Method, system and computer equipment for predicting remaining service life of fan bearings

Abstract

The present invention relates to a method, system and computer equipment for predicting the remaining service life of a fan bearing, the method comprising: firstly obtaining the vibration data of the fan bearing at a preset time; secondly performing feature extraction on the data to obtain energy feature vectors corresponding to various frequency features of the vibration data within a preset time period; then inputting the energy feature vectors corresponding to various frequency features of the vibration data into a pre-trained health index prediction model to obtain the health index value of the fan bearing at a preset time; finally determining the remaining service life of the fan bearing based on the health index value. The technical solution provided by the present invention is based on extracting the features of the vibration data of the bearing to obtain the energy features corresponding to various types of frequencies, then determining the health index value of the fan bearing, and then obtaining the remaining service life of the bearing, so that the prediction of the remaining service life of the bearing is more accurate.

Description

Technical Field

[0001] The present invention relates to the field of bearing life prediction, and in particular to a method, system and computer equipment for predicting the remaining service life of a fan bearing. Background Art

[0002] Wind energy is a renewable energy source with great potential and has received widespread attention from countries around the world. Wind power generation helps to resolve the contradiction between economic and social development and environmental pollution, and therefore plays an important role in adjusting the energy structure. Wind power generation systems generally include wind turbines, gearboxes, blades, power converters and other components, among which wind turbines are the core components of wind power generation systems, realizing the conversion of mechanical energy into electrical energy. Wind turbines are mostly installed in remote and harsh environments such as the sea, plateaus, and mountainous areas. The operating conditions are extremely harsh, causing the performance of wind turbine components to degrade easily. At present, with the continuous increase in the installed capacity of wind power generation, the losses caused by wind turbine bearing failures are also increasing. Ensuring the stable, safe and reliable operation of wind turbines has become the primary issue facing the wind power industry.

[0003] The existing method for predicting the remaining service life of wind turbine bearings is to obtain the vibration signal or electrical signal of the wind turbine bearing through the monitoring system, perform time domain analysis and frequency domain analysis to extract the characteristic parameters of the signal, and use fitting methods to perform life regression analysis. However, the extracted characteristic parameters do not necessarily have a strong correlation with the life of the bearing, and the generalization ability of the wind turbine at different speeds is poor; the life regression model is also relatively simple and cannot discover the relevant features in the bearing vibration data. It is also easily disturbed by external noise signals. There is a large error between the fitted degradation curve and the actual degradation curve, which ultimately leads to low accuracy in the remaining service life prediction value. Summary of the invention

[0004] The present application provides a method, system and computer device for predicting the remaining useful life of a fan bearing, so as to at least solve the technical problem of low accuracy in predicting the remaining useful life of a fan bearing in the related art.

[0005] The first embodiment of the present application provides a method for predicting the remaining service life of a fan bearing, the method comprising:

[0006] Obtain the fan bearing vibration data at a preset time;

[0007] Performing feature extraction on the data to obtain energy feature vectors corresponding to various frequency features of the vibration data within a preset time period;

[0008] Inputting energy feature vectors corresponding to various frequency features of the vibration data into a pre-trained health index prediction model to obtain a health index value of the fan bearing at a preset time;

[0009] determining a remaining service life of the fan bearing based on the health indicator value;

[0010] Among them, the various frequency characteristics include: outer circle ball pass frequency, inner circle ball pass frequency, ball failure frequency, basic training frequency and shaft rotation basic frequency.

[0011] The second embodiment of the present application provides a remaining service life prediction system for a fan bearing, the system comprising:

[0012] A first acquisition module, used to acquire the vibration data of the fan bearing at a preset time;

[0013] A second acquisition module is used to extract features from the data to obtain energy feature vectors corresponding to various frequency features of the vibration data within a preset time period;

[0014] A third acquisition module is used to input the energy feature vectors corresponding to various frequency features of the vibration data into a pre-trained health index prediction model to obtain the health index value of the fan bearing at a preset time;

[0015] A determination module, configured to determine the remaining service life of the fan bearing based on the health indicator value;

[0016] Among them, the various frequency characteristics include: outer circle ball pass frequency, inner circle ball pass frequency, ball failure frequency, basic training frequency and shaft rotation basic frequency.

[0017] The third aspect of the present application provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the prediction method of the first aspect of the present application is implemented.

[0018] The technical solution provided by the embodiments of the present application brings at least the following beneficial effects:

[0019] In the remaining service life prediction method, system and computer device of the fan bearing provided by the present invention, firstly, the vibration data of the fan bearing at a preset time is obtained; secondly, the data is feature extracted to obtain the energy feature vectors corresponding to the various frequency features of the vibration data within a preset time period; then, the energy feature vectors corresponding to the various frequency features of the vibration data are input into a pre-trained health index prediction model to obtain the health index value of the fan bearing at a preset time; finally, the remaining service life of the fan bearing is determined based on the health index value. The technical solution provided by the present invention is based on extracting the energy features corresponding to each type of frequency from the vibration data of the bearing, and then determining the health index value of the fan bearing, thereby obtaining the remaining service life of the bearing, so that the remaining service life prediction of the bearing is more accurate.

[0020] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0022] Figure 1 is a flow chart of a method for predicting the remaining service life of a fan bearing provided according to an embodiment of the present application;

[0023] Figure 2 is a wavelet neural network diagram provided according to an embodiment of the present application;

[0024] Figure 3 It is a structural diagram of a remaining useful life prediction system for a fan bearing provided according to an embodiment of the present application. DETAILED DESCRIPTION

[0025] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0026] In the remaining service life prediction method, system and computer device of the fan bearing proposed in the present application, firstly, the vibration data of the fan bearing at a preset time is obtained; secondly, the data is feature extracted to obtain the energy features corresponding to various frequency features of the vibration data within a preset time period; then, the energy features corresponding to various frequency features of the vibration data are feature smoothed to obtain an energy vector composed of the smoothed energy features; then, the energy vector is input into a pre-trained health index prediction model to obtain the health index value of the fan bearing at a preset time; finally, the remaining service life of the fan bearing is determined based on the health index value. The technical solution provided by the present invention is based on extracting the features of the vibration data of the bearing to obtain the energy features corresponding to various types of frequencies, and then determining the health index value of the fan bearing, thereby obtaining the remaining service life of the bearing, so that the remaining service life prediction of the bearing is more accurate.

[0027] Example 1

[0028] Figure 1 A flowchart of a method for predicting the remaining service life of a fan bearing provided by an embodiment of the present disclosure is shown in FIG. Figure 1 As shown, the method includes:

[0029] Step 1: Obtain the fan bearing vibration data at a preset time;

[0030] It should be noted that after obtaining the data, it is necessary to determine whether the acquired data has missing data, outliers and other phenomena. If so, the least squares, polynomial regression and other methods are used to interpolate the values ​​and remove outliers. The threshold set in advance is used to determine whether the collected data exceeds the threshold, so as to directly determine whether the equipment parameters are faulty. If it is determined that there is a fault, it is directly reported for repair. If it is determined that there is no fault, the health trend of the equipment is predicted.

[0031] Step 2: Extract features from the data to obtain energy feature vectors corresponding to various frequency features of the vibration data within a preset time period;

[0032] In the embodiment of the present disclosure, the feature extraction of the data to obtain energy feature vectors corresponding to various frequency features of the vibration data within a preset time period includes:

[0033] It should be noted that the various frequency characteristics include: outer circle ball pass frequency, inner circle ball pass frequency, ball failure frequency, basic training frequency and shaft rotation basic frequency.

[0034] The outer ring ball pass frequency, inner ring ball pass frequency, ball failure frequency and basic training frequency are all based on the shaft rotation basic frequency f of the fan bearing. N Sure.

[0035] Specifically, the outer ring ball frequency can be 5.6113f N , the outer ball frequency can be 7.3887f N , the ball fault frequency can be 7.1776f N , the basic training frequency can be 0.4316f N .

[0036] First, the data is subjected to a fast Fourier transform to obtain a frequency spectrum corresponding to the data;

[0037] For example, the data is subjected to a fast Fourier transform to obtain a frequency spectrum corresponding to the data, wherein the frequency spectrum has frequency as the horizontal axis and amplitude as the vertical axis.

[0038] Secondly, find the amplitude corresponding to each type of frequency feature in the spectrum diagram corresponding to the data;

[0039] Then, determining energy features corresponding to various frequency features based on the amplitudes corresponding to the various frequency features;

[0040] For example, the energy characteristic corresponding to the outer ball pass frequency characteristic of the vibration data at the preset time t is determined by the following formula:

[0041]

[0042] In the formula, is the amplitude corresponding to the outer ring ball pass frequency of the vibration data at the preset time t, A F,t is the amplitude of the frequency spectrum of the vibration data at the preset time t after fast Fourier transformation, and F is the total number of types of frequency features.

[0043] Finally, the energy characteristics corresponding to various frequency characteristics are used to construct an energy characteristic vector corresponding to the fan bearing vibration data at a preset time.

[0044] For example, the energy characteristic vector corresponding to the wind turbine bearing vibration data at the preset time t may be: in, is the energy characteristic value corresponding to the inner circle ball pass frequency characteristic of the vibration data at the preset time t, is the energy characteristic value corresponding to the ball fault frequency characteristic of the vibration data at the preset time t, is the energy characteristic value corresponding to the basic training frequency characteristic of the vibration data at the preset time t, It is the energy characteristic value corresponding to the shaft rotation fundamental frequency characteristic of the vibration data at the preset time t.

[0045] Step 3: Input the energy feature vectors corresponding to various frequency features of the vibration data into a pre-trained health index prediction model to obtain the health index value of the fan bearing at a preset time;

[0046] It should be noted that the training process of the pre-trained health index prediction model includes:

[0047] First, the energy feature time series data corresponding to various frequency features after feature extraction of the fan bearing vibration data at each moment in the historical period and the health index values ​​corresponding to the fan bearing vibration data at each moment in the historical period are obtained;

[0048] Secondly, the energy feature time series data corresponding to each type of frequency feature is smoothed by using an exponential smoothing method to obtain the energy features corresponding to each type of frequency feature after smoothing;

[0049] For example, the exponential smoothing method is used and the smoothing parameter α=0.9 is selected to smooth the time series data corresponding to the i-th frequency feature, as shown in the following formula:

[0050] y i,0 =s i,0

[0051] y i,t =α·y i,t-1+(1-α)s i,t ,t>0

[0052] y i,0 is the value of the first data in the time series data corresponding to the smoothed i-th frequency feature, s i,0 is the value of the first data in the time series data corresponding to the i-th frequency feature before smoothing, s i,t is the value of the data at time t in the time series data corresponding to the i-th frequency feature before smoothing, y i,t is the value of the data corresponding to time t in the time series data corresponding to the smoothed i-th frequency feature, y i,t-1 is the value of the data corresponding to time t-1 in the time series data corresponding to the i-th frequency feature after smoothing, i∈[1~F].

[0053] Then, the energy features corresponding to the smoothed various frequency features and the time correspond to form a time series, and the time in the time series corresponding to the various frequency features is aligned to form an energy feature matrix corresponding to the various frequency features in the historical period;

[0054] Subsequently, the energy characteristic matrix is ​​subjected to dimension reduction processing by using the main program analysis method, the eigenvalues ​​after dimension reduction are obtained, and an input eigenvector is formed based on the energy eigenvalues;

[0055] Finally, the input feature vector is used as the input of the wavelet neural network, and the health index value corresponding to the input feature vector is used as the output of the wavelet neural network. The wavelet neural network is trained to obtain a trained health index prediction model.

[0056] For example, Figure 2 As shown, the input feature vectors T1, T2, and T3 are used as the input of the wavelet neural network, and the health index values ​​corresponding to the input feature vectors T1, T2, and T3 are used as the output of the wavelet neural network. The wavelet neural network is trained to obtain a trained health index prediction model, wherein the wavelet neural network consists of an input layer, a hidden layer, and an output layer.

[0057] Step 4: Determine the remaining service life of the fan bearing based on the health index value;

[0058] In the disclosed embodiment, a linear regression is performed on the health index value to obtain the fault time of the wind turbine bearing corresponding to the health index value;

[0059] The remaining service life of the fan bearing is determined based on a preset time and the fault time.

[0060] In the disclosed embodiment, in order to ensure that the pre-trained health index prediction model, i.e., the target model, has globally effective results, the model is trained under a joint learning framework.

[0061] Specifically, the target model to be trained is downloaded from the server, and the model is updated based on the parameter data uploaded by each device to obtain an updated target model.

[0062] After each device receives the feedback of the corresponding target prediction model from the server, it updates its own device model of the current state by downloading the corresponding updated target prediction model.

[0063] For example, server A distributes a data set consisting of health indicator values ​​corresponding to each health level to three devices, device a1, device a2 and device a3. For example, the above devices use the data sets b1, data sets b2 and data sets b3 distributed to their own devices to train their respective models. After their respective models are trained, devices a1, device a2 and device a3 upload the parameters c1, parameter c2 and parameter c3 to be updated to server A through encrypted gradients. Server A aggregates the updated model parameters c1, parameter c2 and parameter c3 of each device, and updates the parameters c1, parameter c2 and parameter c3 to the corresponding prediction models, respectively, to obtain target prediction models B1, B2 and B3, respectively, and feeds back the updated target prediction models B1, B2 and B3 to the corresponding devices. Devices a1, device a2 and device a3 download the corresponding target prediction models from server A respectively.

[0064] To summarize, the remaining service life prediction method for a fan bearing provided by the present invention is based on feature extraction of the vibration data of the bearing to obtain the energy characteristics corresponding to each type of frequency, and then determines the health index value of the fan bearing, and then obtains the remaining service life of the bearing, thereby making the remaining service life prediction of the bearing more accurate.

[0065] Example 2

[0066] Figure 3 A structural diagram of a remaining service life prediction system for a fan bearing provided by an embodiment of the present disclosure, such as Figure 3 As shown, the system comprises:

[0067] A first acquisition module, used to acquire the vibration data of the fan bearing at a preset time;

[0068] A second acquisition module is used to extract features from the data to obtain energy feature vectors corresponding to various frequency features of the vibration data within a preset time period;

[0069] A third acquisition module is used to input the energy feature vectors corresponding to various frequency features of the vibration data into a pre-trained health index prediction model to obtain the health index value of the fan bearing at a preset time;

[0070] A determination module, configured to determine the remaining service life of the fan bearing based on the health indicator value;

[0071] Among them, the various frequency characteristics include: outer circle ball pass frequency, inner circle ball pass frequency, ball failure frequency, basic training frequency and shaft rotation basic frequency.

[0072] In the embodiment of the present disclosure, the second acquisition module includes:

[0073] A transform unit, used for performing a fast Fourier transform on the data to obtain a frequency spectrum corresponding to the data;

[0074] A search unit, used to search for amplitudes corresponding to various frequency features in a frequency spectrum corresponding to the data;

[0075] A first determining unit, configured to determine energy features corresponding to various frequency features based on amplitudes corresponding to the various frequency features;

[0076] The composition unit is used to use the energy characteristics corresponding to various frequency characteristics to form an energy characteristic vector corresponding to the vibration data of the fan bearing in a preset time period.

[0077] Furthermore, the first determining unit is specifically configured to:

[0078] The energy characteristic corresponding to the outer ring ball pass frequency characteristic of the vibration data at the preset time t is determined by the following formula:

[0079]

[0080] In the formula, is the amplitude corresponding to the outer ring ball pass frequency of the vibration data at the preset time t, A F,t is the amplitude of the frequency spectrum of the vibration data at the preset time t after fast Fourier transformation, and F is the total number of types of frequency features.

[0081] It should be noted that the training process of the pre-trained health index prediction model includes:

[0082] Obtain energy feature time series data corresponding to various frequency features after feature extraction of the fan bearing vibration data at each moment in the historical period and health index values ​​corresponding to the fan bearing vibration data at each moment in the historical period;

[0083] The energy characteristic time series data corresponding to each type of frequency characteristics are smoothed by using an exponential smoothing method to obtain smoothed energy characteristics corresponding to each type of frequency characteristics;

[0084] The energy features corresponding to the smoothed various frequency features and the time correspond to form a time series, and the time in the time series corresponding to the various frequency features is aligned to form an energy feature matrix corresponding to the various frequency features in the historical period;

[0085] Performing dimension reduction processing on the energy characteristic matrix by using the main program analysis method, obtaining the eigenvalues ​​after dimension reduction and forming an input eigenvector based on the energy eigenvalues;

[0086] The input feature vector is used as the input of the wavelet neural network, and the health index value corresponding to the input feature vector is used as the output of the wavelet neural network. The wavelet neural network is trained to obtain a trained health index prediction model.

[0087] In the embodiment of the present disclosure, the determining module includes:

[0088] A linear regression unit, used to perform linear regression on the health index value to obtain the fault time of the fan bearing corresponding to the health index value;

[0089] The second determining unit is used to determine the remaining service life of the fan bearing based on a preset time and the fault time.

[0090] Example 3

[0091] In order to implement the above embodiments, the present disclosure also provides a computer device.

[0092] The computer device provided in this embodiment includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method in Embodiment 1 is implemented.

[0093] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0094] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0095] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A method for predicting the remaining service life of a fan bearing, characterized in that: The method comprises: Obtain the fan bearing vibration data at a preset time; Perform feature extraction on the data to obtain energy feature vectors corresponding to various frequency features of the vibration data within a preset time period; Inputting energy feature vectors corresponding to various frequency features of the vibration data into a pre-trained health index prediction model to obtain a health index value of the fan bearing at a preset time; determining a remaining service life of the fan bearing based on the health indicator value; Among them, the various frequency characteristics include: outer circle ball pass frequency, inner circle ball pass frequency, ball failure frequency, basic training frequency and shaft rotation basic frequency; The training process of the pre-trained health index prediction model includes: Obtain energy feature time series data corresponding to various frequency features after feature extraction of the fan bearing vibration data at each moment in the historical period and health index values ​​corresponding to the fan bearing vibration data at each moment in the historical period; The energy characteristic time series data corresponding to each type of frequency characteristics are smoothed by using an exponential smoothing method to obtain smoothed energy characteristics corresponding to each type of frequency characteristics; The energy features corresponding to the smoothed various frequency features and the time correspond to form a time series, and the time in the time series corresponding to the various frequency features is aligned to form an energy feature matrix corresponding to the various frequency features in the historical period; Performing dimensionality reduction processing on the energy feature matrix by using principal component analysis, obtaining eigenvalues ​​after dimensionality reduction and forming an input eigenvector based on the energy eigenvalues; The input feature vector is used as the input of the wavelet neural network, and the health index value corresponding to the input feature vector is used as the output of the wavelet neural network. The wavelet neural network is trained to obtain a trained health index prediction model.

2. The method according to claim 1, characterized in that The step of extracting features from the data to obtain energy feature vectors corresponding to various frequency features of the vibration data within a preset time period includes: Performing a fast Fourier transform on the data to obtain a frequency spectrum corresponding to the data; Find the amplitudes corresponding to various frequency features in the frequency spectrum corresponding to the data; Determining energy features corresponding to various frequency features based on the amplitudes corresponding to the various frequency features; The energy characteristics corresponding to various frequency characteristics are used to form an energy characteristic vector corresponding to the vibration data of the fan bearing at a preset time.

3. The method according to claim 2, characterized in that The determining the energy features corresponding to the various frequency features based on the amplitudes corresponding to the various frequency features includes: The energy characteristic corresponding to the outer ring ball pass frequency characteristic of the vibration data at the preset time t is determined by the following formula: : In the formula, is the amplitude corresponding to the outer ring ball pass frequency of the vibration data at the preset time t, is the amplitude of the frequency spectrum of the vibration data at the preset time t after fast Fourier transformation, is the total number of types of frequency features.

4. The method according to claim 1, characterized in that The determining the remaining service life of the fan bearing based on the health indicator value includes: Performing linear regression on the health index value to obtain the fault time of the fan bearing corresponding to the health index value; The remaining service life of the fan bearing is determined based on a preset time and the fault time.

5. A system for predicting the remaining useful life of a fan bearing, characterized in that: The system comprises: A first acquisition module, used to acquire the vibration data of the fan bearing at a preset time; A second acquisition module is used to extract features from the data to obtain energy feature vectors corresponding to various frequency features of the vibration data within a preset time period; A third acquisition module is used to input the energy feature vectors corresponding to various frequency features of the vibration data into a pre-trained health index prediction model to obtain the health index value of the fan bearing at a preset time; A determination module, configured to determine the remaining service life of the fan bearing based on the health indicator value; Among them, the various frequency characteristics include: outer circle ball pass frequency, inner circle ball pass frequency, ball failure frequency, basic training frequency and shaft rotation basic frequency; The training process of the pre-trained health index prediction model includes: Obtain energy feature time series data corresponding to various frequency features after feature extraction of the fan bearing vibration data at each moment in the historical period and health index values ​​corresponding to the fan bearing vibration data at each moment in the historical period; The energy characteristic time series data corresponding to each type of frequency characteristics are smoothed by using an exponential smoothing method to obtain smoothed energy characteristics corresponding to each type of frequency characteristics; The energy features corresponding to the smoothed various frequency features and the time correspond to form a time series, and the time in the time series corresponding to the various frequency features is aligned to form an energy feature matrix corresponding to the various frequency features in the historical period; Performing dimensionality reduction processing on the energy feature matrix by using principal component analysis, obtaining eigenvalues ​​after dimensionality reduction and forming an input eigenvector based on the energy eigenvalues; The input feature vector is used as the input of the wavelet neural network, and the health index value corresponding to the input feature vector is used as the output of the wavelet neural network. The wavelet neural network is trained to obtain a trained health index prediction model.

6. The system according to claim 5, characterized in that The second acquisition module includes: A transform unit, used for performing a fast Fourier transform on the data to obtain a frequency spectrum corresponding to the data; A search unit, used to search for amplitudes corresponding to various frequency features in a frequency spectrum corresponding to the data; A first determining unit, configured to determine energy features corresponding to various frequency features based on amplitudes corresponding to the various frequency features; The composition unit is used to use the energy characteristics corresponding to various frequency characteristics to form an energy characteristic vector corresponding to the vibration data of the fan bearing at a preset time.

7. The system according to claim 5, characterized in that The determining module comprises: A linear regression unit, used to perform linear regression on the health index value to obtain the fault time of the fan bearing corresponding to the health index value; The second determining unit is used to determine the remaining service life of the fan bearing based on a preset time and the fault time.

8. A computer device, characterized in that: The method comprises a memory, a processor and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the method according to any one of claims 1 to 4 is implemented.

Images