Fan online vibration monitoring and fault diagnosis method and system

Abstract

The embodiment of the invention provides a fan online vibration monitoring and fault diagnosis method and system, and belongs to the technical field of fan monitoring. Comprising the following steps: acquiring vibration data of fan operation within a period of time and performing noise reduction processing; windowing and framing are carried out on the vibration data after noise reduction processing, vibration data features are extracted, a training data set is obtained, and a fan vibration fault diagnosis model is constructed and trained. A trained fan vibration fault diagnosis model is deployed in an STM32 chip of an intelligent diagnosis terminal through STM32CUBEMAX software, and terminal edge calculation and real-time fault diagnosis functions are realized. The fan online vibration monitoring system can perform real-time vibration monitoring on the fan, perform online processing and diagnosis on vibration data, and transmit an analysis and diagnosis result to a background in real time, so that a vibration fault and fault early warning of the fan can be found in time, operation and maintenance personnel can conveniently retrieve the vibration analysis result at any time, and the operation and maintenance efficiency is improved. And the vibration state of the fan is compared and analyzed in real time, and the potential fault hidden danger of the fan is found in time.

Description

Technical Field

[0001] The present invention relates to the technical field of fan monitoring, and in particular to a method and system for online vibration monitoring and fault diagnosis of a fan. Background Art

[0002] Fan vibration monitoring plays an important role in identifying and handling hidden faults of fans. Since the fan vibration monitoring data is relatively large, it is impossible to transmit the data to the background in real time and conduct a comprehensive online diagnosis of the possible fault types of the fan. At present, fan vibration monitoring mostly adopts the method of copying data regularly (usually one month or even longer) and then processing it on the computer. This planned maintenance method cannot detect and handle some sudden faults and hidden dangers in time, and cannot truly realize real-time vibration monitoring fault warning, which greatly reduces the reliability and effectiveness of fan vibration fault monitoring. Therefore, a preventive method is needed to monitor and diagnose the vibration of the fan in real time to detect the fault of the fan in time. Summary of the invention

[0003] The purpose of the embodiments of the present invention is to provide a method and system for online vibration monitoring and fault diagnosis of a fan. The method can preventively monitor the vibration of the fan and promptly discover the vibration fault of the fan.

[0004] In order to achieve the above-mentioned object, on the one hand, an embodiment of the present invention provides a method for online vibration monitoring and fault diagnosis of a fan, and the method for online vibration monitoring and fault diagnosis of a fan comprises:

[0005] Acquire vibration data of the fan operation over a period of time, and perform noise reduction processing on the vibration data;

[0006] Performing windowing and framing on the vibration data after the noise reduction process to extract vibration data features, so as to obtain a training data set;

[0007] Constructing a first vibration prediction model, and bringing the training data set into the first vibration prediction model to obtain a predicted value of the vibration data;

[0008] Subtract the corresponding predicted value from the acquired time sample data to obtain the residual data;

[0009] Bringing the residual data into a second vibration prediction model to obtain residual prediction data;

[0010] Determine whether the residual prediction data is greater than a set threshold;

[0011] In the case where the residual prediction data is greater than a set threshold, determining that the working state of the fan within the predicted time period is a predicted fault state;

[0012] Acquire real time sample data within the predicted time period, and bring it into the first vibration prediction model again to perform correction training on the first vibration prediction model and the second vibration prediction model;

[0013] The real time sample data within the predicted time period is transformed by Fourier to obtain a frequency domain signal, and the frequency components contained in the vibration are determined according to the frequency domain signal to determine the fault type.

[0014] Optionally, the first vibration prediction model and the second vibration prediction model are both Attention-BiGRU-GRU network models.

[0015] Optionally, constructing a first vibration prediction model and bringing the training data set into the first vibration prediction model to obtain a predicted value of the vibration data includes:

[0016] Performing bidirectional temporal feature extraction on the training data set through the BiGRU neural network layer, and based on this, using the GRU neural network layer to model the extracted features;

[0017] Combined with the Attention layer, the attention weights of the GRU hidden layer states at different times are calculated to predict vibration data;

[0018] The prediction results of the vibration data are output through the fully connected layer.

[0019] Optionally, the unit structure of GRU is formula (1):

[0020]

[0021] Among them, r t 、z t are the outputs of the reset gate and the update gate at time t, W r 、b r , W z 、b z , W h 、b h are the weights and biases of the update gate, reset gate and candidate hidden layer, respectively, h t-1 、h t are the hidden layer information at time t-1 and time t, is the candidate state information at time t, σ(·) and tanh(·) are the sigmoid function and hyperbolic tangent function respectively.

[0022] Optionally, the calculation process of the Attention mechanism is shown in formula (2):

[0023]

[0024] Among them, h t is the output feature vector of the GRU unit at the tth time step, α is the random weight matrix adaptively updated during training after random initialization, γ is the final output value of the attention layer, and softmax is the normalization function.

[0025] Optionally, windowing and framing the vibration data after noise reduction processing to extract vibration data features to obtain a training data set includes:

[0026] Acquire the vibration data arranged in time sequence;

[0027] Set the sampling time span and sampling time schedule;

[0028] According to the sampling time progress, a certain time is advanced and according to the time span, feature extraction is performed on the vibration data arranged in chronological order to obtain a training data set.

[0029] Optionally, performing noise reduction processing on the vibration data includes:

[0030] Select the set wavelet and determine the level N of wavelet decomposition;

[0031] Performing N-layer wavelet packet decomposition on the vibration data;

[0032] Select a set threshold to perform quantization denoising on the high-frequency coefficients from layers 1 to N and the low-frequency coefficients of the Nth layer obtained through decomposition;

[0033] The low-frequency coefficients of the Nth layer after quantization denoising and the high-frequency coefficients from layers 1-N are reconstructed to obtain the denoised vibration data.

[0034] Optionally, obtaining vibration data of the wind turbine operation over a period of time includes: when the wind turbine is a doubly fed and semi-direct drive unit, sensors are deployed at the following nine positions: 1. radial vertical for the main bearing 2. radial vertical for the gearbox housing - left 3. axial for the gearbox housing 4. radial vertical for the gearbox housing - right 5. radial vertical for the generator housing - left 6. radial vertical for the generator housing - right 7. lateral for the nacelle elbow 8. axial for the nacelle elbow 9. internal bearing sleeve of the generator, to obtain vibration data of the wind turbine operation.

[0035] Optionally, obtaining the vibration data of the fan operation over a period of time includes: when the fan is a direct-drive model, the sensors are installed at the following 7 positions: 1. axial direction of the first measuring point of the main shaft 2. radial horizontal direction of the first measuring point of the main shaft 3. radial vertical direction of the first measuring point of the main shaft 4. radial vertical direction of the second measuring point of the main shaft 5. radial horizontal direction of the third measuring point on the left side of the stator 6. axial direction of the third measuring point on the left side of the stator 7. radial horizontal direction of the fourth measuring point on the right side of the stator, so as to obtain the vibration data of the fan operation.

[0036] In another aspect, the present invention provides a fan online vibration monitoring and fault diagnosis system, comprising:

[0037] A fan vibration sensor is installed on the fan to obtain vibration data of the fan operation;

[0038] An intelligent diagnostic terminal, arranged on the fan, obtains vibration data detected by the fan vibration sensor and executes the fan online vibration monitoring and fault diagnosis method according to any one of claims 1 to 7;

[0039] The backend server performs exception processing according to the diagnosis result calculated by the intelligent diagnosis terminal;

[0040] The fault alarm system performs fault alarm according to the abnormal processing instruction sent by the background server.

[0041] Through the above technical scheme, the present invention provides a method and system for online vibration monitoring and fault diagnosis of a fan, which obtains the vibration data of the fan operation within a period of time, and then obtains a training data set according to the vibration data. After obtaining the training data set, a first vibration prediction model and a second vibration prediction model can be constructed. After the training data set is brought into the first vibration prediction model, the predicted value of the vibration data can be obtained. After the predicted value is obtained, the time sample data can be subtracted from the corresponding predicted value, so that the residual data can be obtained. The residual data can be brought into the second vibration prediction model to obtain the residual prediction data. According to the residual prediction data, it can be judged whether there will be a high probability of failure in a certain period of time in the future, and then prevention can be carried out in advance. Then, fault monitoring can be carried out in advance according to the predicted fault state in the predicted time period, and the real time sample data in the predicted time period can be Fourier transformed to obtain a frequency domain signal, so that the frequency components contained in the vibration can be determined according to the frequency domain signal to determine the fault type. The online vibration monitoring and fault diagnosis method of the fan can preventively monitor the vibration of the fan and timely discover the vibration fault of the fan.

[0042] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings:

[0044] Figure 1 is a flow chart of a method for online vibration monitoring and fault diagnosis of a fan according to an embodiment of the present invention;

[0045] Figure 2 is a flow chart of obtaining prediction values ​​of a method for online vibration monitoring and fault diagnosis of a wind turbine according to an embodiment of the present invention;

[0046] Figure 3 is a flow chart of obtaining a training data set for a method for online vibration monitoring and fault diagnosis of a wind turbine according to an embodiment of the present invention;

[0047] Figure 4 is a flow chart of a noise reduction process of a method for online vibration monitoring and fault diagnosis of a fan according to an embodiment of the present invention;

[0048] Figure 5 is a system connection diagram of a fan online vibration monitoring and fault diagnosis system according to an embodiment of the present invention;

[0049] Figure 6 System diagram of an intelligent diagnosis terminal of a wind turbine online vibration monitoring and fault diagnosis system according to an embodiment of the present invention;

[0050] Figure 7 A schematic diagram of an intelligent diagnosis terminal of an online vibration monitoring and fault diagnosis system for a wind turbine according to an embodiment of the present invention. DETAILED DESCRIPTION

[0051] The specific implementation of the embodiment of the present invention is described in detail below in conjunction with the accompanying drawings. It should be understood that the specific implementation described here is only used to illustrate and explain the embodiment of the present invention, and is not used to limit the embodiment of the present invention.

[0052] Figure 1 1 is a flow chart of a method for online vibration monitoring and fault diagnosis of a wind turbine according to an embodiment of the present invention. In the present invention, the vibration monitoring process may include:

[0053] In step S1, vibration data of the fan operation within a period of time is obtained, and noise reduction processing is performed on the vibration data.

[0054] In step S2, the vibration data after noise reduction processing is subjected to windowing and framing to extract vibration data features, so as to obtain a training data set.

[0055] In step S3, a first vibration prediction model is constructed, and the training data set is introduced into the first vibration prediction model to obtain a predicted value of the vibration data.

[0056] In step S4, the corresponding predicted value is subtracted from the acquired time sample data to obtain residual data.

[0057] In step S5, the residual data is introduced into the second vibration prediction model to obtain residual prediction data.

[0058] In step S6, it is determined whether the residual prediction data is greater than a set threshold.

[0059] In step S7, when the residual prediction data is greater than a set threshold, it is determined that the working state of the wind turbine in the predicted time period is a predicted fault state.

[0060] In step S8, the real time sample data within the predicted time period is obtained and brought into the first vibration prediction model again to perform correction training on the first vibration prediction model and the second vibration prediction model.

[0061] In step S9, the real time sample data within the predicted time period is transformed by Fourier transform to obtain a frequency domain signal, and the frequency components contained in the vibration are determined according to the frequency domain signal to determine the fault type.

[0062] The existing monitoring methods for fan faults generally use time domain spectrum analysis of current data to determine the fan vibration fault. However, during the operation of the fan, the fault time period is shorter than the normal time period, and some analysis methods collect data for analysis at regular intervals. These methods cannot detect and handle some sudden fault hazards in a timely manner, and cannot truly achieve real-time vibration monitoring fault warning. Therefore, a method is needed to predict possible faults within a period of time, so as to warn and monitor fan faults in advance.

[0063] In the present invention, when monitoring the vibration of the fan, the vibration data of the fan within a period of time can be obtained, and the vibration data can be subjected to noise reduction processing, and then the vibration data subjected to noise reduction processing can be subjected to windowing and framing, and the vibration data features can be extracted from the vibration data, so that a training data set can be obtained. After obtaining the training data set, a first vibration prediction model can be constructed, and then the training data set can be brought into the first vibration prediction model, so that the predicted value of the vibration data can be obtained. After obtaining the predicted value, the time sample data of the real fan vibration of the time series of the predicted value can be obtained, and then the corresponding predicted value can be subtracted from the time sample data to obtain the residual data. The residual data is obtained by subtracting the predicted value of the corresponding predicted time from the obtained real fan vibration data, and the residual data can be used to judge whether the result predicted by the first vibration prediction model is accurate. After a large amount of training on the first vibration prediction model, the first vibration prediction model can be made to relatively accurately predict the change of the vibration data within a future end time, so that the residual data can be made as small as possible. The residual data can reflect the change trend of the fan vibration, so the vibration condition of the fan can be judged by the residual data.

[0064] After obtaining the residual data, the residual data can be brought into the second vibration prediction model, so that the residual prediction data can be obtained. According to the residual prediction data, the difference between the possible vibration data and the predicted value in the future period can be predicted, so that it can be judged whether the fan will fail in the future period, and then the staff can be reminded to pay attention in advance. Therefore, after obtaining the residual prediction data, it can be judged whether the residual prediction data is greater than the set threshold. In the case of judging that the residual prediction data is greater than the set threshold, it indicates that the fan has a greater possibility of its working state being the predicted fault state in the predicted time period, so that the staff can be reminded to monitor in advance so that the fault can be found in time. After obtaining the real time sample data in the predicted time period, the time sample data can be brought into the first vibration model, so that the parameters of the first vibration model can be corrected, and then the parameters in the second vibration prediction model can be corrected. After obtaining the real time sample data in the predicted time period, the data can be Fourier transformed to obtain a frequency domain signal, and the frequency components contained in the vibration can be determined according to the frequency domain signal, so that the fault type can be determined to complete the fault monitoring of the fan vibration.

[0065] In one embodiment of the present invention, the first vibration prediction model and the second vibration prediction model may be an Attention-BiGRU-GRU network model. The Attention-BiGRU-GRU network model may have a strong ability to capture nonlinear features, so that its nonlinear laws may be extracted based on existing data to obtain the predicted value of future residual data.

[0066] In one embodiment of the present invention, Figure 2 As shown, the process of obtaining the predicted value may include:

[0067] In step S10, bidirectional time series feature extraction is performed on the training data set through the BiGRU neural network layer, and on this basis, the extracted features are modeled using the GRU neural network layer.

[0068] In step S11, the attention weights of the GRU hidden layer states at different times are calculated in combination with the Attention layer to predict vibration data.

[0069] In step S12, the prediction result of the vibration data is output through the fully connected layer.

[0070] In the present invention, when obtaining the predicted value, an Attention-BiGRU-GRU network model can be constructed, specifically, a BiGRU neural network layer is used to perform bidirectional time series feature extraction on the training data set, and on this basis, a GRU neural network layer is used to model the extracted features, and the attention weights of the GRU hidden layer states at different times are calculated in combination with the Attention layer to predict the vibration data, and finally a fully connected layer is used to output the residual prediction result. When modeling, the model can be trained using the training data set according to the gradient descent rule to obtain the optimal prediction model. Finally, the predicted value of the vibration data is obtained by the optimal prediction model. When the residual prediction data is obtained early, the same method is used to model and obtain the residual prediction data.

[0071] In one embodiment of the present invention, the unit structure of GRU is formula (1):

[0072]

[0073] Among them, r t 、z t are the outputs of the reset gate and the update gate at time t, W r 、b r , W z 、b z , W h 、b h are the weights and biases of the update gate, reset gate and candidate hidden layer, respectively, h t-1 、h t are the hidden layer information at time t-1 and time t, is the candidate state information at time t, σ(·) and tanh(·) are the sigmoid function and hyperbolic tangent function respectively.

[0074] In one embodiment of the present invention, the calculation process of the Attention mechanism is shown in formula (2):

[0075]

[0076] Among them, h t is the output feature vector of the GRU unit at the tth time step, α is the random weight matrix adaptively updated during training after random initialization, γ is the final output value of the attention layer, and softmax is the normalization function.

[0077] In one embodiment of the present invention, Figure 3 As shown, the process of obtaining a training data set may include:

[0078] In step S13, vibration data arranged in time sequence is acquired.

[0079] In step S14, the time span of sampling and the time progress of sampling are set.

[0080] In step S15, a certain time is advanced according to the sampling time progress and features are extracted from the vibration data arranged in chronological order according to the time span to obtain a training data set.

[0081] In the present invention, after the vibration data is obtained, the vibration data can be extracted according to certain feature extraction rules so that the data can better reflect the characteristics of the fan vibration. Therefore, the vibration data can be arranged according to the time sequence, and then the sampling time span and the sampling time progress can be set. The sampling time span can be the range of one sampling. When sampling, it can start from the beginning of the vibration data, and can move forward according to the sampling time progress, and can extract features of the vibration data according to the time span, so that a training data set can be obtained.

[0082] In one embodiment of the present invention, Figure 4 As shown, the noise reduction process may include:

[0083] In step S16, a set wavelet is selected and the level N of wavelet decomposition is determined.

[0084] In step S17, the vibration data is subjected to N-layer wavelet packet decomposition.

[0085] In step S18, a set threshold is selected to perform quantization denoising on the high-frequency coefficients of layers 1-N and the low-frequency coefficients of layer N obtained through decomposition.

[0086] In step S19, the low-frequency coefficients of the Nth layer after quantization and denoising and the high-frequency coefficients of the 1-N layers are reconstructed to obtain the vibration data after denoising.

[0087] In the present invention, when monitoring is performed based on the vibration data of the fan, because the noise is particularly large when the fan is operating normally, the noise is filtered out before extracting the characteristic vector, which is beneficial to the accuracy of vibration monitoring. When the vibration data is subjected to denoising, a set wavelet can be selected according to the characteristics of the vibration data and the level N of the wavelet decomposition can be determined. Then the vibration data can be subjected to N-layer wavelet packet decomposition. After the vibration data is subjected to N-layer wavelet packet decomposition, the vibration data can be decomposed into high-frequency coefficients from layers 1-N and high-frequency coefficients of the Nth layer. Then a set threshold can be selected to perform quantization denoising on the high-frequency coefficients from layers 1-N and the low-frequency coefficients of the Nth layer obtained by decomposition. After the denoising is completed, the low-frequency coefficients of the Nth layer and the high-frequency coefficients of the 1-N layers can be reconstructed to obtain the denoised vibration data.

[0088] In one embodiment of the present invention, when installing sensors on a wind turbine to obtain vibration data of the wind turbine operation, when the wind turbine is a doubly fed and semi-direct drive unit, the sensors can be deployed at the following nine positions: 1. radial vertical for the main bearing 2. radial vertical for the gearbox housing - left 3. axial for the gearbox housing 4. radial vertical for the gearbox housing - right 5. radial vertical for the generator housing - left 6. radial vertical for the generator housing - right 7. lateral for the nacelle elbow 8. axial for the nacelle elbow 9. internal bearing sleeve of the generator, to obtain vibration data of the wind turbine operation.

[0089] In one embodiment of the present invention, when installing sensors on a fan to obtain vibration data of the fan operation, when the fan is a direct-drive model, the sensors can be installed in the following seven positions: 1. Axial direction of the first measuring point of the main shaft 2. Radial horizontal direction of the first measuring point of the main shaft 3. Radial vertical direction of the first measuring point of the main shaft 4. Radial vertical direction of the second measuring point of the main shaft 5. Radial horizontal direction of the third measuring point on the left side of the stator 6. Axial direction of the third measuring point on the left side of the stator 7. Radial horizontal direction of the fourth measuring point on the right side of the stator, so as to obtain vibration data of the fan operation.

[0090] On the other hand, the present invention also provides a fan online vibration monitoring and fault diagnosis system, such as Figure 5 As shown, the fan vibration monitoring system may include: a fan vibration sensor 1, an intelligent diagnosis terminal 2, a background server 3 and a fault alarm system 4. The fan vibration sensor 1 can be set on the fan to obtain the vibration data of the fan operation. The intelligent diagnosis terminal 2 can obtain the vibration data detected by the fan vibration sensor 1 and execute the above-mentioned fan online vibration monitoring and fault diagnosis method. The background server 3 can perform abnormal processing according to the diagnosis result calculated by the intelligent diagnosis terminal 2. The fault alarm system 4 can perform fault alarm according to the abnormal processing instruction sent by the background server 3. The trained fan vibration fault diagnosis model is deployed to the STM32 chip of the intelligent diagnosis terminal through the STM32CUBEMAX software, so that the fan vibration monitoring terminal realizes the edge computing function, greatly reduces the data transmission volume, reduces the network I / O of the remote server, and realizes the real-time diagnosis and processing function of the fan vibration. The fan online vibration monitoring and fault diagnosis method can process and diagnose the fan vibration data in real time, and preventively monitor the vibration of the fan, so as to timely discover the vibration fault of the fan and give an early warning. The intelligent diagnosis terminal can continuously iterate the algorithm in an end-to-end manner, greatly improving the applicability and reliability of the algorithm. Figure 7As shown, at the wind turbine end: the vibration collection terminal collects data when the wind turbine is operating normally, and then the diagnostic algorithm can be trained based on the data to obtain a model file, and then it can be determined whether the model has passed the verification. If it has not passed the verification, the algorithm can be adjusted and retrained. If the model passes the verification, the corresponding C language header file and source file can be generated, the project can be updated and recompiled, and then it can be burned, the device software can be iteratively updated, and the vibration data of the wind turbine can be collected.

[0091] The system of the intelligent diagnosis terminal includes a main control module, a storage module, a communication module, a sampling module, a power module and a sensor module integrated in the terminal box. The terminal draws energy from the secondary side of the cabin cabinet and is equipped with sufficient lithium batteries as a backup power supply. When the fan loses power, it can still provide electricity for vibration data collection and diagnosis and early warning. Figure 6 As shown, the power module may include a lithium battery and a power controller, the main control module may include a sampling module, a storage module and an MCU module, and the measurement and communication module may include a sensor module and a communication module. The energy-taking power line can be drawn from the secondary side of the cabin cabinet to power the system and can power the lithium battery. The sensor module can be extended from the vibration probe to the probe position via the data transmission line to collect vibration data and can transmit it to the sampling module. The sampling module collects data and has a storage module for storage, and then the MCU module can execute the monitoring method described above on the data. The power controller can control the lithium battery to power the MCU module. The communication module can also send the information processed by the MCU to the background server. The intelligent diagnostic terminal is provided with a touch screen, which is embedded in the terminal box and has the functions of viewing real-time curves, historical curves, alarm information and battery power.

[0092] Through the above technical scheme, the present invention provides a method and system for online vibration monitoring and fault diagnosis of a fan, which obtains the vibration data of the fan operation within a period of time, and then obtains a training data set according to the vibration data. After obtaining the training data set, a first vibration prediction model and a second vibration prediction model can be constructed. After the training data set is brought into the first vibration prediction model, the predicted value of the vibration data can be obtained. After the predicted value is obtained, the time sample data can be subtracted from the corresponding predicted value, so that the residual data can be obtained. The residual data can be brought into the second vibration prediction model to obtain the residual prediction data. According to the residual prediction data, it can be judged whether there will be a high probability of failure in a certain period of time in the future, and then prevention can be carried out in advance. Then, fault monitoring can be carried out in advance according to the predicted fault state in the predicted time period, and the real time sample data in the predicted time period can be Fourier transformed to obtain a frequency domain signal, so that the frequency components contained in the vibration can be determined according to the frequency domain signal to determine the fault type. The online vibration monitoring and fault diagnosis method of the fan can preventively monitor the vibration of the fan and timely discover the vibration fault of the fan.

[0093] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment in combination with software and hardware. Moreover, the present application may adopt the form of a computer program product implemented in one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that include computer-usable program code.

[0094] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0095] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0096] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0097] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0098] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0099] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0100] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0101] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A method for online vibration monitoring and fault diagnosis of a fan, characterized in that: The fan online vibration monitoring and fault diagnosis method comprises: Acquire vibration data of the fan operation over a period of time, and perform noise reduction processing on the vibration data; Performing windowing and framing on the vibration data after the noise reduction process to extract vibration data features, so as to obtain a training data set; Constructing a first vibration prediction model, and bringing the training data set into the first vibration prediction model to obtain a predicted value of the vibration data; Subtract the corresponding predicted value from the acquired time sample data to obtain the residual data; Bringing the residual data into a second vibration prediction model to obtain residual prediction data; Determine whether the residual prediction data is greater than a set threshold; In the case where the residual prediction data is greater than a set threshold, determining that the working state of the fan within the predicted time period is a predicted fault state; Acquire real time sample data within the predicted time period, and bring it into the first vibration prediction model again to perform correction training on the first vibration prediction model and the second vibration prediction model; The real time sample data within the predicted time period is transformed by Fourier to obtain a frequency domain signal, and the frequency components contained in the vibration are determined according to the frequency domain signal to determine the fault type.

2. The method for online vibration monitoring and fault diagnosis of a fan according to claim 1, characterized in that: The first vibration prediction model and the second vibration prediction model are both Attention-BiGRU-GRU network models.

3. The method for online vibration monitoring and fault diagnosis of a fan according to claim 2, characterized in that: Constructing a first vibration prediction model and bringing the training data set into the first vibration prediction model to obtain a predicted value of the vibration data includes: Performing bidirectional temporal feature extraction on the training data set through the BiGRU neural network layer, and based on this, using the GRU neural network layer to model the extracted features; Combined with the Attention layer, the attention weights of the GRU hidden layer states at different times are calculated to predict vibration data; The prediction results of the vibration data are output through the fully connected layer.

4. The method for online vibration monitoring and fault diagnosis of a fan according to claim 3, characterized in that: The unit structure of GRU is formula (1): Among them, r t 、z t are the outputs of the reset gate and the update gate at time t, W r , b r , W z , b z , W h , b h are the weights and biases of the update gate, reset gate and candidate hidden layer, respectively, h t-1 、h t are the hidden layer information at time t-1 and time t, is the candidate state information at time t, σ(·) and tanh(·) are the sigmoid function and hyperbolic tangent function respectively.

5. The method for online vibration monitoring and fault diagnosis of a fan according to claim 4, characterized in that: The calculation process of the Attention mechanism is shown in formula (2): Among them, h t is the output feature vector of the GRU unit at the tth time step, α is the random weight matrix adaptively updated during training after random initialization, γ is the final output value of the attention layer, and softmax is the normalization function.

6. The method for online vibration monitoring and fault diagnosis of a fan according to claim 1, characterized in that: The vibration data after noise reduction is subjected to windowing and framing to extract vibration data features, so as to obtain a training data set including: Acquire the vibration data arranged in time sequence; Set the sampling time span and sampling time schedule; According to the sampling time progress, a certain time is advanced and according to the time span, feature extraction is performed on the vibration data arranged in chronological order to obtain a training data set.

7. The method for online vibration monitoring and fault diagnosis of a fan according to claim 1, characterized in that: The noise reduction process for the vibration data includes: Select the set wavelet and determine the level N of wavelet decomposition; Performing N-layer wavelet packet decomposition on the vibration data; Select a set threshold to perform quantization denoising on the high-frequency coefficients from layers 1 to N and the low-frequency coefficients of the Nth layer obtained through decomposition; The low-frequency coefficients of the Nth layer after quantization denoising and the high-frequency coefficients from layers 1-N are reconstructed to obtain the denoised vibration data.

8. The method for online vibration monitoring and fault diagnosis of a fan according to claim 1, characterized in that: Obtaining the vibration data of the wind turbine operation over a period of time includes: when the wind turbine is a doubly fed and semi-direct drive unit, the sensors are deployed at the following nine positions:

1. For the main bearing radial vertical 2. Gearbox housing radial vertical-left 3. Gearbox housing axial 4. Gearbox housing radial vertical-right 5. Generator housing radial vertical-left 6. Generator housing radial vertical-right 7. Nacelle elbow transverse 8. Nacelle elbow axial 9. Generator internal bearing sleeve to obtain the vibration data of the wind turbine operation.

9. The method for online vibration monitoring and fault diagnosis of a fan according to claim 1, characterized in that: Obtaining the vibration data of the fan operation over a period of time includes: when the fan is a direct-drive model, the sensors are installed at the following 7 positions:

1. Axial direction of the first measuring point of the main shaft 2. Radial horizontal direction of the first measuring point of the main shaft 3. Radial vertical direction of the first measuring point of the main shaft 4. Radial vertical direction of the second measuring point of the main shaft 5. Radial horizontal direction of the third measuring point on the left side of the stator 6. Axial direction of the third measuring point on the left side of the stator 7. Radial horizontal direction of the fourth measuring point on the right side of the stator, so as to obtain the vibration data of the fan operation.

10. A fan online vibration monitoring and fault diagnosis system, characterized in that: include: A fan vibration sensor is installed on the fan to obtain vibration data of the fan operation; An intelligent diagnostic terminal, arranged on the fan, obtains vibration data detected by the fan vibration sensor and executes the fan online vibration monitoring and fault diagnosis method according to any one of claims 1 to 7; The backend server performs exception processing according to the diagnosis result calculated by the intelligent diagnosis terminal; The fault alarm system performs fault alarm according to the abnormality handling instruction sent by the background server.

Images