 5 is Figure 4 a flowchart of step S404 in

[0071] Figure 6 is another schematic flowchart of the fan life detection method provided by the embodiments of the present application;

[0072] Figure 7 is Figure 1 a flowchart of step S103 in

[0073] Figure 8 is Figure 7 a flowchart of step S702 in

[0074] Figure 9 For Figure 1 the flowchart of step S104 in

[0075] Figure 10 For Figure 9 the flowchart of step S901 in

[0076] Figure 11 For Figure 10 the flowchart of step S1001 in

[0077] Figure 12 For For Figure 9 the flowchart of step S902 in

[0078] Figure 13 is a schematic structural diagram of the fan life detection device provided by an embodiment of the present application;

[0079] Figure 14 is a schematic hardware structure diagram of the electronic device provided by an embodiment of the present application. Specific Embodiments

[0080] The embodiments of the present application will be described in detail below. The examples of the embodiments are shown in the drawings, where the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the drawings are exemplary and are only used to explain the present application and should not be construed as limiting the present application.

[0081] In the description of the present application, the meaning of several is one or more, the meaning of multiple is two or more, greater than, less than, exceeding, etc. are understood as not including the number itself, and above, below, within, etc. are understood as including the number itself. If there is a description of first and second, it is only for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features or implicitly indicating the sequence of the indicated technical features.

[0082] In the description of the present application, it should be understood that for the orientation description, such as up, down, left, right, front, back, etc., the orientation or position relationship indicated is based on the orientation or position relationship shown in the drawings, and is only for the convenience of describing the present application and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be construed as limiting the present application.

[0083] In the description of this specification, the descriptions referring to terms such as "one embodiment", "some embodiments", "schematic embodiments", "examples", "specific examples", or "some examples" etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0084] In the description of the present application, it should be noted that unless otherwise clearly defined, words such as "set", "installed", "connected", etc. should be understood in a broad sense, and those skilled in the art can reasonably determine the specific meanings of the above words in the present application in combination with the specific content of the technical solution. In addition, the identification of specific steps hereinafter does not represent a limitation on the step sequence and execution logic. The execution sequence and execution logic between each step should be understood and inferred with reference to the content described in the embodiment.

[0085] The detection of the fan life is a key technology for nuclear power plants to improve the wind energy utilization efficiency and ensure the safe operation of fan equipment. However, with the complexity of the fan operation environment, the diversity of fan failure modes, and the increase in fan data, the difficulty of detecting the life of fan equipment in nuclear power plants is constantly increasing, and there may be situations where the safe operation of the fan equipment in nuclear power plants is affected.

[0086] Currently, the detection of fan life usually predicts the remaining service life of fan equipment through machine learning models (such as support vector machines, random forests, or linear regression, etc.). However, this method requires manual selection of the characteristic indicators of fan operation data and also ignores the time characteristics of fan operation, making the accuracy of fan life detection limited by the selection of characteristic indicators and resulting in a relatively low accuracy rate of fan life detection. Therefore, how to improve the accuracy rate of fan life detection remains a difficult problem urgently to be solved in the industry.

[0087] Therefore, using a fan life detection model to perform fan life detection on the target operation state data of the fan to improve the accuracy rate of fan life detection can automatically determine the key characteristic indicators related to fan life detection, without the need for manual selection of the characteristic indicators of fan operation data, and combines the time series of fan operation, which helps to improve the accuracy rate of fan life detection.

[0088] This application aims to solve at least one of the technical problems existing in the prior art. For this purpose, this application proposes a fan life detection method, its device, electronic equipment, and storage medium, which can improve the accuracy rate of fan life detection.

[0089] The following is a further description based on the accompanying drawings:

[0090] Reference Figure 1 , according to the fan life detection method of the embodiments of the present application, it may include, but is not limited to:

[0091] Step S101, obtaining a pre-trained fan life detection model;

[0092] Step S102, obtaining the target operating state data of the fan;

[0093] Step S103, extracting features from the target operating state data to obtain the target operating state features;

[0094] Step S104, performing fan life detection on the target operating state features through the pre-trained fan life detection model to obtain the target fan life data.

[0095] The fan life detection method shown by steps S101 to S104 of the embodiments of the present application needs to obtain a pre-trained fan life detection model; wherein, the pre-training process of the fan life detection model includes: obtaining the prior operating state data of the fan; performing fan life distribution prediction on the prior operating state data through a preset Weibull cumulative distribution function to obtain the prior life distribution data; training the initial fan life detection model based on the prior life distribution data and the prior operating state data to obtain the pre-trained fan life detection model; obtaining the target operating state data of the fan; extracting features from the target operating state data to obtain the target operating state features; performing fan life detection on the target operating state features through the pre-trained fan life detection model to obtain the target fan life data. In this way, the accuracy of fan life detection can be improved.

[0096] In step S101 of some embodiments, specifically, the pre-trained fan life detection model is obtained by training with the prior life distribution data and the prior operating state data, and the fan life detection model can be a neural network model based on the attention mechanism.

[0097] Specifically, the pre-training process of the fan life detection model includes: obtaining the prior operating state data of the fan; performing fan life distribution prediction on the prior operating state data through a preset Weibull cumulative distribution function to obtain the prior life distribution data; training the initial fan life detection model based on the prior life distribution data and the prior operating state data to obtain the pre-trained fan life detection model.

[0098] Further, the prior operating state data of the fan can be the prior operating knowledge of the fan collected from the sensors of the fan generator set in the nuclear power plant. This prior operating knowledge can be derived from a large number of experimental data, industry standards, or long-term operating experience summaries of the same type of fan in the past. These data reflect the operating conditions and performance indicators of the fan at different time points, and the prior operating state data can include, but is not limited to: fan bearing temperature, fan motor temperature, fan vibration data, wind speed, fan blade speed, internal pressure data of the fan, torque and force on the fan shaft, lubricating oil viscosity, ambient humidity, and ambient temperature, etc.

[0099] Referring to Figure 2 , according to some embodiments of the present application, in step S101, the prior operating state data is used to predict the fan life distribution through a preset Weibull cumulative distribution function, and prior life distribution data can be obtained, including, but not limited to:

[0100] Step S201, determining the fan shape characterization parameter based on the prior operating state data;

[0101] Step S202, determining the fan scale characterization parameter according to the fan shape characterization parameter and the prior operating state data;

[0102] Step S203, predicting the fan life distribution of the prior operating state data according to the fan shape characterization parameter and the fan scale characterization parameter, and obtaining prior life distribution data.

[0103] In some embodiments of the present application, the Weibull cumulative distribution function is a commonly used statistical tool, which can predict the life distribution of the fan according to the prior operating state data of the fan. In step 101, this calculation process includes several key operations:

[0104] In step S201, specifically, the fan shape characterization parameter is one of the key parameters of the Weibull distribution, which describes the fan failure trend changing with time.

[0105] Further, when the fan shape characterization parameter k = 1, it means that the fan failure rate remains constant over time, and the fan may have random failures; when the fan shape characterization parameter k > 1, it means that the fan failure rate increases with time, and the fan failure may be related to wear or aging; when the fan shape characterization parameter k < 1, it means that the fan failure rate decreases with time, which may indicate that the fan belongs to initial failure.

[0106] For example, for type A fans, through historical data statistical analysis, it is known that its fan shape characterization parameter is usually between 1.5 - 2.5, indicating that the fan failure rate increases or decreases with time, and the fan failure may be related to wear or aging.

[0107] In this embodiment, the shape characterization parameters of the fan are determined based on the prior operating state data, which facilitates predicting when the fan in the nuclear power plant needs to be repaired or replaced subsequently, and helps improve the efficiency and reliability of wind power generation.

[0108] Step S202 of some embodiments refers to Figure 3 , according to some embodiments of the present application, in step S202, determining the scale characterization parameters of the fan based on the shape characterization parameters of the fan and the prior operating state data may include, but is not limited to:

[0109] Step S301, obtaining the fan failure time, the number of fan failures, and the fan failure variables of the prior operating state data;

[0110] Step S302, performing scale characterization calculations on the fan failure time, the number of fan failures, the fan failure variables, and the shape characterization parameters of the fan to obtain the scale characterization parameters of the fan.

[0111] In some embodiments of the present application, when determining the scale characterization parameters of the fan, it is necessary to obtain the fan failure time, the number of fan failures, and the fan failure variables in the prior operating state data to obtain the operating conditions of the fan, which is the basis for predicting the fan life distribution subsequently. In step 202, this calculation process includes several key operations:

[0112] In step S301, specifically, the fan failure time refers to the specific time point or time period from when the fan starts running to when it fails.

[0113] Specifically, the number of fan failures records the total number of fan failure devices that occurred within a specific time period.

[0114] Specifically, the fan failure variable refers to the number of fan failures plus incomplete operating failure samples.

[0115] Specifically, the fan failure time, the number of fan failures, and the fan failure variables are all obtained from the prior operating state data.

[0116] In step S302, specifically, the scale characterization parameters of the fan represent the scale of the fan life distribution or the fan failure time in the Weibull distribution. The scale characterization parameters of the fan play a role in magnifying or shrinking the Weibull distribution curve, but it does not affect the shape of the distribution, and the smaller the scale characterization parameters of the fan, the faster the failure rate of the fan increases.

[0117] Specifically, the fan failure time records the specific time points and time periods from the installation and operation of the fan to the occurrence of a failure. The distribution of fan failures can be observed. The number of fan failures helps to understand the failure frequency of the fan, and the fan failure variable helps us to more comprehensively understand the failure mode of the fan. Together, these data reflect the actual lifespan of the fan under different operating conditions.

[0118] Specifically, scale characterization calculations are performed on the fan failure time, the number of fan failures, the fan failure variable, and the fan shape characterization parameter through the fan scale characterization parameter formula. The weighted average of all fan failure times is considered, which is applicable to processing incomplete fan data, that is, there may not be enough data at the end of the last observation period to determine the failure status of all fan components, improving the accuracy of subsequent fan lifespan detection.

[0119] The embodiments of the present application shown in steps S301 to S302 can be used together with the fan shape characterization parameter in the subsequent Weibull distribution function to predict the remaining service life of the fan, realizing the conversion of the fan operation state data into a scientific prediction of the fan lifespan, providing important information support for the operation and management of the fans in nuclear power plants.

[0120] In step S203, specifically, the prior lifespan distribution data refers to the shape and scale of the prior lifespan Weibull distribution diagram of the fan that jointly determines the prior operation state data through the fan shape characterization parameter and the fan scale characterization parameter.

[0121] Specifically, the cumulative distribution function of the Weibull distribution can predict the lifespan distribution based on the fan operation state data. This function gives the probability that the fan will fail before a given time point or time period, realizing the prediction of the failure probability of different fans at different time points, and thus realizing the prediction of the remaining service life of the fan.

[0122] Furthermore, during the prediction process, considering various conditions of fan operation, such as temperature, vibration, rotational speed, etc., which may affect the lifespan of the fan. By combining these conditions with the cumulative distribution function of the Weibull distribution, the lifespan distribution of the fan under specific operating conditions can be predicted more accurately.

[0123] In this embodiment, the fan lifespan distribution is predicted based on the fan shape characterization parameter and the fan scale characterization parameter for the prior operation state data, reflecting the expected lifespan of the prior fan under different operating conditions, realizing the quantitative assessment of the health status of the fan, and facilitating the improvement of the accuracy of subsequent fan lifespan detection.

[0124] The embodiments of the present application shown in steps S201 to S203 can provide a scientific basis for the life prediction of the fan, help the maintenance team better plan the maintenance strategy, thereby improving the operation efficiency and reliability of the fan. Finally, these prior life distribution data will be used to train and verify the fan life prediction model to ensure that the model can accurately predict the remaining service life of the fan.

[0125] In some more specific embodiments of the present application, the fan scale characterization parameter can be expressed by the following formula:

[0126]

[0127] where η represents the fan scale characterization parameter, n represents the fan failure variable, t m represents the fan failure time when the mth fan failure occurs, β represents the fan shape characterization parameter, and r represents the number of fan failures.

[0128] Furthermore, the prior life distribution data can be expressed by the following formula:

[0129]

[0130] where F(t) represents the prior life distribution data of the fan failure time t, t represents the fan failure time, η represents the fan scale characterization parameter, and β represents the fan shape characterization parameter.

[0131] In step S101, the embodiments of the present application train the initial fan life detection model based on the prior life distribution data and the prior operating state data, which can continuously improve the prediction effect of the fan life detection model by combining the prior data, and help improve the accuracy of the subsequent model in predicting the remaining life of the fan. This process can include the following sub-steps.

[0132] Referring to Figure 4 , according to some embodiments of the present application, step S101 trains the initial fan life detection model based on the prior life distribution data and the prior operating state data to obtain a pre-trained fan life detection model, which can include, but is not limited to:

[0133] Step S401, obtaining the initial fan life detection model;

[0134] Step S402, predicting the fan life through the fan life detection model for the prior operating state data to obtain the predicted fan life data;

[0135] Step S403, predicting the fan life distribution for the predicted fan life data through the Weibull cumulative distribution function to obtain the predicted life distribution data;

[0136] Step S404: Obtain the true fan life data of the prior operating state data, and calculate the target loss value of the predicted fan life data, the predicted life distribution data, the true fan life data, and the prior life distribution data according to a preset loss function;

[0137] Step S405: Update the model parameters of the fan life detection model based on the target loss value, and return to perform fan life prediction on the prior operating state data through the fan life detection model until the fan life detection model meets the preset training conditions, and obtain a pre-trained fan life detection model.

[0138] In step S401 of some embodiments, specifically, the initial fan life detection model refers to an unoptimized neural network model, and this initial fan life detection model can be a neural network model based on an attention mechanism.

[0139] Specifically, through the initial fan life detection model, the life prediction of the prior operating state data of the fan can be performed, and the life prediction result of the fan can be output.

[0140] In step S402 of some embodiments, specifically, use this model to perform attention mechanism processing on the prior operating state data to achieve the extraction of prior health features of the prior operating data, and evaluate the health degree of the fan based on the prior health features, and predict the fan life based on the health degree of the fan.

[0141] Specifically, through the fan life detection model, the relationship between the prior operating state data and the fan life can be learned, and the key features affecting the fan life can be identified, thereby improving the accuracy of the fan life distribution prediction.

[0142] For example, the fan life detection model may learn that at specific temperatures and vibration levels, the failure risk of the fan increases, or when the wind speed is too high, the life of the fan may be shortened. Through this information, the fan life detection model can predict the remaining service life of the fan and generate predicted fan life data.

[0143] In step S403 of some embodiments, specifically, the predicted life distribution data refers to jointly determining the shape and scale of the predicted prior fan life Weibull distribution diagram of the prior operating state data through the fan shape characterization parameter and the fan scale characterization parameter.

[0144] Specifically, performing fan life distribution prediction on the predicted fan life data through the Weibull cumulative distribution function helps to more comprehensively understand the statistical characteristics of the fan life, including key indicators such as the average life and the failure rate, thereby realizing the scoring of the predicted fan life data, and determining whether the initial fan life detection model needs to be trained according to this score.

[0145] In step S404 of some embodiments, specifically, the real fan life data refers to the specific operation time data of the fan during the entire period from installation, commissioning to final failure or retirement during actual operation.

[0146] For example, when the fan has been operating for 5 years and 3 months after installation and fails due to bearing damage, the time length of 5 years and 3 months is the real life data of the fan; for a fan that has not failed but is retired due to other reasons (such as technological update, efficiency reduction), its total operation time from installation to retirement is also real life data; when the power generation of the fan drops from 10 million kWh per year initially to 9 million kWh in the 3rd year, the change data of the power generation can also reflect the aging process of the fan.

[0147] Referring to Figure 5 , according to some embodiments of the present application, step S404 calculates the target loss value of the predicted fan life data, the predicted life distribution data, the real fan life data and the prior life distribution data according to a preset loss function, which may include, but is not limited to:

[0148] Step S501, calculating the life loss value between the predicted fan life data and the real fan life data according to the neural network loss function;

[0149] Step S502, calculating the life distribution loss value between the predicted life distribution data and the prior life distribution data according to the Weibull loss function;

[0150] Step S503, integrating the life loss value and the life distribution loss value through a preset weight hyperparameter to obtain the target loss value.

[0151] In some embodiments of the present application, the target loss value is a key indicator for measuring the difference between the predicted value and the real value, and is used to guide the parameter optimization during the model training process. During the training process of the fan life detection model, it is not only necessary to measure the accuracy of a single predicted value through the loss function, but also necessary to measure the situation of the entire fan life distribution through the loss function. Therefore, the loss function includes the neural network loss function and the Weibull loss function. In step S404, this calculation process includes several key operations:

[0152] In step S501 of some embodiments, specifically, the life loss value refers to the loss value between the predicted fan life data and the real fan life data.

[0153] Specifically, the neural network loss function can be the MES (Mean Squared Error) function.

[0154] In this embodiment, the life loss value between the predicted fan life data and the actual fan life data is calculated according to the neural network loss function, which facilitates continuously adjusting the parameters of the fan life detection model based on the life loss function to minimize the life loss value, and helps improve the accuracy of fan life detection.

[0155] In step S502 of some embodiments, specifically, the life distribution loss value refers to the loss value between the predicted life distribution data and the prior life distribution data.

[0156] Specifically, the Weibull loss function, as a prior constraint condition, is used to measure the consistency between the predicted life distribution and the prior life distribution based on historical data, and can further evaluate the accuracy of the fan life predicted by the life distribution model.

[0157] In this embodiment, the life distribution loss value between the predicted life distribution data and the prior life distribution data is calculated according to the Weibull loss function, which helps ensure that the model not only performs well in individual predictions, but also can accurately reflect the life characteristics of the fan in the overall distribution.

[0158] In step S503 of some embodiments, specifically, the target loss function includes the neural network loss function and the Weibull loss function.

[0159] Specifically, the life loss value and the life distribution loss value can be integrated through preset weight hyperparameters to balance the contributions of the two loss values to the target loss value and obtain the final target loss value. The integration of the loss function enables the fan life detection model to simultaneously consider the accuracy of individual fan life predictions and the consistency of the overall fan life distribution during optimization. Among them, the setting of the weight hyperparameters can be adjusted according to the specific requirements of the actual task.

[0160] The embodiment of the present application provided via steps S501 to S503 comprehensively considers the performance of the fan life detection model in predicting the life of an individual fan and the overall fan life distribution, so that during the training process of the fan life detection model, an optimization target is provided to the model, and the model will continuously update its parameters to minimize the target loss value, reduce the error of the fan life detection model prediction, and thus improve the accuracy of fan life prediction.

[0161] In step S405 of some embodiments, specifically, the connection weights between the neurons of the fan life detection model can be adjusted through the backpropagation algorithm to achieve adjusting the model parameters by minimizing the loss function. After updating the parameters, the model continuously learns based on the running state characteristic data and the known life labels, and repeats the above process until the performance of the model meets the preset training conditions (such as reaching a certain accuracy rate or the target loss value is less than the preset loss threshold), and a pre-trained fan life detection model is obtained.

[0162] For example, the predicted fan life data corresponding to the operating state characteristic data of the fan (such as high rotational speed, large vibration amplitude, high temperature) is relatively short. The model adjusts the connection weights so that when the model encounters similar operating state characteristic data, it can predict a shorter fan life.

[0163] In this embodiment, through multiple iterations and parameter updates of the model, a pre-trained fan life detection model is obtained. This model can accurately predict the remaining service life of the fan, and its prediction result matches the prior fan life data, ensuring the accuracy of fan life detection, thereby providing support for the sustainable development of the nuclear power plant wind power industry.

[0164] Through the embodiments of the present application provided by steps S401 to S405, the fan life detection model can be continuously adjusted and optimized during the training process, reducing the prediction error. And through the use of the comprehensive loss function, the fan life detection model can more comprehensively learn the fan fault characteristics related to the fan life, improving the accuracy of fan life detection.

[0165] In some more specific embodiments of the present application, the loss function can be expressed by the following formula:

[0166]

[0167] Among them, represents the loss function of the predicted fan life data and the true fan life data y. n represents the number of samples of the prior operating state data, y i represents the i-th true fan life data, represents the i-th predicted fan life data, λ represents the weight hyperparameter, used to balance the contributions of the neural network loss function and the Weibull loss function in the loss function, F(y i ) represents the prior life distribution data of the i-th true fan life data, F(y i ) represents the predicted life distribution data of the i-th predicted fan life data.

[0168] In step S102 of some embodiments, specifically, the target operating state data refers to the operating state data of the fan that needs to be detected for the fan life in the nuclear power plant, and can be the bearing temperature, motor temperature, vibration data, wind speed, fan blade rotational speed, internal pressure data of the fan, torque and force on the fan shaft, lubricating oil viscosity, ambient humidity, and ambient temperature of the fan collected from the sensors of the nuclear power plant fan generator set, etc.

[0169] Refer to Figure 6, according to some embodiments of the present application, before step S103 extracts features from the target operating state data to obtain the target operating state features, the fan life detection method may further include, but is not limited to:

[0170] Step S601, obtain the initial operating state data, and perform data cleaning on the initial operating state data to obtain the cleaned operating state data;

[0171] Step S602, perform normalization processing on the cleaned operating state data to obtain the normalized operating state data;

[0172] Step S603, divide the normalized operating state data through a preset fan operating time period to obtain the target operating state data.

[0173] In some embodiments of the present application, by preprocessing the operating state data of the fan, it is possible to clean and normalize the operating state data, thereby improving the data quality of the operating state data of the fan. Before step S103, the cleaning and normalization process includes the following key operations:

[0174] In step S601 of some embodiments, specifically, the initial operating state data refers to the data that has not been preprocessed, and the initial operating state data can be sourced from various sensors of the fan (such as temperature sensors, vibration sensors, speed sensors, etc.).

[0175] Specifically, data cleaning aims to remove errors and inconsistencies in the data to improve data quality. It includes identifying and processing missing values in the initial operating state data, which can be completed by interpolation or deleting fan operation records containing missing values. At the same time, identify and process outliers in the initial operating state. These values may be extreme values caused by sensor failures or external interferences and can be identified and processed through statistical methods. Further, operations such as removing duplicate records, correcting error data, and formatting the initial operating state data are performed to obtain the cleaned operating state data.

[0176] In this embodiment, by performing data cleaning on the initial operating state data, it is ensured that the operating state data used for analysis is accurate and reliable, facilitating the subsequent improvement of the accuracy of fan life detection.

[0177] In step S602 of some embodiments, specifically, normalization processing refers to scaling the cleaned operating state data to a specific range, usually 0 to 1 or -1 to 1, to obtain the normalized operating state data.

[0178] In this embodiment, by normalizing the cleaning operation status data, it is possible to avoid the influence of data with different dimensions and magnitudes on the detection results of the subsequent fan life detection model, and ensure that each data has the same weight during the life detection process.

[0179] In step S603 of some embodiments, specifically, the fan operation time period is an analysis method based on the operation data of the fan at different time series. By dividing the operation data of the fan in different time periods into windows of a fixed size, each window contains the operation status data of the fan for a period of time.

[0180] In this embodiment, by using the preset fan operation time period to divide the normalized operation status data, it is convenient to perform Fourier transform and frequency band division on the operation status data of the fan in different time periods subsequently, improving the efficiency of feature extraction. It also helps the subsequent fan life detection model capture the characteristics of the fan operation data in different time periods, thereby capturing the dynamic changes of the fan operation status and contributing to improving the efficiency of fan life prediction.

[0181] The embodiment of the present application provided via steps S601 to S603 provides a high-quality data basis for subsequent feature extraction and the fan life detection model to perform remaining life detection on the fan, contributing to improving the accuracy and reliability of fan life detection.

[0182] Refer to Figure 7 , according to some embodiments of the present application, step S103 extracts features from the target operation status data to obtain the target operation status features, which may include, but are not limited to:

[0183] Step S701, perform Fourier transform processing on the target operation status features to obtain at least one operation status spectrum feature;

[0184] Step S702, screen out the operation status spectrum feature corresponding to the maximum amplitude from at least one operation status spectrum feature as the target operation status feature.

[0185] In some embodiments of the present application, the target operation status features not only contain the detailed information of the fan operation status, but also achieve the extraction of the fan operation data features in an efficient manner by focusing on the most significant part of the fan amplitude signal. In step S103, this process includes the following key operations:

[0186] In step S701 of some embodiments, specifically, the operation status spectrum feature refers to the feature extracted from the target operation status data of the fan through Fourier transform processing technology, which represents the vibration or fluctuation of the fan at different frequencies, and the operation status spectrum feature includes, but is not limited to, the vibration frequencies of the fan bearing, blade, and gear and the amplitudes corresponding to the vibration frequencies, etc.

[0187] Specifically, by performing a Fourier Transform on the target operating state data, the time-domain signal is converted into a frequency-domain signal.

[0188] In this embodiment, by performing Fourier transform processing on the target operating state characteristics, at least one operating state spectrum characteristic is obtained. The operating state spectrum characteristic can be obtained from the operating data of the fan and can reflect the dynamic characteristics of the fan during operation, which is the key information for identifying the health state of the fan and predicting potential faults.

[0189] Refer to Figure 8 , according to some embodiments of the present application, step S702 of screening out the operating state spectrum characteristic corresponding to the maximum amplitude from at least one operating state spectrum characteristic as the target operating state characteristic may include, but is not limited to:

[0190] Step S801, dividing the frequency band of at least one operating state spectrum characteristic to obtain at least one operating state frequency band;

[0191] Step S802, screening out the operating state frequency band corresponding to the maximum amplitude from at least one operating state frequency band as the target operating state characteristic.

[0192] In some embodiments of the present application, screening out the operating state spectrum characteristic corresponding to the maximum amplitude from at least one operating state spectrum characteristic as the target operating state characteristic enables the model to focus only on the most significant part of the fan amplitude signal during the fan life detection process, improving the efficiency of subsequent fan life detection. In step S702, this process includes the following key operations:

[0193] In step S801 of some embodiments, specifically, the operating state frequency band refers to dividing the frequency range of the signal into several continuous intervals or bands in spectral analysis, and each interval or band represents a specific frequency range.

[0194] For example, the operating state spectrum characteristic can be divided into an operating state low-frequency band, an operating state medium-frequency band, and an operating state high-frequency band, etc.

[0195] Furthermore, dividing the frequency band of at least one operating state spectrum characteristic means splitting the frequency range in the spectrum into several frequency bands, and each frequency band represents a specific frequency interval of the fan operating state. The division of the frequency band can be achieved based on the physical characteristics of the fan, common fault modes, or the requirements of spectral analysis, with the aim of simplifying the complex spectral data into a more easily analyzable and understandable frequency band structure.

[0196] Step S802 of some embodiments. Specifically, the target operating state feature refers to the key information extracted from the operating data of the fan that can represent the current operating state of the fan.

[0197] For example, the target operating state features can be the vibration level, temperature, pressure, rotational speed, etc. of each fan in the nuclear power plant.

[0198] Specifically, the operating state frequency band corresponding to the maximum amplitude is selected from at least one operating state frequency band as the target operating state feature. The purpose of this step is to identify the frequency intervals where the energy is most concentrated in the spectrum, and these intervals may be closely related to the key operating data or potential faults of the fan. The frequency band with the maximum amplitude may indicate the most significant vibration signals during the operation of the fan, and these signals may be caused by the normal operation or potential faults of the fan.

[0199] The embodiments of the present application shown via steps S801 to S802 provide key data support for the subsequent fault diagnosis and life prediction of the fan by highlighting the most significant vibration signals in each operating state frequency band.

[0200] The embodiments of the present application shown via steps S701 to S702. The target operating state feature not only contains the time-domain information of the fan operation but also contains the frequency-domain information, facilitating the subsequent fan life detection model to learn the operating state features of the fan from multiple perspectives, thereby improving the accuracy and reliability of the fan life prediction.

[0201] In some more specific embodiments of the present application, the operating state spectrum can be divided into an operating state low-frequency band (0 - 10 Hz), an operating state medium-frequency band (10 - 50 Hz), and an operating state high-frequency band (50 - 100 Hz). If in the medium-frequency band, the amplitude at 30 Hz is significantly higher than the amplitudes of other frequencies, it may indicate that there are significant vibration signals in the bearing components of the fan at this frequency, which may be caused by the specific operating mechanism of the fan or potential bearing faults. Taking the maximum amplitude corresponding to each frequency band as the target operating state feature, for example, in the subsequent training of the fan life detection model, if the amplitude at 30 Hz has a high correlation with the early bearing faults of the fan, then this feature will be used to predict the maintenance requirements and remaining service life of the fan.

[0202] Refer to Figure 9 , according to some embodiments of the present application, step S104 performs fan life detection on the target operating state feature through a pre-trained fan life detection model to obtain target fan life data, which may include, but is not limited to:

[0203] Step S901, performing fan health state detection on the target operating state feature through the fan life detection model to obtain fan health degree data;

[0204] Step S902: Detect the remaining life of the target fan based on the fan health degree data to obtain the target fan life data.

[0205] In some embodiments of the present application, by using a pre-trained fan life detection model to detect the fan life of the target operating state characteristics, it is possible to avoid manually selecting the characteristic indicators of the fan operation data, and combine the operating states of the fan at different times in the fan life detection, effectively improving the accuracy of the fan life detection. In step S104, this process includes the following key operations:

[0206] In step S901 of some embodiments, refer to Figure 10 , according to some embodiments of the present application, in step S901, the fan health state of the target operating state characteristics is detected by the fan life detection model to obtain the fan health degree data, which may include, but is not limited to:

[0207] Step S1001: Extract the fan operation health characteristics of the target operating state characteristics through the fan life detection model to obtain the fan operation health characteristics;

[0208] Step S1002: Evaluate the fan health degree according to the preset linear function for the fan operation health characteristics to obtain the fan health evaluation score;

[0209] Step S1003: Determine the fan health degree data according to the fan health evaluation score.

[0210] In some embodiments of the present application, by using the fan life detection model to detect the fan health state of the target operating state characteristics, it is possible to quantify the fan health degree detection process and effectively measure the remaining life of the subsequent fan. In step S901, this process includes the following key operations:

[0211] In step S1001 of some embodiments, refer to Figure 11 , according to some embodiments of the present application, in step S1001, the fan operation health characteristics of the target operating state characteristics are extracted through the fan life detection model to obtain the fan operation health characteristics, which may include, but is not limited to:

[0212] Step S1101: Perform feature encoding on the target operating state characteristics through the fan life detection model to obtain the encoded operation characteristics;

[0213] Step S1102: Perform health feature weighting processing on the encoded operation characteristics to obtain the weighted health operation characteristics;

[0214] Step S1103: Decode the weighted health operation characteristics to obtain the fan operation health characteristics.

[0215] In some embodiments of the present application, by extracting the health characteristics of the fan operation, it is possible to remove the characteristics irrelevant to the life detection in the fan, extract the characteristics indicative of the health state of the fan, and subsequently improve the accuracy of the fan life detection. In step S1001, this process includes the following key operations:

[0216] In step S1101 of some embodiments, specifically, encoding the operation characteristics refers to the target operation state characteristics represented by a fixed-dimensional vector space.

[0217] Specifically, the fan life detection model includes an embedding layer and a bidirectional GRU (Bidirectional Gated Recurrent Unit) network layer. Through the embedding layer and the bidirectional GRU network layer, it is possible to encode the target operation state characteristics. Among them, each GRU includes an update gate, a reset gate, and a candidate mechanism.

[0218] Furthermore, first, the target operation state characteristics at each time step are converted into fixed-dimensional vectors through an embedding layer to map the target operation state characteristics to a high-dimensional space; second, in the forward GRU, the embedded vector data is processed according to the time series from the past to the future. The update gate determines the vector data to be retained in the past, the reset gate determines the vector data to be discarded in the past, and the candidate mechanism is used to generate a new candidate state of the target operation state characteristics. The new candidate state combines the input of the current time series and the gated vector data to form the forward operation characteristics; furthermore, in the backward GRU, for the forward operation characteristics, according to the time series from the future to the past, combined with the update gate, the reset gate, and the candidate mechanism, the forward operation characteristics to be retained or discarded are determined, and the backward operation characteristics are output; finally, the forward operation characteristics and the backward operation characteristics are integrated to output the encoded operation characteristics.

[0219] In this embodiment, the fan life detection model encodes the target operation state characteristics, providing a standardized input representation for capturing time series-related information in the subsequent process. Combined with the bidirectional GRU, it ensures that the model can capture the operation characteristics of the time series from two directions, can consider the front and back information of each operation characteristic in the time series simultaneously, provides a rich information basis for the subsequent health state assessment and fan life prediction, and enables the fan life detection model to more accurately predict the health condition and remaining service life of the fan.

[0220] Step S1102 of some embodiments. Specifically, the weighted healthy operation features reflect the influence degree of each coded operation feature on the identification of the fan health state. The weighted healthy operation features not only contain the information of the coded operation features, but also integrate the importance of each time step in different time series.

[0221] Specifically, the fan life distribution model identifies the coded operation features that make important contributions to the evaluation of the fan health state by using the attention mechanism, assigns different weights to different coded operation features, reflects the contribution degree of the coded operation features to the fan health condition, and realizes the combination of the attention weights and the coded operation features to generate the weighted healthy operation features.

[0222] For example, if the fan life detection model identifies that the vibration peak value of the fan is closely related to the fan health state within a period of time, the coded operation features of this period will be assigned a greater weight.

[0223] In this embodiment, by performing health feature weighting processing on the coded operation features, the model can focus on the coded operation features that have a significant impact on the fan health state, can more accurately identify the potential faults of the fan, and is convenient for improving the accuracy of the fan life detection subsequently.

[0224] Step S1103 of some embodiments. Specifically, the fan operation health features refer to the features related to the fan operation health degree.

[0225] For example, the vibration frequency during the fan operation, the fan speed, the transmitter power, the fan operation sound signal, etc.

[0226] Specifically, the fan life detection model includes a decoding layer. The decoding layer determines the hidden state of the current time step based on the hidden state of the previous time step output by the bidirectional GRU, the output of the previous time step, and the weighted healthy operation features, and activates the hidden state through the ReLU function to obtain the fan operation health features.

[0227] In this embodiment, by decoding the weighted healthy operation features to obtain the fan operation health features, it can convert the weighted healthy operation features into fan health state features, can also highlight the features that have the greatest impact on the fan health state, and considers the influence of each time step, which is convenient for subsequently determining the specific time point when the fan fails and improves the accuracy of the fan life detection.

[0228] Through steps S1101 to S1103 shown in the embodiments of the present application, the fan life detection model can extract features indicative of the health state from the original target operating state features, transform these features into a form that the model can utilize, and also consider the influence of each time step, facilitating the subsequent determination of the specific time point when the fan fails, thereby providing a scientific basis for the maintenance and life prediction of the fan.

[0229] In step S1002 of some embodiments, specifically, the fan health assessment score is used to measure the operating health of the fan at different time points.

[0230] Specifically, this linear function can calculate weights for each fan operating health feature, and finally combine with the function of the bias term to transform multiple fan operating health features into a quantifiable health assessment score.

[0231] Furthermore, by linearly weighting and summing the vibration levels, operating temperatures, and rotational speeds of each component of the fan, the comprehensive fan health assessment score for each component is calculated. The fan health assessment score can intuitively reflect the health state of the fan. The higher the score, the lower the probability that the fan fails.

[0232] The fan health assessment score provides a quantitative reference for the maintenance and life prediction of the fan, enabling the operator to optimize the operation and maintenance strategies of the fan based on a data-driven method. This method combines data analysis and domain knowledge, providing an effective tool for the intelligent management of the fan.

[0233] In this embodiment, according to the preset linear function, the fan health degree can be evaluated for each fan operating health feature to obtain the fan health assessment score. The fan health assessment score provides a quantitative index for the life prediction of the fan. The higher the score, the better the health state of the fan. The lower the score, it intuitively indicates the possibility of the fan failing, which helps to improve the efficiency of subsequent fan life prediction.

[0234] In step S1003 of some embodiments, specifically, the fan health degree data reflects the current health features of each component of the fan and is used to identify the possibility of the fan failing.

[0235] Specifically, by comparing the fan health assessment score with a preset health threshold, the degree of failure of the features of different components of the fan can be determined. Further, if the fan health assessment score is less than the health threshold, it indicates that there is a risk of failure in the features of different components of the fan. If the fan health assessment score is greater than or equal to the health threshold, it indicates that the fan can continue to operate normally.

[0236] In this embodiment, the health degree data of the fan is determined according to the fan health assessment score, realizing the quantitative evaluation of fan fault detection, being able to intuitively monitor the operation status of the fan, and providing an effective method for the intelligent management of the fan.

[0237] Via steps S1001 to S1003 shown in the embodiments of the present application, the fan life detection model can intuitively display the health degree of the fan and provide a data basis for predicting the remaining service life of the fan subsequently, which helps to improve the efficiency of subsequent fan life detection.

[0238] Step S902 of some embodiments refers to Figure 12 According to some embodiments of the present application, in step S902, the remaining life of the fan is detected for the target operating state characteristics according to the fan health degree data, and the target fan life data can be obtained, including but not limited to:

[0239] Step S1201, obtaining the health degree weight and the health degree offset term of the fan health degree data;

[0240] Step S1202, performing fan operation fault detection on the target operating state characteristics according to the health degree weight, the health degree offset term and the fan health degree data to obtain fan operation fault data;

[0241] Step S1203, determining the target fan life data according to the fan operation fault data.

[0242] In some embodiments of the present application, detecting the remaining life of the fan for the target operating state characteristics according to the fan health degree data involves fan fault analysis of the target operating state characteristics to determine the health condition of the fan and predict the remaining life of the fan. In step S902, this process includes the following key operations:

[0243] In step S1201 of some embodiments, specifically, the health degree weight reflects the influence degree of different component characteristics of the fan on the overall health condition of the fan, and the offset term is used to adjust the influence degree of the fan health degree data on the overall health condition of the fan to ensure that the fan health degree data can accurately reflect the actual health state of the whole fan.

[0244] In this embodiment, by obtaining the health degree weight and the health degree offset term of the fan health degree data, the fan health degree data can be quantified, and a comprehensive evaluation of the fan health degree data can be realized by combining the weight and the offset term, which helps to improve the accuracy of subsequent fan life prediction.

[0245] In step S1202 of some embodiments, specifically, the fan operation fault data refers to the fault characteristics that there are fault risks in the fan operation.

[0246] Specifically, by performing a weighted sum of the health degree weights and health degree bias terms on the health degree data of each fan during operation, when signs of a fan failure risk or performance degradation are identified, data characteristics related to the fan failure can be determined.

[0247] It involves analyzing the operation data of the fan to identify possible failure modes or signs of performance degradation. For example, if the vibration level of the fan suddenly increases, this may be an early signal of a failure. Through this analysis, fan operation failure data can be obtained, which reveals the problems that may occur during the operation of the fan.

[0248] In this embodiment, fan operation failure detection is performed on the target operation state characteristics based on the health degree weight, health degree bias term, and fan health degree data, combining the current operation health state of the fan and the failure data during operation, providing comprehensive data support for fan life prediction.

[0249] In step S1203 of some embodiments, specifically, the target fan life data refers to the remaining life data of the fan currently.

[0250] Specifically, the sigmoid function of the fan life detection model can be used to activate the fan operation failure data, thereby determining the target life data of the fan.

[0251] Furthermore, if the fan operation failure data shows multiple signs of performance degradation in the fan, it may mean that the remaining life of the fan is short, and more frequent maintenance or possible replacement is required.

[0252] In this embodiment, determining the target fan life data based on the fan operation failure data can provide important information for the health management and life prediction of the fan, facilitating the nuclear power plant to make more accurate fan maintenance or fan replacement decisions based on the target fan life data.

[0253] Through steps S1201 to S1203 provided by the embodiments of the present application, considering the current fan health state and the failure characteristics during fan operation, comprehensive data support is provided for fan life prediction, effectively improving the accuracy of fan life prediction, and facilitating the nuclear power plant to make more accurate fan maintenance or fan replacement decisions based on the target fan life data.

[0254] Through steps S901 to S902 provided by the embodiments of the present application, the fan health state data is automatically identified by the life detection model, without the need to manually select the characteristic indexes of the fan operation data, and the operation states of the fan at different times are combined in the fan life detection, effectively improving the accuracy rate of the fan life detection.

[0255] In some more specific embodiments of the present application, if the target operating state features are the temperature, vibration, speed, wind speed, power, etc. of each component during the operation of the fan, through health feature weighting processing and decoding processing, the features related to the health state of the fan operation output are the fan vibration level, bearing temperature, fan speed, etc. These features are scored for the degree of health. If the vibration level is 0.05g, the bearing temperature is 85°C, and the fan speed is 1500 RPM, these features are weighted by the fan life detection model to evaluate the current health score of the fan as 85 points. Based on this current health score and the operating health features, the weight and bias term of the health degree are calculated. Assuming the weight is 0.7 and the bias term is 5, the bias term and weight are activated through the Sigmoid function, and the remaining life of the fan can be obtained as 8 years.

[0256] It should be noted that in the embodiments of the present application, first, a pre-trained fan life detection model is obtained, which can understand and identify the complex relationship between the fan operation state and the fan life. And the pre-trained model does not need to collect and analyze data from scratch every time when predicting the fan life, which improves the efficiency of fan life detection. And the target operating state data of the fan is obtained, and the actual data of the current fan operation is extracted, providing data support for subsequent fan life detection. Secondly, by extracting features from the target operating state data, the target operating state features are obtained, and the key features related to fan life detection can be selected, which is convenient for the subsequent support model to improve the accuracy of fan life detection. Finally, through the pre-trained fan life detection model, the fan life detection is carried out on the target operating state features to obtain the target fan life data, without manually selecting the feature indexes of the fan operation data, and the operation state of the fan at different times is combined in the fan life detection, effectively improving the accuracy of fan life detection.

[0257] Refer to Figure 13 , according to the fan life detection device of the second aspect embodiment of the present application, it may include, but is not limited to:

[0258] The fan life detection model acquisition module 1301 is used to acquire a pre-trained fan life detection model;

[0259] The fan operating state data acquisition module 1302 is used to acquire the target operating state data of the fan;

[0260] The feature extraction module 1303 is used to extract features from the target operating state data to obtain the target operating state features;

[0261] The fan life detection module 1304 is used to perform fan life detection on the target operating state features through the pre-trained fan life detection model to obtain the target fan life data.

[0262] It can be seen that the content in the embodiments of the above-mentioned fan life detection method is applicable to the embodiments of this fan life detection device. The functions specifically implemented by the embodiments of this fan life detection device are the same as those of the embodiments of the above-mentioned fan life detection method, and the beneficial effects achieved are also the same as those of the embodiments of the above-mentioned fan life detection method.

[0263] Refer to Figure 14 , Figure 14 which schematically shows the hardware structure of an electronic device in another embodiment. The electronic device includes:

[0264] A processor 1401, which can be implemented in ways such as a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided by the embodiments of this application;

[0265] A memory 1402, which can be implemented in forms such as a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 1402 can store an operating system and other application programs. When implementing the technical solutions provided by the embodiments of this specification through software or firmware, the relevant program codes are stored in the memory 1402 and are called by the processor 1401 to execute the fan life detection method of the embodiments of this application;

[0266] An input / output interface 1403, which is used to implement information input and output;

[0267] A communication interface 1404, which is used to implement communication interaction between this device and other devices. It can communicate through wired means (such as USB, network cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.);

[0268] A bus 1405, which transmits information between various components of the device (such as the processor 1401, the memory 1402, the input / output interface 1403, and the communication interface 1404);

[0269] Among them, the processor 1401, the memory 1402, the input / output interface 1403, and the communication interface 1404 are communicatively connected to each other inside the device through the bus 1405.

[0270] An embodiment of the present application also provides a computer program product, which includes a computer program. The processor of the computer device reads and executes the computer program, so that the computer device implements the above-mentioned fan life detection method.

[0271] Terms such as "first", "second", "third", "fourth", etc. (if any) in the specification of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "comprising" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0272] It should be understood that in the present disclosure, "at least one (item)" means one or more, and "multiple" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships can exist. For example, "A and / or B" can mean: only A exists, only B exists, and both A and B exist at the same time. Among them, A and B can be singular or plural. The character " / " generally means that the associated objects before and after are in an "or" relationship. "At least one (item) of the following" or its similar expression means any combination of these items, which can include, but is not limited to, any combination of single item (s) or plural item (s). For example, at least one (item) of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0273] It should be understood that in the description of the embodiments of the present application, the meaning of multiple (or multiple items) is more than two. Understandings such as greater than, less than, exceeding, etc. do not include the present number, and understandings such as above, below, within, etc. include the present number.

[0274] In several embodiments provided by the present disclosure, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces, and the indirect couplings or communication connections of devices or units can be in electrical, mechanical, or other forms.

[0275] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0276] In addition, each functional unit in various embodiments of the present disclosure can be integrated in a processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit. The above-mentioned integrated units can be implemented in the form of hardware or in the form of software functional units.

[0277] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present disclosure, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and can include, but is not limited to, several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods in various embodiments of the present disclosure. The foregoing storage medium includes: USB flash drives, mobile hard disks, read-only memories (ROM for short), random access memories (RAM for short), magnetic disks, or optical discs, and other media that can store program codes.

[0278] It should also be understood that the various embodiments provided by the embodiments of the present application can be combined arbitrarily to achieve different technical effects.

[0279] The above is a specific description of the embodiments of the present disclosure. However, the present disclosure is not limited to the above embodiments. Those skilled in the art can also make various equivalent deformations or substitutions without departing from the spirit of the present disclosure, and these equivalent deformations or substitutions are all included within the scope defined by the claims of the present disclosure.

Claims

1. A method for detecting the life of a fan, characterized in that: include: Obtain a pre-trained wind turbine life detection model; wherein the pre-training process of the wind turbine life detection model includes: Obtaining prior operating status data of the fan; The fan life distribution is predicted by using a preset Weibull cumulative distribution function on the prior operating state data to obtain prior life distribution data; Training the initial wind turbine life detection model based on the prior life distribution data and the prior operating state data to obtain the pre-trained wind turbine life detection model; Obtain target operating status data of the wind turbine; Extracting features from the target operating state data to obtain target operating state features; The fan life detection is performed on the target operating state characteristics through the pre-trained fan life detection model to obtain target fan life data.

2. The method according to claim 1, characterized in that The initial wind turbine life detection model is trained based on the prior life distribution data and the prior operating state data to obtain the pre-trained wind turbine life detection model, including: Obtaining the initial wind turbine life detection model; Performing fan life prediction on the priori operating status data by using the fan life detection model to obtain predicted fan life data; Performing fan life distribution prediction on the predicted fan life data by using the Weibull cumulative distribution function to obtain predicted life distribution data; Acquire the real wind turbine life data of the priori operating state data, and calculate the target loss value of the predicted wind turbine life data, the predicted life distribution data, the real wind turbine life data and the priori life distribution data according to a preset loss function; The model parameters of the fan life detection model are updated based on the target loss value, and the fan life detection model is returned to be executed to predict the fan life of the prior operating status data until the fan life detection model meets the preset training conditions, thereby obtaining the pre-trained fan life detection model.

3. The method according to claim 2, characterized in that The loss function includes a neural network loss function and a Weibull loss function; The calculating, according to a preset loss function, the target loss value of the predicted wind turbine life data, the predicted life distribution data, the real wind turbine life data and the priori life distribution data includes: Calculating the life loss value between the predicted wind turbine life data and the actual wind turbine life data according to the neural network loss function; Calculating a life distribution loss value between the predicted life distribution data and the prior life distribution data according to the Weibull loss function; The life loss value and the life distribution loss value are integrated through preset weight hyperparameters to obtain the target loss value.

4. The method according to claim 1, characterized in that: The method of predicting the life distribution of the wind turbine by using a preset Weibull cumulative distribution function on the prior operating state data to obtain the prior life distribution data includes: Determining a fan shape characterization parameter based on the priori operating state data; Determining a fan scale characterization parameter according to the fan shape characterization parameter and the priori operating state data; The fan life distribution is predicted based on the priori operating state data according to the fan shape characterization parameters and the fan scale characterization parameters to obtain the priori life distribution data.

5. The method according to claim 4, characterized in that The step of determining the fan scale characterization parameter according to the fan shape characterization parameter and the priori operating state data includes: Obtain the fan failure time, fan failure quantity and fan failure variable of the priori operating status data; The fan fault time, the fan fault quantity, the fan fault variable and the fan shape characterization parameter are subjected to scale characterization calculation to obtain the fan scale characterization parameter.

6. The method according to claim 1, characterized in that The fan life detection is performed on the target operating state characteristics by the pre-trained fan life detection model to obtain target fan life data, including: Performing a fan health status detection on the target operating status characteristics through the fan life detection model to obtain fan health status data; The remaining life of the fan is detected based on the target operating state characteristics according to the fan health data to obtain the target fan life data.

7. The method according to claim 6, characterized in that The step of performing a fan health status detection on the target operating status characteristics through the fan life detection model to obtain fan health status data includes: Extracting the healthy operation characteristics of the fan from the target operation state characteristics through the fan life detection model to obtain the healthy operation characteristics of the fan; Performing a fan health evaluation on the fan operation health characteristics according to a preset linear function to obtain a fan health evaluation score; The wind turbine health data is determined according to the wind turbine health assessment score.

8. The method according to claim 7, characterized in that The step of extracting the healthy operation characteristics of the fan from the target operation state characteristics through the fan life detection model to obtain the healthy operation characteristics of the fan includes: Characteristic encoding is performed on the target operating state characteristics by using the wind turbine life detection model to obtain a coded operating characteristic; Performing health feature weighting processing on the coded operation feature to obtain a weighted health operation feature; The weighted healthy operation feature is decoded to obtain the wind turbine healthy operation feature.

9. The method according to claim 6, characterized in that The performing a fan remaining life detection on the target operating state characteristics according to the fan health data to obtain the target fan life data includes: Obtaining a health weight and a health bias item of the wind turbine health data; Performing fan operation fault detection on the target operation state feature according to the health weight, the health bias item and the fan health data to obtain fan operation fault data; The target fan life data is determined according to the fan operation fault data.

10. The method according to claim 1, characterized in that The step of extracting features from the target operating state data to obtain target operating state features includes: Performing Fourier transform processing on the target operating state feature to obtain at least one operating state spectrum feature; The operating state spectrum feature corresponding to the maximum amplitude is selected from at least one of the operating state spectrum features as the target operating state feature.

11. The method according to claim 10, characterized in that The step of selecting the operating state spectrum feature corresponding to the maximum amplitude from at least one of the operating state spectrum features as the target operating state feature comprises: Dividing at least one of the operating state frequency spectrum characteristics into frequency bands to obtain at least one operating state frequency band; The operating state frequency band corresponding to the maximum amplitude is selected from at least one of the operating state frequency bands as the target operating state feature.

12. The method according to claim 10, characterized in that Before extracting features from the target operating state data to obtain target operating state features, the method further includes: Acquiring initial operation status data, and performing data cleaning on the initial operation status data to obtain cleaned operation status data; Normalizing the cleaning operation status data to obtain normalized operation status data; The normalized operating status data is divided according to a preset fan operating time period to obtain the target operating status data.

13. A fan life detection device, characterized in that: include: A fan life detection model acquisition module is used to obtain a pre-trained fan life detection model; A fan operation status data acquisition module is used to acquire target operation status data of the fan; A feature extraction module is used to extract features from the target operating state data to obtain target operating state features; The fan life detection module is used to perform fan life detection on the target operating state characteristics through the pre-trained fan life detection model to obtain target fan life data.

14. An electronic device, characterized in that: include: A memory and a processor, wherein the memory stores a computer program, and the processor implements the wind turbine life detection method as described in any one of claims 1 to 12 when executing the computer program.

15. A computer-readable storage medium, characterized in that: The storage medium stores a program, and the program is executed by a processor to implement the wind turbine life detection method according to any one of claims 1 to 12.

Images