Real-time vibration data acquisition and analysis method and related system

Abstract

The invention belongs to the technical field of data analysis, discloses a real-time vibration data acquisition and analysis method and a related system, and adopts an LSTM (Long Short Term Memory) neural network to capture time sequence characteristics in vibration signals and reserve time domain information, so that the change of the running state of a generator set can be reflected more accurately. Through the combination of the LSTM neural network and the K-means clustering algorithm, the model can deeply learn complex features of vibration signals and distinguish different modes, thereby improving the recognition precision of abnormal data. The K-means clustering can classify the signals in the data processing process, so that the trained LSTM model is more accurate and effective. The LSTM neural network has the capability of processing long sequence data, can adaptively analyze different vibration state signals in multiple modes, is not limited to the operation state of a specific mode, and thus is suitable for a complex generator set operation environment.

Description

Technical Field

[0001] The present invention belongs to the technical field of data analysis, and in particular relates to a real-time vibration data acquisition and analysis method and a related system. Background Art

[0002] At present, the collection of vibration data of generators is a key technical activity for judging the working status of generators, which is mainly achieved through a specific vibration monitoring system. This system consists of vibration sensors, data collectors and computers, etc. It can monitor the operating status of generators in real time, detect and diagnose faults in time, thereby extending the life of the equipment and reducing power generation losses.

[0003] Chinese patent CN115434870B discloses a method, device, equipment and storage medium for processing vibration data of a wind power generation group, the method comprising: obtaining original vibration evaluation data of the wind power generation group, obtaining current vibration data and current speed data according to the original vibration evaluation data; calculating the current vibration data and current speed data respectively through a target mean square value strategy; performing data extrapolation on the root mean square value of the vibration velocity and the root mean square value of the acceleration according to a preset root mean square standard rule; preprocessing the current vibration data and the current speed data according to a target parameter format rule; in the above manner, calculating the root mean square value of the wind power generation group data meter, then performing data extrapolation according to a preset root mean square standard rule, and preprocessing according to the target parameter format rule, so that the final file and data meet the diagnostic requirements, but the existing method is insufficient in information and lacks time domain information when extrapolating the root mean square value of the vibration velocity and the root mean square value of the acceleration, and thus cannot accurately evaluate the abnormal data and operating status of the generator set. Summary of the invention

[0004] The purpose of the present invention is to overcome the shortcomings of existing methods in extrapolating root mean square values ​​of vibration velocity and root mean square values ​​of acceleration, such as insufficient information and lack of time domain information, and thus inability to accurately evaluate abnormal data and operating status of generator sets, and to provide a real-time vibration data collection and analysis method and related systems.

[0005] In order to achieve the above object, the present invention adopts the following technical solution: In a first aspect, the present invention provides a real-time vibration data collection and analysis method, comprising the following steps: Acquire the original vibration signal of the generator set under different operating modes, and preprocess the original vibration signal to obtain a signal preprocessing set; The signal preprocessing set is divided into a training set and a test set; The training set is used to train the abnormal data recognition model of the LSTM neural network model, and the K-means clustering algorithm is used to process the LSTM neural network model until the training of the abnormal data recognition model is completed to obtain the trained LSTM neural network model; The trained LSTM neural network model is tested using the test set until the test reaches the required accuracy, the test is completed, and the final LSTM neural network model is obtained; The original vibration signals of the generator set under different operating modes are input into the final LSTM neural network model to obtain the abnormal signal set.

[0006] A further improvement of the present invention is that the specific method for preprocessing the original vibration signal is as follows: Add white noise to the original vibration signal; Decomposing the original vibration signal with white noise added into at least one set of intrinsic mode function components based on the empirical mode decomposition algorithm; Calculate the overall mean of the intrinsic mode function components as the intrinsic mode function component output; Loading the intrinsic mode function component output, capturing the frequency and time characteristics of the intrinsic mode function component output based on continuous wavelet transform; The signal wavelet coefficients are obtained by using the inner product of the wavelet function and the intrinsic mode function component output. The signal wavelet coefficients, frequency and time characteristics are plotted into a composite time-frequency diagram in combination with short-time Fourier transform. The composite time-frequency diagram is used as the signal preprocessing set to complete the preprocessing of the original vibration signal.

[0007] A further improvement of the present invention is that when the abnormal data recognition model of the LSTM neural network model is trained using a training set, the specific method of using the K-means clustering algorithm to process the abnormal data recognition model in training is as follows: When the training set is used to train the abnormal data recognition model of the LSTM neural network model, the lightweight detection structure VoV-GSCSP in the K-means clustering algorithm is used to replace the hidden door of the LSTM neural network model, and the activation function, loss function and learning rate hyperparameters of the LSTM neural network model are set to complete the processing of the abnormal data recognition model.

[0008] A further improvement of the present invention is that the conditions until the training of the abnormal data recognition model is completed are as follows: The trained abnormal data recognition model is iteratively trained using the training set until the abnormal data recognition model converges.

[0009] A further improvement of the present invention is that the trained LSTM neural network model is tested using a test set until the test reaches the required accuracy. The specific method for completing the test is as follows: Use the test set to test the trained LSTM neural network model to determine whether the test results meet the preset accuracy threshold. If so, output the trained abnormal data recognition model; otherwise, re-use the K-means clustering algorithm to process the LSTM neural network model until the test results meet the preset accuracy threshold.

[0010] A further improvement of the present invention is that the original vibration signal of the generator set under different operating modes is input into the final LSTM neural network model, and the specific method of obtaining the abnormal signal set is as follows: Perform information conversion on the original vibration signal input into the final LSTM neural network model to obtain a signal dimension reduction set; The abnormal data recognition model in the LSTM neural network model normalizes the signal dimension reduction set and performs one-hot encoding to obtain the encoding matrix; The coding matrix is ​​input into the K-means clustering algorithm, the K-means clustering algorithm identifies at least one set of cluster centers of the coding matrix, traverses the coding matrix, and calculates the cluster membership of the cluster center corresponding to a single set of data vectors of the coding matrix; The lightweight detection structure VoV-GSCSP is used to read the cluster membership of the cluster center corresponding to a single group of data vectors. Based on the set membership threshold, it is judged whether the current single group of data vectors is an abnormal signal. If the membership threshold is exceeded, the abnormal signal corresponding to the current single group of data vectors is output.

[0011] In a second aspect, the present invention provides a real-time vibration data acquisition and analysis system, comprising: A preprocessing module is used to obtain the original vibration signal of the generator set under different operating modes, and preprocess the original vibration signal to obtain a signal preprocessing set; A signal classification module is used to divide the signal preprocessing set into a training set and a test set; The model training module is used to train the abnormal data recognition model of the LSTM neural network model using the training set, and at the same time use the K-means clustering algorithm to process the LSTM neural network model until the training of the abnormal data recognition model is completed to obtain the trained LSTM neural network model; The model testing module is used to test the trained LSTM neural network model using the test set until the test reaches the required accuracy, completes the test, and obtains the final LSTM neural network model; The output module is used to input the original vibration signals of the generator set under different operating modes into the final LSTM neural network model to obtain an abnormal signal set.

[0012] A further improvement of the present invention is that the function of the model training module is implemented by the following method: When the abnormal data recognition model of the LSTM neural network model is trained using the training set, the lightweight detection structure VoV-GSCSP in the K-means clustering algorithm is used to replace the hidden door of the LSTM neural network model, and the activation function, loss function and learning rate hyperparameters of the LSTM neural network model are set to complete the processing of the abnormal data recognition model; The trained abnormal data recognition model is iteratively trained using the training set until the abnormal data recognition model converges.

[0013] In a third aspect, the present invention provides an electronic device including a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of a real-time vibration data collection and analysis method when executing the computer program.

[0014] In a fourth aspect, the present invention provides a storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of a real-time vibration data collection and analysis method.

[0015] Compared with the prior art, the present invention has the following beneficial effects: The present invention combines LSTM neural network and K-means clustering algorithm to deeply analyze and identify abnormalities of vibration data of generator sets. When analyzing vibration data, traditional methods usually only rely on the root mean square value of vibration velocity and the root mean square value of acceleration, ignoring the time domain dynamic changes in vibration signals. The present invention adopts LSTM neural network to capture the time series characteristics in vibration signals and retain the time domain information, so as to more accurately reflect the changes in the operating state of the generator set. Through the combination of LSTM neural network and K-means clustering algorithm, the model can deeply learn the complex characteristics of vibration signals and distinguish different modes, thereby improving the recognition accuracy of abnormal data. K-means clustering can classify signals during data processing, making the trained LSTM model more accurate and effective. LSTM neural network has the ability to process long sequence data, can adaptively analyze different vibration state signals in multiple modes, and is not limited to the operating state of a specific mode, so it is suitable for complex generator set operating environment. The present invention reduces the need for human intervention through automated preprocessing, training and testing processes. The LSTM model can automatically identify abnormal signals, reduce the dependence of experts on data, and can respond to abnormal situations more quickly. The present invention is not limited to a single statistical indicator (such as the root mean square value) in the traditional method, but also makes full use of a large amount of original vibration signal data, improves the information utilization rate, and thus improves the ability to evaluate the operating status of the generator set. In summary, when analyzing the vibration data of the generator set, the present invention can more comprehensively and accurately identify abnormal signals, thereby improving the safety and reliability of equipment operation. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a flow chart of the present invention; Figure 2 is a system diagram of the present invention; Figure 3 This is a system diagram of Example 4. DETAILED DESCRIPTION

[0017] In order to further understand the content of the present invention, the present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the embodiments are only for explaining the present invention and are not intended to limit it.

[0018] Embodiment 1: See also Figure 1 , the real-time vibration data collection and analysis method comprises the following steps: S1, obtaining the original vibration signal of the generator set under different operating modes, and preprocessing the original vibration signal to obtain a signal preprocessing set.

[0019] S2, divide the signal preprocessing set into a training set and a test set.

[0020] S3, using the training set to train the abnormal data recognition model of the LSTM neural network model, and at the same time using the K-means clustering algorithm to process the LSTM neural network model until the training of the abnormal data recognition model is completed, and the trained LSTM neural network model is obtained.

[0021] S4, use the test set to test the trained LSTM neural network model until the test reaches the required accuracy, complete the test, and obtain the final LSTM neural network model. S5, input the original vibration signals of the generator set under different operating modes into the final LSTM neural network model to obtain the abnormal signal set.

[0022] Embodiment 2: See also Figure 2 , real-time vibration data acquisition and analysis system, including: A preprocessing module is used to obtain the original vibration signal of the generator set under different operating modes, and preprocess the original vibration signal to obtain a signal preprocessing set; A signal classification module is used to divide the signal preprocessing set into a training set and a test set; The model training module is used to train the abnormal data recognition model of the LSTM neural network model using the training set, and at the same time use the K-means clustering algorithm to process the LSTM neural network model until the training of the abnormal data recognition model is completed to obtain the trained LSTM neural network model; The model testing module is used to test the trained LSTM neural network model using the test set until the test reaches the required accuracy, completes the test, and obtains the final LSTM neural network model; The output module is used to input the original vibration signals of the generator set under different operating modes into the final LSTM neural network model to obtain an abnormal signal set.

[0023] Embodiment 3: Real-time vibration data collection and analysis method, specifically including: Step S10, collecting original vibration signals of the generator set in different operating modes based on the vibration sensor, preprocessing the original vibration signals to obtain a signal preprocessing set; It should be noted that the vibration sensor can be a displacement sensor, a velocity sensor or an acceleration sensor, and the vibration sensor is deployed in key components such as the wind wheel and gearbox of the generator set. The vibration sensor data transmission method can choose wired Ethernet or wireless GRPS / 3G network.

[0024] Step S20, obtaining a signal preprocessing set, pre-building an abnormal data recognition model combining a K-means clustering algorithm and an LSTM neural network model, dividing the signal preprocessing set into a training set and a test set, and using the training set to pre-train the abnormal data recognition model; Step S30, loading the pre-trained LSTM neural network model, introducing the K-means clustering algorithm to a group of LSTM blocks of the LSTM neural network model, using the lightweight detection structure VoV-GSCSP to replace the hidden door of the LSTM neural network model, and setting the activation function, loss function and learning rate hyperparameters of the LSTM neural network model; Step S40, obtaining a training set, and iteratively training the improved LSTM neural network model using the training set until the model converges; Step S50, using the test set to test the converged LSTM neural network model to determine whether the test result meets the preset accuracy threshold; Step S60: if the preset accuracy threshold is met, output the trained abnormal data recognition model; In this embodiment, the accuracy threshold may be set to 0.95-0.98.

[0025] If the accuracy threshold is not met, the optimizer Adam is used to optimize the loss function of the abnormal data recognition model, and the process returns to step S40 to restart training.

[0026] Step S70, taking the signal preprocessing set as input, executing the abnormal data recognition model, the abnormal data recognition model identifies and determines the abnormal signal, and generates an abnormal signal set.

[0027] Step S80, obtaining an abnormal signal set, and the abnormal data recognition model calculates the operating state of the generator set based on the cascading failure evaluation.

[0028] In an embodiment of the present invention, a real-time vibration data collection and analysis method is provided. When analyzing the real-time vibration data, the vibration data is decomposed and reconstructed based on the empirical mode decomposition algorithm and the continuous wavelet transform, so as to ensure that the reconstructed original data is complete and distortion-free, and effectively remove the random noise in the vibration signal, thereby improving the signal-to-noise ratio of the signal. By constructing an abnormal data recognition model that introduces the K-means clustering algorithm and the lightweight detection structure VoV-GSCSP, the recognition accuracy of the abnormal data recognition model for the abnormal data can be guaranteed, and the recognition speed and accuracy of the abnormal data recognition model are significantly improved, which effectively overcomes the problem that the existing method is insufficient in information and lacks time domain information when extrapolating the root mean square value of the vibration velocity and the root mean square value of the acceleration, and thus cannot accurately evaluate the abnormal data and operating status of the generator set.

[0029] The method for preprocessing the original vibration signal specifically comprises: Step S101, obtaining the original vibration signal and adding white noise to the original vibration signal , the amplitude of the white noise is , the amplitude value range is the positive and negative three standard deviations of the original vibration signal, which can be 0.2-0.8; It should be noted that since the power spectral density of white noise is uniformly distributed in the entire frequency domain, adding it to the vibration signal can help analyze and understand the response of the signal at each frequency component, which has a positive effect on model analysis to identify the characteristics and dynamic behavior of the signal.

[0030] Step S102, decomposing the original vibration signal with white noise added into at least one group of intrinsic mode function components based on an empirical mode decomposition algorithm; Step S103, using multiple groups of white noises of different levels Repeat the above steps, and then calculate the overall mean of the intrinsic mode function components as the intrinsic mode function component output; Step S104, loading the intrinsic mode function component output, and capturing the frequency and time characteristics of the intrinsic mode function component output based on continuous wavelet transform; Step S105, using the inner product of the wavelet function and the intrinsic mode function component output to obtain the signal wavelet coefficient, combining the short-time Fourier transform with the signal wavelet coefficient, frequency and time characteristics to draw a composite time-frequency diagram, and using the composite time-frequency diagram as a signal preprocessing set; The wavelet function is expressed as: (1) in, is the frequency characteristic of the intrinsic mode function component, is the time feature, It is the intrinsic mode function component output.

[0031] In the embodiment of the present invention, the empirical mode decomposition algorithm, short-time Fourier transform, and continuous wavelet transform algorithm are effectively combined when preprocessing the original vibration signal. Compared with the traditional Fourier transform or short-time Fourier transform, while maintaining high time-frequency resolution, it can also reduce the computational complexity through algorithm optimization, thereby improving the robustness of the data and facilitating the construction and training of the abnormal data recognition model.

[0032] The method for pre-building an abnormal data recognition model combining a K-means clustering algorithm and an LSTM neural network model specifically includes: Step S201, introducing the K-means clustering algorithm mechanism, mining all possible cluster centers of the training set based on the K-means clustering algorithm membership; It should be noted that the K-means clustering algorithm mechanism consists of an initialization layer, an allocation layer, and an update layer. The initialization layer is used to mine all possible cluster centers of the training set based on the K-means clustering algorithm membership, and the allocation layer is used to assign each data point to the nearest cluster center to form K clusters.

[0033] The abnormal data recognition model consists of an input gate, an encoder, a K-means clustering algorithm mechanism, an LSTM block and a lightweight detection structure VoV-GSCSP. The input gate is connected to the encoder, the encoder is connected to the K-means clustering algorithm mechanism, the K-means clustering algorithm mechanism is connected to the LSTM block, and the LSTM block is connected to the lightweight detection structure VoV-GSCSP.

[0034] Step S202, freezing the LSTM block of the LSTM neural network model, and replacing the LSTM block of the LSTM neural network model with a K-means clustering algorithm mechanism; Step S203, set the parameters of the K-means clustering algorithm mechanism, the range of the number of clusters, and the iteration error to 10 -5 , the maximum number of iterations is 1200, and the current number of iterations is 1; Step S204, obtaining a training set, pre-training the K-means clustering algorithm mechanism based on the training set, calculating the cluster center corresponding to the training set based on the K-means clustering algorithm, calculating the clustering criterion function value according to the cluster center, and selecting the optimal clustering criterion function value.

[0035] Step S205, using the test set to verify the pre-trained K-means clustering algorithm mechanism, and selecting the mean absolute ratio error to determine the deviation between the clustering prediction value of the K-means clustering algorithm mechanism and the clustering true value; The mean absolute proportional error is calculated by the following formula: (2) in, Get the frequency for the vibration signal, Get the time for the vibration signal, The period for obtaining the mark of the vibration signal can be 1 day, 1 week, or 30 days. is the clustering true value, To predict the cluster value; Step S206, if the mean absolute ratio error is less than a preset deviation threshold, in this embodiment, the preset deviation threshold may be 0.05, then the pre-trained K-means clustering algorithm mechanism is output, and the K-means clustering algorithm mechanism is nested in the LSTM neural network model; Step S207, if the mean absolute ratio error is greater than the preset deviation threshold, fine-tune the K-means clustering algorithm mechanism through supervised training, return to step S204, and continue to repeat the training.

[0036] In this embodiment, the loss function of the LSTM neural network model is a cross entropy loss function, which is expressed as follows: (3) in, is the true value, is the predicted value, and is the sample size, Indicates Sample of the moment.

[0037] In an embodiment of the present invention, an abnormal data recognition model and a construction method thereof are provided. The abnormal data recognition model is embedded with a K-means clustering algorithm mechanism and a lightweight detection structure VoV-GSCSP. The synergistic cooperation of the K-means clustering algorithm mechanism and the lightweight detection structure VoV-GSCSP can improve feature recognition accuracy, enhance the generalization ability and robustness of the model, and the K-means clustering result can be used as an auxiliary tool to explain LSTM predictions, thereby increasing the transparency and credibility of the model.

[0038] The method of identifying and judging abnormal signals by the abnormal data recognition model and generating an abnormal signal set specifically includes: Step S301, obtain a signal preprocessing set, and the input gate of the abnormal data recognition model performs information conversion on the data of the signal preprocessing set to obtain a signal dimension reduction set. The information conversion is converted by the following formula: (4) in, is the signal dimension reduction vector, is the dimension of data dimensionality reduction. In this embodiment, the dimension of data dimensionality reduction can be 1-4. for The data dimension reduction weight parameter of the moment signal preprocessing set, is the input representation of the signal preprocessing set; (5) in, is the input representation of the signal preprocessing set, are respectively the pre-coefficient and the post-coefficient of the local linear embedding dimensionality reduction function. In this embodiment, the pre-coefficient and the post-coefficient can be set to 0.01-0.05; Step S302, obtaining a signal dimension reduction set, the encoder of the abnormal data recognition model normalizes the signal dimension reduction set and performs one-hot encoding, and inputs the encoding matrix into the K-means clustering algorithm mechanism; It should be noted that the encoder normalizes the signal dimensionality reduction set and performs one-hot encoding to convert the dimensionality reduction variables in the normalized signal dimensionality reduction set into a format suitable for processing by a machine learning algorithm. The specific method is to create a new binary feature (i.e., a "one-hot" feature) for each possible value of each categorical feature, where only one feature is activated (marked as 1) at any given time, and all other features are marked as 0.

[0039] Step S303, the K-means clustering algorithm mechanism identifies at least one group of cluster centers of the coding matrix, traverses the coding matrix, and calculates the cluster membership of the cluster center corresponding to a single group of data vectors in the coding matrix; The cluster membership is calculated by the following formula: (6) in, is the cluster membership of the cluster center corresponding to a single group of data vectors in the encoding matrix, represents the code of the cluster center, is the number of cluster centers, is the distance between a single set of data vectors and the corresponding cluster center, is the clustering partial derivative corresponding to a single set of data vectors, is the clustering fuzziness. In this embodiment, the clustering fuzziness may be 0.3-0.6; Step S304, use the lightweight detection structure VoV-GSCSP to read the cluster membership of the cluster center corresponding to the single group of data vectors, and determine whether the current single group of data vectors is an abnormal signal based on the set membership threshold. If it exceeds the membership threshold, output the abnormal signal corresponding to the current single group of data vectors.

[0040] The method for calculating the operating status of a generator set based on the abnormal data identification model and the cascading failure evaluation specifically includes: Step S401, obtain an abnormal signal set, and the fully connected layer of the abnormal data recognition model determines the weight corresponding to a single group of data vectors in the abnormal signal set. The weight is calculated by the following formula: (7) in, for The weight corresponding to a single set of data vectors at a certain moment, is the initial weight, is the number of vibration monitoring points, is the weight coefficient. In this embodiment, the initial weight can be set by He initialization, constant initialization or Xavier initialization. It should be noted that the initial weight setting is crucial, which is directly related to the training effect and final performance of the model. By selecting a suitable weight initialization method, the gradient vanishing and explosion problems can be effectively avoided, and the learning and generalization capabilities of the network can be improved.

[0041] Step S402, calculating the operating status of the generator set based on the cascading failure assessment combined with the weight; (8) in, is the operating status value of the generator set, for The weight corresponding to a single set of data vectors at a certain moment, is the number of cluster centers.

[0042] Embodiment 4: See also Figure 3 As shown, the present invention also provides an electronic device 100 for real-time vibration data acquisition and analysis method; the electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on the at least one processor 102, and at least one communication bus 104.

[0043] The memory 101 can be used to store the computer program 103. The processor 102 implements the steps of the real-time vibration data acquisition and analysis method described in Example 1 by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101. The memory 101 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area can store data (such as audio data) created according to the use of the electronic device 100, etc. In addition, the memory 101 can include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (FlashCard), at least one disk storage device, a flash memory device, or other non-volatile solid-state storage devices.

[0044] The at least one processor 102 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor 102 may be a microprocessor or any conventional processor, etc. The processor 102 is the control center of the electronic device 100, and uses various interfaces and lines to connect various parts of the entire electronic device 100.

[0045] The memory 101 in the electronic device 100 stores a plurality of instructions to implement a real-time vibration data collection and analysis method, and the processor 102 can execute the plurality of instructions to implement: Acquire the original vibration signal of the generator set under different operating modes, and preprocess the original vibration signal to obtain a signal preprocessing set; The signal preprocessing set is divided into a training set and a test set; The training set is used to train the abnormal data recognition model of the LSTM neural network model, and the K-means clustering algorithm is used to process the LSTM neural network model until the training of the abnormal data recognition model is completed to obtain the trained LSTM neural network model; The trained LSTM neural network model is tested using the test set until the test reaches the required accuracy, the test is completed, and the final LSTM neural network model is obtained; The original vibration signals of the generator set under different operating modes are input into the final LSTM neural network model to obtain the abnormal signal set.

[0046] Embodiment 5: If the module / unit integrated in the electronic device 100 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 present invention implements all or part of the process in the above-mentioned embodiment method, and can also be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of the above-mentioned method embodiments when executed by the processor. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, disk, optical disk, computer memory and read-only memory (ROM, Read-Only Memory).

[0047] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0048] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks 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 produce 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.

[0049] 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 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0050] 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.

[0051] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A real-time vibration data acquisition and analysis method, characterized in that: The following steps are involved: Acquire the original vibration signal of the generator set under different operating modes, and preprocess the original vibration signal to obtain a signal preprocessing set; The signal preprocessing set is divided into a training set and a test set; The training set is used to train the abnormal data recognition model of the LSTM neural network model, and the K-means clustering algorithm is used to process the LSTM neural network model until the training of the abnormal data recognition model is completed to obtain the trained LSTM neural network model; The trained LSTM neural network model is tested using the test set until the test reaches the required accuracy, the test is completed, and the final LSTM neural network model is obtained; The original vibration signals of the generator set under different operating modes are input into the final LSTM neural network model to obtain the abnormal signal set.

2. The real-time vibration data acquisition and analysis method according to claim 1, characterized in that: The specific method for preprocessing the original vibration signal is as follows: Add white noise to the original vibration signal; Decomposing the original vibration signal with white noise added into at least one set of intrinsic mode function components based on the empirical mode decomposition algorithm; Calculate the overall mean of the intrinsic mode function components as the intrinsic mode function component output; Loading the intrinsic mode function component output, capturing the frequency and time characteristics of the intrinsic mode function component output based on continuous wavelet transform; The signal wavelet coefficients are obtained by using the inner product of the wavelet function and the intrinsic mode function component output. The signal wavelet coefficients, frequency and time characteristics are plotted into a composite time-frequency diagram in combination with short-time Fourier transform. The composite time-frequency diagram is used as the signal preprocessing set to complete the preprocessing of the original vibration signal.

3. The real-time vibration data acquisition and analysis method according to claim 1, characterized in that: When the training set is used to train the abnormal data recognition model of the LSTM neural network model, the specific method of using the K-means clustering algorithm to process the abnormal data recognition model in training is as follows: When the training set is used to train the abnormal data recognition model of the LSTM neural network model, the lightweight detection structure VoV-GSCSP in the K-means clustering algorithm is used to replace the hidden door of the LSTM neural network model, and the activation function, loss function and learning rate hyperparameters of the LSTM neural network model are set to complete the processing of the abnormal data recognition model.

4. The real-time vibration data acquisition and analysis method according to claim 1, characterized in that: The conditions for completing the training of the abnormal data recognition model are as follows: The trained abnormal data recognition model is iteratively trained using the training set until the abnormal data recognition model converges.

5. The real-time vibration data acquisition and analysis method according to claim 1, characterized in that: The trained LSTM neural network model is tested using the test set until the test reaches the required accuracy. The specific method to complete the test is as follows: Use the test set to test the trained LSTM neural network model to determine whether the test result meets the preset accuracy threshold. If it meets the preset accuracy threshold, output the trained abnormal data recognition model; Otherwise, the K-means clustering algorithm is used again to process the LSTM neural network model until the test result meets the preset accuracy threshold.

6. The real-time vibration data acquisition and analysis method according to claim 1, characterized in that: The specific method of inputting the original vibration signals of the generator set under different operating modes into the final LSTM neural network model to obtain the abnormal signal set is as follows: Perform information conversion on the original vibration signal input into the final LSTM neural network model to obtain a signal dimension reduction set; The abnormal data recognition model in the LSTM neural network model normalizes the signal dimension reduction set and performs one-hot encoding to obtain the encoding matrix; The coding matrix is ​​input into the K-means clustering algorithm, the K-means clustering algorithm identifies at least one set of cluster centers of the coding matrix, traverses the coding matrix, and calculates the cluster membership of the cluster center corresponding to a single set of data vectors of the coding matrix; The lightweight detection structure VoV-GSCSP is used to read the cluster membership of the cluster center corresponding to a single group of data vectors. Based on the set membership threshold, it is judged whether the current single group of data vectors is an abnormal signal. If the membership threshold is exceeded, the abnormal signal corresponding to the current single group of data vectors is output.

7. Real-time vibration data acquisition and analysis system, characterized in that: include: A preprocessing module is used to obtain the original vibration signal of the generator set under different operating modes, and preprocess the original vibration signal to obtain a signal preprocessing set; A signal classification module is used to divide the signal preprocessing set into a training set and a test set; The model training module is used to train the abnormal data recognition model of the LSTM neural network model using the training set, and at the same time use the K-means clustering algorithm to process the LSTM neural network model until the training of the abnormal data recognition model is completed to obtain the trained LSTM neural network model; The model testing module is used to test the trained LSTM neural network model using the test set until the test reaches the required accuracy, completes the test, and obtains the final LSTM neural network model; The output module is used to input the original vibration signals of the generator set under different operating modes into the final LSTM neural network model to obtain an abnormal signal set.

8. The real-time vibration data acquisition and analysis system according to claim 7, characterized in that: The functions of the model training module are implemented through the following methods: When the abnormal data recognition model of the LSTM neural network model is trained using the training set, the lightweight detection structure VoV-GSCSP in the K-means clustering algorithm is used to replace the hidden door of the LSTM neural network model, and the activation function, loss function and learning rate hyperparameters of the LSTM neural network model are set to complete the processing of the abnormal data recognition model; The trained abnormal data recognition model is iteratively trained using the training set until the abnormal data recognition model converges.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the real-time vibration data collection and analysis method according to any one of claims 1 to 6 are implemented.

10. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the real-time vibration data collection and analysis method according to any one of claims 1 to 6 are implemented.

Images